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/10.0/sys/ufs/ufs/ufs_vnops.c 254627 2013-08-21 23:04:48Z ken $");
   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 *);
  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);
  208         if (error)
  209                 return (error);
  210         return (0);
  211 }
  212 
  213 /*
  214  * Mknod vnode call
  215  */
  216 /* ARGSUSED */
  217 static int
  218 ufs_mknod(ap)
  219         struct vop_mknod_args /* {
  220                 struct vnode *a_dvp;
  221                 struct vnode **a_vpp;
  222                 struct componentname *a_cnp;
  223                 struct vattr *a_vap;
  224         } */ *ap;
  225 {
  226         struct vattr *vap = ap->a_vap;
  227         struct vnode **vpp = ap->a_vpp;
  228         struct inode *ip;
  229         ino_t ino;
  230         int error;
  231 
  232         error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
  233             ap->a_dvp, vpp, ap->a_cnp);
  234         if (error)
  235                 return (error);
  236         ip = VTOI(*vpp);
  237         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
  238         if (vap->va_rdev != VNOVAL) {
  239                 /*
  240                  * Want to be able to use this to make badblock
  241                  * inodes, so don't truncate the dev number.
  242                  */
  243                 DIP_SET(ip, i_rdev, vap->va_rdev);
  244         }
  245         /*
  246          * Remove inode, then reload it through VFS_VGET so it is
  247          * checked to see if it is an alias of an existing entry in
  248          * the inode cache.  XXX I don't believe this is necessary now.
  249          */
  250         (*vpp)->v_type = VNON;
  251         ino = ip->i_number;     /* Save this before vgone() invalidates ip. */
  252         vgone(*vpp);
  253         vput(*vpp);
  254         error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
  255         if (error) {
  256                 *vpp = NULL;
  257                 return (error);
  258         }
  259         return (0);
  260 }
  261 
  262 /*
  263  * Open called.
  264  */
  265 /* ARGSUSED */
  266 static int
  267 ufs_open(struct vop_open_args *ap)
  268 {
  269         struct vnode *vp = ap->a_vp;
  270         struct inode *ip;
  271 
  272         if (vp->v_type == VCHR || vp->v_type == VBLK)
  273                 return (EOPNOTSUPP);
  274 
  275         ip = VTOI(vp);
  276         /*
  277          * Files marked append-only must be opened for appending.
  278          */
  279         if ((ip->i_flags & APPEND) &&
  280             (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
  281                 return (EPERM);
  282         vnode_create_vobject(vp, DIP(ip, i_size), ap->a_td);
  283         return (0);
  284 }
  285 
  286 /*
  287  * Close called.
  288  *
  289  * Update the times on the inode.
  290  */
  291 /* ARGSUSED */
  292 static int
  293 ufs_close(ap)
  294         struct vop_close_args /* {
  295                 struct vnode *a_vp;
  296                 int  a_fflag;
  297                 struct ucred *a_cred;
  298                 struct thread *a_td;
  299         } */ *ap;
  300 {
  301         struct vnode *vp = ap->a_vp;
  302         int usecount;
  303 
  304         VI_LOCK(vp);
  305         usecount = vp->v_usecount;
  306         if (usecount > 1)
  307                 ufs_itimes_locked(vp);
  308         VI_UNLOCK(vp);
  309         return (0);
  310 }
  311 
  312 static int
  313 ufs_accessx(ap)
  314         struct vop_accessx_args /* {
  315                 struct vnode *a_vp;
  316                 accmode_t a_accmode;
  317                 struct ucred *a_cred;
  318                 struct thread *a_td;
  319         } */ *ap;
  320 {
  321         struct vnode *vp = ap->a_vp;
  322         struct inode *ip = VTOI(vp);
  323         accmode_t accmode = ap->a_accmode;
  324         int error;
  325 #ifdef QUOTA
  326         int relocked;
  327 #endif
  328 #ifdef UFS_ACL
  329         struct acl *acl;
  330         acl_type_t type;
  331 #endif
  332 
  333         /*
  334          * Disallow write attempts on read-only filesystems;
  335          * unless the file is a socket, fifo, or a block or
  336          * character device resident on the filesystem.
  337          */
  338         if (accmode & VMODIFY_PERMS) {
  339                 switch (vp->v_type) {
  340                 case VDIR:
  341                 case VLNK:
  342                 case VREG:
  343                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
  344                                 return (EROFS);
  345 #ifdef QUOTA
  346                         /*
  347                          * Inode is accounted in the quotas only if struct
  348                          * dquot is attached to it. VOP_ACCESS() is called
  349                          * from vn_open_cred() and provides a convenient
  350                          * point to call getinoquota().
  351                          */
  352                         if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
  353 
  354                                 /*
  355                                  * Upgrade vnode lock, since getinoquota()
  356                                  * requires exclusive lock to modify inode.
  357                                  */
  358                                 relocked = 1;
  359                                 vhold(vp);
  360                                 vn_lock(vp, LK_UPGRADE | LK_RETRY);
  361                                 VI_LOCK(vp);
  362                                 if (vp->v_iflag & VI_DOOMED) {
  363                                         vdropl(vp);
  364                                         error = ENOENT;
  365                                         goto relock;
  366                                 }
  367                                 vdropl(vp);
  368                         } else
  369                                 relocked = 0;
  370                         error = getinoquota(ip);
  371 relock:
  372                         if (relocked)
  373                                 vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
  374                         if (error != 0)
  375                                 return (error);
  376 #endif
  377                         break;
  378                 default:
  379                         break;
  380                 }
  381         }
  382 
  383         /*
  384          * If immutable bit set, nobody gets to write it.  "& ~VADMIN_PERMS"
  385          * is here, because without it, * it would be impossible for the owner
  386          * to remove the IMMUTABLE flag.
  387          */
  388         if ((accmode & (VMODIFY_PERMS & ~VADMIN_PERMS)) &&
  389             (ip->i_flags & (IMMUTABLE | SF_SNAPSHOT)))
  390                 return (EPERM);
  391 
  392 #ifdef UFS_ACL
  393         if ((vp->v_mount->mnt_flag & (MNT_ACLS | MNT_NFS4ACLS)) != 0) {
  394                 if (vp->v_mount->mnt_flag & MNT_NFS4ACLS)
  395                         type = ACL_TYPE_NFS4;
  396                 else
  397                         type = ACL_TYPE_ACCESS;
  398 
  399                 acl = acl_alloc(M_WAITOK);
  400                 if (type == ACL_TYPE_NFS4)
  401                         error = ufs_getacl_nfs4_internal(vp, acl, ap->a_td);
  402                 else
  403                         error = VOP_GETACL(vp, type, acl, ap->a_cred, ap->a_td);
  404                 switch (error) {
  405                 case 0:
  406                         if (type == ACL_TYPE_NFS4) {
  407                                 error = vaccess_acl_nfs4(vp->v_type, ip->i_uid,
  408                                     ip->i_gid, acl, accmode, ap->a_cred, NULL);
  409                         } else {
  410                                 error = vfs_unixify_accmode(&accmode);
  411                                 if (error == 0)
  412                                         error = vaccess_acl_posix1e(vp->v_type, ip->i_uid,
  413                                             ip->i_gid, acl, accmode, ap->a_cred, NULL);
  414                         }
  415                         break;
  416                 default:
  417                         if (error != EOPNOTSUPP)
  418                                 printf(
  419 "ufs_accessx(): Error retrieving ACL on object (%d).\n",
  420                                     error);
  421                         /*
  422                          * XXX: Fall back until debugged.  Should
  423                          * eventually possibly log an error, and return
  424                          * EPERM for safety.
  425                          */
  426                         error = vfs_unixify_accmode(&accmode);
  427                         if (error == 0)
  428                                 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid,
  429                                     ip->i_gid, accmode, ap->a_cred, NULL);
  430                 }
  431                 acl_free(acl);
  432 
  433                 return (error);
  434         }
  435 #endif /* !UFS_ACL */
  436         error = vfs_unixify_accmode(&accmode);
  437         if (error == 0)
  438                 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
  439                     accmode, ap->a_cred, NULL);
  440         return (error);
  441 }
  442 
  443 /* ARGSUSED */
  444 static int
  445 ufs_getattr(ap)
  446         struct vop_getattr_args /* {
  447                 struct vnode *a_vp;
  448                 struct vattr *a_vap;
  449                 struct ucred *a_cred;
  450         } */ *ap;
  451 {
  452         struct vnode *vp = ap->a_vp;
  453         struct inode *ip = VTOI(vp);
  454         struct vattr *vap = ap->a_vap;
  455 
  456         VI_LOCK(vp);
  457         ufs_itimes_locked(vp);
  458         if (ip->i_ump->um_fstype == UFS1) {
  459                 vap->va_atime.tv_sec = ip->i_din1->di_atime;
  460                 vap->va_atime.tv_nsec = ip->i_din1->di_atimensec;
  461         } else {
  462                 vap->va_atime.tv_sec = ip->i_din2->di_atime;
  463                 vap->va_atime.tv_nsec = ip->i_din2->di_atimensec;
  464         }
  465         VI_UNLOCK(vp);
  466         /*
  467          * Copy from inode table
  468          */
  469         vap->va_fsid = dev2udev(ip->i_dev);
  470         vap->va_fileid = ip->i_number;
  471         vap->va_mode = ip->i_mode & ~IFMT;
  472         vap->va_nlink = ip->i_effnlink;
  473         vap->va_uid = ip->i_uid;
  474         vap->va_gid = ip->i_gid;
  475         if (ip->i_ump->um_fstype == UFS1) {
  476                 vap->va_rdev = ip->i_din1->di_rdev;
  477                 vap->va_size = ip->i_din1->di_size;
  478                 vap->va_mtime.tv_sec = ip->i_din1->di_mtime;
  479                 vap->va_mtime.tv_nsec = ip->i_din1->di_mtimensec;
  480                 vap->va_ctime.tv_sec = ip->i_din1->di_ctime;
  481                 vap->va_ctime.tv_nsec = ip->i_din1->di_ctimensec;
  482                 vap->va_bytes = dbtob((u_quad_t)ip->i_din1->di_blocks);
  483                 vap->va_filerev = ip->i_din1->di_modrev;
  484         } else {
  485                 vap->va_rdev = ip->i_din2->di_rdev;
  486                 vap->va_size = ip->i_din2->di_size;
  487                 vap->va_mtime.tv_sec = ip->i_din2->di_mtime;
  488                 vap->va_mtime.tv_nsec = ip->i_din2->di_mtimensec;
  489                 vap->va_ctime.tv_sec = ip->i_din2->di_ctime;
  490                 vap->va_ctime.tv_nsec = ip->i_din2->di_ctimensec;
  491                 vap->va_birthtime.tv_sec = ip->i_din2->di_birthtime;
  492                 vap->va_birthtime.tv_nsec = ip->i_din2->di_birthnsec;
  493                 vap->va_bytes = dbtob((u_quad_t)ip->i_din2->di_blocks);
  494                 vap->va_filerev = ip->i_din2->di_modrev;
  495         }
  496         vap->va_flags = ip->i_flags;
  497         vap->va_gen = ip->i_gen;
  498         vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
  499         vap->va_type = IFTOVT(ip->i_mode);
  500         return (0);
  501 }
  502 
  503 /*
  504  * Set attribute vnode op. called from several syscalls
  505  */
  506 static int
  507 ufs_setattr(ap)
  508         struct vop_setattr_args /* {
  509                 struct vnode *a_vp;
  510                 struct vattr *a_vap;
  511                 struct ucred *a_cred;
  512         } */ *ap;
  513 {
  514         struct vattr *vap = ap->a_vap;
  515         struct vnode *vp = ap->a_vp;
  516         struct inode *ip = VTOI(vp);
  517         struct ucred *cred = ap->a_cred;
  518         struct thread *td = curthread;
  519         int error;
  520 
  521         /*
  522          * Check for unsettable attributes.
  523          */
  524         if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
  525             (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
  526             (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
  527             ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
  528                 return (EINVAL);
  529         }
  530         if (vap->va_flags != VNOVAL) {
  531                 if ((vap->va_flags & ~(SF_APPEND | SF_ARCHIVED | SF_IMMUTABLE |
  532                     SF_NOUNLINK | SF_SNAPSHOT | UF_APPEND | UF_ARCHIVE |
  533                     UF_HIDDEN | UF_IMMUTABLE | UF_NODUMP | UF_NOUNLINK |
  534                     UF_OFFLINE | UF_OPAQUE | UF_READONLY | UF_REPARSE |
  535                     UF_SPARSE | UF_SYSTEM)) != 0)
  536                         return (EOPNOTSUPP);
  537                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  538                         return (EROFS);
  539                 /*
  540                  * Callers may only modify the file flags on objects they
  541                  * have VADMIN rights for.
  542                  */
  543                 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
  544                         return (error);
  545                 /*
  546                  * Unprivileged processes are not permitted to unset system
  547                  * flags, or modify flags if any system flags are set.
  548                  * Privileged non-jail processes may not modify system flags
  549                  * if securelevel > 0 and any existing system flags are set.
  550                  * Privileged jail processes behave like privileged non-jail
  551                  * processes if the security.jail.chflags_allowed sysctl is
  552                  * is non-zero; otherwise, they behave like unprivileged
  553                  * processes.
  554                  */
  555                 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
  556                         if (ip->i_flags &
  557                             (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) {
  558                                 error = securelevel_gt(cred, 0);
  559                                 if (error)
  560                                         return (error);
  561                         }
  562                         /* The snapshot flag cannot be toggled. */
  563                         if ((vap->va_flags ^ ip->i_flags) & SF_SNAPSHOT)
  564                                 return (EPERM);
  565                 } else {
  566                         if (ip->i_flags &
  567                             (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
  568                             ((vap->va_flags ^ ip->i_flags) & SF_SETTABLE))
  569                                 return (EPERM);
  570                 }
  571                 ip->i_flags = vap->va_flags;
  572                 DIP_SET(ip, i_flags, vap->va_flags);
  573                 ip->i_flag |= IN_CHANGE;
  574                 error = UFS_UPDATE(vp, 0);
  575                 if (ip->i_flags & (IMMUTABLE | APPEND))
  576                         return (error);
  577         }
  578         /*
  579          * If immutable or append, no one can change any of its attributes
  580          * except the ones already handled (in some cases, file flags
  581          * including the immutability flags themselves for the superuser).
  582          */
  583         if (ip->i_flags & (IMMUTABLE | APPEND))
  584                 return (EPERM);
  585         /*
  586          * Go through the fields and update iff not VNOVAL.
  587          */
  588         if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
  589                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  590                         return (EROFS);
  591                 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred,
  592                     td)) != 0)
  593                         return (error);
  594         }
  595         if (vap->va_size != VNOVAL) {
  596                 /*
  597                  * XXX most of the following special cases should be in
  598                  * callers instead of in N filesystems.  The VDIR check
  599                  * mostly already is.
  600                  */
  601                 switch (vp->v_type) {
  602                 case VDIR:
  603                         return (EISDIR);
  604                 case VLNK:
  605                 case VREG:
  606                         /*
  607                          * Truncation should have an effect in these cases.
  608                          * Disallow it if the filesystem is read-only or
  609                          * the file is being snapshotted.
  610                          */
  611                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
  612                                 return (EROFS);
  613                         if ((ip->i_flags & SF_SNAPSHOT) != 0)
  614                                 return (EPERM);
  615                         break;
  616                 default:
  617                         /*
  618                          * According to POSIX, the result is unspecified
  619                          * for file types other than regular files,
  620                          * directories and shared memory objects.  We
  621                          * don't support shared memory objects in the file
  622                          * system, and have dubious support for truncating
  623                          * symlinks.  Just ignore the request in other cases.
  624                          */
  625                         return (0);
  626                 }
  627                 if ((error = UFS_TRUNCATE(vp, vap->va_size, IO_NORMAL,
  628                     cred)) != 0)
  629                         return (error);
  630         }
  631         if (vap->va_atime.tv_sec != VNOVAL ||
  632             vap->va_mtime.tv_sec != VNOVAL ||
  633             vap->va_birthtime.tv_sec != VNOVAL) {
  634                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  635                         return (EROFS);
  636                 if ((ip->i_flags & SF_SNAPSHOT) != 0)
  637                         return (EPERM);
  638                 /*
  639                  * From utimes(2):
  640                  * If times is NULL, ... The caller must be the owner of
  641                  * the file, have permission to write the file, or be the
  642                  * super-user.
  643                  * If times is non-NULL, ... The caller must be the owner of
  644                  * the file or be the super-user.
  645                  *
  646                  * Possibly for historical reasons, try to use VADMIN in
  647                  * preference to VWRITE for a NULL timestamp.  This means we
  648                  * will return EACCES in preference to EPERM if neither
  649                  * check succeeds.
  650                  */
  651                 if (vap->va_vaflags & VA_UTIMES_NULL) {
  652                         /*
  653                          * NFSv4.1, draft 21, 6.2.1.3.1, Discussion of Mask Attributes
  654                          *
  655                          * "A user having ACL_WRITE_DATA or ACL_WRITE_ATTRIBUTES
  656                          * will be allowed to set the times [..] to the current
  657                          * server time."
  658                          *
  659                          * XXX: Calling it four times seems a little excessive.
  660                          */
  661                         error = VOP_ACCESSX(vp, VWRITE_ATTRIBUTES, cred, td);
  662                         if (error)
  663                                 error = VOP_ACCESS(vp, VWRITE, cred, td);
  664                 } else
  665                         error = VOP_ACCESSX(vp, VWRITE_ATTRIBUTES, cred, td);
  666                 if (error)
  667                         return (error);
  668                 if (vap->va_atime.tv_sec != VNOVAL)
  669                         ip->i_flag |= IN_ACCESS;
  670                 if (vap->va_mtime.tv_sec != VNOVAL)
  671                         ip->i_flag |= IN_CHANGE | IN_UPDATE;
  672                 if (vap->va_birthtime.tv_sec != VNOVAL &&
  673                     ip->i_ump->um_fstype == UFS2)
  674                         ip->i_flag |= IN_MODIFIED;
  675                 ufs_itimes(vp);
  676                 if (vap->va_atime.tv_sec != VNOVAL) {
  677                         DIP_SET(ip, i_atime, vap->va_atime.tv_sec);
  678                         DIP_SET(ip, i_atimensec, vap->va_atime.tv_nsec);
  679                 }
  680                 if (vap->va_mtime.tv_sec != VNOVAL) {
  681                         DIP_SET(ip, i_mtime, vap->va_mtime.tv_sec);
  682                         DIP_SET(ip, i_mtimensec, vap->va_mtime.tv_nsec);
  683                 }
  684                 if (vap->va_birthtime.tv_sec != VNOVAL &&
  685                     ip->i_ump->um_fstype == UFS2) {
  686                         ip->i_din2->di_birthtime = vap->va_birthtime.tv_sec;
  687                         ip->i_din2->di_birthnsec = vap->va_birthtime.tv_nsec;
  688                 }
  689                 error = UFS_UPDATE(vp, 0);
  690                 if (error)
  691                         return (error);
  692         }
  693         error = 0;
  694         if (vap->va_mode != (mode_t)VNOVAL) {
  695                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  696                         return (EROFS);
  697                 if ((ip->i_flags & SF_SNAPSHOT) != 0 && (vap->va_mode &
  698                    (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP | S_IXOTH | S_IWOTH)))
  699                         return (EPERM);
  700                 error = ufs_chmod(vp, (int)vap->va_mode, cred, td);
  701         }
  702         return (error);
  703 }
  704 
  705 #ifdef UFS_ACL
  706 static int
  707 ufs_update_nfs4_acl_after_mode_change(struct vnode *vp, int mode,
  708     int file_owner_id, struct ucred *cred, struct thread *td)
  709 {
  710         int error;
  711         struct acl *aclp;
  712 
  713         aclp = acl_alloc(M_WAITOK);
  714         error = ufs_getacl_nfs4_internal(vp, aclp, td);
  715         /*
  716          * We don't have to handle EOPNOTSUPP here, as the filesystem claims
  717          * it supports ACLs.
  718          */
  719         if (error)
  720                 goto out;
  721 
  722         acl_nfs4_sync_acl_from_mode(aclp, mode, file_owner_id);
  723         error = ufs_setacl_nfs4_internal(vp, aclp, td);
  724 
  725 out:
  726         acl_free(aclp);
  727         return (error);
  728 }
  729 #endif /* UFS_ACL */
  730 
  731 /*
  732  * Mark this file's access time for update for vfs_mark_atime().  This
  733  * is called from execve() and mmap().
  734  */
  735 static int
  736 ufs_markatime(ap)
  737         struct vop_markatime_args /* {
  738                 struct vnode *a_vp;
  739         } */ *ap;
  740 {
  741         struct vnode *vp = ap->a_vp;
  742         struct inode *ip = VTOI(vp);
  743 
  744         VI_LOCK(vp);
  745         ip->i_flag |= IN_ACCESS;
  746         VI_UNLOCK(vp);
  747         /*
  748          * XXXKIB No UFS_UPDATE(ap->a_vp, 0) there.
  749          */
  750         return (0);
  751 }
  752 
  753 /*
  754  * Change the mode on a file.
  755  * Inode must be locked before calling.
  756  */
  757 static int
  758 ufs_chmod(vp, mode, cred, td)
  759         struct vnode *vp;
  760         int mode;
  761         struct ucred *cred;
  762         struct thread *td;
  763 {
  764         struct inode *ip = VTOI(vp);
  765         int error;
  766 
  767         /*
  768          * To modify the permissions on a file, must possess VADMIN
  769          * for that file.
  770          */
  771         if ((error = VOP_ACCESSX(vp, VWRITE_ACL, cred, td)))
  772                 return (error);
  773         /*
  774          * Privileged processes may set the sticky bit on non-directories,
  775          * as well as set the setgid bit on a file with a group that the
  776          * process is not a member of.  Both of these are allowed in
  777          * jail(8).
  778          */
  779         if (vp->v_type != VDIR && (mode & S_ISTXT)) {
  780                 if (priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0))
  781                         return (EFTYPE);
  782         }
  783         if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
  784                 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0);
  785                 if (error)
  786                         return (error);
  787         }
  788 
  789         /*
  790          * Deny setting setuid if we are not the file owner.
  791          */
  792         if ((mode & ISUID) && ip->i_uid != cred->cr_uid) {
  793                 error = priv_check_cred(cred, PRIV_VFS_ADMIN, 0);
  794                 if (error)
  795                         return (error);
  796         }
  797 
  798         ip->i_mode &= ~ALLPERMS;
  799         ip->i_mode |= (mode & ALLPERMS);
  800         DIP_SET(ip, i_mode, ip->i_mode);
  801         ip->i_flag |= IN_CHANGE;
  802 #ifdef UFS_ACL
  803         if ((vp->v_mount->mnt_flag & MNT_NFS4ACLS) != 0)
  804                 error = ufs_update_nfs4_acl_after_mode_change(vp, mode, ip->i_uid, cred, td);
  805 #endif
  806         if (error == 0 && (ip->i_flag & IN_CHANGE) != 0)
  807                 error = UFS_UPDATE(vp, 0);
  808 
  809         return (error);
  810 }
  811 
  812 /*
  813  * Perform chown operation on inode ip;
  814  * inode must be locked prior to call.
  815  */
  816 static int
  817 ufs_chown(vp, uid, gid, cred, td)
  818         struct vnode *vp;
  819         uid_t uid;
  820         gid_t gid;
  821         struct ucred *cred;
  822         struct thread *td;
  823 {
  824         struct inode *ip = VTOI(vp);
  825         uid_t ouid;
  826         gid_t ogid;
  827         int error = 0;
  828 #ifdef QUOTA
  829         int i;
  830         ufs2_daddr_t change;
  831 #endif
  832 
  833         if (uid == (uid_t)VNOVAL)
  834                 uid = ip->i_uid;
  835         if (gid == (gid_t)VNOVAL)
  836                 gid = ip->i_gid;
  837         /*
  838          * To modify the ownership of a file, must possess VADMIN for that
  839          * file.
  840          */
  841         if ((error = VOP_ACCESSX(vp, VWRITE_OWNER, cred, td)))
  842                 return (error);
  843         /*
  844          * To change the owner of a file, or change the group of a file to a
  845          * group of which we are not a member, the caller must have
  846          * privilege.
  847          */
  848         if (((uid != ip->i_uid && uid != cred->cr_uid) || 
  849             (gid != ip->i_gid && !groupmember(gid, cred))) &&
  850             (error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0)))
  851                 return (error);
  852         ogid = ip->i_gid;
  853         ouid = ip->i_uid;
  854 #ifdef QUOTA
  855         if ((error = getinoquota(ip)) != 0)
  856                 return (error);
  857         if (ouid == uid) {
  858                 dqrele(vp, ip->i_dquot[USRQUOTA]);
  859                 ip->i_dquot[USRQUOTA] = NODQUOT;
  860         }
  861         if (ogid == gid) {
  862                 dqrele(vp, ip->i_dquot[GRPQUOTA]);
  863                 ip->i_dquot[GRPQUOTA] = NODQUOT;
  864         }
  865         change = DIP(ip, i_blocks);
  866         (void) chkdq(ip, -change, cred, CHOWN);
  867         (void) chkiq(ip, -1, cred, CHOWN);
  868         for (i = 0; i < MAXQUOTAS; i++) {
  869                 dqrele(vp, ip->i_dquot[i]);
  870                 ip->i_dquot[i] = NODQUOT;
  871         }
  872 #endif
  873         ip->i_gid = gid;
  874         DIP_SET(ip, i_gid, gid);
  875         ip->i_uid = uid;
  876         DIP_SET(ip, i_uid, uid);
  877 #ifdef QUOTA
  878         if ((error = getinoquota(ip)) == 0) {
  879                 if (ouid == uid) {
  880                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  881                         ip->i_dquot[USRQUOTA] = NODQUOT;
  882                 }
  883                 if (ogid == gid) {
  884                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  885                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  886                 }
  887                 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
  888                         if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
  889                                 goto good;
  890                         else
  891                                 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
  892                 }
  893                 for (i = 0; i < MAXQUOTAS; i++) {
  894                         dqrele(vp, ip->i_dquot[i]);
  895                         ip->i_dquot[i] = NODQUOT;
  896                 }
  897         }
  898         ip->i_gid = ogid;
  899         DIP_SET(ip, i_gid, ogid);
  900         ip->i_uid = ouid;
  901         DIP_SET(ip, i_uid, ouid);
  902         if (getinoquota(ip) == 0) {
  903                 if (ouid == uid) {
  904                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  905                         ip->i_dquot[USRQUOTA] = NODQUOT;
  906                 }
  907                 if (ogid == gid) {
  908                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  909                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  910                 }
  911                 (void) chkdq(ip, change, cred, FORCE|CHOWN);
  912                 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
  913                 (void) getinoquota(ip);
  914         }
  915         return (error);
  916 good:
  917         if (getinoquota(ip))
  918                 panic("ufs_chown: lost quota");
  919 #endif /* QUOTA */
  920         ip->i_flag |= IN_CHANGE;
  921         if ((ip->i_mode & (ISUID | ISGID)) && (ouid != uid || ogid != gid)) {
  922                 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0)) {
  923                         ip->i_mode &= ~(ISUID | ISGID);
  924                         DIP_SET(ip, i_mode, ip->i_mode);
  925                 }
  926         }
  927         error = UFS_UPDATE(vp, 0);
  928         return (error);
  929 }
  930 
  931 static int
  932 ufs_remove(ap)
  933         struct vop_remove_args /* {
  934                 struct vnode *a_dvp;
  935                 struct vnode *a_vp;
  936                 struct componentname *a_cnp;
  937         } */ *ap;
  938 {
  939         struct inode *ip;
  940         struct vnode *vp = ap->a_vp;
  941         struct vnode *dvp = ap->a_dvp;
  942         int error;
  943         struct thread *td;
  944 
  945         td = curthread;
  946         ip = VTOI(vp);
  947         if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
  948             (VTOI(dvp)->i_flags & APPEND)) {
  949                 error = EPERM;
  950                 goto out;
  951         }
  952 #ifdef UFS_GJOURNAL
  953         ufs_gjournal_orphan(vp);
  954 #endif
  955         error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
  956         if (ip->i_nlink <= 0)
  957                 vp->v_vflag |= VV_NOSYNC;
  958         if ((ip->i_flags & SF_SNAPSHOT) != 0) {
  959                 /*
  960                  * Avoid deadlock where another thread is trying to
  961                  * update the inodeblock for dvp and is waiting on
  962                  * snaplk.  Temporary unlock the vnode lock for the
  963                  * unlinked file and sync the directory.  This should
  964                  * allow vput() of the directory to not block later on
  965                  * while holding the snapshot vnode locked, assuming
  966                  * that the directory hasn't been unlinked too.
  967                  */
  968                 VOP_UNLOCK(vp, 0);
  969                 (void) VOP_FSYNC(dvp, MNT_WAIT, td);
  970                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  971         }
  972 out:
  973         return (error);
  974 }
  975 
  976 /*
  977  * link vnode call
  978  */
  979 static int
  980 ufs_link(ap)
  981         struct vop_link_args /* {
  982                 struct vnode *a_tdvp;
  983                 struct vnode *a_vp;
  984                 struct componentname *a_cnp;
  985         } */ *ap;
  986 {
  987         struct vnode *vp = ap->a_vp;
  988         struct vnode *tdvp = ap->a_tdvp;
  989         struct componentname *cnp = ap->a_cnp;
  990         struct inode *ip;
  991         struct direct newdir;
  992         int error;
  993 
  994 #ifdef INVARIANTS
  995         if ((cnp->cn_flags & HASBUF) == 0)
  996                 panic("ufs_link: no name");
  997 #endif
  998         if (tdvp->v_mount != vp->v_mount) {
  999                 error = EXDEV;
 1000                 goto out;
 1001         }
 1002         if (VTOI(tdvp)->i_effnlink < 2)
 1003                 panic("ufs_link: Bad link count %d on parent",
 1004                     VTOI(tdvp)->i_effnlink);
 1005         ip = VTOI(vp);
 1006         if ((nlink_t)ip->i_nlink >= LINK_MAX) {
 1007                 error = EMLINK;
 1008                 goto out;
 1009         }
 1010         /*
 1011          * The file may have been removed after namei droped the original
 1012          * lock.
 1013          */
 1014         if (ip->i_effnlink == 0) {
 1015                 error = ENOENT;
 1016                 goto out;
 1017         }
 1018         if (ip->i_flags & (IMMUTABLE | APPEND)) {
 1019                 error = EPERM;
 1020                 goto out;
 1021         }
 1022         ip->i_effnlink++;
 1023         ip->i_nlink++;
 1024         DIP_SET(ip, i_nlink, ip->i_nlink);
 1025         ip->i_flag |= IN_CHANGE;
 1026         if (DOINGSOFTDEP(vp))
 1027                 softdep_setup_link(VTOI(tdvp), ip);
 1028         error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
 1029         if (!error) {
 1030                 ufs_makedirentry(ip, cnp, &newdir);
 1031                 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL, 0);
 1032         }
 1033 
 1034         if (error) {
 1035                 ip->i_effnlink--;
 1036                 ip->i_nlink--;
 1037                 DIP_SET(ip, i_nlink, ip->i_nlink);
 1038                 ip->i_flag |= IN_CHANGE;
 1039                 if (DOINGSOFTDEP(vp))
 1040                         softdep_revert_link(VTOI(tdvp), ip);
 1041         }
 1042 out:
 1043         return (error);
 1044 }
 1045 
 1046 /*
 1047  * whiteout vnode call
 1048  */
 1049 static int
 1050 ufs_whiteout(ap)
 1051         struct vop_whiteout_args /* {
 1052                 struct vnode *a_dvp;
 1053                 struct componentname *a_cnp;
 1054                 int a_flags;
 1055         } */ *ap;
 1056 {
 1057         struct vnode *dvp = ap->a_dvp;
 1058         struct componentname *cnp = ap->a_cnp;
 1059         struct direct newdir;
 1060         int error = 0;
 1061 
 1062         switch (ap->a_flags) {
 1063         case LOOKUP:
 1064                 /* 4.4 format directories support whiteout operations */
 1065                 if (dvp->v_mount->mnt_maxsymlinklen > 0)
 1066                         return (0);
 1067                 return (EOPNOTSUPP);
 1068 
 1069         case CREATE:
 1070                 /* create a new directory whiteout */
 1071 #ifdef INVARIANTS
 1072                 if ((cnp->cn_flags & SAVENAME) == 0)
 1073                         panic("ufs_whiteout: missing name");
 1074                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
 1075                         panic("ufs_whiteout: old format filesystem");
 1076 #endif
 1077 
 1078                 newdir.d_ino = WINO;
 1079                 newdir.d_namlen = cnp->cn_namelen;
 1080                 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
 1081                 newdir.d_type = DT_WHT;
 1082                 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL, 0);
 1083                 break;
 1084 
 1085         case DELETE:
 1086                 /* remove an existing directory whiteout */
 1087 #ifdef INVARIANTS
 1088                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
 1089                         panic("ufs_whiteout: old format filesystem");
 1090 #endif
 1091 
 1092                 cnp->cn_flags &= ~DOWHITEOUT;
 1093                 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
 1094                 break;
 1095         default:
 1096                 panic("ufs_whiteout: unknown op");
 1097         }
 1098         return (error);
 1099 }
 1100 
 1101 static volatile int rename_restarts;
 1102 SYSCTL_INT(_vfs_ufs, OID_AUTO, rename_restarts, CTLFLAG_RD,
 1103     __DEVOLATILE(int *, &rename_restarts), 0,
 1104     "Times rename had to restart due to lock contention");
 1105 
 1106 /*
 1107  * Rename system call.
 1108  *      rename("foo", "bar");
 1109  * is essentially
 1110  *      unlink("bar");
 1111  *      link("foo", "bar");
 1112  *      unlink("foo");
 1113  * but ``atomically''.  Can't do full commit without saving state in the
 1114  * inode on disk which isn't feasible at this time.  Best we can do is
 1115  * always guarantee the target exists.
 1116  *
 1117  * Basic algorithm is:
 1118  *
 1119  * 1) Bump link count on source while we're linking it to the
 1120  *    target.  This also ensure the inode won't be deleted out
 1121  *    from underneath us while we work (it may be truncated by
 1122  *    a concurrent `trunc' or `open' for creation).
 1123  * 2) Link source to destination.  If destination already exists,
 1124  *    delete it first.
 1125  * 3) Unlink source reference to inode if still around. If a
 1126  *    directory was moved and the parent of the destination
 1127  *    is different from the source, patch the ".." entry in the
 1128  *    directory.
 1129  */
 1130 static int
 1131 ufs_rename(ap)
 1132         struct vop_rename_args  /* {
 1133                 struct vnode *a_fdvp;
 1134                 struct vnode *a_fvp;
 1135                 struct componentname *a_fcnp;
 1136                 struct vnode *a_tdvp;
 1137                 struct vnode *a_tvp;
 1138                 struct componentname *a_tcnp;
 1139         } */ *ap;
 1140 {
 1141         struct vnode *tvp = ap->a_tvp;
 1142         struct vnode *tdvp = ap->a_tdvp;
 1143         struct vnode *fvp = ap->a_fvp;
 1144         struct vnode *fdvp = ap->a_fdvp;
 1145         struct vnode *nvp;
 1146         struct componentname *tcnp = ap->a_tcnp;
 1147         struct componentname *fcnp = ap->a_fcnp;
 1148         struct thread *td = fcnp->cn_thread;
 1149         struct inode *fip, *tip, *tdp, *fdp;
 1150         struct direct newdir;
 1151         off_t endoff;
 1152         int doingdirectory, newparent;
 1153         int error = 0;
 1154         struct mount *mp;
 1155         ino_t ino;
 1156 
 1157 #ifdef INVARIANTS
 1158         if ((tcnp->cn_flags & HASBUF) == 0 ||
 1159             (fcnp->cn_flags & HASBUF) == 0)
 1160                 panic("ufs_rename: no name");
 1161 #endif
 1162         endoff = 0;
 1163         mp = tdvp->v_mount;
 1164         VOP_UNLOCK(tdvp, 0);
 1165         if (tvp && tvp != tdvp)
 1166                 VOP_UNLOCK(tvp, 0);
 1167         /*
 1168          * Check for cross-device rename.
 1169          */
 1170         if ((fvp->v_mount != tdvp->v_mount) ||
 1171             (tvp && (fvp->v_mount != tvp->v_mount))) {
 1172                 error = EXDEV;
 1173                 mp = NULL;
 1174                 goto releout;
 1175         }
 1176         error = vfs_busy(mp, 0);
 1177         if (error) {
 1178                 mp = NULL;
 1179                 goto releout;
 1180         }
 1181 relock:
 1182         /* 
 1183          * We need to acquire 2 to 4 locks depending on whether tvp is NULL
 1184          * and fdvp and tdvp are the same directory.  Subsequently we need
 1185          * to double-check all paths and in the directory rename case we
 1186          * need to verify that we are not creating a directory loop.  To
 1187          * handle this we acquire all but fdvp using non-blocking
 1188          * acquisitions.  If we fail to acquire any lock in the path we will
 1189          * drop all held locks, acquire the new lock in a blocking fashion,
 1190          * and then release it and restart the rename.  This acquire/release
 1191          * step ensures that we do not spin on a lock waiting for release.
 1192          */
 1193         error = vn_lock(fdvp, LK_EXCLUSIVE);
 1194         if (error)
 1195                 goto releout;
 1196         if (vn_lock(tdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
 1197                 VOP_UNLOCK(fdvp, 0);
 1198                 error = vn_lock(tdvp, LK_EXCLUSIVE);
 1199                 if (error)
 1200                         goto releout;
 1201                 VOP_UNLOCK(tdvp, 0);
 1202                 atomic_add_int(&rename_restarts, 1);
 1203                 goto relock;
 1204         }
 1205         /*
 1206          * Re-resolve fvp to be certain it still exists and fetch the
 1207          * correct vnode.
 1208          */
 1209         error = ufs_lookup_ino(fdvp, NULL, fcnp, &ino);
 1210         if (error) {
 1211                 VOP_UNLOCK(fdvp, 0);
 1212                 VOP_UNLOCK(tdvp, 0);
 1213                 goto releout;
 1214         }
 1215         error = VFS_VGET(mp, ino, LK_EXCLUSIVE | LK_NOWAIT, &nvp);
 1216         if (error) {
 1217                 VOP_UNLOCK(fdvp, 0);
 1218                 VOP_UNLOCK(tdvp, 0);
 1219                 if (error != EBUSY)
 1220                         goto releout;
 1221                 error = VFS_VGET(mp, ino, LK_EXCLUSIVE, &nvp);
 1222                 if (error != 0)
 1223                         goto releout;
 1224                 VOP_UNLOCK(nvp, 0);
 1225                 vrele(fvp);
 1226                 fvp = nvp;
 1227                 atomic_add_int(&rename_restarts, 1);
 1228                 goto relock;
 1229         }
 1230         vrele(fvp);
 1231         fvp = nvp;
 1232         /*
 1233          * Re-resolve tvp and acquire the vnode lock if present.
 1234          */
 1235         error = ufs_lookup_ino(tdvp, NULL, tcnp, &ino);
 1236         if (error != 0 && error != EJUSTRETURN) {
 1237                 VOP_UNLOCK(fdvp, 0);
 1238                 VOP_UNLOCK(tdvp, 0);
 1239                 VOP_UNLOCK(fvp, 0);
 1240                 goto releout;
 1241         }
 1242         /*
 1243          * If tvp disappeared we just carry on.
 1244          */
 1245         if (error == EJUSTRETURN && tvp != NULL) {
 1246                 vrele(tvp);
 1247                 tvp = NULL;
 1248         }
 1249         /*
 1250          * Get the tvp ino if the lookup succeeded.  We may have to restart
 1251          * if the non-blocking acquire fails.
 1252          */
 1253         if (error == 0) {
 1254                 nvp = NULL;
 1255                 error = VFS_VGET(mp, ino, LK_EXCLUSIVE | LK_NOWAIT, &nvp);
 1256                 if (tvp)
 1257                         vrele(tvp);
 1258                 tvp = nvp;
 1259                 if (error) {
 1260                         VOP_UNLOCK(fdvp, 0);
 1261                         VOP_UNLOCK(tdvp, 0);
 1262                         VOP_UNLOCK(fvp, 0);
 1263                         if (error != EBUSY)
 1264                                 goto releout;
 1265                         error = VFS_VGET(mp, ino, LK_EXCLUSIVE, &nvp);
 1266                         if (error != 0)
 1267                                 goto releout;
 1268                         vput(nvp);
 1269                         atomic_add_int(&rename_restarts, 1);
 1270                         goto relock;
 1271                 }
 1272         }
 1273         fdp = VTOI(fdvp);
 1274         fip = VTOI(fvp);
 1275         tdp = VTOI(tdvp);
 1276         tip = NULL;
 1277         if (tvp)
 1278                 tip = VTOI(tvp);
 1279         if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
 1280             (VTOI(tdvp)->i_flags & APPEND))) {
 1281                 error = EPERM;
 1282                 goto unlockout;
 1283         }
 1284         /*
 1285          * Renaming a file to itself has no effect.  The upper layers should
 1286          * not call us in that case.  However, things could change after
 1287          * we drop the locks above.
 1288          */
 1289         if (fvp == tvp) {
 1290                 error = 0;
 1291                 goto unlockout;
 1292         }
 1293         doingdirectory = 0;
 1294         newparent = 0;
 1295         ino = fip->i_number;
 1296         if (fip->i_nlink >= LINK_MAX) {
 1297                 error = EMLINK;
 1298                 goto unlockout;
 1299         }
 1300         if ((fip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
 1301             || (fdp->i_flags & APPEND)) {
 1302                 error = EPERM;
 1303                 goto unlockout;
 1304         }
 1305         if ((fip->i_mode & IFMT) == IFDIR) {
 1306                 /*
 1307                  * Avoid ".", "..", and aliases of "." for obvious reasons.
 1308                  */
 1309                 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
 1310                     fdp == fip ||
 1311                     (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT) {
 1312                         error = EINVAL;
 1313                         goto unlockout;
 1314                 }
 1315                 if (fdp->i_number != tdp->i_number)
 1316                         newparent = tdp->i_number;
 1317                 doingdirectory = 1;
 1318         }
 1319         if ((fvp->v_type == VDIR && fvp->v_mountedhere != NULL) ||
 1320             (tvp != NULL && tvp->v_type == VDIR &&
 1321             tvp->v_mountedhere != NULL)) {
 1322                 error = EXDEV;
 1323                 goto unlockout;
 1324         }
 1325 
 1326         /*
 1327          * If ".." must be changed (ie the directory gets a new
 1328          * parent) then the source directory must not be in the
 1329          * directory hierarchy above the target, as this would
 1330          * orphan everything below the source directory. Also
 1331          * the user must have write permission in the source so
 1332          * as to be able to change "..".
 1333          */
 1334         if (doingdirectory && newparent) {
 1335                 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
 1336                 if (error)
 1337                         goto unlockout;
 1338                 error = ufs_checkpath(ino, fdp->i_number, tdp, tcnp->cn_cred,
 1339                     &ino);
 1340                 /*
 1341                  * We encountered a lock that we have to wait for.  Unlock
 1342                  * everything else and VGET before restarting.
 1343                  */
 1344                 if (ino) {
 1345                         VOP_UNLOCK(fdvp, 0);
 1346                         VOP_UNLOCK(fvp, 0);
 1347                         VOP_UNLOCK(tdvp, 0);
 1348                         if (tvp)
 1349                                 VOP_UNLOCK(tvp, 0);
 1350                         error = VFS_VGET(mp, ino, LK_SHARED, &nvp);
 1351                         if (error == 0)
 1352                                 vput(nvp);
 1353                         atomic_add_int(&rename_restarts, 1);
 1354                         goto relock;
 1355                 }
 1356                 if (error)
 1357                         goto unlockout;
 1358                 if ((tcnp->cn_flags & SAVESTART) == 0)
 1359                         panic("ufs_rename: lost to startdir");
 1360         }
 1361         if (fip->i_effnlink == 0 || fdp->i_effnlink == 0 ||
 1362             tdp->i_effnlink == 0)
 1363                 panic("Bad effnlink fip %p, fdp %p, tdp %p", fip, fdp, tdp);
 1364 
 1365         /*
 1366          * 1) Bump link count while we're moving stuff
 1367          *    around.  If we crash somewhere before
 1368          *    completing our work, the link count
 1369          *    may be wrong, but correctable.
 1370          */
 1371         fip->i_effnlink++;
 1372         fip->i_nlink++;
 1373         DIP_SET(fip, i_nlink, fip->i_nlink);
 1374         fip->i_flag |= IN_CHANGE;
 1375         if (DOINGSOFTDEP(fvp))
 1376                 softdep_setup_link(tdp, fip);
 1377         error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) | DOINGASYNC(fvp)));
 1378         if (error)
 1379                 goto bad;
 1380 
 1381         /*
 1382          * 2) If target doesn't exist, link the target
 1383          *    to the source and unlink the source.
 1384          *    Otherwise, rewrite the target directory
 1385          *    entry to reference the source inode and
 1386          *    expunge the original entry's existence.
 1387          */
 1388         if (tip == NULL) {
 1389                 if (tdp->i_dev != fip->i_dev)
 1390                         panic("ufs_rename: EXDEV");
 1391                 if (doingdirectory && newparent) {
 1392                         /*
 1393                          * Account for ".." in new directory.
 1394                          * When source and destination have the same
 1395                          * parent we don't adjust the link count.  The
 1396                          * actual link modification is completed when
 1397                          * .. is rewritten below.
 1398                          */
 1399                         if ((nlink_t)tdp->i_nlink >= LINK_MAX) {
 1400                                 error = EMLINK;
 1401                                 goto bad;
 1402                         }
 1403                 }
 1404                 ufs_makedirentry(fip, tcnp, &newdir);
 1405                 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL, 1);
 1406                 if (error)
 1407                         goto bad;
 1408                 /* Setup tdvp for directory compaction if needed. */
 1409                 if (tdp->i_count && tdp->i_endoff &&
 1410                     tdp->i_endoff < tdp->i_size)
 1411                         endoff = tdp->i_endoff;
 1412         } else {
 1413                 if (tip->i_dev != tdp->i_dev || tip->i_dev != fip->i_dev)
 1414                         panic("ufs_rename: EXDEV");
 1415                 /*
 1416                  * Short circuit rename(foo, foo).
 1417                  */
 1418                 if (tip->i_number == fip->i_number)
 1419                         panic("ufs_rename: same file");
 1420                 /*
 1421                  * If the parent directory is "sticky", then the caller
 1422                  * must possess VADMIN for the parent directory, or the
 1423                  * destination of the rename.  This implements append-only
 1424                  * directories.
 1425                  */
 1426                 if ((tdp->i_mode & S_ISTXT) &&
 1427                     VOP_ACCESS(tdvp, VADMIN, tcnp->cn_cred, td) &&
 1428                     VOP_ACCESS(tvp, VADMIN, tcnp->cn_cred, td)) {
 1429                         error = EPERM;
 1430                         goto bad;
 1431                 }
 1432                 /*
 1433                  * Target must be empty if a directory and have no links
 1434                  * to it. Also, ensure source and target are compatible
 1435                  * (both directories, or both not directories).
 1436                  */
 1437                 if ((tip->i_mode & IFMT) == IFDIR) {
 1438                         if ((tip->i_effnlink > 2) ||
 1439                             !ufs_dirempty(tip, tdp->i_number, tcnp->cn_cred)) {
 1440                                 error = ENOTEMPTY;
 1441                                 goto bad;
 1442                         }
 1443                         if (!doingdirectory) {
 1444                                 error = ENOTDIR;
 1445                                 goto bad;
 1446                         }
 1447                         cache_purge(tdvp);
 1448                 } else if (doingdirectory) {
 1449                         error = EISDIR;
 1450                         goto bad;
 1451                 }
 1452                 if (doingdirectory) {
 1453                         if (!newparent) {
 1454                                 tdp->i_effnlink--;
 1455                                 if (DOINGSOFTDEP(tdvp))
 1456                                         softdep_change_linkcnt(tdp);
 1457                         }
 1458                         tip->i_effnlink--;
 1459                         if (DOINGSOFTDEP(tvp))
 1460                                 softdep_change_linkcnt(tip);
 1461                 }
 1462                 error = ufs_dirrewrite(tdp, tip, fip->i_number,
 1463                     IFTODT(fip->i_mode),
 1464                     (doingdirectory && newparent) ? newparent : doingdirectory);
 1465                 if (error) {
 1466                         if (doingdirectory) {
 1467                                 if (!newparent) {
 1468                                         tdp->i_effnlink++;
 1469                                         if (DOINGSOFTDEP(tdvp))
 1470                                                 softdep_change_linkcnt(tdp);
 1471                                 }
 1472                                 tip->i_effnlink++;
 1473                                 if (DOINGSOFTDEP(tvp))
 1474                                         softdep_change_linkcnt(tip);
 1475                         }
 1476                 }
 1477                 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
 1478                         /*
 1479                          * The only stuff left in the directory is "."
 1480                          * and "..". The "." reference is inconsequential
 1481                          * since we are quashing it. We have removed the "."
 1482                          * reference and the reference in the parent directory,
 1483                          * but there may be other hard links. The soft
 1484                          * dependency code will arrange to do these operations
 1485                          * after the parent directory entry has been deleted on
 1486                          * disk, so when running with that code we avoid doing
 1487                          * them now.
 1488                          */
 1489                         if (!newparent) {
 1490                                 tdp->i_nlink--;
 1491                                 DIP_SET(tdp, i_nlink, tdp->i_nlink);
 1492                                 tdp->i_flag |= IN_CHANGE;
 1493                         }
 1494                         tip->i_nlink--;
 1495                         DIP_SET(tip, i_nlink, tip->i_nlink);
 1496                         tip->i_flag |= IN_CHANGE;
 1497                 }
 1498         }
 1499 
 1500         /*
 1501          * 3) Unlink the source.  We have to resolve the path again to
 1502          * fixup the directory offset and count for ufs_dirremove.
 1503          */
 1504         if (fdvp == tdvp) {
 1505                 error = ufs_lookup_ino(fdvp, NULL, fcnp, &ino);
 1506                 if (error)
 1507                         panic("ufs_rename: from entry went away!");
 1508                 if (ino != fip->i_number)
 1509                         panic("ufs_rename: ino mismatch %ju != %ju\n",
 1510                             (uintmax_t)ino, (uintmax_t)fip->i_number);
 1511         }
 1512         /*
 1513          * If the source is a directory with a
 1514          * new parent, the link count of the old
 1515          * parent directory must be decremented
 1516          * and ".." set to point to the new parent.
 1517          */
 1518         if (doingdirectory && newparent) {
 1519                 /*
 1520                  * If tip exists we simply use its link, otherwise we must
 1521                  * add a new one.
 1522                  */
 1523                 if (tip == NULL) {
 1524                         tdp->i_effnlink++;
 1525                         tdp->i_nlink++;
 1526                         DIP_SET(tdp, i_nlink, tdp->i_nlink);
 1527                         tdp->i_flag |= IN_CHANGE;
 1528                         if (DOINGSOFTDEP(tdvp))
 1529                                 softdep_setup_dotdot_link(tdp, fip);
 1530                         error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
 1531                                                    DOINGASYNC(tdvp)));
 1532                         /* Don't go to bad here as the new link exists. */
 1533                         if (error)
 1534                                 goto unlockout;
 1535                 } else if (DOINGSUJ(tdvp))
 1536                         /* Journal must account for each new link. */
 1537                         softdep_setup_dotdot_link(tdp, fip);
 1538                 fip->i_offset = mastertemplate.dot_reclen;
 1539                 ufs_dirrewrite(fip, fdp, newparent, DT_DIR, 0);
 1540                 cache_purge(fdvp);
 1541         }
 1542         error = ufs_dirremove(fdvp, fip, fcnp->cn_flags, 0);
 1543         /*
 1544          * The kern_renameat() looks up the fvp using the DELETE flag, which
 1545          * causes the removal of the name cache entry for fvp.
 1546          * As the relookup of the fvp is done in two steps:
 1547          * ufs_lookup_ino() and then VFS_VGET(), another thread might do a
 1548          * normal lookup of the from name just before the VFS_VGET() call,
 1549          * causing the cache entry to be re-instantiated.
 1550          *
 1551          * The same issue also applies to tvp if it exists as
 1552          * otherwise we may have a stale name cache entry for the new
 1553          * name that references the old i-node if it has other links
 1554          * or open file descriptors.
 1555          */
 1556         cache_purge(fvp);
 1557         if (tvp)
 1558                 cache_purge(tvp);
 1559         cache_purge_negative(tdvp);
 1560 
 1561 unlockout:
 1562         vput(fdvp);
 1563         vput(fvp);
 1564         if (tvp)
 1565                 vput(tvp);
 1566         /*
 1567          * If compaction or fsync was requested do it now that other locks
 1568          * are no longer needed.
 1569          */
 1570         if (error == 0 && endoff != 0) {
 1571 #ifdef UFS_DIRHASH
 1572                 if (tdp->i_dirhash != NULL)
 1573                         ufsdirhash_dirtrunc(tdp, endoff);
 1574 #endif
 1575                 UFS_TRUNCATE(tdvp, endoff, IO_NORMAL | IO_SYNC, tcnp->cn_cred);
 1576         }
 1577         if (error == 0 && tdp->i_flag & IN_NEEDSYNC)
 1578                 error = VOP_FSYNC(tdvp, MNT_WAIT, td);
 1579         vput(tdvp);
 1580         if (mp)
 1581                 vfs_unbusy(mp);
 1582         return (error);
 1583 
 1584 bad:
 1585         fip->i_effnlink--;
 1586         fip->i_nlink--;
 1587         DIP_SET(fip, i_nlink, fip->i_nlink);
 1588         fip->i_flag |= IN_CHANGE;
 1589         if (DOINGSOFTDEP(fvp))
 1590                 softdep_revert_link(tdp, fip);
 1591         goto unlockout;
 1592 
 1593 releout:
 1594         vrele(fdvp);
 1595         vrele(fvp);
 1596         vrele(tdvp);
 1597         if (tvp)
 1598                 vrele(tvp);
 1599         if (mp)
 1600                 vfs_unbusy(mp);
 1601 
 1602         return (error);
 1603 }
 1604 
 1605 #ifdef UFS_ACL
 1606 static int
 1607 ufs_do_posix1e_acl_inheritance_dir(struct vnode *dvp, struct vnode *tvp,
 1608     mode_t dmode, struct ucred *cred, struct thread *td)
 1609 {
 1610         int error;
 1611         struct inode *ip = VTOI(tvp);
 1612         struct acl *dacl, *acl;
 1613 
 1614         acl = acl_alloc(M_WAITOK);
 1615         dacl = acl_alloc(M_WAITOK);
 1616 
 1617         /*
 1618          * Retrieve default ACL from parent, if any.
 1619          */
 1620         error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
 1621         switch (error) {
 1622         case 0:
 1623                 /*
 1624                  * Retrieved a default ACL, so merge mode and ACL if
 1625                  * necessary.  If the ACL is empty, fall through to
 1626                  * the "not defined or available" case.
 1627                  */
 1628                 if (acl->acl_cnt != 0) {
 1629                         dmode = acl_posix1e_newfilemode(dmode, acl);
 1630                         ip->i_mode = dmode;
 1631                         DIP_SET(ip, i_mode, dmode);
 1632                         *dacl = *acl;
 1633                         ufs_sync_acl_from_inode(ip, acl);
 1634                         break;
 1635                 }
 1636                 /* FALLTHROUGH */
 1637 
 1638         case EOPNOTSUPP:
 1639                 /*
 1640                  * Just use the mode as-is.
 1641                  */
 1642                 ip->i_mode = dmode;
 1643                 DIP_SET(ip, i_mode, dmode);
 1644                 error = 0;
 1645                 goto out;
 1646         
 1647         default:
 1648                 goto out;
 1649         }
 1650 
 1651         /*
 1652          * XXX: If we abort now, will Soft Updates notify the extattr
 1653          * code that the EAs for the file need to be released?
 1654          */
 1655         error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
 1656         if (error == 0)
 1657                 error = VOP_SETACL(tvp, ACL_TYPE_DEFAULT, dacl, cred, td);
 1658         switch (error) {
 1659         case 0:
 1660                 break;
 1661 
 1662         case EOPNOTSUPP:
 1663                 /*
 1664                  * XXX: This should not happen, as EOPNOTSUPP above
 1665                  * was supposed to free acl.
 1666                  */
 1667                 printf("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()\n");
 1668                 /*
 1669                 panic("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()");
 1670                  */
 1671                 break;
 1672 
 1673         default:
 1674                 goto out;
 1675         }
 1676 
 1677 out:
 1678         acl_free(acl);
 1679         acl_free(dacl);
 1680 
 1681         return (error);
 1682 }
 1683 
 1684 static int
 1685 ufs_do_posix1e_acl_inheritance_file(struct vnode *dvp, struct vnode *tvp,
 1686     mode_t mode, struct ucred *cred, struct thread *td)
 1687 {
 1688         int error;
 1689         struct inode *ip = VTOI(tvp);
 1690         struct acl *acl;
 1691 
 1692         acl = acl_alloc(M_WAITOK);
 1693 
 1694         /*
 1695          * Retrieve default ACL for parent, if any.
 1696          */
 1697         error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
 1698         switch (error) {
 1699         case 0:
 1700                 /*
 1701                  * Retrieved a default ACL, so merge mode and ACL if
 1702                  * necessary.
 1703                  */
 1704                 if (acl->acl_cnt != 0) {
 1705                         /*
 1706                          * Two possible ways for default ACL to not
 1707                          * be present.  First, the EA can be
 1708                          * undefined, or second, the default ACL can
 1709                          * be blank.  If it's blank, fall through to
 1710                          * the it's not defined case.
 1711                          */
 1712                         mode = acl_posix1e_newfilemode(mode, acl);
 1713                         ip->i_mode = mode;
 1714                         DIP_SET(ip, i_mode, mode);
 1715                         ufs_sync_acl_from_inode(ip, acl);
 1716                         break;
 1717                 }
 1718                 /* FALLTHROUGH */
 1719 
 1720         case EOPNOTSUPP:
 1721                 /*
 1722                  * Just use the mode as-is.
 1723                  */
 1724                 ip->i_mode = mode;
 1725                 DIP_SET(ip, i_mode, mode);
 1726                 error = 0;
 1727                 goto out;
 1728 
 1729         default:
 1730                 goto out;
 1731         }
 1732 
 1733         /*
 1734          * XXX: If we abort now, will Soft Updates notify the extattr
 1735          * code that the EAs for the file need to be released?
 1736          */
 1737         error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
 1738         switch (error) {
 1739         case 0:
 1740                 break;
 1741 
 1742         case EOPNOTSUPP:
 1743                 /*
 1744                  * XXX: This should not happen, as EOPNOTSUPP above was
 1745                  * supposed to free acl.
 1746                  */
 1747                 printf("ufs_makeinode: VOP_GETACL() but no "
 1748                     "VOP_SETACL()\n");
 1749                 /* panic("ufs_makeinode: VOP_GETACL() but no "
 1750                     "VOP_SETACL()"); */
 1751                 break;
 1752 
 1753         default:
 1754                 goto out;
 1755         }
 1756 
 1757 out:
 1758         acl_free(acl);
 1759 
 1760         return (error);
 1761 }
 1762 
 1763 static int
 1764 ufs_do_nfs4_acl_inheritance(struct vnode *dvp, struct vnode *tvp,
 1765     mode_t child_mode, struct ucred *cred, struct thread *td)
 1766 {
 1767         int error;
 1768         struct acl *parent_aclp, *child_aclp;
 1769 
 1770         parent_aclp = acl_alloc(M_WAITOK);
 1771         child_aclp = acl_alloc(M_WAITOK | M_ZERO);
 1772 
 1773         error = ufs_getacl_nfs4_internal(dvp, parent_aclp, td);
 1774         if (error)
 1775                 goto out;
 1776         acl_nfs4_compute_inherited_acl(parent_aclp, child_aclp,
 1777             child_mode, VTOI(tvp)->i_uid, tvp->v_type == VDIR);
 1778         error = ufs_setacl_nfs4_internal(tvp, child_aclp, td);
 1779         if (error)
 1780                 goto out;
 1781 out:
 1782         acl_free(parent_aclp);
 1783         acl_free(child_aclp);
 1784 
 1785         return (error);
 1786 }
 1787 #endif
 1788 
 1789 /*
 1790  * Mkdir system call
 1791  */
 1792 static int
 1793 ufs_mkdir(ap)
 1794         struct vop_mkdir_args /* {
 1795                 struct vnode *a_dvp;
 1796                 struct vnode **a_vpp;
 1797                 struct componentname *a_cnp;
 1798                 struct vattr *a_vap;
 1799         } */ *ap;
 1800 {
 1801         struct vnode *dvp = ap->a_dvp;
 1802         struct vattr *vap = ap->a_vap;
 1803         struct componentname *cnp = ap->a_cnp;
 1804         struct inode *ip, *dp;
 1805         struct vnode *tvp;
 1806         struct buf *bp;
 1807         struct dirtemplate dirtemplate, *dtp;
 1808         struct direct newdir;
 1809         int error, dmode;
 1810         long blkoff;
 1811 
 1812 #ifdef INVARIANTS
 1813         if ((cnp->cn_flags & HASBUF) == 0)
 1814                 panic("ufs_mkdir: no name");
 1815 #endif
 1816         dp = VTOI(dvp);
 1817         if ((nlink_t)dp->i_nlink >= LINK_MAX) {
 1818                 error = EMLINK;
 1819                 goto out;
 1820         }
 1821         dmode = vap->va_mode & 0777;
 1822         dmode |= IFDIR;
 1823         /*
 1824          * Must simulate part of ufs_makeinode here to acquire the inode,
 1825          * but not have it entered in the parent directory. The entry is
 1826          * made later after writing "." and ".." entries.
 1827          */
 1828         error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
 1829         if (error)
 1830                 goto out;
 1831         ip = VTOI(tvp);
 1832         ip->i_gid = dp->i_gid;
 1833         DIP_SET(ip, i_gid, dp->i_gid);
 1834 #ifdef SUIDDIR
 1835         {
 1836 #ifdef QUOTA
 1837                 struct ucred ucred, *ucp;
 1838                 gid_t ucred_group;
 1839                 ucp = cnp->cn_cred;
 1840 #endif
 1841                 /*
 1842                  * If we are hacking owners here, (only do this where told to)
 1843                  * and we are not giving it TO root, (would subvert quotas)
 1844                  * then go ahead and give it to the other user.
 1845                  * The new directory also inherits the SUID bit.
 1846                  * If user's UID and dir UID are the same,
 1847                  * 'give it away' so that the SUID is still forced on.
 1848                  */
 1849                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 1850                     (dp->i_mode & ISUID) && dp->i_uid) {
 1851                         dmode |= ISUID;
 1852                         ip->i_uid = dp->i_uid;
 1853                         DIP_SET(ip, i_uid, dp->i_uid);
 1854 #ifdef QUOTA
 1855                         if (dp->i_uid != cnp->cn_cred->cr_uid) {
 1856                                 /*
 1857                                  * Make sure the correct user gets charged
 1858                                  * for the space.
 1859                                  * Make a dummy credential for the victim.
 1860                                  * XXX This seems to never be accessed out of
 1861                                  * our context so a stack variable is ok.
 1862                                  */
 1863                                 refcount_init(&ucred.cr_ref, 1);
 1864                                 ucred.cr_uid = ip->i_uid;
 1865                                 ucred.cr_ngroups = 1;
 1866                                 ucred.cr_groups = &ucred_group;
 1867                                 ucred.cr_groups[0] = dp->i_gid;
 1868                                 ucp = &ucred;
 1869                         }
 1870 #endif
 1871                 } else {
 1872                         ip->i_uid = cnp->cn_cred->cr_uid;
 1873                         DIP_SET(ip, i_uid, ip->i_uid);
 1874                 }
 1875 #ifdef QUOTA
 1876                 if ((error = getinoquota(ip)) ||
 1877                     (error = chkiq(ip, 1, ucp, 0))) {
 1878                         if (DOINGSOFTDEP(tvp))
 1879                                 softdep_revert_link(dp, ip);
 1880                         UFS_VFREE(tvp, ip->i_number, dmode);
 1881                         vput(tvp);
 1882                         return (error);
 1883                 }
 1884 #endif
 1885         }
 1886 #else   /* !SUIDDIR */
 1887         ip->i_uid = cnp->cn_cred->cr_uid;
 1888         DIP_SET(ip, i_uid, ip->i_uid);
 1889 #ifdef QUOTA
 1890         if ((error = getinoquota(ip)) ||
 1891             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 1892                 if (DOINGSOFTDEP(tvp))
 1893                         softdep_revert_link(dp, ip);
 1894                 UFS_VFREE(tvp, ip->i_number, dmode);
 1895                 vput(tvp);
 1896                 return (error);
 1897         }
 1898 #endif
 1899 #endif  /* !SUIDDIR */
 1900         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 1901         ip->i_mode = dmode;
 1902         DIP_SET(ip, i_mode, dmode);
 1903         tvp->v_type = VDIR;     /* Rest init'd in getnewvnode(). */
 1904         ip->i_effnlink = 2;
 1905         ip->i_nlink = 2;
 1906         DIP_SET(ip, i_nlink, 2);
 1907 
 1908         if (cnp->cn_flags & ISWHITEOUT) {
 1909                 ip->i_flags |= UF_OPAQUE;
 1910                 DIP_SET(ip, i_flags, ip->i_flags);
 1911         }
 1912 
 1913         /*
 1914          * Bump link count in parent directory to reflect work done below.
 1915          * Should be done before reference is created so cleanup is
 1916          * possible if we crash.
 1917          */
 1918         dp->i_effnlink++;
 1919         dp->i_nlink++;
 1920         DIP_SET(dp, i_nlink, dp->i_nlink);
 1921         dp->i_flag |= IN_CHANGE;
 1922         if (DOINGSOFTDEP(dvp))
 1923                 softdep_setup_mkdir(dp, ip);
 1924         error = UFS_UPDATE(dvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
 1925         if (error)
 1926                 goto bad;
 1927 #ifdef MAC
 1928         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 1929                 error = mac_vnode_create_extattr(cnp->cn_cred, dvp->v_mount,
 1930                     dvp, tvp, cnp);
 1931                 if (error)
 1932                         goto bad;
 1933         }
 1934 #endif
 1935 #ifdef UFS_ACL
 1936         if (dvp->v_mount->mnt_flag & MNT_ACLS) {
 1937                 error = ufs_do_posix1e_acl_inheritance_dir(dvp, tvp, dmode,
 1938                     cnp->cn_cred, cnp->cn_thread);
 1939                 if (error)
 1940                         goto bad;
 1941         } else if (dvp->v_mount->mnt_flag & MNT_NFS4ACLS) {
 1942                 error = ufs_do_nfs4_acl_inheritance(dvp, tvp, dmode,
 1943                     cnp->cn_cred, cnp->cn_thread);
 1944                 if (error)
 1945                         goto bad;
 1946         }
 1947 #endif /* !UFS_ACL */
 1948 
 1949         /*
 1950          * Initialize directory with "." and ".." from static template.
 1951          */
 1952         if (dvp->v_mount->mnt_maxsymlinklen > 0)
 1953                 dtp = &mastertemplate;
 1954         else
 1955                 dtp = (struct dirtemplate *)&omastertemplate;
 1956         dirtemplate = *dtp;
 1957         dirtemplate.dot_ino = ip->i_number;
 1958         dirtemplate.dotdot_ino = dp->i_number;
 1959         if ((error = UFS_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
 1960             BA_CLRBUF, &bp)) != 0)
 1961                 goto bad;
 1962         ip->i_size = DIRBLKSIZ;
 1963         DIP_SET(ip, i_size, DIRBLKSIZ);
 1964         ip->i_flag |= IN_CHANGE | IN_UPDATE;
 1965         vnode_pager_setsize(tvp, (u_long)ip->i_size);
 1966         bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
 1967         if (DOINGSOFTDEP(tvp)) {
 1968                 /*
 1969                  * Ensure that the entire newly allocated block is a
 1970                  * valid directory so that future growth within the
 1971                  * block does not have to ensure that the block is
 1972                  * written before the inode.
 1973                  */
 1974                 blkoff = DIRBLKSIZ;
 1975                 while (blkoff < bp->b_bcount) {
 1976                         ((struct direct *)
 1977                            (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
 1978                         blkoff += DIRBLKSIZ;
 1979                 }
 1980         }
 1981         if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
 1982                                        DOINGASYNC(tvp)))) != 0) {
 1983                 (void)bwrite(bp);
 1984                 goto bad;
 1985         }
 1986         /*
 1987          * Directory set up, now install its entry in the parent directory.
 1988          *
 1989          * If we are not doing soft dependencies, then we must write out the
 1990          * buffer containing the new directory body before entering the new 
 1991          * name in the parent. If we are doing soft dependencies, then the
 1992          * buffer containing the new directory body will be passed to and
 1993          * released in the soft dependency code after the code has attached
 1994          * an appropriate ordering dependency to the buffer which ensures that
 1995          * the buffer is written before the new name is written in the parent.
 1996          */
 1997         if (DOINGASYNC(dvp))
 1998                 bdwrite(bp);
 1999         else if (!DOINGSOFTDEP(dvp) && ((error = bwrite(bp))))
 2000                 goto bad;
 2001         ufs_makedirentry(ip, cnp, &newdir);
 2002         error = ufs_direnter(dvp, tvp, &newdir, cnp, bp, 0);
 2003         
 2004 bad:
 2005         if (error == 0) {
 2006                 *ap->a_vpp = tvp;
 2007         } else {
 2008                 dp->i_effnlink--;
 2009                 dp->i_nlink--;
 2010                 DIP_SET(dp, i_nlink, dp->i_nlink);
 2011                 dp->i_flag |= IN_CHANGE;
 2012                 /*
 2013                  * No need to do an explicit VOP_TRUNCATE here, vrele will
 2014                  * do this for us because we set the link count to 0.
 2015                  */
 2016                 ip->i_effnlink = 0;
 2017                 ip->i_nlink = 0;
 2018                 DIP_SET(ip, i_nlink, 0);
 2019                 ip->i_flag |= IN_CHANGE;
 2020                 if (DOINGSOFTDEP(tvp))
 2021                         softdep_revert_mkdir(dp, ip);
 2022 
 2023                 vput(tvp);
 2024         }
 2025 out:
 2026         return (error);
 2027 }
 2028 
 2029 /*
 2030  * Rmdir system call.
 2031  */
 2032 static int
 2033 ufs_rmdir(ap)
 2034         struct vop_rmdir_args /* {
 2035                 struct vnode *a_dvp;
 2036                 struct vnode *a_vp;
 2037                 struct componentname *a_cnp;
 2038         } */ *ap;
 2039 {
 2040         struct vnode *vp = ap->a_vp;
 2041         struct vnode *dvp = ap->a_dvp;
 2042         struct componentname *cnp = ap->a_cnp;
 2043         struct inode *ip, *dp;
 2044         int error;
 2045 
 2046         ip = VTOI(vp);
 2047         dp = VTOI(dvp);
 2048 
 2049         /*
 2050          * Do not remove a directory that is in the process of being renamed.
 2051          * Verify the directory is empty (and valid). Rmdir ".." will not be
 2052          * valid since ".." will contain a reference to the current directory
 2053          * and thus be non-empty. Do not allow the removal of mounted on
 2054          * directories (this can happen when an NFS exported filesystem
 2055          * tries to remove a locally mounted on directory).
 2056          */
 2057         error = 0;
 2058         if (ip->i_effnlink < 2) {
 2059                 error = EINVAL;
 2060                 goto out;
 2061         }
 2062         if (dp->i_effnlink < 3)
 2063                 panic("ufs_dirrem: Bad link count %d on parent",
 2064                     dp->i_effnlink);
 2065         if (!ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
 2066                 error = ENOTEMPTY;
 2067                 goto out;
 2068         }
 2069         if ((dp->i_flags & APPEND)
 2070             || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
 2071                 error = EPERM;
 2072                 goto out;
 2073         }
 2074         if (vp->v_mountedhere != 0) {
 2075                 error = EINVAL;
 2076                 goto out;
 2077         }
 2078 #ifdef UFS_GJOURNAL
 2079         ufs_gjournal_orphan(vp);
 2080 #endif
 2081         /*
 2082          * Delete reference to directory before purging
 2083          * inode.  If we crash in between, the directory
 2084          * will be reattached to lost+found,
 2085          */
 2086         dp->i_effnlink--;
 2087         ip->i_effnlink--;
 2088         if (DOINGSOFTDEP(vp))
 2089                 softdep_setup_rmdir(dp, ip);
 2090         error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
 2091         if (error) {
 2092                 dp->i_effnlink++;
 2093                 ip->i_effnlink++;
 2094                 if (DOINGSOFTDEP(vp))
 2095                         softdep_revert_rmdir(dp, ip);
 2096                 goto out;
 2097         }
 2098         cache_purge(dvp);
 2099         /*
 2100          * The only stuff left in the directory is "." and "..". The "."
 2101          * reference is inconsequential since we are quashing it. The soft
 2102          * dependency code will arrange to do these operations after
 2103          * the parent directory entry has been deleted on disk, so
 2104          * when running with that code we avoid doing them now.
 2105          */
 2106         if (!DOINGSOFTDEP(vp)) {
 2107                 dp->i_nlink--;
 2108                 DIP_SET(dp, i_nlink, dp->i_nlink);
 2109                 dp->i_flag |= IN_CHANGE;
 2110                 error = UFS_UPDATE(dvp, 0);
 2111                 ip->i_nlink--;
 2112                 DIP_SET(ip, i_nlink, ip->i_nlink);
 2113                 ip->i_flag |= IN_CHANGE;
 2114         }
 2115         cache_purge(vp);
 2116 #ifdef UFS_DIRHASH
 2117         /* Kill any active hash; i_effnlink == 0, so it will not come back. */
 2118         if (ip->i_dirhash != NULL)
 2119                 ufsdirhash_free(ip);
 2120 #endif
 2121 out:
 2122         return (error);
 2123 }
 2124 
 2125 /*
 2126  * symlink -- make a symbolic link
 2127  */
 2128 static int
 2129 ufs_symlink(ap)
 2130         struct vop_symlink_args /* {
 2131                 struct vnode *a_dvp;
 2132                 struct vnode **a_vpp;
 2133                 struct componentname *a_cnp;
 2134                 struct vattr *a_vap;
 2135                 char *a_target;
 2136         } */ *ap;
 2137 {
 2138         struct vnode *vp, **vpp = ap->a_vpp;
 2139         struct inode *ip;
 2140         int len, error;
 2141 
 2142         error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
 2143             vpp, ap->a_cnp);
 2144         if (error)
 2145                 return (error);
 2146         vp = *vpp;
 2147         len = strlen(ap->a_target);
 2148         if (len < vp->v_mount->mnt_maxsymlinklen) {
 2149                 ip = VTOI(vp);
 2150                 bcopy(ap->a_target, SHORTLINK(ip), len);
 2151                 ip->i_size = len;
 2152                 DIP_SET(ip, i_size, len);
 2153                 ip->i_flag |= IN_CHANGE | IN_UPDATE;
 2154                 error = UFS_UPDATE(vp, 0);
 2155         } else
 2156                 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
 2157                     UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
 2158                     ap->a_cnp->cn_cred, NOCRED, NULL, NULL);
 2159         if (error)
 2160                 vput(vp);
 2161         return (error);
 2162 }
 2163 
 2164 /*
 2165  * Vnode op for reading directories.
 2166  */
 2167 int
 2168 ufs_readdir(ap)
 2169         struct vop_readdir_args /* {
 2170                 struct vnode *a_vp;
 2171                 struct uio *a_uio;
 2172                 struct ucred *a_cred;
 2173                 int *a_eofflag;
 2174                 int *a_ncookies;
 2175                 u_long **a_cookies;
 2176         } */ *ap;
 2177 {
 2178         struct vnode *vp = ap->a_vp;
 2179         struct uio *uio = ap->a_uio;
 2180         struct buf *bp;
 2181         struct inode *ip;
 2182         struct direct *dp, *edp;
 2183         u_long *cookies;
 2184         struct dirent dstdp;
 2185         off_t offset, startoffset;
 2186         size_t readcnt, skipcnt;
 2187         ssize_t startresid;
 2188         int ncookies;
 2189         int error;
 2190 
 2191         if (uio->uio_offset < 0)
 2192                 return (EINVAL);
 2193         ip = VTOI(vp);
 2194         if (ip->i_effnlink == 0)
 2195                 return (0);
 2196         if (ap->a_ncookies != NULL) {
 2197                 ncookies = uio->uio_resid;
 2198                 if (uio->uio_offset >= ip->i_size)
 2199                         ncookies = 0;
 2200                 else if (ip->i_size - uio->uio_offset < ncookies)
 2201                         ncookies = ip->i_size - uio->uio_offset;
 2202                 ncookies = ncookies / (offsetof(struct direct, d_name) + 4) + 1;
 2203                 cookies = malloc(ncookies * sizeof(*cookies), M_TEMP, M_WAITOK);
 2204                 *ap->a_ncookies = ncookies;
 2205                 *ap->a_cookies = cookies;
 2206         } else {
 2207                 ncookies = 0;
 2208                 cookies = NULL;
 2209         }
 2210         offset = startoffset = uio->uio_offset;
 2211         startresid = uio->uio_resid;
 2212         error = 0;
 2213         while (error == 0 && uio->uio_resid > 0 &&
 2214             uio->uio_offset < ip->i_size) {
 2215                 error = ffs_blkatoff(vp, uio->uio_offset, NULL, &bp);
 2216                 if (error)
 2217                         break;
 2218                 if (bp->b_offset + bp->b_bcount > ip->i_size)
 2219                         readcnt = ip->i_size - bp->b_offset;
 2220                 else
 2221                         readcnt = bp->b_bcount;
 2222                 skipcnt = (size_t)(uio->uio_offset - bp->b_offset) &
 2223                     ~(size_t)(DIRBLKSIZ - 1);
 2224                 offset = bp->b_offset + skipcnt;
 2225                 dp = (struct direct *)&bp->b_data[skipcnt];
 2226                 edp = (struct direct *)&bp->b_data[readcnt];
 2227                 while (error == 0 && uio->uio_resid > 0 && dp < edp) {
 2228                         if (dp->d_reclen <= offsetof(struct direct, d_name) ||
 2229                             (caddr_t)dp + dp->d_reclen > (caddr_t)edp) {
 2230                                 error = EIO;
 2231                                 break;
 2232                         }
 2233 #if BYTE_ORDER == LITTLE_ENDIAN
 2234                         /* Old filesystem format. */
 2235                         if (vp->v_mount->mnt_maxsymlinklen <= 0) {
 2236                                 dstdp.d_namlen = dp->d_type;
 2237                                 dstdp.d_type = dp->d_namlen;
 2238                         } else
 2239 #endif
 2240                         {
 2241                                 dstdp.d_namlen = dp->d_namlen;
 2242                                 dstdp.d_type = dp->d_type;
 2243                         }
 2244                         if (offsetof(struct direct, d_name) + dstdp.d_namlen >
 2245                             dp->d_reclen) {
 2246                                 error = EIO;
 2247                                 break;
 2248                         }
 2249                         if (offset < startoffset || dp->d_ino == 0)
 2250                                 goto nextentry;
 2251                         dstdp.d_fileno = dp->d_ino;
 2252                         dstdp.d_reclen = GENERIC_DIRSIZ(&dstdp);
 2253                         bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
 2254                         dstdp.d_name[dstdp.d_namlen] = '\0';
 2255                         if (dstdp.d_reclen > uio->uio_resid) {
 2256                                 if (uio->uio_resid == startresid)
 2257                                         error = EINVAL;
 2258                                 else
 2259                                         error = EJUSTRETURN;
 2260                                 break;
 2261                         }
 2262                         /* Advance dp. */
 2263                         error = uiomove((caddr_t)&dstdp, dstdp.d_reclen, uio);
 2264                         if (error)
 2265                                 break;
 2266                         if (cookies != NULL) {
 2267                                 KASSERT(ncookies > 0,
 2268                                     ("ufs_readdir: cookies buffer too small"));
 2269                                 *cookies = offset + dp->d_reclen;
 2270                                 cookies++;
 2271                                 ncookies--;
 2272                         }
 2273 nextentry:
 2274                         offset += dp->d_reclen;
 2275                         dp = (struct direct *)((caddr_t)dp + dp->d_reclen);
 2276                 }
 2277                 bqrelse(bp);
 2278                 uio->uio_offset = offset;
 2279         }
 2280         /* We need to correct uio_offset. */
 2281         uio->uio_offset = offset;
 2282         if (error == EJUSTRETURN)
 2283                 error = 0;
 2284         if (ap->a_ncookies != NULL) {
 2285                 if (error == 0) {
 2286                         ap->a_ncookies -= ncookies;
 2287                 } else {
 2288                         free(*ap->a_cookies, M_TEMP);
 2289                         *ap->a_ncookies = 0;
 2290                         *ap->a_cookies = NULL;
 2291                 }
 2292         }
 2293         if (error == 0 && ap->a_eofflag)
 2294                 *ap->a_eofflag = ip->i_size <= uio->uio_offset;
 2295         return (error);
 2296 }
 2297 
 2298 /*
 2299  * Return target name of a symbolic link
 2300  */
 2301 static int
 2302 ufs_readlink(ap)
 2303         struct vop_readlink_args /* {
 2304                 struct vnode *a_vp;
 2305                 struct uio *a_uio;
 2306                 struct ucred *a_cred;
 2307         } */ *ap;
 2308 {
 2309         struct vnode *vp = ap->a_vp;
 2310         struct inode *ip = VTOI(vp);
 2311         doff_t isize;
 2312 
 2313         isize = ip->i_size;
 2314         if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
 2315             DIP(ip, i_blocks) == 0) { /* XXX - for old fastlink support */
 2316                 return (uiomove(SHORTLINK(ip), isize, ap->a_uio));
 2317         }
 2318         return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
 2319 }
 2320 
 2321 /*
 2322  * Calculate the logical to physical mapping if not done already,
 2323  * then call the device strategy routine.
 2324  *
 2325  * In order to be able to swap to a file, the ufs_bmaparray() operation may not
 2326  * deadlock on memory.  See ufs_bmap() for details.
 2327  */
 2328 static int
 2329 ufs_strategy(ap)
 2330         struct vop_strategy_args /* {
 2331                 struct vnode *a_vp;
 2332                 struct buf *a_bp;
 2333         } */ *ap;
 2334 {
 2335         struct buf *bp = ap->a_bp;
 2336         struct vnode *vp = ap->a_vp;
 2337         struct bufobj *bo;
 2338         struct inode *ip;
 2339         ufs2_daddr_t blkno;
 2340         int error;
 2341 
 2342         ip = VTOI(vp);
 2343         if (bp->b_blkno == bp->b_lblkno) {
 2344                 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, bp, NULL, NULL);
 2345                 bp->b_blkno = blkno;
 2346                 if (error) {
 2347                         bp->b_error = error;
 2348                         bp->b_ioflags |= BIO_ERROR;
 2349                         bufdone(bp);
 2350                         return (0);
 2351                 }
 2352                 if ((long)bp->b_blkno == -1)
 2353                         vfs_bio_clrbuf(bp);
 2354         }
 2355         if ((long)bp->b_blkno == -1) {
 2356                 bufdone(bp);
 2357                 return (0);
 2358         }
 2359         bp->b_iooffset = dbtob(bp->b_blkno);
 2360         bo = ip->i_umbufobj;
 2361         BO_STRATEGY(bo, bp);
 2362         return (0);
 2363 }
 2364 
 2365 /*
 2366  * Print out the contents of an inode.
 2367  */
 2368 static int
 2369 ufs_print(ap)
 2370         struct vop_print_args /* {
 2371                 struct vnode *a_vp;
 2372         } */ *ap;
 2373 {
 2374         struct vnode *vp = ap->a_vp;
 2375         struct inode *ip = VTOI(vp);
 2376 
 2377         printf("\tino %lu, on dev %s", (u_long)ip->i_number,
 2378             devtoname(ip->i_dev));
 2379         if (vp->v_type == VFIFO)
 2380                 fifo_printinfo(vp);
 2381         printf("\n");
 2382         return (0);
 2383 }
 2384 
 2385 /*
 2386  * Close wrapper for fifos.
 2387  *
 2388  * Update the times on the inode then do device close.
 2389  */
 2390 static int
 2391 ufsfifo_close(ap)
 2392         struct vop_close_args /* {
 2393                 struct vnode *a_vp;
 2394                 int  a_fflag;
 2395                 struct ucred *a_cred;
 2396                 struct thread *a_td;
 2397         } */ *ap;
 2398 {
 2399         struct vnode *vp = ap->a_vp;
 2400         int usecount;
 2401 
 2402         VI_LOCK(vp);
 2403         usecount = vp->v_usecount;
 2404         if (usecount > 1)
 2405                 ufs_itimes_locked(vp);
 2406         VI_UNLOCK(vp);
 2407         return (fifo_specops.vop_close(ap));
 2408 }
 2409 
 2410 /*
 2411  * Kqfilter wrapper for fifos.
 2412  *
 2413  * Fall through to ufs kqfilter routines if needed 
 2414  */
 2415 static int
 2416 ufsfifo_kqfilter(ap)
 2417         struct vop_kqfilter_args *ap;
 2418 {
 2419         int error;
 2420 
 2421         error = fifo_specops.vop_kqfilter(ap);
 2422         if (error)
 2423                 error = vfs_kqfilter(ap);
 2424         return (error);
 2425 }
 2426 
 2427 /*
 2428  * Return POSIX pathconf information applicable to fifos.
 2429  */
 2430 static int
 2431 ufsfifo_pathconf(ap)
 2432         struct vop_pathconf_args /* {
 2433                 struct vnode *a_vp;
 2434                 int a_name;
 2435                 int *a_retval;
 2436         } */ *ap;
 2437 {
 2438 
 2439         switch (ap->a_name) {
 2440         case _PC_ACL_EXTENDED:
 2441         case _PC_ACL_NFS4:
 2442         case _PC_ACL_PATH_MAX:
 2443         case _PC_MAC_PRESENT:
 2444                 return (ufs_pathconf(ap));
 2445         default:
 2446                 return (fifo_specops.vop_pathconf(ap));
 2447         }
 2448         /* NOTREACHED */
 2449 }
 2450 
 2451 /*
 2452  * Return POSIX pathconf information applicable to ufs filesystems.
 2453  */
 2454 static int
 2455 ufs_pathconf(ap)
 2456         struct vop_pathconf_args /* {
 2457                 struct vnode *a_vp;
 2458                 int a_name;
 2459                 int *a_retval;
 2460         } */ *ap;
 2461 {
 2462         int error;
 2463 
 2464         error = 0;
 2465         switch (ap->a_name) {
 2466         case _PC_LINK_MAX:
 2467                 *ap->a_retval = LINK_MAX;
 2468                 break;
 2469         case _PC_NAME_MAX:
 2470                 *ap->a_retval = NAME_MAX;
 2471                 break;
 2472         case _PC_PATH_MAX:
 2473                 *ap->a_retval = PATH_MAX;
 2474                 break;
 2475         case _PC_PIPE_BUF:
 2476                 *ap->a_retval = PIPE_BUF;
 2477                 break;
 2478         case _PC_CHOWN_RESTRICTED:
 2479                 *ap->a_retval = 1;
 2480                 break;
 2481         case _PC_NO_TRUNC:
 2482                 *ap->a_retval = 1;
 2483                 break;
 2484         case _PC_ACL_EXTENDED:
 2485 #ifdef UFS_ACL
 2486                 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
 2487                         *ap->a_retval = 1;
 2488                 else
 2489                         *ap->a_retval = 0;
 2490 #else
 2491                 *ap->a_retval = 0;
 2492 #endif
 2493                 break;
 2494 
 2495         case _PC_ACL_NFS4:
 2496 #ifdef UFS_ACL
 2497                 if (ap->a_vp->v_mount->mnt_flag & MNT_NFS4ACLS)
 2498                         *ap->a_retval = 1;
 2499                 else
 2500                         *ap->a_retval = 0;
 2501 #else
 2502                 *ap->a_retval = 0;
 2503 #endif
 2504                 break;
 2505 
 2506         case _PC_ACL_PATH_MAX:
 2507 #ifdef UFS_ACL
 2508                 if (ap->a_vp->v_mount->mnt_flag & (MNT_ACLS | MNT_NFS4ACLS))
 2509                         *ap->a_retval = ACL_MAX_ENTRIES;
 2510                 else
 2511                         *ap->a_retval = 3;
 2512 #else
 2513                 *ap->a_retval = 3;
 2514 #endif
 2515                 break;
 2516         case _PC_MAC_PRESENT:
 2517 #ifdef MAC
 2518                 if (ap->a_vp->v_mount->mnt_flag & MNT_MULTILABEL)
 2519                         *ap->a_retval = 1;
 2520                 else
 2521                         *ap->a_retval = 0;
 2522 #else
 2523                 *ap->a_retval = 0;
 2524 #endif
 2525                 break;
 2526         case _PC_MIN_HOLE_SIZE:
 2527                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2528                 break;
 2529         case _PC_ASYNC_IO:
 2530                 /* _PC_ASYNC_IO should have been handled by upper layers. */
 2531                 KASSERT(0, ("_PC_ASYNC_IO should not get here"));
 2532                 error = EINVAL;
 2533                 break;
 2534         case _PC_PRIO_IO:
 2535                 *ap->a_retval = 0;
 2536                 break;
 2537         case _PC_SYNC_IO:
 2538                 *ap->a_retval = 0;
 2539                 break;
 2540         case _PC_ALLOC_SIZE_MIN:
 2541                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
 2542                 break;
 2543         case _PC_FILESIZEBITS:
 2544                 *ap->a_retval = 64;
 2545                 break;
 2546         case _PC_REC_INCR_XFER_SIZE:
 2547                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2548                 break;
 2549         case _PC_REC_MAX_XFER_SIZE:
 2550                 *ap->a_retval = -1; /* means ``unlimited'' */
 2551                 break;
 2552         case _PC_REC_MIN_XFER_SIZE:
 2553                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2554                 break;
 2555         case _PC_REC_XFER_ALIGN:
 2556                 *ap->a_retval = PAGE_SIZE;
 2557                 break;
 2558         case _PC_SYMLINK_MAX:
 2559                 *ap->a_retval = MAXPATHLEN;
 2560                 break;
 2561 
 2562         default:
 2563                 error = EINVAL;
 2564                 break;
 2565         }
 2566         return (error);
 2567 }
 2568 
 2569 /*
 2570  * Initialize the vnode associated with a new inode, handle aliased
 2571  * vnodes.
 2572  */
 2573 int
 2574 ufs_vinit(mntp, fifoops, vpp)
 2575         struct mount *mntp;
 2576         struct vop_vector *fifoops;
 2577         struct vnode **vpp;
 2578 {
 2579         struct inode *ip;
 2580         struct vnode *vp;
 2581 
 2582         vp = *vpp;
 2583         ip = VTOI(vp);
 2584         vp->v_type = IFTOVT(ip->i_mode);
 2585         if (vp->v_type == VFIFO)
 2586                 vp->v_op = fifoops;
 2587         ASSERT_VOP_LOCKED(vp, "ufs_vinit");
 2588         if (ip->i_number == ROOTINO)
 2589                 vp->v_vflag |= VV_ROOT;
 2590         *vpp = vp;
 2591         return (0);
 2592 }
 2593 
 2594 /*
 2595  * Allocate a new inode.
 2596  * Vnode dvp must be locked.
 2597  */
 2598 static int
 2599 ufs_makeinode(mode, dvp, vpp, cnp)
 2600         int mode;
 2601         struct vnode *dvp;
 2602         struct vnode **vpp;
 2603         struct componentname *cnp;
 2604 {
 2605         struct inode *ip, *pdir;
 2606         struct direct newdir;
 2607         struct vnode *tvp;
 2608         int error;
 2609 
 2610         pdir = VTOI(dvp);
 2611 #ifdef INVARIANTS
 2612         if ((cnp->cn_flags & HASBUF) == 0)
 2613                 panic("ufs_makeinode: no name");
 2614 #endif
 2615         *vpp = NULL;
 2616         if ((mode & IFMT) == 0)
 2617                 mode |= IFREG;
 2618 
 2619         if (VTOI(dvp)->i_effnlink < 2)
 2620                 panic("ufs_makeinode: Bad link count %d on parent",
 2621                     VTOI(dvp)->i_effnlink);
 2622         error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
 2623         if (error)
 2624                 return (error);
 2625         ip = VTOI(tvp);
 2626         ip->i_gid = pdir->i_gid;
 2627         DIP_SET(ip, i_gid, pdir->i_gid);
 2628 #ifdef SUIDDIR
 2629         {
 2630 #ifdef QUOTA
 2631                 struct ucred ucred, *ucp;
 2632                 gid_t ucred_group;
 2633                 ucp = cnp->cn_cred;
 2634 #endif
 2635                 /*
 2636                  * If we are not the owner of the directory,
 2637                  * and we are hacking owners here, (only do this where told to)
 2638                  * and we are not giving it TO root, (would subvert quotas)
 2639                  * then go ahead and give it to the other user.
 2640                  * Note that this drops off the execute bits for security.
 2641                  */
 2642                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 2643                     (pdir->i_mode & ISUID) &&
 2644                     (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
 2645                         ip->i_uid = pdir->i_uid;
 2646                         DIP_SET(ip, i_uid, ip->i_uid);
 2647                         mode &= ~07111;
 2648 #ifdef QUOTA
 2649                         /*
 2650                          * Make sure the correct user gets charged
 2651                          * for the space.
 2652                          * Quickly knock up a dummy credential for the victim.
 2653                          * XXX This seems to never be accessed out of our
 2654                          * context so a stack variable is ok.
 2655                          */
 2656                         refcount_init(&ucred.cr_ref, 1);
 2657                         ucred.cr_uid = ip->i_uid;
 2658                         ucred.cr_ngroups = 1;
 2659                         ucred.cr_groups = &ucred_group;
 2660                         ucred.cr_groups[0] = pdir->i_gid;
 2661                         ucp = &ucred;
 2662 #endif
 2663                 } else {
 2664                         ip->i_uid = cnp->cn_cred->cr_uid;
 2665                         DIP_SET(ip, i_uid, ip->i_uid);
 2666                 }
 2667 
 2668 #ifdef QUOTA
 2669                 if ((error = getinoquota(ip)) ||
 2670                     (error = chkiq(ip, 1, ucp, 0))) {
 2671                         if (DOINGSOFTDEP(tvp))
 2672                                 softdep_revert_link(pdir, ip);
 2673                         UFS_VFREE(tvp, ip->i_number, mode);
 2674                         vput(tvp);
 2675                         return (error);
 2676                 }
 2677 #endif
 2678         }
 2679 #else   /* !SUIDDIR */
 2680         ip->i_uid = cnp->cn_cred->cr_uid;
 2681         DIP_SET(ip, i_uid, ip->i_uid);
 2682 #ifdef QUOTA
 2683         if ((error = getinoquota(ip)) ||
 2684             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 2685                 if (DOINGSOFTDEP(tvp))
 2686                         softdep_revert_link(pdir, ip);
 2687                 UFS_VFREE(tvp, ip->i_number, mode);
 2688                 vput(tvp);
 2689                 return (error);
 2690         }
 2691 #endif
 2692 #endif  /* !SUIDDIR */
 2693         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 2694         ip->i_mode = mode;
 2695         DIP_SET(ip, i_mode, mode);
 2696         tvp->v_type = IFTOVT(mode);     /* Rest init'd in getnewvnode(). */
 2697         ip->i_effnlink = 1;
 2698         ip->i_nlink = 1;
 2699         DIP_SET(ip, i_nlink, 1);
 2700         if (DOINGSOFTDEP(tvp))
 2701                 softdep_setup_create(VTOI(dvp), ip);
 2702         if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
 2703             priv_check_cred(cnp->cn_cred, PRIV_VFS_SETGID, 0)) {
 2704                 ip->i_mode &= ~ISGID;
 2705                 DIP_SET(ip, i_mode, ip->i_mode);
 2706         }
 2707 
 2708         if (cnp->cn_flags & ISWHITEOUT) {
 2709                 ip->i_flags |= UF_OPAQUE;
 2710                 DIP_SET(ip, i_flags, ip->i_flags);
 2711         }
 2712 
 2713         /*
 2714          * Make sure inode goes to disk before directory entry.
 2715          */
 2716         error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
 2717         if (error)
 2718                 goto bad;
 2719 #ifdef MAC
 2720         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 2721                 error = mac_vnode_create_extattr(cnp->cn_cred, dvp->v_mount,
 2722                     dvp, tvp, cnp);
 2723                 if (error)
 2724                         goto bad;
 2725         }
 2726 #endif
 2727 #ifdef UFS_ACL
 2728         if (dvp->v_mount->mnt_flag & MNT_ACLS) {
 2729                 error = ufs_do_posix1e_acl_inheritance_file(dvp, tvp, mode,
 2730                     cnp->cn_cred, cnp->cn_thread);
 2731                 if (error)
 2732                         goto bad;
 2733         } else if (dvp->v_mount->mnt_flag & MNT_NFS4ACLS) {
 2734                 error = ufs_do_nfs4_acl_inheritance(dvp, tvp, mode,
 2735                     cnp->cn_cred, cnp->cn_thread);
 2736                 if (error)
 2737                         goto bad;
 2738         }
 2739 #endif /* !UFS_ACL */
 2740         ufs_makedirentry(ip, cnp, &newdir);
 2741         error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL, 0);
 2742         if (error)
 2743                 goto bad;
 2744         *vpp = tvp;
 2745         return (0);
 2746 
 2747 bad:
 2748         /*
 2749          * Write error occurred trying to update the inode
 2750          * or the directory so must deallocate the inode.
 2751          */
 2752         ip->i_effnlink = 0;
 2753         ip->i_nlink = 0;
 2754         DIP_SET(ip, i_nlink, 0);
 2755         ip->i_flag |= IN_CHANGE;
 2756         if (DOINGSOFTDEP(tvp))
 2757                 softdep_revert_create(VTOI(dvp), ip);
 2758         vput(tvp);
 2759         return (error);
 2760 }
 2761 
 2762 static int
 2763 ufs_ioctl(struct vop_ioctl_args *ap)
 2764 {
 2765 
 2766         switch (ap->a_command) {
 2767         case FIOSEEKDATA:
 2768         case FIOSEEKHOLE:
 2769                 return (vn_bmap_seekhole(ap->a_vp, ap->a_command,
 2770                     (off_t *)ap->a_data, ap->a_cred));
 2771         default:
 2772                 return (ENOTTY);
 2773         }
 2774 }
 2775 
 2776 /* Global vfs data structures for ufs. */
 2777 struct vop_vector ufs_vnodeops = {
 2778         .vop_default =          &default_vnodeops,
 2779         .vop_fsync =            VOP_PANIC,
 2780         .vop_read =             VOP_PANIC,
 2781         .vop_reallocblks =      VOP_PANIC,
 2782         .vop_write =            VOP_PANIC,
 2783         .vop_accessx =          ufs_accessx,
 2784         .vop_bmap =             ufs_bmap,
 2785         .vop_cachedlookup =     ufs_lookup,
 2786         .vop_close =            ufs_close,
 2787         .vop_create =           ufs_create,
 2788         .vop_getattr =          ufs_getattr,
 2789         .vop_inactive =         ufs_inactive,
 2790         .vop_ioctl =            ufs_ioctl,
 2791         .vop_link =             ufs_link,
 2792         .vop_lookup =           vfs_cache_lookup,
 2793         .vop_markatime =        ufs_markatime,
 2794         .vop_mkdir =            ufs_mkdir,
 2795         .vop_mknod =            ufs_mknod,
 2796         .vop_open =             ufs_open,
 2797         .vop_pathconf =         ufs_pathconf,
 2798         .vop_poll =             vop_stdpoll,
 2799         .vop_print =            ufs_print,
 2800         .vop_readdir =          ufs_readdir,
 2801         .vop_readlink =         ufs_readlink,
 2802         .vop_reclaim =          ufs_reclaim,
 2803         .vop_remove =           ufs_remove,
 2804         .vop_rename =           ufs_rename,
 2805         .vop_rmdir =            ufs_rmdir,
 2806         .vop_setattr =          ufs_setattr,
 2807 #ifdef MAC
 2808         .vop_setlabel =         vop_stdsetlabel_ea,
 2809 #endif
 2810         .vop_strategy =         ufs_strategy,
 2811         .vop_symlink =          ufs_symlink,
 2812         .vop_whiteout =         ufs_whiteout,
 2813 #ifdef UFS_EXTATTR
 2814         .vop_getextattr =       ufs_getextattr,
 2815         .vop_deleteextattr =    ufs_deleteextattr,
 2816         .vop_setextattr =       ufs_setextattr,
 2817 #endif
 2818 #ifdef UFS_ACL
 2819         .vop_getacl =           ufs_getacl,
 2820         .vop_setacl =           ufs_setacl,
 2821         .vop_aclcheck =         ufs_aclcheck,
 2822 #endif
 2823 };
 2824 
 2825 struct vop_vector ufs_fifoops = {
 2826         .vop_default =          &fifo_specops,
 2827         .vop_fsync =            VOP_PANIC,
 2828         .vop_accessx =          ufs_accessx,
 2829         .vop_close =            ufsfifo_close,
 2830         .vop_getattr =          ufs_getattr,
 2831         .vop_inactive =         ufs_inactive,
 2832         .vop_kqfilter =         ufsfifo_kqfilter,
 2833         .vop_markatime =        ufs_markatime,
 2834         .vop_pathconf =         ufsfifo_pathconf,
 2835         .vop_print =            ufs_print,
 2836         .vop_read =             VOP_PANIC,
 2837         .vop_reclaim =          ufs_reclaim,
 2838         .vop_setattr =          ufs_setattr,
 2839 #ifdef MAC
 2840         .vop_setlabel =         vop_stdsetlabel_ea,
 2841 #endif
 2842         .vop_write =            VOP_PANIC,
 2843 #ifdef UFS_EXTATTR
 2844         .vop_getextattr =       ufs_getextattr,
 2845         .vop_deleteextattr =    ufs_deleteextattr,
 2846         .vop_setextattr =       ufs_setextattr,
 2847 #endif
 2848 #ifdef UFS_ACL
 2849         .vop_getacl =           ufs_getacl,
 2850         .vop_setacl =           ufs_setacl,
 2851         .vop_aclcheck =         ufs_aclcheck,
 2852 #endif
 2853 };

Cache object: 1fb55987d23f000b8247f661acaae59f


[ 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.