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/10.1/sys/ufs/ufs/ufs_vnops.c 270695 2014-08-27 01:38:26Z kib $");
   39 
   40 #include "opt_quota.h"
   41 #include "opt_suiddir.h"
   42 #include "opt_ufs.h"
   43 #include "opt_ffs.h"
   44 
   45 #include <sys/param.h>
   46 #include <sys/systm.h>
   47 #include <sys/malloc.h>
   48 #include <sys/namei.h>
   49 #include <sys/kernel.h>
   50 #include <sys/fcntl.h>
   51 #include <sys/filio.h>
   52 #include <sys/stat.h>
   53 #include <sys/bio.h>
   54 #include <sys/buf.h>
   55 #include <sys/mount.h>
   56 #include <sys/priv.h>
   57 #include <sys/refcount.h>
   58 #include <sys/unistd.h>
   59 #include <sys/vnode.h>
   60 #include <sys/dirent.h>
   61 #include <sys/lockf.h>
   62 #include <sys/conf.h>
   63 #include <sys/acl.h>
   64 
   65 #include <security/mac/mac_framework.h>
   66 
   67 #include <sys/file.h>           /* XXX */
   68 
   69 #include <vm/vm.h>
   70 #include <vm/vm_extern.h>
   71 
   72 #include <ufs/ufs/acl.h>
   73 #include <ufs/ufs/extattr.h>
   74 #include <ufs/ufs/quota.h>
   75 #include <ufs/ufs/inode.h>
   76 #include <ufs/ufs/dir.h>
   77 #include <ufs/ufs/ufsmount.h>
   78 #include <ufs/ufs/ufs_extern.h>
   79 #ifdef UFS_DIRHASH
   80 #include <ufs/ufs/dirhash.h>
   81 #endif
   82 #ifdef UFS_GJOURNAL
   83 #include <ufs/ufs/gjournal.h>
   84 FEATURE(ufs_gjournal, "Journaling support through GEOM for UFS");
   85 #endif
   86 
   87 #ifdef QUOTA
   88 FEATURE(ufs_quota, "UFS disk quotas support");
   89 FEATURE(ufs_quota64, "64bit UFS disk quotas support");
   90 #endif
   91 
   92 #ifdef SUIDDIR
   93 FEATURE(suiddir,
   94     "Give all new files in directory the same ownership as the directory");
   95 #endif
   96 
   97 
   98 #include <ufs/ffs/ffs_extern.h>
   99 
  100 static vop_accessx_t    ufs_accessx;
  101 static int ufs_chmod(struct vnode *, int, struct ucred *, struct thread *);
  102 static int ufs_chown(struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
  103 static vop_close_t      ufs_close;
  104 static vop_create_t     ufs_create;
  105 static vop_getattr_t    ufs_getattr;
  106 static vop_ioctl_t      ufs_ioctl;
  107 static vop_link_t       ufs_link;
  108 static int ufs_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *);
  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                 error = vn_utimes_perm(vp, vap, cred, td);
  639                 if (error != 0)
  640                         return (error);
  641                 if (vap->va_atime.tv_sec != VNOVAL)
  642                         ip->i_flag |= IN_ACCESS;
  643                 if (vap->va_mtime.tv_sec != VNOVAL)
  644                         ip->i_flag |= IN_CHANGE | IN_UPDATE;
  645                 if (vap->va_birthtime.tv_sec != VNOVAL &&
  646                     ip->i_ump->um_fstype == UFS2)
  647                         ip->i_flag |= IN_MODIFIED;
  648                 ufs_itimes(vp);
  649                 if (vap->va_atime.tv_sec != VNOVAL) {
  650                         DIP_SET(ip, i_atime, vap->va_atime.tv_sec);
  651                         DIP_SET(ip, i_atimensec, vap->va_atime.tv_nsec);
  652                 }
  653                 if (vap->va_mtime.tv_sec != VNOVAL) {
  654                         DIP_SET(ip, i_mtime, vap->va_mtime.tv_sec);
  655                         DIP_SET(ip, i_mtimensec, vap->va_mtime.tv_nsec);
  656                 }
  657                 if (vap->va_birthtime.tv_sec != VNOVAL &&
  658                     ip->i_ump->um_fstype == UFS2) {
  659                         ip->i_din2->di_birthtime = vap->va_birthtime.tv_sec;
  660                         ip->i_din2->di_birthnsec = vap->va_birthtime.tv_nsec;
  661                 }
  662                 error = UFS_UPDATE(vp, 0);
  663                 if (error)
  664                         return (error);
  665         }
  666         error = 0;
  667         if (vap->va_mode != (mode_t)VNOVAL) {
  668                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  669                         return (EROFS);
  670                 if ((ip->i_flags & SF_SNAPSHOT) != 0 && (vap->va_mode &
  671                    (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP | S_IXOTH | S_IWOTH)))
  672                         return (EPERM);
  673                 error = ufs_chmod(vp, (int)vap->va_mode, cred, td);
  674         }
  675         return (error);
  676 }
  677 
  678 #ifdef UFS_ACL
  679 static int
  680 ufs_update_nfs4_acl_after_mode_change(struct vnode *vp, int mode,
  681     int file_owner_id, struct ucred *cred, struct thread *td)
  682 {
  683         int error;
  684         struct acl *aclp;
  685 
  686         aclp = acl_alloc(M_WAITOK);
  687         error = ufs_getacl_nfs4_internal(vp, aclp, td);
  688         /*
  689          * We don't have to handle EOPNOTSUPP here, as the filesystem claims
  690          * it supports ACLs.
  691          */
  692         if (error)
  693                 goto out;
  694 
  695         acl_nfs4_sync_acl_from_mode(aclp, mode, file_owner_id);
  696         error = ufs_setacl_nfs4_internal(vp, aclp, td);
  697 
  698 out:
  699         acl_free(aclp);
  700         return (error);
  701 }
  702 #endif /* UFS_ACL */
  703 
  704 /*
  705  * Mark this file's access time for update for vfs_mark_atime().  This
  706  * is called from execve() and mmap().
  707  */
  708 static int
  709 ufs_markatime(ap)
  710         struct vop_markatime_args /* {
  711                 struct vnode *a_vp;
  712         } */ *ap;
  713 {
  714         struct vnode *vp = ap->a_vp;
  715         struct inode *ip = VTOI(vp);
  716 
  717         VI_LOCK(vp);
  718         ip->i_flag |= IN_ACCESS;
  719         VI_UNLOCK(vp);
  720         /*
  721          * XXXKIB No UFS_UPDATE(ap->a_vp, 0) there.
  722          */
  723         return (0);
  724 }
  725 
  726 /*
  727  * Change the mode on a file.
  728  * Inode must be locked before calling.
  729  */
  730 static int
  731 ufs_chmod(vp, mode, cred, td)
  732         struct vnode *vp;
  733         int mode;
  734         struct ucred *cred;
  735         struct thread *td;
  736 {
  737         struct inode *ip = VTOI(vp);
  738         int error;
  739 
  740         /*
  741          * To modify the permissions on a file, must possess VADMIN
  742          * for that file.
  743          */
  744         if ((error = VOP_ACCESSX(vp, VWRITE_ACL, cred, td)))
  745                 return (error);
  746         /*
  747          * Privileged processes may set the sticky bit on non-directories,
  748          * as well as set the setgid bit on a file with a group that the
  749          * process is not a member of.  Both of these are allowed in
  750          * jail(8).
  751          */
  752         if (vp->v_type != VDIR && (mode & S_ISTXT)) {
  753                 if (priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0))
  754                         return (EFTYPE);
  755         }
  756         if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
  757                 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0);
  758                 if (error)
  759                         return (error);
  760         }
  761 
  762         /*
  763          * Deny setting setuid if we are not the file owner.
  764          */
  765         if ((mode & ISUID) && ip->i_uid != cred->cr_uid) {
  766                 error = priv_check_cred(cred, PRIV_VFS_ADMIN, 0);
  767                 if (error)
  768                         return (error);
  769         }
  770 
  771         ip->i_mode &= ~ALLPERMS;
  772         ip->i_mode |= (mode & ALLPERMS);
  773         DIP_SET(ip, i_mode, ip->i_mode);
  774         ip->i_flag |= IN_CHANGE;
  775 #ifdef UFS_ACL
  776         if ((vp->v_mount->mnt_flag & MNT_NFS4ACLS) != 0)
  777                 error = ufs_update_nfs4_acl_after_mode_change(vp, mode, ip->i_uid, cred, td);
  778 #endif
  779         if (error == 0 && (ip->i_flag & IN_CHANGE) != 0)
  780                 error = UFS_UPDATE(vp, 0);
  781 
  782         return (error);
  783 }
  784 
  785 /*
  786  * Perform chown operation on inode ip;
  787  * inode must be locked prior to call.
  788  */
  789 static int
  790 ufs_chown(vp, uid, gid, cred, td)
  791         struct vnode *vp;
  792         uid_t uid;
  793         gid_t gid;
  794         struct ucred *cred;
  795         struct thread *td;
  796 {
  797         struct inode *ip = VTOI(vp);
  798         uid_t ouid;
  799         gid_t ogid;
  800         int error = 0;
  801 #ifdef QUOTA
  802         int i;
  803         ufs2_daddr_t change;
  804 #endif
  805 
  806         if (uid == (uid_t)VNOVAL)
  807                 uid = ip->i_uid;
  808         if (gid == (gid_t)VNOVAL)
  809                 gid = ip->i_gid;
  810         /*
  811          * To modify the ownership of a file, must possess VADMIN for that
  812          * file.
  813          */
  814         if ((error = VOP_ACCESSX(vp, VWRITE_OWNER, cred, td)))
  815                 return (error);
  816         /*
  817          * To change the owner of a file, or change the group of a file to a
  818          * group of which we are not a member, the caller must have
  819          * privilege.
  820          */
  821         if (((uid != ip->i_uid && uid != cred->cr_uid) || 
  822             (gid != ip->i_gid && !groupmember(gid, cred))) &&
  823             (error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0)))
  824                 return (error);
  825         ogid = ip->i_gid;
  826         ouid = ip->i_uid;
  827 #ifdef QUOTA
  828         if ((error = getinoquota(ip)) != 0)
  829                 return (error);
  830         if (ouid == uid) {
  831                 dqrele(vp, ip->i_dquot[USRQUOTA]);
  832                 ip->i_dquot[USRQUOTA] = NODQUOT;
  833         }
  834         if (ogid == gid) {
  835                 dqrele(vp, ip->i_dquot[GRPQUOTA]);
  836                 ip->i_dquot[GRPQUOTA] = NODQUOT;
  837         }
  838         change = DIP(ip, i_blocks);
  839         (void) chkdq(ip, -change, cred, CHOWN);
  840         (void) chkiq(ip, -1, cred, CHOWN);
  841         for (i = 0; i < MAXQUOTAS; i++) {
  842                 dqrele(vp, ip->i_dquot[i]);
  843                 ip->i_dquot[i] = NODQUOT;
  844         }
  845 #endif
  846         ip->i_gid = gid;
  847         DIP_SET(ip, i_gid, gid);
  848         ip->i_uid = uid;
  849         DIP_SET(ip, i_uid, uid);
  850 #ifdef QUOTA
  851         if ((error = getinoquota(ip)) == 0) {
  852                 if (ouid == uid) {
  853                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  854                         ip->i_dquot[USRQUOTA] = NODQUOT;
  855                 }
  856                 if (ogid == gid) {
  857                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  858                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  859                 }
  860                 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
  861                         if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
  862                                 goto good;
  863                         else
  864                                 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
  865                 }
  866                 for (i = 0; i < MAXQUOTAS; i++) {
  867                         dqrele(vp, ip->i_dquot[i]);
  868                         ip->i_dquot[i] = NODQUOT;
  869                 }
  870         }
  871         ip->i_gid = ogid;
  872         DIP_SET(ip, i_gid, ogid);
  873         ip->i_uid = ouid;
  874         DIP_SET(ip, i_uid, ouid);
  875         if (getinoquota(ip) == 0) {
  876                 if (ouid == uid) {
  877                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  878                         ip->i_dquot[USRQUOTA] = NODQUOT;
  879                 }
  880                 if (ogid == gid) {
  881                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  882                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  883                 }
  884                 (void) chkdq(ip, change, cred, FORCE|CHOWN);
  885                 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
  886                 (void) getinoquota(ip);
  887         }
  888         return (error);
  889 good:
  890         if (getinoquota(ip))
  891                 panic("ufs_chown: lost quota");
  892 #endif /* QUOTA */
  893         ip->i_flag |= IN_CHANGE;
  894         if ((ip->i_mode & (ISUID | ISGID)) && (ouid != uid || ogid != gid)) {
  895                 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0)) {
  896                         ip->i_mode &= ~(ISUID | ISGID);
  897                         DIP_SET(ip, i_mode, ip->i_mode);
  898                 }
  899         }
  900         error = UFS_UPDATE(vp, 0);
  901         return (error);
  902 }
  903 
  904 static int
  905 ufs_remove(ap)
  906         struct vop_remove_args /* {
  907                 struct vnode *a_dvp;
  908                 struct vnode *a_vp;
  909                 struct componentname *a_cnp;
  910         } */ *ap;
  911 {
  912         struct inode *ip;
  913         struct vnode *vp = ap->a_vp;
  914         struct vnode *dvp = ap->a_dvp;
  915         int error;
  916         struct thread *td;
  917 
  918         td = curthread;
  919         ip = VTOI(vp);
  920         if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
  921             (VTOI(dvp)->i_flags & APPEND)) {
  922                 error = EPERM;
  923                 goto out;
  924         }
  925 #ifdef UFS_GJOURNAL
  926         ufs_gjournal_orphan(vp);
  927 #endif
  928         error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
  929         if (ip->i_nlink <= 0)
  930                 vp->v_vflag |= VV_NOSYNC;
  931         if ((ip->i_flags & SF_SNAPSHOT) != 0) {
  932                 /*
  933                  * Avoid deadlock where another thread is trying to
  934                  * update the inodeblock for dvp and is waiting on
  935                  * snaplk.  Temporary unlock the vnode lock for the
  936                  * unlinked file and sync the directory.  This should
  937                  * allow vput() of the directory to not block later on
  938                  * while holding the snapshot vnode locked, assuming
  939                  * that the directory hasn't been unlinked too.
  940                  */
  941                 VOP_UNLOCK(vp, 0);
  942                 (void) VOP_FSYNC(dvp, MNT_WAIT, td);
  943                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  944         }
  945 out:
  946         return (error);
  947 }
  948 
  949 /*
  950  * link vnode call
  951  */
  952 static int
  953 ufs_link(ap)
  954         struct vop_link_args /* {
  955                 struct vnode *a_tdvp;
  956                 struct vnode *a_vp;
  957                 struct componentname *a_cnp;
  958         } */ *ap;
  959 {
  960         struct vnode *vp = ap->a_vp;
  961         struct vnode *tdvp = ap->a_tdvp;
  962         struct componentname *cnp = ap->a_cnp;
  963         struct inode *ip;
  964         struct direct newdir;
  965         int error;
  966 
  967 #ifdef INVARIANTS
  968         if ((cnp->cn_flags & HASBUF) == 0)
  969                 panic("ufs_link: no name");
  970 #endif
  971         if (VTOI(tdvp)->i_effnlink < 2)
  972                 panic("ufs_link: Bad link count %d on parent",
  973                     VTOI(tdvp)->i_effnlink);
  974         ip = VTOI(vp);
  975         if ((nlink_t)ip->i_nlink >= LINK_MAX) {
  976                 error = EMLINK;
  977                 goto out;
  978         }
  979         /*
  980          * The file may have been removed after namei droped the original
  981          * lock.
  982          */
  983         if (ip->i_effnlink == 0) {
  984                 error = ENOENT;
  985                 goto out;
  986         }
  987         if (ip->i_flags & (IMMUTABLE | APPEND)) {
  988                 error = EPERM;
  989                 goto out;
  990         }
  991         ip->i_effnlink++;
  992         ip->i_nlink++;
  993         DIP_SET(ip, i_nlink, ip->i_nlink);
  994         ip->i_flag |= IN_CHANGE;
  995         if (DOINGSOFTDEP(vp))
  996                 softdep_setup_link(VTOI(tdvp), ip);
  997         error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
  998         if (!error) {
  999                 ufs_makedirentry(ip, cnp, &newdir);
 1000                 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL, 0);
 1001         }
 1002 
 1003         if (error) {
 1004                 ip->i_effnlink--;
 1005                 ip->i_nlink--;
 1006                 DIP_SET(ip, i_nlink, ip->i_nlink);
 1007                 ip->i_flag |= IN_CHANGE;
 1008                 if (DOINGSOFTDEP(vp))
 1009                         softdep_revert_link(VTOI(tdvp), ip);
 1010         }
 1011 out:
 1012         return (error);
 1013 }
 1014 
 1015 /*
 1016  * whiteout vnode call
 1017  */
 1018 static int
 1019 ufs_whiteout(ap)
 1020         struct vop_whiteout_args /* {
 1021                 struct vnode *a_dvp;
 1022                 struct componentname *a_cnp;
 1023                 int a_flags;
 1024         } */ *ap;
 1025 {
 1026         struct vnode *dvp = ap->a_dvp;
 1027         struct componentname *cnp = ap->a_cnp;
 1028         struct direct newdir;
 1029         int error = 0;
 1030 
 1031         switch (ap->a_flags) {
 1032         case LOOKUP:
 1033                 /* 4.4 format directories support whiteout operations */
 1034                 if (dvp->v_mount->mnt_maxsymlinklen > 0)
 1035                         return (0);
 1036                 return (EOPNOTSUPP);
 1037 
 1038         case CREATE:
 1039                 /* create a new directory whiteout */
 1040 #ifdef INVARIANTS
 1041                 if ((cnp->cn_flags & SAVENAME) == 0)
 1042                         panic("ufs_whiteout: missing name");
 1043                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
 1044                         panic("ufs_whiteout: old format filesystem");
 1045 #endif
 1046 
 1047                 newdir.d_ino = WINO;
 1048                 newdir.d_namlen = cnp->cn_namelen;
 1049                 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
 1050                 newdir.d_type = DT_WHT;
 1051                 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL, 0);
 1052                 break;
 1053 
 1054         case DELETE:
 1055                 /* remove an existing directory whiteout */
 1056 #ifdef INVARIANTS
 1057                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
 1058                         panic("ufs_whiteout: old format filesystem");
 1059 #endif
 1060 
 1061                 cnp->cn_flags &= ~DOWHITEOUT;
 1062                 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
 1063                 break;
 1064         default:
 1065                 panic("ufs_whiteout: unknown op");
 1066         }
 1067         return (error);
 1068 }
 1069 
 1070 static volatile int rename_restarts;
 1071 SYSCTL_INT(_vfs_ufs, OID_AUTO, rename_restarts, CTLFLAG_RD,
 1072     __DEVOLATILE(int *, &rename_restarts), 0,
 1073     "Times rename had to restart due to lock contention");
 1074 
 1075 /*
 1076  * Rename system call.
 1077  *      rename("foo", "bar");
 1078  * is essentially
 1079  *      unlink("bar");
 1080  *      link("foo", "bar");
 1081  *      unlink("foo");
 1082  * but ``atomically''.  Can't do full commit without saving state in the
 1083  * inode on disk which isn't feasible at this time.  Best we can do is
 1084  * always guarantee the target exists.
 1085  *
 1086  * Basic algorithm is:
 1087  *
 1088  * 1) Bump link count on source while we're linking it to the
 1089  *    target.  This also ensure the inode won't be deleted out
 1090  *    from underneath us while we work (it may be truncated by
 1091  *    a concurrent `trunc' or `open' for creation).
 1092  * 2) Link source to destination.  If destination already exists,
 1093  *    delete it first.
 1094  * 3) Unlink source reference to inode if still around. If a
 1095  *    directory was moved and the parent of the destination
 1096  *    is different from the source, patch the ".." entry in the
 1097  *    directory.
 1098  */
 1099 static int
 1100 ufs_rename(ap)
 1101         struct vop_rename_args  /* {
 1102                 struct vnode *a_fdvp;
 1103                 struct vnode *a_fvp;
 1104                 struct componentname *a_fcnp;
 1105                 struct vnode *a_tdvp;
 1106                 struct vnode *a_tvp;
 1107                 struct componentname *a_tcnp;
 1108         } */ *ap;
 1109 {
 1110         struct vnode *tvp = ap->a_tvp;
 1111         struct vnode *tdvp = ap->a_tdvp;
 1112         struct vnode *fvp = ap->a_fvp;
 1113         struct vnode *fdvp = ap->a_fdvp;
 1114         struct vnode *nvp;
 1115         struct componentname *tcnp = ap->a_tcnp;
 1116         struct componentname *fcnp = ap->a_fcnp;
 1117         struct thread *td = fcnp->cn_thread;
 1118         struct inode *fip, *tip, *tdp, *fdp;
 1119         struct direct newdir;
 1120         off_t endoff;
 1121         int doingdirectory, newparent;
 1122         int error = 0;
 1123         struct mount *mp;
 1124         ino_t ino;
 1125 
 1126 #ifdef INVARIANTS
 1127         if ((tcnp->cn_flags & HASBUF) == 0 ||
 1128             (fcnp->cn_flags & HASBUF) == 0)
 1129                 panic("ufs_rename: no name");
 1130 #endif
 1131         endoff = 0;
 1132         mp = tdvp->v_mount;
 1133         VOP_UNLOCK(tdvp, 0);
 1134         if (tvp && tvp != tdvp)
 1135                 VOP_UNLOCK(tvp, 0);
 1136         /*
 1137          * Check for cross-device rename.
 1138          */
 1139         if ((fvp->v_mount != tdvp->v_mount) ||
 1140             (tvp && (fvp->v_mount != tvp->v_mount))) {
 1141                 error = EXDEV;
 1142                 mp = NULL;
 1143                 goto releout;
 1144         }
 1145 relock:
 1146         /* 
 1147          * We need to acquire 2 to 4 locks depending on whether tvp is NULL
 1148          * and fdvp and tdvp are the same directory.  Subsequently we need
 1149          * to double-check all paths and in the directory rename case we
 1150          * need to verify that we are not creating a directory loop.  To
 1151          * handle this we acquire all but fdvp using non-blocking
 1152          * acquisitions.  If we fail to acquire any lock in the path we will
 1153          * drop all held locks, acquire the new lock in a blocking fashion,
 1154          * and then release it and restart the rename.  This acquire/release
 1155          * step ensures that we do not spin on a lock waiting for release.
 1156          */
 1157         error = vn_lock(fdvp, LK_EXCLUSIVE);
 1158         if (error)
 1159                 goto releout;
 1160         if (vn_lock(tdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
 1161                 VOP_UNLOCK(fdvp, 0);
 1162                 error = vn_lock(tdvp, LK_EXCLUSIVE);
 1163                 if (error)
 1164                         goto releout;
 1165                 VOP_UNLOCK(tdvp, 0);
 1166                 atomic_add_int(&rename_restarts, 1);
 1167                 goto relock;
 1168         }
 1169         /*
 1170          * Re-resolve fvp to be certain it still exists and fetch the
 1171          * correct vnode.
 1172          */
 1173         error = ufs_lookup_ino(fdvp, NULL, fcnp, &ino);
 1174         if (error) {
 1175                 VOP_UNLOCK(fdvp, 0);
 1176                 VOP_UNLOCK(tdvp, 0);
 1177                 goto releout;
 1178         }
 1179         error = VFS_VGET(mp, ino, LK_EXCLUSIVE | LK_NOWAIT, &nvp);
 1180         if (error) {
 1181                 VOP_UNLOCK(fdvp, 0);
 1182                 VOP_UNLOCK(tdvp, 0);
 1183                 if (error != EBUSY)
 1184                         goto releout;
 1185                 error = VFS_VGET(mp, ino, LK_EXCLUSIVE, &nvp);
 1186                 if (error != 0)
 1187                         goto releout;
 1188                 VOP_UNLOCK(nvp, 0);
 1189                 vrele(fvp);
 1190                 fvp = nvp;
 1191                 atomic_add_int(&rename_restarts, 1);
 1192                 goto relock;
 1193         }
 1194         vrele(fvp);
 1195         fvp = nvp;
 1196         /*
 1197          * Re-resolve tvp and acquire the vnode lock if present.
 1198          */
 1199         error = ufs_lookup_ino(tdvp, NULL, tcnp, &ino);
 1200         if (error != 0 && error != EJUSTRETURN) {
 1201                 VOP_UNLOCK(fdvp, 0);
 1202                 VOP_UNLOCK(tdvp, 0);
 1203                 VOP_UNLOCK(fvp, 0);
 1204                 goto releout;
 1205         }
 1206         /*
 1207          * If tvp disappeared we just carry on.
 1208          */
 1209         if (error == EJUSTRETURN && tvp != NULL) {
 1210                 vrele(tvp);
 1211                 tvp = NULL;
 1212         }
 1213         /*
 1214          * Get the tvp ino if the lookup succeeded.  We may have to restart
 1215          * if the non-blocking acquire fails.
 1216          */
 1217         if (error == 0) {
 1218                 nvp = NULL;
 1219                 error = VFS_VGET(mp, ino, LK_EXCLUSIVE | LK_NOWAIT, &nvp);
 1220                 if (tvp)
 1221                         vrele(tvp);
 1222                 tvp = nvp;
 1223                 if (error) {
 1224                         VOP_UNLOCK(fdvp, 0);
 1225                         VOP_UNLOCK(tdvp, 0);
 1226                         VOP_UNLOCK(fvp, 0);
 1227                         if (error != EBUSY)
 1228                                 goto releout;
 1229                         error = VFS_VGET(mp, ino, LK_EXCLUSIVE, &nvp);
 1230                         if (error != 0)
 1231                                 goto releout;
 1232                         vput(nvp);
 1233                         atomic_add_int(&rename_restarts, 1);
 1234                         goto relock;
 1235                 }
 1236         }
 1237         fdp = VTOI(fdvp);
 1238         fip = VTOI(fvp);
 1239         tdp = VTOI(tdvp);
 1240         tip = NULL;
 1241         if (tvp)
 1242                 tip = VTOI(tvp);
 1243         if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
 1244             (VTOI(tdvp)->i_flags & APPEND))) {
 1245                 error = EPERM;
 1246                 goto unlockout;
 1247         }
 1248         /*
 1249          * Renaming a file to itself has no effect.  The upper layers should
 1250          * not call us in that case.  However, things could change after
 1251          * we drop the locks above.
 1252          */
 1253         if (fvp == tvp) {
 1254                 error = 0;
 1255                 goto unlockout;
 1256         }
 1257         doingdirectory = 0;
 1258         newparent = 0;
 1259         ino = fip->i_number;
 1260         if (fip->i_nlink >= LINK_MAX) {
 1261                 error = EMLINK;
 1262                 goto unlockout;
 1263         }
 1264         if ((fip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
 1265             || (fdp->i_flags & APPEND)) {
 1266                 error = EPERM;
 1267                 goto unlockout;
 1268         }
 1269         if ((fip->i_mode & IFMT) == IFDIR) {
 1270                 /*
 1271                  * Avoid ".", "..", and aliases of "." for obvious reasons.
 1272                  */
 1273                 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
 1274                     fdp == fip ||
 1275                     (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT) {
 1276                         error = EINVAL;
 1277                         goto unlockout;
 1278                 }
 1279                 if (fdp->i_number != tdp->i_number)
 1280                         newparent = tdp->i_number;
 1281                 doingdirectory = 1;
 1282         }
 1283         if ((fvp->v_type == VDIR && fvp->v_mountedhere != NULL) ||
 1284             (tvp != NULL && tvp->v_type == VDIR &&
 1285             tvp->v_mountedhere != NULL)) {
 1286                 error = EXDEV;
 1287                 goto unlockout;
 1288         }
 1289 
 1290         /*
 1291          * If ".." must be changed (ie the directory gets a new
 1292          * parent) then the source directory must not be in the
 1293          * directory hierarchy above the target, as this would
 1294          * orphan everything below the source directory. Also
 1295          * the user must have write permission in the source so
 1296          * as to be able to change "..".
 1297          */
 1298         if (doingdirectory && newparent) {
 1299                 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
 1300                 if (error)
 1301                         goto unlockout;
 1302                 error = ufs_checkpath(ino, fdp->i_number, tdp, tcnp->cn_cred,
 1303                     &ino);
 1304                 /*
 1305                  * We encountered a lock that we have to wait for.  Unlock
 1306                  * everything else and VGET before restarting.
 1307                  */
 1308                 if (ino) {
 1309                         VOP_UNLOCK(fdvp, 0);
 1310                         VOP_UNLOCK(fvp, 0);
 1311                         VOP_UNLOCK(tdvp, 0);
 1312                         if (tvp)
 1313                                 VOP_UNLOCK(tvp, 0);
 1314                         error = VFS_VGET(mp, ino, LK_SHARED, &nvp);
 1315                         if (error == 0)
 1316                                 vput(nvp);
 1317                         atomic_add_int(&rename_restarts, 1);
 1318                         goto relock;
 1319                 }
 1320                 if (error)
 1321                         goto unlockout;
 1322                 if ((tcnp->cn_flags & SAVESTART) == 0)
 1323                         panic("ufs_rename: lost to startdir");
 1324         }
 1325         if (fip->i_effnlink == 0 || fdp->i_effnlink == 0 ||
 1326             tdp->i_effnlink == 0)
 1327                 panic("Bad effnlink fip %p, fdp %p, tdp %p", fip, fdp, tdp);
 1328 
 1329         /*
 1330          * 1) Bump link count while we're moving stuff
 1331          *    around.  If we crash somewhere before
 1332          *    completing our work, the link count
 1333          *    may be wrong, but correctable.
 1334          */
 1335         fip->i_effnlink++;
 1336         fip->i_nlink++;
 1337         DIP_SET(fip, i_nlink, fip->i_nlink);
 1338         fip->i_flag |= IN_CHANGE;
 1339         if (DOINGSOFTDEP(fvp))
 1340                 softdep_setup_link(tdp, fip);
 1341         error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) | DOINGASYNC(fvp)));
 1342         if (error)
 1343                 goto bad;
 1344 
 1345         /*
 1346          * 2) If target doesn't exist, link the target
 1347          *    to the source and unlink the source.
 1348          *    Otherwise, rewrite the target directory
 1349          *    entry to reference the source inode and
 1350          *    expunge the original entry's existence.
 1351          */
 1352         if (tip == NULL) {
 1353                 if (tdp->i_dev != fip->i_dev)
 1354                         panic("ufs_rename: EXDEV");
 1355                 if (doingdirectory && newparent) {
 1356                         /*
 1357                          * Account for ".." in new directory.
 1358                          * When source and destination have the same
 1359                          * parent we don't adjust the link count.  The
 1360                          * actual link modification is completed when
 1361                          * .. is rewritten below.
 1362                          */
 1363                         if ((nlink_t)tdp->i_nlink >= LINK_MAX) {
 1364                                 error = EMLINK;
 1365                                 goto bad;
 1366                         }
 1367                 }
 1368                 ufs_makedirentry(fip, tcnp, &newdir);
 1369                 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL, 1);
 1370                 if (error)
 1371                         goto bad;
 1372                 /* Setup tdvp for directory compaction if needed. */
 1373                 if (tdp->i_count && tdp->i_endoff &&
 1374                     tdp->i_endoff < tdp->i_size)
 1375                         endoff = tdp->i_endoff;
 1376         } else {
 1377                 if (tip->i_dev != tdp->i_dev || tip->i_dev != fip->i_dev)
 1378                         panic("ufs_rename: EXDEV");
 1379                 /*
 1380                  * Short circuit rename(foo, foo).
 1381                  */
 1382                 if (tip->i_number == fip->i_number)
 1383                         panic("ufs_rename: same file");
 1384                 /*
 1385                  * If the parent directory is "sticky", then the caller
 1386                  * must possess VADMIN for the parent directory, or the
 1387                  * destination of the rename.  This implements append-only
 1388                  * directories.
 1389                  */
 1390                 if ((tdp->i_mode & S_ISTXT) &&
 1391                     VOP_ACCESS(tdvp, VADMIN, tcnp->cn_cred, td) &&
 1392                     VOP_ACCESS(tvp, VADMIN, tcnp->cn_cred, td)) {
 1393                         error = EPERM;
 1394                         goto bad;
 1395                 }
 1396                 /*
 1397                  * Target must be empty if a directory and have no links
 1398                  * to it. Also, ensure source and target are compatible
 1399                  * (both directories, or both not directories).
 1400                  */
 1401                 if ((tip->i_mode & IFMT) == IFDIR) {
 1402                         if ((tip->i_effnlink > 2) ||
 1403                             !ufs_dirempty(tip, tdp->i_number, tcnp->cn_cred)) {
 1404                                 error = ENOTEMPTY;
 1405                                 goto bad;
 1406                         }
 1407                         if (!doingdirectory) {
 1408                                 error = ENOTDIR;
 1409                                 goto bad;
 1410                         }
 1411                         cache_purge(tdvp);
 1412                 } else if (doingdirectory) {
 1413                         error = EISDIR;
 1414                         goto bad;
 1415                 }
 1416                 if (doingdirectory) {
 1417                         if (!newparent) {
 1418                                 tdp->i_effnlink--;
 1419                                 if (DOINGSOFTDEP(tdvp))
 1420                                         softdep_change_linkcnt(tdp);
 1421                         }
 1422                         tip->i_effnlink--;
 1423                         if (DOINGSOFTDEP(tvp))
 1424                                 softdep_change_linkcnt(tip);
 1425                 }
 1426                 error = ufs_dirrewrite(tdp, tip, fip->i_number,
 1427                     IFTODT(fip->i_mode),
 1428                     (doingdirectory && newparent) ? newparent : doingdirectory);
 1429                 if (error) {
 1430                         if (doingdirectory) {
 1431                                 if (!newparent) {
 1432                                         tdp->i_effnlink++;
 1433                                         if (DOINGSOFTDEP(tdvp))
 1434                                                 softdep_change_linkcnt(tdp);
 1435                                 }
 1436                                 tip->i_effnlink++;
 1437                                 if (DOINGSOFTDEP(tvp))
 1438                                         softdep_change_linkcnt(tip);
 1439                         }
 1440                 }
 1441                 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
 1442                         /*
 1443                          * The only stuff left in the directory is "."
 1444                          * and "..". The "." reference is inconsequential
 1445                          * since we are quashing it. We have removed the "."
 1446                          * reference and the reference in the parent directory,
 1447                          * but there may be other hard links. The soft
 1448                          * dependency code will arrange to do these operations
 1449                          * after the parent directory entry has been deleted on
 1450                          * disk, so when running with that code we avoid doing
 1451                          * them now.
 1452                          */
 1453                         if (!newparent) {
 1454                                 tdp->i_nlink--;
 1455                                 DIP_SET(tdp, i_nlink, tdp->i_nlink);
 1456                                 tdp->i_flag |= IN_CHANGE;
 1457                         }
 1458                         tip->i_nlink--;
 1459                         DIP_SET(tip, i_nlink, tip->i_nlink);
 1460                         tip->i_flag |= IN_CHANGE;
 1461                 }
 1462         }
 1463 
 1464         /*
 1465          * 3) Unlink the source.  We have to resolve the path again to
 1466          * fixup the directory offset and count for ufs_dirremove.
 1467          */
 1468         if (fdvp == tdvp) {
 1469                 error = ufs_lookup_ino(fdvp, NULL, fcnp, &ino);
 1470                 if (error)
 1471                         panic("ufs_rename: from entry went away!");
 1472                 if (ino != fip->i_number)
 1473                         panic("ufs_rename: ino mismatch %ju != %ju\n",
 1474                             (uintmax_t)ino, (uintmax_t)fip->i_number);
 1475         }
 1476         /*
 1477          * If the source is a directory with a
 1478          * new parent, the link count of the old
 1479          * parent directory must be decremented
 1480          * and ".." set to point to the new parent.
 1481          */
 1482         if (doingdirectory && newparent) {
 1483                 /*
 1484                  * If tip exists we simply use its link, otherwise we must
 1485                  * add a new one.
 1486                  */
 1487                 if (tip == NULL) {
 1488                         tdp->i_effnlink++;
 1489                         tdp->i_nlink++;
 1490                         DIP_SET(tdp, i_nlink, tdp->i_nlink);
 1491                         tdp->i_flag |= IN_CHANGE;
 1492                         if (DOINGSOFTDEP(tdvp))
 1493                                 softdep_setup_dotdot_link(tdp, fip);
 1494                         error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
 1495                                                    DOINGASYNC(tdvp)));
 1496                         /* Don't go to bad here as the new link exists. */
 1497                         if (error)
 1498                                 goto unlockout;
 1499                 } else if (DOINGSUJ(tdvp))
 1500                         /* Journal must account for each new link. */
 1501                         softdep_setup_dotdot_link(tdp, fip);
 1502                 fip->i_offset = mastertemplate.dot_reclen;
 1503                 ufs_dirrewrite(fip, fdp, newparent, DT_DIR, 0);
 1504                 cache_purge(fdvp);
 1505         }
 1506         error = ufs_dirremove(fdvp, fip, fcnp->cn_flags, 0);
 1507         /*
 1508          * The kern_renameat() looks up the fvp using the DELETE flag, which
 1509          * causes the removal of the name cache entry for fvp.
 1510          * As the relookup of the fvp is done in two steps:
 1511          * ufs_lookup_ino() and then VFS_VGET(), another thread might do a
 1512          * normal lookup of the from name just before the VFS_VGET() call,
 1513          * causing the cache entry to be re-instantiated.
 1514          *
 1515          * The same issue also applies to tvp if it exists as
 1516          * otherwise we may have a stale name cache entry for the new
 1517          * name that references the old i-node if it has other links
 1518          * or open file descriptors.
 1519          */
 1520         cache_purge(fvp);
 1521         if (tvp)
 1522                 cache_purge(tvp);
 1523         cache_purge_negative(tdvp);
 1524 
 1525 unlockout:
 1526         vput(fdvp);
 1527         vput(fvp);
 1528         if (tvp)
 1529                 vput(tvp);
 1530         /*
 1531          * If compaction or fsync was requested do it now that other locks
 1532          * are no longer needed.
 1533          */
 1534         if (error == 0 && endoff != 0) {
 1535 #ifdef UFS_DIRHASH
 1536                 if (tdp->i_dirhash != NULL)
 1537                         ufsdirhash_dirtrunc(tdp, endoff);
 1538 #endif
 1539                 UFS_TRUNCATE(tdvp, endoff, IO_NORMAL | IO_SYNC, tcnp->cn_cred);
 1540         }
 1541         if (error == 0 && tdp->i_flag & IN_NEEDSYNC)
 1542                 error = VOP_FSYNC(tdvp, MNT_WAIT, td);
 1543         vput(tdvp);
 1544         return (error);
 1545 
 1546 bad:
 1547         fip->i_effnlink--;
 1548         fip->i_nlink--;
 1549         DIP_SET(fip, i_nlink, fip->i_nlink);
 1550         fip->i_flag |= IN_CHANGE;
 1551         if (DOINGSOFTDEP(fvp))
 1552                 softdep_revert_link(tdp, fip);
 1553         goto unlockout;
 1554 
 1555 releout:
 1556         vrele(fdvp);
 1557         vrele(fvp);
 1558         vrele(tdvp);
 1559         if (tvp)
 1560                 vrele(tvp);
 1561 
 1562         return (error);
 1563 }
 1564 
 1565 #ifdef UFS_ACL
 1566 static int
 1567 ufs_do_posix1e_acl_inheritance_dir(struct vnode *dvp, struct vnode *tvp,
 1568     mode_t dmode, struct ucred *cred, struct thread *td)
 1569 {
 1570         int error;
 1571         struct inode *ip = VTOI(tvp);
 1572         struct acl *dacl, *acl;
 1573 
 1574         acl = acl_alloc(M_WAITOK);
 1575         dacl = acl_alloc(M_WAITOK);
 1576 
 1577         /*
 1578          * Retrieve default ACL from parent, if any.
 1579          */
 1580         error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
 1581         switch (error) {
 1582         case 0:
 1583                 /*
 1584                  * Retrieved a default ACL, so merge mode and ACL if
 1585                  * necessary.  If the ACL is empty, fall through to
 1586                  * the "not defined or available" case.
 1587                  */
 1588                 if (acl->acl_cnt != 0) {
 1589                         dmode = acl_posix1e_newfilemode(dmode, acl);
 1590                         ip->i_mode = dmode;
 1591                         DIP_SET(ip, i_mode, dmode);
 1592                         *dacl = *acl;
 1593                         ufs_sync_acl_from_inode(ip, acl);
 1594                         break;
 1595                 }
 1596                 /* FALLTHROUGH */
 1597 
 1598         case EOPNOTSUPP:
 1599                 /*
 1600                  * Just use the mode as-is.
 1601                  */
 1602                 ip->i_mode = dmode;
 1603                 DIP_SET(ip, i_mode, dmode);
 1604                 error = 0;
 1605                 goto out;
 1606         
 1607         default:
 1608                 goto out;
 1609         }
 1610 
 1611         /*
 1612          * XXX: If we abort now, will Soft Updates notify the extattr
 1613          * code that the EAs for the file need to be released?
 1614          */
 1615         error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
 1616         if (error == 0)
 1617                 error = VOP_SETACL(tvp, ACL_TYPE_DEFAULT, dacl, cred, td);
 1618         switch (error) {
 1619         case 0:
 1620                 break;
 1621 
 1622         case EOPNOTSUPP:
 1623                 /*
 1624                  * XXX: This should not happen, as EOPNOTSUPP above
 1625                  * was supposed to free acl.
 1626                  */
 1627                 printf("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()\n");
 1628                 /*
 1629                 panic("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()");
 1630                  */
 1631                 break;
 1632 
 1633         default:
 1634                 goto out;
 1635         }
 1636 
 1637 out:
 1638         acl_free(acl);
 1639         acl_free(dacl);
 1640 
 1641         return (error);
 1642 }
 1643 
 1644 static int
 1645 ufs_do_posix1e_acl_inheritance_file(struct vnode *dvp, struct vnode *tvp,
 1646     mode_t mode, struct ucred *cred, struct thread *td)
 1647 {
 1648         int error;
 1649         struct inode *ip = VTOI(tvp);
 1650         struct acl *acl;
 1651 
 1652         acl = acl_alloc(M_WAITOK);
 1653 
 1654         /*
 1655          * Retrieve default ACL for parent, if any.
 1656          */
 1657         error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
 1658         switch (error) {
 1659         case 0:
 1660                 /*
 1661                  * Retrieved a default ACL, so merge mode and ACL if
 1662                  * necessary.
 1663                  */
 1664                 if (acl->acl_cnt != 0) {
 1665                         /*
 1666                          * Two possible ways for default ACL to not
 1667                          * be present.  First, the EA can be
 1668                          * undefined, or second, the default ACL can
 1669                          * be blank.  If it's blank, fall through to
 1670                          * the it's not defined case.
 1671                          */
 1672                         mode = acl_posix1e_newfilemode(mode, acl);
 1673                         ip->i_mode = mode;
 1674                         DIP_SET(ip, i_mode, mode);
 1675                         ufs_sync_acl_from_inode(ip, acl);
 1676                         break;
 1677                 }
 1678                 /* FALLTHROUGH */
 1679 
 1680         case EOPNOTSUPP:
 1681                 /*
 1682                  * Just use the mode as-is.
 1683                  */
 1684                 ip->i_mode = mode;
 1685                 DIP_SET(ip, i_mode, mode);
 1686                 error = 0;
 1687                 goto out;
 1688 
 1689         default:
 1690                 goto out;
 1691         }
 1692 
 1693         /*
 1694          * XXX: If we abort now, will Soft Updates notify the extattr
 1695          * code that the EAs for the file need to be released?
 1696          */
 1697         error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
 1698         switch (error) {
 1699         case 0:
 1700                 break;
 1701 
 1702         case EOPNOTSUPP:
 1703                 /*
 1704                  * XXX: This should not happen, as EOPNOTSUPP above was
 1705                  * supposed to free acl.
 1706                  */
 1707                 printf("ufs_makeinode: VOP_GETACL() but no "
 1708                     "VOP_SETACL()\n");
 1709                 /* panic("ufs_makeinode: VOP_GETACL() but no "
 1710                     "VOP_SETACL()"); */
 1711                 break;
 1712 
 1713         default:
 1714                 goto out;
 1715         }
 1716 
 1717 out:
 1718         acl_free(acl);
 1719 
 1720         return (error);
 1721 }
 1722 
 1723 static int
 1724 ufs_do_nfs4_acl_inheritance(struct vnode *dvp, struct vnode *tvp,
 1725     mode_t child_mode, struct ucred *cred, struct thread *td)
 1726 {
 1727         int error;
 1728         struct acl *parent_aclp, *child_aclp;
 1729 
 1730         parent_aclp = acl_alloc(M_WAITOK);
 1731         child_aclp = acl_alloc(M_WAITOK | M_ZERO);
 1732 
 1733         error = ufs_getacl_nfs4_internal(dvp, parent_aclp, td);
 1734         if (error)
 1735                 goto out;
 1736         acl_nfs4_compute_inherited_acl(parent_aclp, child_aclp,
 1737             child_mode, VTOI(tvp)->i_uid, tvp->v_type == VDIR);
 1738         error = ufs_setacl_nfs4_internal(tvp, child_aclp, td);
 1739         if (error)
 1740                 goto out;
 1741 out:
 1742         acl_free(parent_aclp);
 1743         acl_free(child_aclp);
 1744 
 1745         return (error);
 1746 }
 1747 #endif
 1748 
 1749 /*
 1750  * Mkdir system call
 1751  */
 1752 static int
 1753 ufs_mkdir(ap)
 1754         struct vop_mkdir_args /* {
 1755                 struct vnode *a_dvp;
 1756                 struct vnode **a_vpp;
 1757                 struct componentname *a_cnp;
 1758                 struct vattr *a_vap;
 1759         } */ *ap;
 1760 {
 1761         struct vnode *dvp = ap->a_dvp;
 1762         struct vattr *vap = ap->a_vap;
 1763         struct componentname *cnp = ap->a_cnp;
 1764         struct inode *ip, *dp;
 1765         struct vnode *tvp;
 1766         struct buf *bp;
 1767         struct dirtemplate dirtemplate, *dtp;
 1768         struct direct newdir;
 1769         int error, dmode;
 1770         long blkoff;
 1771 
 1772 #ifdef INVARIANTS
 1773         if ((cnp->cn_flags & HASBUF) == 0)
 1774                 panic("ufs_mkdir: no name");
 1775 #endif
 1776         dp = VTOI(dvp);
 1777         if ((nlink_t)dp->i_nlink >= LINK_MAX) {
 1778                 error = EMLINK;
 1779                 goto out;
 1780         }
 1781         dmode = vap->va_mode & 0777;
 1782         dmode |= IFDIR;
 1783         /*
 1784          * Must simulate part of ufs_makeinode here to acquire the inode,
 1785          * but not have it entered in the parent directory. The entry is
 1786          * made later after writing "." and ".." entries.
 1787          */
 1788         error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
 1789         if (error)
 1790                 goto out;
 1791         ip = VTOI(tvp);
 1792         ip->i_gid = dp->i_gid;
 1793         DIP_SET(ip, i_gid, dp->i_gid);
 1794 #ifdef SUIDDIR
 1795         {
 1796 #ifdef QUOTA
 1797                 struct ucred ucred, *ucp;
 1798                 gid_t ucred_group;
 1799                 ucp = cnp->cn_cred;
 1800 #endif
 1801                 /*
 1802                  * If we are hacking owners here, (only do this where told to)
 1803                  * and we are not giving it TO root, (would subvert quotas)
 1804                  * then go ahead and give it to the other user.
 1805                  * The new directory also inherits the SUID bit.
 1806                  * If user's UID and dir UID are the same,
 1807                  * 'give it away' so that the SUID is still forced on.
 1808                  */
 1809                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 1810                     (dp->i_mode & ISUID) && dp->i_uid) {
 1811                         dmode |= ISUID;
 1812                         ip->i_uid = dp->i_uid;
 1813                         DIP_SET(ip, i_uid, dp->i_uid);
 1814 #ifdef QUOTA
 1815                         if (dp->i_uid != cnp->cn_cred->cr_uid) {
 1816                                 /*
 1817                                  * Make sure the correct user gets charged
 1818                                  * for the space.
 1819                                  * Make a dummy credential for the victim.
 1820                                  * XXX This seems to never be accessed out of
 1821                                  * our context so a stack variable is ok.
 1822                                  */
 1823                                 refcount_init(&ucred.cr_ref, 1);
 1824                                 ucred.cr_uid = ip->i_uid;
 1825                                 ucred.cr_ngroups = 1;
 1826                                 ucred.cr_groups = &ucred_group;
 1827                                 ucred.cr_groups[0] = dp->i_gid;
 1828                                 ucp = &ucred;
 1829                         }
 1830 #endif
 1831                 } else {
 1832                         ip->i_uid = cnp->cn_cred->cr_uid;
 1833                         DIP_SET(ip, i_uid, ip->i_uid);
 1834                 }
 1835 #ifdef QUOTA
 1836                 if ((error = getinoquota(ip)) ||
 1837                     (error = chkiq(ip, 1, ucp, 0))) {
 1838                         if (DOINGSOFTDEP(tvp))
 1839                                 softdep_revert_link(dp, ip);
 1840                         UFS_VFREE(tvp, ip->i_number, dmode);
 1841                         vput(tvp);
 1842                         return (error);
 1843                 }
 1844 #endif
 1845         }
 1846 #else   /* !SUIDDIR */
 1847         ip->i_uid = cnp->cn_cred->cr_uid;
 1848         DIP_SET(ip, i_uid, ip->i_uid);
 1849 #ifdef QUOTA
 1850         if ((error = getinoquota(ip)) ||
 1851             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 1852                 if (DOINGSOFTDEP(tvp))
 1853                         softdep_revert_link(dp, ip);
 1854                 UFS_VFREE(tvp, ip->i_number, dmode);
 1855                 vput(tvp);
 1856                 return (error);
 1857         }
 1858 #endif
 1859 #endif  /* !SUIDDIR */
 1860         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 1861         ip->i_mode = dmode;
 1862         DIP_SET(ip, i_mode, dmode);
 1863         tvp->v_type = VDIR;     /* Rest init'd in getnewvnode(). */
 1864         ip->i_effnlink = 2;
 1865         ip->i_nlink = 2;
 1866         DIP_SET(ip, i_nlink, 2);
 1867 
 1868         if (cnp->cn_flags & ISWHITEOUT) {
 1869                 ip->i_flags |= UF_OPAQUE;
 1870                 DIP_SET(ip, i_flags, ip->i_flags);
 1871         }
 1872 
 1873         /*
 1874          * Bump link count in parent directory to reflect work done below.
 1875          * Should be done before reference is created so cleanup is
 1876          * possible if we crash.
 1877          */
 1878         dp->i_effnlink++;
 1879         dp->i_nlink++;
 1880         DIP_SET(dp, i_nlink, dp->i_nlink);
 1881         dp->i_flag |= IN_CHANGE;
 1882         if (DOINGSOFTDEP(dvp))
 1883                 softdep_setup_mkdir(dp, ip);
 1884         error = UFS_UPDATE(dvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
 1885         if (error)
 1886                 goto bad;
 1887 #ifdef MAC
 1888         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 1889                 error = mac_vnode_create_extattr(cnp->cn_cred, dvp->v_mount,
 1890                     dvp, tvp, cnp);
 1891                 if (error)
 1892                         goto bad;
 1893         }
 1894 #endif
 1895 #ifdef UFS_ACL
 1896         if (dvp->v_mount->mnt_flag & MNT_ACLS) {
 1897                 error = ufs_do_posix1e_acl_inheritance_dir(dvp, tvp, dmode,
 1898                     cnp->cn_cred, cnp->cn_thread);
 1899                 if (error)
 1900                         goto bad;
 1901         } else if (dvp->v_mount->mnt_flag & MNT_NFS4ACLS) {
 1902                 error = ufs_do_nfs4_acl_inheritance(dvp, tvp, dmode,
 1903                     cnp->cn_cred, cnp->cn_thread);
 1904                 if (error)
 1905                         goto bad;
 1906         }
 1907 #endif /* !UFS_ACL */
 1908 
 1909         /*
 1910          * Initialize directory with "." and ".." from static template.
 1911          */
 1912         if (dvp->v_mount->mnt_maxsymlinklen > 0)
 1913                 dtp = &mastertemplate;
 1914         else
 1915                 dtp = (struct dirtemplate *)&omastertemplate;
 1916         dirtemplate = *dtp;
 1917         dirtemplate.dot_ino = ip->i_number;
 1918         dirtemplate.dotdot_ino = dp->i_number;
 1919         if ((error = UFS_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
 1920             BA_CLRBUF, &bp)) != 0)
 1921                 goto bad;
 1922         ip->i_size = DIRBLKSIZ;
 1923         DIP_SET(ip, i_size, DIRBLKSIZ);
 1924         ip->i_flag |= IN_CHANGE | IN_UPDATE;
 1925         vnode_pager_setsize(tvp, (u_long)ip->i_size);
 1926         bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
 1927         if (DOINGSOFTDEP(tvp)) {
 1928                 /*
 1929                  * Ensure that the entire newly allocated block is a
 1930                  * valid directory so that future growth within the
 1931                  * block does not have to ensure that the block is
 1932                  * written before the inode.
 1933                  */
 1934                 blkoff = DIRBLKSIZ;
 1935                 while (blkoff < bp->b_bcount) {
 1936                         ((struct direct *)
 1937                            (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
 1938                         blkoff += DIRBLKSIZ;
 1939                 }
 1940         }
 1941         if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
 1942                                        DOINGASYNC(tvp)))) != 0) {
 1943                 (void)bwrite(bp);
 1944                 goto bad;
 1945         }
 1946         /*
 1947          * Directory set up, now install its entry in the parent directory.
 1948          *
 1949          * If we are not doing soft dependencies, then we must write out the
 1950          * buffer containing the new directory body before entering the new 
 1951          * name in the parent. If we are doing soft dependencies, then the
 1952          * buffer containing the new directory body will be passed to and
 1953          * released in the soft dependency code after the code has attached
 1954          * an appropriate ordering dependency to the buffer which ensures that
 1955          * the buffer is written before the new name is written in the parent.
 1956          */
 1957         if (DOINGASYNC(dvp))
 1958                 bdwrite(bp);
 1959         else if (!DOINGSOFTDEP(dvp) && ((error = bwrite(bp))))
 1960                 goto bad;
 1961         ufs_makedirentry(ip, cnp, &newdir);
 1962         error = ufs_direnter(dvp, tvp, &newdir, cnp, bp, 0);
 1963         
 1964 bad:
 1965         if (error == 0) {
 1966                 *ap->a_vpp = tvp;
 1967         } else {
 1968                 dp->i_effnlink--;
 1969                 dp->i_nlink--;
 1970                 DIP_SET(dp, i_nlink, dp->i_nlink);
 1971                 dp->i_flag |= IN_CHANGE;
 1972                 /*
 1973                  * No need to do an explicit VOP_TRUNCATE here, vrele will
 1974                  * do this for us because we set the link count to 0.
 1975                  */
 1976                 ip->i_effnlink = 0;
 1977                 ip->i_nlink = 0;
 1978                 DIP_SET(ip, i_nlink, 0);
 1979                 ip->i_flag |= IN_CHANGE;
 1980                 if (DOINGSOFTDEP(tvp))
 1981                         softdep_revert_mkdir(dp, ip);
 1982 
 1983                 vput(tvp);
 1984         }
 1985 out:
 1986         return (error);
 1987 }
 1988 
 1989 /*
 1990  * Rmdir system call.
 1991  */
 1992 static int
 1993 ufs_rmdir(ap)
 1994         struct vop_rmdir_args /* {
 1995                 struct vnode *a_dvp;
 1996                 struct vnode *a_vp;
 1997                 struct componentname *a_cnp;
 1998         } */ *ap;
 1999 {
 2000         struct vnode *vp = ap->a_vp;
 2001         struct vnode *dvp = ap->a_dvp;
 2002         struct componentname *cnp = ap->a_cnp;
 2003         struct inode *ip, *dp;
 2004         int error;
 2005 
 2006         ip = VTOI(vp);
 2007         dp = VTOI(dvp);
 2008 
 2009         /*
 2010          * Do not remove a directory that is in the process of being renamed.
 2011          * Verify the directory is empty (and valid). Rmdir ".." will not be
 2012          * valid since ".." will contain a reference to the current directory
 2013          * and thus be non-empty. Do not allow the removal of mounted on
 2014          * directories (this can happen when an NFS exported filesystem
 2015          * tries to remove a locally mounted on directory).
 2016          */
 2017         error = 0;
 2018         if (ip->i_effnlink < 2) {
 2019                 error = EINVAL;
 2020                 goto out;
 2021         }
 2022         if (dp->i_effnlink < 3)
 2023                 panic("ufs_dirrem: Bad link count %d on parent",
 2024                     dp->i_effnlink);
 2025         if (!ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
 2026                 error = ENOTEMPTY;
 2027                 goto out;
 2028         }
 2029         if ((dp->i_flags & APPEND)
 2030             || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
 2031                 error = EPERM;
 2032                 goto out;
 2033         }
 2034         if (vp->v_mountedhere != 0) {
 2035                 error = EINVAL;
 2036                 goto out;
 2037         }
 2038 #ifdef UFS_GJOURNAL
 2039         ufs_gjournal_orphan(vp);
 2040 #endif
 2041         /*
 2042          * Delete reference to directory before purging
 2043          * inode.  If we crash in between, the directory
 2044          * will be reattached to lost+found,
 2045          */
 2046         dp->i_effnlink--;
 2047         ip->i_effnlink--;
 2048         if (DOINGSOFTDEP(vp))
 2049                 softdep_setup_rmdir(dp, ip);
 2050         error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
 2051         if (error) {
 2052                 dp->i_effnlink++;
 2053                 ip->i_effnlink++;
 2054                 if (DOINGSOFTDEP(vp))
 2055                         softdep_revert_rmdir(dp, ip);
 2056                 goto out;
 2057         }
 2058         cache_purge(dvp);
 2059         /*
 2060          * The only stuff left in the directory is "." and "..". The "."
 2061          * reference is inconsequential since we are quashing it. The soft
 2062          * dependency code will arrange to do these operations after
 2063          * the parent directory entry has been deleted on disk, so
 2064          * when running with that code we avoid doing them now.
 2065          */
 2066         if (!DOINGSOFTDEP(vp)) {
 2067                 dp->i_nlink--;
 2068                 DIP_SET(dp, i_nlink, dp->i_nlink);
 2069                 dp->i_flag |= IN_CHANGE;
 2070                 error = UFS_UPDATE(dvp, 0);
 2071                 ip->i_nlink--;
 2072                 DIP_SET(ip, i_nlink, ip->i_nlink);
 2073                 ip->i_flag |= IN_CHANGE;
 2074         }
 2075         cache_purge(vp);
 2076 #ifdef UFS_DIRHASH
 2077         /* Kill any active hash; i_effnlink == 0, so it will not come back. */
 2078         if (ip->i_dirhash != NULL)
 2079                 ufsdirhash_free(ip);
 2080 #endif
 2081 out:
 2082         return (error);
 2083 }
 2084 
 2085 /*
 2086  * symlink -- make a symbolic link
 2087  */
 2088 static int
 2089 ufs_symlink(ap)
 2090         struct vop_symlink_args /* {
 2091                 struct vnode *a_dvp;
 2092                 struct vnode **a_vpp;
 2093                 struct componentname *a_cnp;
 2094                 struct vattr *a_vap;
 2095                 char *a_target;
 2096         } */ *ap;
 2097 {
 2098         struct vnode *vp, **vpp = ap->a_vpp;
 2099         struct inode *ip;
 2100         int len, error;
 2101 
 2102         error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
 2103             vpp, ap->a_cnp);
 2104         if (error)
 2105                 return (error);
 2106         vp = *vpp;
 2107         len = strlen(ap->a_target);
 2108         if (len < vp->v_mount->mnt_maxsymlinklen) {
 2109                 ip = VTOI(vp);
 2110                 bcopy(ap->a_target, SHORTLINK(ip), len);
 2111                 ip->i_size = len;
 2112                 DIP_SET(ip, i_size, len);
 2113                 ip->i_flag |= IN_CHANGE | IN_UPDATE;
 2114                 error = UFS_UPDATE(vp, 0);
 2115         } else
 2116                 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
 2117                     UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
 2118                     ap->a_cnp->cn_cred, NOCRED, NULL, NULL);
 2119         if (error)
 2120                 vput(vp);
 2121         return (error);
 2122 }
 2123 
 2124 /*
 2125  * Vnode op for reading directories.
 2126  */
 2127 int
 2128 ufs_readdir(ap)
 2129         struct vop_readdir_args /* {
 2130                 struct vnode *a_vp;
 2131                 struct uio *a_uio;
 2132                 struct ucred *a_cred;
 2133                 int *a_eofflag;
 2134                 int *a_ncookies;
 2135                 u_long **a_cookies;
 2136         } */ *ap;
 2137 {
 2138         struct vnode *vp = ap->a_vp;
 2139         struct uio *uio = ap->a_uio;
 2140         struct buf *bp;
 2141         struct inode *ip;
 2142         struct direct *dp, *edp;
 2143         u_long *cookies;
 2144         struct dirent dstdp;
 2145         off_t offset, startoffset;
 2146         size_t readcnt, skipcnt;
 2147         ssize_t startresid;
 2148         int ncookies;
 2149         int error;
 2150 
 2151         if (uio->uio_offset < 0)
 2152                 return (EINVAL);
 2153         ip = VTOI(vp);
 2154         if (ip->i_effnlink == 0)
 2155                 return (0);
 2156         if (ap->a_ncookies != NULL) {
 2157                 ncookies = uio->uio_resid;
 2158                 if (uio->uio_offset >= ip->i_size)
 2159                         ncookies = 0;
 2160                 else if (ip->i_size - uio->uio_offset < ncookies)
 2161                         ncookies = ip->i_size - uio->uio_offset;
 2162                 ncookies = ncookies / (offsetof(struct direct, d_name) + 4) + 1;
 2163                 cookies = malloc(ncookies * sizeof(*cookies), M_TEMP, M_WAITOK);
 2164                 *ap->a_ncookies = ncookies;
 2165                 *ap->a_cookies = cookies;
 2166         } else {
 2167                 ncookies = 0;
 2168                 cookies = NULL;
 2169         }
 2170         offset = startoffset = uio->uio_offset;
 2171         startresid = uio->uio_resid;
 2172         error = 0;
 2173         while (error == 0 && uio->uio_resid > 0 &&
 2174             uio->uio_offset < ip->i_size) {
 2175                 error = ffs_blkatoff(vp, uio->uio_offset, NULL, &bp);
 2176                 if (error)
 2177                         break;
 2178                 if (bp->b_offset + bp->b_bcount > ip->i_size)
 2179                         readcnt = ip->i_size - bp->b_offset;
 2180                 else
 2181                         readcnt = bp->b_bcount;
 2182                 skipcnt = (size_t)(uio->uio_offset - bp->b_offset) &
 2183                     ~(size_t)(DIRBLKSIZ - 1);
 2184                 offset = bp->b_offset + skipcnt;
 2185                 dp = (struct direct *)&bp->b_data[skipcnt];
 2186                 edp = (struct direct *)&bp->b_data[readcnt];
 2187                 while (error == 0 && uio->uio_resid > 0 && dp < edp) {
 2188                         if (dp->d_reclen <= offsetof(struct direct, d_name) ||
 2189                             (caddr_t)dp + dp->d_reclen > (caddr_t)edp) {
 2190                                 error = EIO;
 2191                                 break;
 2192                         }
 2193 #if BYTE_ORDER == LITTLE_ENDIAN
 2194                         /* Old filesystem format. */
 2195                         if (vp->v_mount->mnt_maxsymlinklen <= 0) {
 2196                                 dstdp.d_namlen = dp->d_type;
 2197                                 dstdp.d_type = dp->d_namlen;
 2198                         } else
 2199 #endif
 2200                         {
 2201                                 dstdp.d_namlen = dp->d_namlen;
 2202                                 dstdp.d_type = dp->d_type;
 2203                         }
 2204                         if (offsetof(struct direct, d_name) + dstdp.d_namlen >
 2205                             dp->d_reclen) {
 2206                                 error = EIO;
 2207                                 break;
 2208                         }
 2209                         if (offset < startoffset || dp->d_ino == 0)
 2210                                 goto nextentry;
 2211                         dstdp.d_fileno = dp->d_ino;
 2212                         dstdp.d_reclen = GENERIC_DIRSIZ(&dstdp);
 2213                         bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
 2214                         dstdp.d_name[dstdp.d_namlen] = '\0';
 2215                         if (dstdp.d_reclen > uio->uio_resid) {
 2216                                 if (uio->uio_resid == startresid)
 2217                                         error = EINVAL;
 2218                                 else
 2219                                         error = EJUSTRETURN;
 2220                                 break;
 2221                         }
 2222                         /* Advance dp. */
 2223                         error = uiomove((caddr_t)&dstdp, dstdp.d_reclen, uio);
 2224                         if (error)
 2225                                 break;
 2226                         if (cookies != NULL) {
 2227                                 KASSERT(ncookies > 0,
 2228                                     ("ufs_readdir: cookies buffer too small"));
 2229                                 *cookies = offset + dp->d_reclen;
 2230                                 cookies++;
 2231                                 ncookies--;
 2232                         }
 2233 nextentry:
 2234                         offset += dp->d_reclen;
 2235                         dp = (struct direct *)((caddr_t)dp + dp->d_reclen);
 2236                 }
 2237                 bqrelse(bp);
 2238                 uio->uio_offset = offset;
 2239         }
 2240         /* We need to correct uio_offset. */
 2241         uio->uio_offset = offset;
 2242         if (error == EJUSTRETURN)
 2243                 error = 0;
 2244         if (ap->a_ncookies != NULL) {
 2245                 if (error == 0) {
 2246                         ap->a_ncookies -= ncookies;
 2247                 } else {
 2248                         free(*ap->a_cookies, M_TEMP);
 2249                         *ap->a_ncookies = 0;
 2250                         *ap->a_cookies = NULL;
 2251                 }
 2252         }
 2253         if (error == 0 && ap->a_eofflag)
 2254                 *ap->a_eofflag = ip->i_size <= uio->uio_offset;
 2255         return (error);
 2256 }
 2257 
 2258 /*
 2259  * Return target name of a symbolic link
 2260  */
 2261 static int
 2262 ufs_readlink(ap)
 2263         struct vop_readlink_args /* {
 2264                 struct vnode *a_vp;
 2265                 struct uio *a_uio;
 2266                 struct ucred *a_cred;
 2267         } */ *ap;
 2268 {
 2269         struct vnode *vp = ap->a_vp;
 2270         struct inode *ip = VTOI(vp);
 2271         doff_t isize;
 2272 
 2273         isize = ip->i_size;
 2274         if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
 2275             DIP(ip, i_blocks) == 0) { /* XXX - for old fastlink support */
 2276                 return (uiomove(SHORTLINK(ip), isize, ap->a_uio));
 2277         }
 2278         return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
 2279 }
 2280 
 2281 /*
 2282  * Calculate the logical to physical mapping if not done already,
 2283  * then call the device strategy routine.
 2284  *
 2285  * In order to be able to swap to a file, the ufs_bmaparray() operation may not
 2286  * deadlock on memory.  See ufs_bmap() for details.
 2287  */
 2288 static int
 2289 ufs_strategy(ap)
 2290         struct vop_strategy_args /* {
 2291                 struct vnode *a_vp;
 2292                 struct buf *a_bp;
 2293         } */ *ap;
 2294 {
 2295         struct buf *bp = ap->a_bp;
 2296         struct vnode *vp = ap->a_vp;
 2297         struct bufobj *bo;
 2298         struct inode *ip;
 2299         ufs2_daddr_t blkno;
 2300         int error;
 2301 
 2302         ip = VTOI(vp);
 2303         if (bp->b_blkno == bp->b_lblkno) {
 2304                 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, bp, NULL, NULL);
 2305                 bp->b_blkno = blkno;
 2306                 if (error) {
 2307                         bp->b_error = error;
 2308                         bp->b_ioflags |= BIO_ERROR;
 2309                         bufdone(bp);
 2310                         return (0);
 2311                 }
 2312                 if ((long)bp->b_blkno == -1)
 2313                         vfs_bio_clrbuf(bp);
 2314         }
 2315         if ((long)bp->b_blkno == -1) {
 2316                 bufdone(bp);
 2317                 return (0);
 2318         }
 2319         bp->b_iooffset = dbtob(bp->b_blkno);
 2320         bo = ip->i_umbufobj;
 2321         BO_STRATEGY(bo, bp);
 2322         return (0);
 2323 }
 2324 
 2325 /*
 2326  * Print out the contents of an inode.
 2327  */
 2328 static int
 2329 ufs_print(ap)
 2330         struct vop_print_args /* {
 2331                 struct vnode *a_vp;
 2332         } */ *ap;
 2333 {
 2334         struct vnode *vp = ap->a_vp;
 2335         struct inode *ip = VTOI(vp);
 2336 
 2337         printf("\tino %lu, on dev %s", (u_long)ip->i_number,
 2338             devtoname(ip->i_dev));
 2339         if (vp->v_type == VFIFO)
 2340                 fifo_printinfo(vp);
 2341         printf("\n");
 2342         return (0);
 2343 }
 2344 
 2345 /*
 2346  * Close wrapper for fifos.
 2347  *
 2348  * Update the times on the inode then do device close.
 2349  */
 2350 static int
 2351 ufsfifo_close(ap)
 2352         struct vop_close_args /* {
 2353                 struct vnode *a_vp;
 2354                 int  a_fflag;
 2355                 struct ucred *a_cred;
 2356                 struct thread *a_td;
 2357         } */ *ap;
 2358 {
 2359         struct vnode *vp = ap->a_vp;
 2360         int usecount;
 2361 
 2362         VI_LOCK(vp);
 2363         usecount = vp->v_usecount;
 2364         if (usecount > 1)
 2365                 ufs_itimes_locked(vp);
 2366         VI_UNLOCK(vp);
 2367         return (fifo_specops.vop_close(ap));
 2368 }
 2369 
 2370 /*
 2371  * Kqfilter wrapper for fifos.
 2372  *
 2373  * Fall through to ufs kqfilter routines if needed 
 2374  */
 2375 static int
 2376 ufsfifo_kqfilter(ap)
 2377         struct vop_kqfilter_args *ap;
 2378 {
 2379         int error;
 2380 
 2381         error = fifo_specops.vop_kqfilter(ap);
 2382         if (error)
 2383                 error = vfs_kqfilter(ap);
 2384         return (error);
 2385 }
 2386 
 2387 /*
 2388  * Return POSIX pathconf information applicable to fifos.
 2389  */
 2390 static int
 2391 ufsfifo_pathconf(ap)
 2392         struct vop_pathconf_args /* {
 2393                 struct vnode *a_vp;
 2394                 int a_name;
 2395                 int *a_retval;
 2396         } */ *ap;
 2397 {
 2398 
 2399         switch (ap->a_name) {
 2400         case _PC_ACL_EXTENDED:
 2401         case _PC_ACL_NFS4:
 2402         case _PC_ACL_PATH_MAX:
 2403         case _PC_MAC_PRESENT:
 2404                 return (ufs_pathconf(ap));
 2405         default:
 2406                 return (fifo_specops.vop_pathconf(ap));
 2407         }
 2408         /* NOTREACHED */
 2409 }
 2410 
 2411 /*
 2412  * Return POSIX pathconf information applicable to ufs filesystems.
 2413  */
 2414 static int
 2415 ufs_pathconf(ap)
 2416         struct vop_pathconf_args /* {
 2417                 struct vnode *a_vp;
 2418                 int a_name;
 2419                 int *a_retval;
 2420         } */ *ap;
 2421 {
 2422         int error;
 2423 
 2424         error = 0;
 2425         switch (ap->a_name) {
 2426         case _PC_LINK_MAX:
 2427                 *ap->a_retval = LINK_MAX;
 2428                 break;
 2429         case _PC_NAME_MAX:
 2430                 *ap->a_retval = NAME_MAX;
 2431                 break;
 2432         case _PC_PATH_MAX:
 2433                 *ap->a_retval = PATH_MAX;
 2434                 break;
 2435         case _PC_PIPE_BUF:
 2436                 *ap->a_retval = PIPE_BUF;
 2437                 break;
 2438         case _PC_CHOWN_RESTRICTED:
 2439                 *ap->a_retval = 1;
 2440                 break;
 2441         case _PC_NO_TRUNC:
 2442                 *ap->a_retval = 1;
 2443                 break;
 2444         case _PC_ACL_EXTENDED:
 2445 #ifdef UFS_ACL
 2446                 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
 2447                         *ap->a_retval = 1;
 2448                 else
 2449                         *ap->a_retval = 0;
 2450 #else
 2451                 *ap->a_retval = 0;
 2452 #endif
 2453                 break;
 2454 
 2455         case _PC_ACL_NFS4:
 2456 #ifdef UFS_ACL
 2457                 if (ap->a_vp->v_mount->mnt_flag & MNT_NFS4ACLS)
 2458                         *ap->a_retval = 1;
 2459                 else
 2460                         *ap->a_retval = 0;
 2461 #else
 2462                 *ap->a_retval = 0;
 2463 #endif
 2464                 break;
 2465 
 2466         case _PC_ACL_PATH_MAX:
 2467 #ifdef UFS_ACL
 2468                 if (ap->a_vp->v_mount->mnt_flag & (MNT_ACLS | MNT_NFS4ACLS))
 2469                         *ap->a_retval = ACL_MAX_ENTRIES;
 2470                 else
 2471                         *ap->a_retval = 3;
 2472 #else
 2473                 *ap->a_retval = 3;
 2474 #endif
 2475                 break;
 2476         case _PC_MAC_PRESENT:
 2477 #ifdef MAC
 2478                 if (ap->a_vp->v_mount->mnt_flag & MNT_MULTILABEL)
 2479                         *ap->a_retval = 1;
 2480                 else
 2481                         *ap->a_retval = 0;
 2482 #else
 2483                 *ap->a_retval = 0;
 2484 #endif
 2485                 break;
 2486         case _PC_MIN_HOLE_SIZE:
 2487                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2488                 break;
 2489         case _PC_ASYNC_IO:
 2490                 /* _PC_ASYNC_IO should have been handled by upper layers. */
 2491                 KASSERT(0, ("_PC_ASYNC_IO should not get here"));
 2492                 error = EINVAL;
 2493                 break;
 2494         case _PC_PRIO_IO:
 2495                 *ap->a_retval = 0;
 2496                 break;
 2497         case _PC_SYNC_IO:
 2498                 *ap->a_retval = 0;
 2499                 break;
 2500         case _PC_ALLOC_SIZE_MIN:
 2501                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
 2502                 break;
 2503         case _PC_FILESIZEBITS:
 2504                 *ap->a_retval = 64;
 2505                 break;
 2506         case _PC_REC_INCR_XFER_SIZE:
 2507                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2508                 break;
 2509         case _PC_REC_MAX_XFER_SIZE:
 2510                 *ap->a_retval = -1; /* means ``unlimited'' */
 2511                 break;
 2512         case _PC_REC_MIN_XFER_SIZE:
 2513                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2514                 break;
 2515         case _PC_REC_XFER_ALIGN:
 2516                 *ap->a_retval = PAGE_SIZE;
 2517                 break;
 2518         case _PC_SYMLINK_MAX:
 2519                 *ap->a_retval = MAXPATHLEN;
 2520                 break;
 2521 
 2522         default:
 2523                 error = EINVAL;
 2524                 break;
 2525         }
 2526         return (error);
 2527 }
 2528 
 2529 /*
 2530  * Initialize the vnode associated with a new inode, handle aliased
 2531  * vnodes.
 2532  */
 2533 int
 2534 ufs_vinit(mntp, fifoops, vpp)
 2535         struct mount *mntp;
 2536         struct vop_vector *fifoops;
 2537         struct vnode **vpp;
 2538 {
 2539         struct inode *ip;
 2540         struct vnode *vp;
 2541 
 2542         vp = *vpp;
 2543         ip = VTOI(vp);
 2544         vp->v_type = IFTOVT(ip->i_mode);
 2545         if (vp->v_type == VFIFO)
 2546                 vp->v_op = fifoops;
 2547         ASSERT_VOP_LOCKED(vp, "ufs_vinit");
 2548         if (ip->i_number == ROOTINO)
 2549                 vp->v_vflag |= VV_ROOT;
 2550         *vpp = vp;
 2551         return (0);
 2552 }
 2553 
 2554 /*
 2555  * Allocate a new inode.
 2556  * Vnode dvp must be locked.
 2557  */
 2558 static int
 2559 ufs_makeinode(mode, dvp, vpp, cnp)
 2560         int mode;
 2561         struct vnode *dvp;
 2562         struct vnode **vpp;
 2563         struct componentname *cnp;
 2564 {
 2565         struct inode *ip, *pdir;
 2566         struct direct newdir;
 2567         struct vnode *tvp;
 2568         int error;
 2569 
 2570         pdir = VTOI(dvp);
 2571 #ifdef INVARIANTS
 2572         if ((cnp->cn_flags & HASBUF) == 0)
 2573                 panic("ufs_makeinode: no name");
 2574 #endif
 2575         *vpp = NULL;
 2576         if ((mode & IFMT) == 0)
 2577                 mode |= IFREG;
 2578 
 2579         if (VTOI(dvp)->i_effnlink < 2)
 2580                 panic("ufs_makeinode: Bad link count %d on parent",
 2581                     VTOI(dvp)->i_effnlink);
 2582         error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
 2583         if (error)
 2584                 return (error);
 2585         ip = VTOI(tvp);
 2586         ip->i_gid = pdir->i_gid;
 2587         DIP_SET(ip, i_gid, pdir->i_gid);
 2588 #ifdef SUIDDIR
 2589         {
 2590 #ifdef QUOTA
 2591                 struct ucred ucred, *ucp;
 2592                 gid_t ucred_group;
 2593                 ucp = cnp->cn_cred;
 2594 #endif
 2595                 /*
 2596                  * If we are not the owner of the directory,
 2597                  * and we are hacking owners here, (only do this where told to)
 2598                  * and we are not giving it TO root, (would subvert quotas)
 2599                  * then go ahead and give it to the other user.
 2600                  * Note that this drops off the execute bits for security.
 2601                  */
 2602                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 2603                     (pdir->i_mode & ISUID) &&
 2604                     (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
 2605                         ip->i_uid = pdir->i_uid;
 2606                         DIP_SET(ip, i_uid, ip->i_uid);
 2607                         mode &= ~07111;
 2608 #ifdef QUOTA
 2609                         /*
 2610                          * Make sure the correct user gets charged
 2611                          * for the space.
 2612                          * Quickly knock up a dummy credential for the victim.
 2613                          * XXX This seems to never be accessed out of our
 2614                          * context so a stack variable is ok.
 2615                          */
 2616                         refcount_init(&ucred.cr_ref, 1);
 2617                         ucred.cr_uid = ip->i_uid;
 2618                         ucred.cr_ngroups = 1;
 2619                         ucred.cr_groups = &ucred_group;
 2620                         ucred.cr_groups[0] = pdir->i_gid;
 2621                         ucp = &ucred;
 2622 #endif
 2623                 } else {
 2624                         ip->i_uid = cnp->cn_cred->cr_uid;
 2625                         DIP_SET(ip, i_uid, ip->i_uid);
 2626                 }
 2627 
 2628 #ifdef QUOTA
 2629                 if ((error = getinoquota(ip)) ||
 2630                     (error = chkiq(ip, 1, ucp, 0))) {
 2631                         if (DOINGSOFTDEP(tvp))
 2632                                 softdep_revert_link(pdir, ip);
 2633                         UFS_VFREE(tvp, ip->i_number, mode);
 2634                         vput(tvp);
 2635                         return (error);
 2636                 }
 2637 #endif
 2638         }
 2639 #else   /* !SUIDDIR */
 2640         ip->i_uid = cnp->cn_cred->cr_uid;
 2641         DIP_SET(ip, i_uid, ip->i_uid);
 2642 #ifdef QUOTA
 2643         if ((error = getinoquota(ip)) ||
 2644             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 2645                 if (DOINGSOFTDEP(tvp))
 2646                         softdep_revert_link(pdir, ip);
 2647                 UFS_VFREE(tvp, ip->i_number, mode);
 2648                 vput(tvp);
 2649                 return (error);
 2650         }
 2651 #endif
 2652 #endif  /* !SUIDDIR */
 2653         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 2654         ip->i_mode = mode;
 2655         DIP_SET(ip, i_mode, mode);
 2656         tvp->v_type = IFTOVT(mode);     /* Rest init'd in getnewvnode(). */
 2657         ip->i_effnlink = 1;
 2658         ip->i_nlink = 1;
 2659         DIP_SET(ip, i_nlink, 1);
 2660         if (DOINGSOFTDEP(tvp))
 2661                 softdep_setup_create(VTOI(dvp), ip);
 2662         if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
 2663             priv_check_cred(cnp->cn_cred, PRIV_VFS_SETGID, 0)) {
 2664                 ip->i_mode &= ~ISGID;
 2665                 DIP_SET(ip, i_mode, ip->i_mode);
 2666         }
 2667 
 2668         if (cnp->cn_flags & ISWHITEOUT) {
 2669                 ip->i_flags |= UF_OPAQUE;
 2670                 DIP_SET(ip, i_flags, ip->i_flags);
 2671         }
 2672 
 2673         /*
 2674          * Make sure inode goes to disk before directory entry.
 2675          */
 2676         error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
 2677         if (error)
 2678                 goto bad;
 2679 #ifdef MAC
 2680         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 2681                 error = mac_vnode_create_extattr(cnp->cn_cred, dvp->v_mount,
 2682                     dvp, tvp, cnp);
 2683                 if (error)
 2684                         goto bad;
 2685         }
 2686 #endif
 2687 #ifdef UFS_ACL
 2688         if (dvp->v_mount->mnt_flag & MNT_ACLS) {
 2689                 error = ufs_do_posix1e_acl_inheritance_file(dvp, tvp, mode,
 2690                     cnp->cn_cred, cnp->cn_thread);
 2691                 if (error)
 2692                         goto bad;
 2693         } else if (dvp->v_mount->mnt_flag & MNT_NFS4ACLS) {
 2694                 error = ufs_do_nfs4_acl_inheritance(dvp, tvp, mode,
 2695                     cnp->cn_cred, cnp->cn_thread);
 2696                 if (error)
 2697                         goto bad;
 2698         }
 2699 #endif /* !UFS_ACL */
 2700         ufs_makedirentry(ip, cnp, &newdir);
 2701         error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL, 0);
 2702         if (error)
 2703                 goto bad;
 2704         *vpp = tvp;
 2705         return (0);
 2706 
 2707 bad:
 2708         /*
 2709          * Write error occurred trying to update the inode
 2710          * or the directory so must deallocate the inode.
 2711          */
 2712         ip->i_effnlink = 0;
 2713         ip->i_nlink = 0;
 2714         DIP_SET(ip, i_nlink, 0);
 2715         ip->i_flag |= IN_CHANGE;
 2716         if (DOINGSOFTDEP(tvp))
 2717                 softdep_revert_create(VTOI(dvp), ip);
 2718         vput(tvp);
 2719         return (error);
 2720 }
 2721 
 2722 static int
 2723 ufs_ioctl(struct vop_ioctl_args *ap)
 2724 {
 2725 
 2726         switch (ap->a_command) {
 2727         case FIOSEEKDATA:
 2728         case FIOSEEKHOLE:
 2729                 return (vn_bmap_seekhole(ap->a_vp, ap->a_command,
 2730                     (off_t *)ap->a_data, ap->a_cred));
 2731         default:
 2732                 return (ENOTTY);
 2733         }
 2734 }
 2735 
 2736 /* Global vfs data structures for ufs. */
 2737 struct vop_vector ufs_vnodeops = {
 2738         .vop_default =          &default_vnodeops,
 2739         .vop_fsync =            VOP_PANIC,
 2740         .vop_read =             VOP_PANIC,
 2741         .vop_reallocblks =      VOP_PANIC,
 2742         .vop_write =            VOP_PANIC,
 2743         .vop_accessx =          ufs_accessx,
 2744         .vop_bmap =             ufs_bmap,
 2745         .vop_cachedlookup =     ufs_lookup,
 2746         .vop_close =            ufs_close,
 2747         .vop_create =           ufs_create,
 2748         .vop_getattr =          ufs_getattr,
 2749         .vop_inactive =         ufs_inactive,
 2750         .vop_ioctl =            ufs_ioctl,
 2751         .vop_link =             ufs_link,
 2752         .vop_lookup =           vfs_cache_lookup,
 2753         .vop_markatime =        ufs_markatime,
 2754         .vop_mkdir =            ufs_mkdir,
 2755         .vop_mknod =            ufs_mknod,
 2756         .vop_open =             ufs_open,
 2757         .vop_pathconf =         ufs_pathconf,
 2758         .vop_poll =             vop_stdpoll,
 2759         .vop_print =            ufs_print,
 2760         .vop_readdir =          ufs_readdir,
 2761         .vop_readlink =         ufs_readlink,
 2762         .vop_reclaim =          ufs_reclaim,
 2763         .vop_remove =           ufs_remove,
 2764         .vop_rename =           ufs_rename,
 2765         .vop_rmdir =            ufs_rmdir,
 2766         .vop_setattr =          ufs_setattr,
 2767 #ifdef MAC
 2768         .vop_setlabel =         vop_stdsetlabel_ea,
 2769 #endif
 2770         .vop_strategy =         ufs_strategy,
 2771         .vop_symlink =          ufs_symlink,
 2772         .vop_whiteout =         ufs_whiteout,
 2773 #ifdef UFS_EXTATTR
 2774         .vop_getextattr =       ufs_getextattr,
 2775         .vop_deleteextattr =    ufs_deleteextattr,
 2776         .vop_setextattr =       ufs_setextattr,
 2777 #endif
 2778 #ifdef UFS_ACL
 2779         .vop_getacl =           ufs_getacl,
 2780         .vop_setacl =           ufs_setacl,
 2781         .vop_aclcheck =         ufs_aclcheck,
 2782 #endif
 2783 };
 2784 
 2785 struct vop_vector ufs_fifoops = {
 2786         .vop_default =          &fifo_specops,
 2787         .vop_fsync =            VOP_PANIC,
 2788         .vop_accessx =          ufs_accessx,
 2789         .vop_close =            ufsfifo_close,
 2790         .vop_getattr =          ufs_getattr,
 2791         .vop_inactive =         ufs_inactive,
 2792         .vop_kqfilter =         ufsfifo_kqfilter,
 2793         .vop_markatime =        ufs_markatime,
 2794         .vop_pathconf =         ufsfifo_pathconf,
 2795         .vop_print =            ufs_print,
 2796         .vop_read =             VOP_PANIC,
 2797         .vop_reclaim =          ufs_reclaim,
 2798         .vop_setattr =          ufs_setattr,
 2799 #ifdef MAC
 2800         .vop_setlabel =         vop_stdsetlabel_ea,
 2801 #endif
 2802         .vop_write =            VOP_PANIC,
 2803 #ifdef UFS_EXTATTR
 2804         .vop_getextattr =       ufs_getextattr,
 2805         .vop_deleteextattr =    ufs_deleteextattr,
 2806         .vop_setextattr =       ufs_setextattr,
 2807 #endif
 2808 #ifdef UFS_ACL
 2809         .vop_getacl =           ufs_getacl,
 2810         .vop_setacl =           ufs_setacl,
 2811         .vop_aclcheck =         ufs_aclcheck,
 2812 #endif
 2813 };

Cache object: a2ec5501d5ede51b50d46be5ffd48539


[ 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.