sys/vfs/gnu/ext2fs/ext2_vnops.c

   1 /*
   2  *  modified for EXT2FS support in Lites 1.1
   3  *
   4  *  Aug 1995, Godmar Back (gback@cs.utah.edu)
   5  *  University of Utah, Department of Computer Science
   6  */
   7 /*
   8  * Copyright (c) 1982, 1986, 1989, 1993
   9  *      The Regents of the University of California.  All rights reserved.
  10  * (c) UNIX System Laboratories, Inc.
  11  * All or some portions of this file are derived from material licensed
  12  * to the University of California by American Telephone and Telegraph
  13  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
  14  * the permission of UNIX System Laboratories, Inc.
  15  *
  16  * Redistribution and use in source and binary forms, with or without
  17  * modification, are permitted provided that the following conditions
  18  * are met:
  19  * 1. Redistributions of source code must retain the above copyright
  20  *    notice, this list of conditions and the following disclaimer.
  21  * 2. Redistributions in binary form must reproduce the above copyright
  22  *    notice, this list of conditions and the following disclaimer in the
  23  *    documentation and/or other materials provided with the distribution.
  24  * 3. All advertising materials mentioning features or use of this software
  25  *    must display the following acknowledgement:
  26  *      This product includes software developed by the University of
  27  *      California, Berkeley and its contributors.
  28  * 4. Neither the name of the University nor the names of its contributors
  29  *    may be used to endorse or promote products derived from this software
  30  *    without specific prior written permission.
  31  *
  32  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  33  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  34  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  35  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  36  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  37  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  38  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  39  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  40  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  41  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  42  * SUCH DAMAGE.
  43  *
  44  *      @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
  45  *      @(#)ext2_vnops.c        8.7 (Berkeley) 2/3/94
  46  * $FreeBSD: src/sys/gnu/ext2fs/ext2_vnops.c,v 1.51.2.2 2003/01/02 17:26:18 bde Exp $
  47  * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_vnops.c,v 1.43 2008/06/19 23:27:39 dillon Exp $
  48  */
  49
  50 #include "opt_quota.h"
  51 #include "opt_suiddir.h"
  52
  53 #include <sys/param.h>
  54 #include <sys/systm.h>
  55 #include <sys/resourcevar.h>
  56 #include <sys/kernel.h>
  57 #include <sys/stat.h>
  58 #include <sys/fcntl.h>
  59 #include <sys/buf.h>
  60 #include <sys/stat.h>
  61 #include <sys/proc.h>
  62 #include <sys/priv.h>
  63 #include <sys/mount.h>
  64 #include <sys/time.h>
  65 #include <sys/vnode.h>
  66 #include <sys/dirent.h>
  67 #include <sys/lockf.h>
  68 #include <sys/event.h>
  69 #include <sys/conf.h>
  70 #include <sys/file.h>
  71 #include <sys/jail.h>
  72 #include <sys/namei.h>
  73 #include <sys/signalvar.h>
  74 #include <sys/unistd.h>
  75
  76 #include <vm/vm.h>
  77 #include <vm/vm_extern.h>
  78 #include <vm/vm_zone.h>
  79 #include <vm/vnode_pager.h>
  80
  81 #include <sys/buf2.h>
  82 #include <sys/thread2.h>
  83
  84 #include <vfs/fifofs/fifo.h>
  85
  86 #include "dir.h"
  87 #include "quota.h"
  88 #include "inode.h"
  89 #include "ext2mount.h"
  90 #include "ext2_fs_sb.h"
  91 #include "fs.h"
  92 #include "ext2_extern.h"
  93 #include "ext2_fs.h"
  94
  95 static int ext2_access (struct vop_access_args *);
  96 static int ext2_advlock (struct vop_advlock_args *);
  97 static int ext2_chmod (struct vnode *, int, struct ucred *);
  98 static int ext2_chown (struct vnode *, uid_t, gid_t, struct ucred *);
  99 static int ext2_close (struct vop_close_args *);
 100 static int ext2_getattr (struct vop_getattr_args *);
 101 static int ext2_makeinode (int mode, struct vnode *, struct vnode **, struct componentname *);
 102 static int ext2_mmap (struct vop_mmap_args *);
 103 static int ext2_open (struct vop_open_args *);
 104 static int ext2_pathconf (struct vop_pathconf_args *);
 105 static int ext2_print (struct vop_print_args *);
 106 static int ext2_readlink (struct vop_readlink_args *);
 107 static int ext2_setattr (struct vop_setattr_args *);
 108 static int ext2_strategy (struct vop_strategy_args *);
 109 static int ext2_whiteout (struct vop_old_whiteout_args *);
 110 static int filt_ext2read (struct knote *kn, long hint);
 111 static int filt_ext2write (struct knote *kn, long hint);
 112 static int filt_ext2vnode (struct knote *kn, long hint);
 113 static void filt_ext2detach (struct knote *kn);
 114 static int ext2_kqfilter (struct vop_kqfilter_args *ap);
 115 static int ext2spec_close (struct vop_close_args *);
 116 static int ext2spec_read (struct vop_read_args *);
 117 static int ext2spec_write (struct vop_write_args *);
 118 static int ext2fifo_close (struct vop_close_args *);
 119 static int ext2fifo_kqfilter (struct vop_kqfilter_args *);
 120 static int ext2fifo_read (struct vop_read_args *);
 121 static int ext2fifo_write (struct vop_write_args *);
 122
 123 static int ext2_fsync (struct vop_fsync_args *);
 124 static int ext2_read (struct vop_read_args *);
 125 static int ext2_write (struct vop_write_args *);
 126 static int ext2_remove (struct vop_old_remove_args *);
 127 static int ext2_link (struct vop_old_link_args *);
 128 static int ext2_rename (struct vop_old_rename_args *);
 129 static int ext2_mkdir (struct vop_old_mkdir_args *);
 130 static int ext2_rmdir (struct vop_old_rmdir_args *);
 131 static int ext2_create (struct vop_old_create_args *);
 132 static int ext2_mknod (struct vop_old_mknod_args *);
 133 static int ext2_symlink (struct vop_old_symlink_args *);
 134 static int ext2_getpages (struct vop_getpages_args *);
 135
 136 #include "ext2_readwrite.c"
 137
 138 union _qcvt {
 139         int64_t qcvt;
 140         int32_t val[2];
 141 };
 142 #define SETHIGH(q, h) { \
 143         union _qcvt tmp; \
 144         tmp.qcvt = (q); \
 145         tmp.val[_QUAD_HIGHWORD] = (h); \
 146         (q) = tmp.qcvt; \
 147 }
 148 #define SETLOW(q, l) { \
 149         union _qcvt tmp; \
 150         tmp.qcvt = (q); \
 151         tmp.val[_QUAD_LOWWORD] = (l); \
 152         (q) = tmp.qcvt; \
 153 }
 154 #define VN_KNOTE(vp, b) \
 155         KNOTE(&vp->v_pollinfo.vpi_selinfo.si_note, (b))
 156
 157 #define OFSFMT(vp)              ((vp)->v_mount->mnt_maxsymlinklen <= 0)
 158
 159 /*
 160  * A virgin directory (no blushing please).
 161  * Note that the type and namlen fields are reversed relative to ufs.
 162  * Also, we don't use `struct odirtemplate', since it would just cause
 163  * endianness problems.
 164  */
 165 static struct dirtemplate ext2_mastertemplate = {
 166         0, 12, 1, EXT2_FT_DIR, ".",
 167         0, DIRBLKSIZ - 12, 2, EXT2_FT_DIR, ".."
 168 };
 169 static struct dirtemplate ext2_omastertemplate = {
 170         0, 12, 1, EXT2_FT_UNKNOWN, ".",
 171         0, DIRBLKSIZ - 12, 2, EXT2_FT_UNKNOWN, ".."
 172 };
 173
 174 /*
 175  * Create a regular file
 176  *
 177  * ext2_create(struct vnode *a_dvp, struct vnode **a_vpp,
 178  *             struct componentname *a_cnp, struct vattr *a_vap)
 179  */
 180 static int
 181 ext2_create(struct vop_old_create_args *ap)
 182 {
 183         int error;
 184
 185         error =
 186             ext2_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
 187             ap->a_dvp, ap->a_vpp, ap->a_cnp);
 188         if (error)
 189                 return (error);
 190         return (0);
 191 }
 192
 193 /*
 194  * Synch an open file.
 195  *
 196  * ext2_fsync(struct vnode *a_vp, struct ucred *a_cred, int a_waitfor,
 197  *            struct proc *a_p)
 198  */
 199 /* ARGSUSED */
 200
 201 static int ext2_fsync_bp(struct buf *bp, void *data);
 202
 203 struct ext2_fsync_bp_info {
 204         struct vnode *vp;
 205         int waitfor;
 206 };
 207
 208 static int
 209 ext2_fsync(struct vop_fsync_args *ap)
 210 {
 211         struct ext2_fsync_bp_info info;
 212         struct vnode *vp = ap->a_vp;
 213         int count;
 214
 215         /*
 216          * XXX why is all this fs specific?
 217          */
 218
 219         /*
 220          * Flush all dirty buffers associated with a vnode.
 221          */
 222         ext2_discard_prealloc(VTOI(vp));
 223
 224         crit_enter();
 225         info.vp = vp;
 226 loop:
 227         info.waitfor = ap->a_waitfor;
 228         count = RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
 229                         ext2_fsync_bp, &info);
 230         if (count)
 231                 goto loop;
 232
 233         if (ap->a_waitfor == MNT_WAIT) {
 234                 while (vp->v_track_write.bk_active) {
 235                         vp->v_track_write.bk_waitflag = 1;
 236                         tsleep(&vp->v_track_write, 0, "e2fsyn", 0);
 237                 }
 238 #if DIAGNOSTIC
 239                 if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
 240                         vprint("ext2_fsync: dirty", vp);
 241                         goto loop;
 242                 }
 243 #endif
 244         }
 245         crit_exit();
 246         return (EXT2_UPDATE(ap->a_vp, ap->a_waitfor == MNT_WAIT));
 247 }
 248
 249 static int
 250 ext2_fsync_bp(struct buf *bp, void *data)
 251 {
 252         struct ext2_fsync_bp_info *info = data;
 253
 254         if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
 255                 return(0);
 256         if ((bp->b_flags & B_DELWRI) == 0)
 257                 panic("ext2_fsync: not dirty");
 258         bremfree(bp);
 259         crit_exit();
 260
 261         /*
 262          * Wait for I/O associated with indirect blocks to complete,
 263          * since there is no way to quickly wait for them below.
 264          */
 265         if (bp->b_vp == info->vp || info->waitfor == MNT_NOWAIT)
 266                 bawrite(bp);
 267         else
 268                 bwrite(bp);
 269         crit_enter();
 270         return(1);
 271 }
 272
 273 /*
 274  * Mknod vnode call
 275  *
 276  * ext2_mknod(struct vnode *a_dvp, struct vnode **a_vpp,
 277  *            struct componentname *a_cnp, struct vattr *a_vap)
 278  */
 279 /* ARGSUSED */
 280 static int
 281 ext2_mknod(struct vop_old_mknod_args *ap)
 282 {
 283         struct vattr *vap = ap->a_vap;
 284         struct vnode **vpp = ap->a_vpp;
 285         struct inode *ip;
 286         ino_t ino;
 287         int error;
 288
 289         if (vap->va_rmajor != VNOVAL &&
 290             makeudev(vap->va_rmajor, vap->va_rminor) == NOUDEV) {
 291                 return (EINVAL);
 292         }
 293
 294         error = ext2_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
 295             ap->a_dvp, vpp, ap->a_cnp);
 296         if (error)
 297                 return (error);
 298         ip = VTOI(*vpp);
 299         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 300         if (vap->va_rmajor != VNOVAL) {
 301                 /*
 302                  * Want to be able to use this to make badblock
 303                  * inodes, so don't truncate the dev number.
 304                  */
 305                 ip->i_rdev = makeudev(vap->va_rmajor, vap->va_rminor);
 306         }
 307         /*
 308          * Remove inode, then reload it through VFS_VGET so it is
 309          * checked to see if it is an alias of an existing entry in
 310          * the inode cache.
 311          */
 312         (*vpp)->v_type = VNON;
 313         ino = ip->i_number;     /* Save this before vgone() invalidates ip. */
 314         vgone_vxlocked(*vpp);
 315         vput(*vpp);
 316         error = VFS_VGET(ap->a_dvp->v_mount, ino, vpp);
 317         if (error) {
 318                 *vpp = NULL;
 319                 return (error);
 320         }
 321         return (0);
 322 }
 323
 324 /*
 325  * ext2_remove(struct vnode *a_dvp, struct vnode *a_vp,
 326  *             struct componentname *a_cnp)
 327  */
 328 static int
 329 ext2_remove(struct vop_old_remove_args *ap)
 330 {
 331         struct inode *ip;
 332         struct vnode *vp = ap->a_vp;
 333         struct vnode *dvp = ap->a_dvp;
 334         int error;
 335
 336         ip = VTOI(vp);
 337         if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
 338             (VTOI(dvp)->i_flags & APPEND)) {
 339                 error = EPERM;
 340                 goto out;
 341         }
 342         error = ext2_dirremove(dvp, ap->a_cnp);
 343         if (error == 0) {
 344                 ip->i_nlink--;
 345                 ip->i_flag |= IN_CHANGE;
 346         }
 347 out:
 348         return (error);
 349 }
 350
 351 /*
 352  * link vnode call
 353  *
 354  * ext2_link(struct vnode *a_tdvp, struct vnode *a_vp,
 355  *           struct componentname *a_cnp)
 356  */
 357 static int
 358 ext2_link(struct vop_old_link_args *ap)
 359 {
 360         struct vnode *vp = ap->a_vp;
 361         struct vnode *tdvp = ap->a_tdvp;
 362         struct componentname *cnp = ap->a_cnp;
 363         struct inode *ip;
 364         int error;
 365
 366         if (tdvp->v_mount != vp->v_mount) {
 367                 error = EXDEV;
 368                 goto out2;
 369         }
 370         if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE))) {
 371                 goto out2;
 372         }
 373         ip = VTOI(vp);
 374         if ((nlink_t)ip->i_nlink >= LINK_MAX) {
 375                 error = EMLINK;
 376                 goto out1;
 377         }
 378         if (ip->i_flags & (IMMUTABLE | APPEND)) {
 379                 error = EPERM;
 380                 goto out1;
 381         }
 382         ip->i_nlink++;
 383         ip->i_flag |= IN_CHANGE;
 384         error = EXT2_UPDATE(vp, 1);
 385         if (!error)
 386                 error = ext2_direnter(ip, tdvp, cnp);
 387         if (error) {
 388                 ip->i_nlink--;
 389                 ip->i_flag |= IN_CHANGE;
 390         }
 391 out1:
 392         if (tdvp != vp)
 393                 vn_unlock(vp);
 394 out2:
 395         return (error);
 396 }
 397
 398 /*
 399  * Rename system call.  fdvp, fvp are ref'd.  tvp, tdvp are ref'd and locked.
 400  * all vp's are released and must be in an unlocked state on return.
 401  *
 402  * ext2_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
 403  *              struct componentname *a_fcnp, struct vnode *a_tdvp,
 404  *              struct vnode *a_tvp, struct componentname *a_tcnp)
 405  */
 406 static int
 407 ext2_rename(struct vop_old_rename_args *ap)
 408 {
 409         struct vnode *tvp = ap->a_tvp;
 410         struct vnode *tdvp = ap->a_tdvp;
 411         struct vnode *fvp = ap->a_fvp;
 412         struct vnode *fdvp = ap->a_fdvp;
 413         struct componentname *tcnp = ap->a_tcnp;
 414         struct componentname *fcnp = ap->a_fcnp;
 415         struct inode *ip, *xp, *dp;
 416         struct dirtemplate dirbuf;
 417         int doingdirectory = 0, oldparent = 0, newparent = 0;
 418         int error = 0;
 419         u_char namlen;
 420
 421         /*
 422          * Check for cross-device rename.
 423          */
 424         if ((fvp->v_mount != tdvp->v_mount) ||
 425             (tvp && (fvp->v_mount != tvp->v_mount)) ||
 426             tvp == tdvp) {
 427                 error = EXDEV;
 428 abortit:
 429                 if (tdvp == tvp)
 430                         vrele(tdvp);
 431                 else
 432                         vput(tdvp);
 433                 if (tvp)
 434                         vput(tvp);
 435                 vrele(fdvp);
 436                 vrele(fvp);
 437                 return (error);
 438         }
 439
 440         if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
 441             (VTOI(tdvp)->i_flags & APPEND))) {
 442                 error = EPERM;
 443                 goto abortit;
 444         }
 445
 446         /*
 447          * Renaming a file to itself has no effect.  The upper layers should
 448          * not call us in that case.  Temporarily just warn if they do.
 449          */
 450         if (fvp == tvp) {
 451                 error = 0;
 452                 goto abortit;
 453         }
 454
 455         if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
 456                 goto abortit;
 457
 458         /*
 459          * fvp, tvp, tdvp locked.  fdvp not locked but note that fdvp may
 460          * be equal to tdvp.
 461          */
 462         dp = VTOI(fdvp);
 463         ip = VTOI(fvp);
 464         if (ip->i_nlink >= LINK_MAX) {
 465                 vn_unlock(fvp);
 466                 error = EMLINK;
 467                 goto abortit;
 468         }
 469         if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
 470             || (dp->i_flags & APPEND)) {
 471                 vn_unlock(fvp);
 472                 error = EPERM;
 473                 goto abortit;
 474         }
 475         if ((ip->i_mode & IFMT) == IFDIR) {
 476                 /*
 477                  * Avoid ".", "..", and aliases of "." for obvious reasons.
 478                  */
 479                 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
 480                     dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & CNP_ISDOTDOT ||
 481                     (ip->i_flag & IN_RENAME)) {
 482                         vn_unlock(fvp);
 483                         error = EINVAL;
 484                         goto abortit;
 485                 }
 486                 ip->i_flag |= IN_RENAME;
 487                 oldparent = dp->i_number;
 488                 doingdirectory++;
 489         }
 490
 491         /*
 492          * tvp is non-NULL if the target exists.   fvp is still locked but
 493          * we will unlock it soon.  The 'bad' goto target requires dp and
 494          * xp to be correctly assigned.
 495          */
 496         dp = VTOI(tdvp);
 497         if (tvp)
 498                 xp = VTOI(tvp);
 499         else
 500                 xp = NULL;
 501
 502         /*
 503          * 1) Bump link count while we're moving stuff
 504          *    around.  If we crash somewhere before
 505          *    completing our work, the link count
 506          *    may be wrong, but correctable.
 507          */
 508         ip->i_nlink++;
 509         ip->i_flag |= IN_CHANGE;
 510         if ((error = EXT2_UPDATE(fvp, 1)) != 0) {
 511                 vn_unlock(fvp);
 512                 goto bad;
 513         }
 514
 515         /*
 516          * If ".." must be changed (ie the directory gets a new
 517          * parent) then the source directory must not be in the
 518          * directory heirarchy above the target, as this would
 519          * orphan everything below the source directory. Also
 520          * the user must have write permission in the source so
 521          * as to be able to change "..". We must repeat the call
 522          * to namei, as the parent directory is unlocked by the
 523          * call to checkpath().
 524          */
 525         error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred);
 526         vn_unlock(fvp);
 527
 528         /*
 529          * tvp (if not NULL) and tdvp are locked.  fvp and fdvp are not.
 530          * dp and xp are set according to tdvp and tvp.
 531          */
 532         if (oldparent != dp->i_number)
 533                 newparent = dp->i_number;
 534         if (doingdirectory && newparent) {
 535                 if (error)      /* write access check above */
 536                         goto bad;
 537
 538                 /*
 539                  * Prepare for relookup, get rid of xp
 540                  */
 541                 if (xp != NULL) {
 542                         vput(tvp);
 543                         xp = NULL;
 544                 }
 545
 546                 /*
 547                  * checkpath vput()'s tdvp (VTOI(dp)) on return no matter what,
 548                  * get an extra ref so we wind up with just an unlocked, ref'd
 549                  * tdvp.  The 'out' target skips xp and tdvp cleanups.  Our
 550                  * tdvp is now unlocked so we have to clean it up ourselves.
 551                  */
 552                 vref(tdvp);
 553                 error = ext2_checkpath(ip, dp, tcnp->cn_cred);
 554                 tcnp->cn_flags |= CNP_PDIRUNLOCK;
 555                 if (error) {
 556                         vrele(tdvp);
 557                         goto out;
 558                 }
 559                 /*
 560                  * relookup no longer messes with the ref count.  An unlocked
 561                  * tdvp must be passed and if no error occurs a locked tdvp
 562                  * will be returned.  We have to use the out target again.
 563                  */
 564                 error = relookup(tdvp, &tvp, tcnp);
 565                 if (error) {
 566                         if (tcnp->cn_flags & CNP_PDIRUNLOCK)
 567                                 vrele(tdvp);
 568                         else
 569                                 vput(tdvp);
 570                         goto out;
 571                 }
 572
 573                 /*
 574                  * tdvp is locked at this point.  in the RENAME case tvp may
 575                  * be NULL without an error, assign xp accordingly.  The
 576                  * 'bad' target can be used again after this.
 577                  */
 578                 dp = VTOI(tdvp);
 579                 if (tvp)
 580                         xp = VTOI(tvp);
 581         }
 582         /*
 583          * 2) If target doesn't exist, link the target
 584          *    to the source and unlink the source.
 585          *    Otherwise, rewrite the target directory
 586          *    entry to reference the source inode and
 587          *    expunge the original entry's existence.
 588          *
 589          * tdvp and tvp are cleaned up by this code.  tvp is only good if
 590          * xp is not NULL.
 591          */
 592         if (xp == NULL) {
 593                 if (dp->i_dev != ip->i_dev)
 594                         panic("ext2_rename: EXDEV");
 595                 /*
 596                  * Account for ".." in new directory.
 597                  * When source and destination have the same
 598                  * parent we don't fool with the link count.
 599                  */
 600                 if (doingdirectory && newparent) {
 601                         if ((nlink_t)dp->i_nlink >= LINK_MAX) {
 602                                 error = EMLINK;
 603                                 goto bad;
 604                         }
 605                         dp->i_nlink++;
 606                         dp->i_flag |= IN_CHANGE;
 607                         error = EXT2_UPDATE(tdvp, 1);
 608                         if (error)
 609                                 goto bad;
 610                 }
 611                 error = ext2_direnter(ip, tdvp, tcnp);
 612                 if (error) {
 613                         if (doingdirectory && newparent) {
 614                                 dp->i_nlink--;
 615                                 dp->i_flag |= IN_CHANGE;
 616                                 EXT2_UPDATE(tdvp, 1);
 617                         }
 618                         goto bad;
 619                 }
 620
 621                 /*
 622                  * manual cleanup, we can't use the bad or out target after
 623                  * this.
 624                  */
 625                 vput(tdvp);
 626         } else {
 627                 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
 628                         panic("ext2_rename: EXDEV");
 629                 /*
 630                  * Short circuit rename(foo, foo).
 631                  */
 632                 if (xp->i_number == ip->i_number)
 633                         panic("ext2_rename: same file");
 634                 /*
 635                  * If the parent directory is "sticky", then the user must
 636                  * own the parent directory, or the destination of the rename,
 637                  * otherwise the destination may not be changed (except by
 638                  * root). This implements append-only directories.
 639                  */
 640                 if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
 641                     tcnp->cn_cred->cr_uid != dp->i_uid &&
 642                     xp->i_uid != tcnp->cn_cred->cr_uid) {
 643                         error = EPERM;
 644                         goto bad;
 645                 }
 646                 /*
 647                  * Target must be empty if a directory and have no links
 648                  * to it. Also, ensure source and target are compatible
 649                  * (both directories, or both not directories).
 650                  */
 651                 if ((xp->i_mode&IFMT) == IFDIR) {
 652                         if (! ext2_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
 653                             xp->i_nlink > 2) {
 654                                 error = ENOTEMPTY;
 655                                 goto bad;
 656                         }
 657                         if (!doingdirectory) {
 658                                 error = ENOTDIR;
 659                                 goto bad;
 660                         }
 661                 } else if (doingdirectory) {
 662                         error = EISDIR;
 663                         goto bad;
 664                 }
 665                 error = ext2_dirrewrite(dp, ip, tcnp);
 666                 if (error)
 667                         goto bad;
 668                 /*
 669                  * If the target directory is in the same
 670                  * directory as the source directory,
 671                  * decrement the link count on the parent
 672                  * of the target directory.
 673                  */
 674                  if (doingdirectory && !newparent) {
 675                         dp->i_nlink--;
 676                         dp->i_flag |= IN_CHANGE;
 677                 }
 678
 679                 /*
 680                  * manual cleanup, we can't use the bad or out target after
 681                  * this.
 682                  */
 683                 vput(tdvp);
 684
 685                 /*
 686                  * Adjust the link count of the target to
 687                  * reflect the dirrewrite above.  If this is
 688                  * a directory it is empty and there are
 689                  * no links to it, so we can squash the inode and
 690                  * any space associated with it.  We disallowed
 691                  * renaming over top of a directory with links to
 692                  * it above, as the remaining link would point to
 693                  * a directory without "." or ".." entries.
 694                  */
 695                 xp->i_nlink--;
 696                 if (doingdirectory) {
 697                         if (--xp->i_nlink != 0)
 698                                 panic("ext2_rename: linked directory");
 699                         error = EXT2_TRUNCATE(tvp, (off_t)0, IO_SYNC,
 700                                               tcnp->cn_cred);
 701                 }
 702                 xp->i_flag |= IN_CHANGE;
 703                 vput(tvp);
 704                 xp = NULL;
 705         }
 706
 707         /*
 708          * tvp and tdvp have been cleaned up.  The bad and out targets may
 709          * not be used.  fvp and fdvp are ref'd but not locked.  ip
 710          * still represents the old fvp and ip->i_flag may still have IN_RENAME
 711          * set (if doingdirectory).
 712          */
 713
 714         /*
 715          * 3) Unlink the source.
 716          *
 717          * fdvp is locked and ref'd. ap->a_fvp holds the old lookup unlocked
 718          * and ref'd, fvp will hold the new lookup locked and ref'd.
 719          *
 720          * After the relookup ap->a_fvp must be released as part of our
 721          * cleanup, not just fdvp and fvp.  And, on success, fdvp and
 722          * fvp will be locked so the bad and out targets cannot be used.
 723          */
 724         fcnp->cn_flags &= ~CNP_MODMASK;
 725         fcnp->cn_flags |= CNP_LOCKPARENT;
 726         KKASSERT(fcnp->cn_flags & CNP_PDIRUNLOCK);
 727         error = relookup(fdvp, &fvp, fcnp);
 728         if (error) {
 729                 /*
 730                  * From name has disappeared.
 731                  */
 732                 if (doingdirectory)
 733                         panic("ext2_rename: lost dir entry");
 734                 /* ip->i_flag only sets IN_RENAME if doingdirectory */
 735                 vrele(ap->a_fvp);
 736                 if (fcnp->cn_flags & CNP_PDIRUNLOCK)
 737                         vrele(fdvp);
 738                 else
 739                         vput(fdvp);
 740                 return (0);
 741         }
 742         KKASSERT((fcnp->cn_flags & CNP_PDIRUNLOCK) == 0);
 743
 744         /*
 745          * This case shouldn't occur
 746          */
 747         if (fvp == NULL) {
 748                 /*
 749                  * From name has disappeared.
 750                  */
 751                 if (doingdirectory)
 752                         panic("ext2_rename: lost dir entry");
 753                 /* ip->i_flag only sets IN_RENAME if doingdirectory */
 754                 vrele(ap->a_fvp);
 755                 vput(fvp);
 756                 vput(fdvp);
 757                 return (0);
 758         }
 759
 760         /*
 761          * fvp and fdvp are both ref'd and locked.
 762          */
 763         xp = VTOI(fvp);
 764         dp = VTOI(fdvp);
 765
 766         /*
 767          * Ensure that the directory entry still exists and has not
 768          * changed while the new name has been entered. If the source is
 769          * a file then the entry may have been unlinked or renamed. In
 770          * either case there is no further work to be done. If the source
 771          * is a directory then it cannot have been rmdir'ed; its link
 772          * count of three would cause a rmdir to fail with ENOTEMPTY.
 773          * The IN_RENAME flag ensures that it cannot be moved by another
 774          * rename.
 775          */
 776         if (xp != ip) {
 777                 if (doingdirectory)
 778                         panic("ext2_rename: lost dir entry");
 779                 /* ip->i_flag only sets IN_RENAME if doingdirectory */
 780         } else {
 781                 /*
 782                  * If the source is a directory with a
 783                  * new parent, the link count of the old
 784                  * parent directory must be decremented
 785                  * and ".." set to point to the new parent.
 786                  */
 787                 if (doingdirectory && newparent) {
 788                         dp->i_nlink--;
 789                         dp->i_flag |= IN_CHANGE;
 790                         error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
 791                                         sizeof (struct dirtemplate), (off_t)0,
 792                                         UIO_SYSSPACE, IO_NODELOCKED,
 793                                         tcnp->cn_cred, (int *)0);
 794                         if (error == 0) {
 795                                 /* Like ext2 little-endian: */
 796                                 namlen = dirbuf.dotdot_type;
 797                                 if (namlen != 2 ||
 798                                     dirbuf.dotdot_name[0] != '.' ||
 799                                     dirbuf.dotdot_name[1] != '.') {
 800                                         ext2_dirbad(xp, (doff_t)12,
 801                                             "rename: mangled dir");
 802                                 } else {
 803                                         dirbuf.dotdot_ino = newparent;
 804                                         vn_rdwr(UIO_WRITE, fvp,
 805                                                 (caddr_t)&dirbuf,
 806                                                 sizeof (struct dirtemplate),
 807                                                 (off_t)0, UIO_SYSSPACE,
 808                                                 IO_NODELOCKED|IO_SYNC,
 809                                                 tcnp->cn_cred, (int *)0);
 810                                 }
 811                         }
 812                 }
 813                 error = ext2_dirremove(fdvp, fcnp);
 814                 if (!error) {
 815                         xp->i_nlink--;
 816                         xp->i_flag |= IN_CHANGE;
 817                 }
 818                 xp->i_flag &= ~IN_RENAME;
 819         }
 820         vput(fdvp);
 821         vput(fvp);
 822         vrele(ap->a_fvp);
 823         return (error);
 824
 825 bad:
 826         if (xp)
 827                 vput(ITOV(xp));
 828         if (dp)
 829                 vput(ITOV(dp));
 830 out:
 831         if (doingdirectory)
 832                 ip->i_flag &= ~IN_RENAME;
 833         if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
 834                 ip->i_nlink--;
 835                 ip->i_flag |= IN_CHANGE;
 836                 ip->i_flag &= ~IN_RENAME;
 837                 vput(fvp);
 838         } else {
 839                 vrele(fvp);
 840         }
 841         return (error);
 842 }
 843
 844 /*
 845  * Mkdir system call
 846  *
 847  * ext2_mkdir(struct vnode *a_dvp, struct vnode **a_vpp,
 848  *            struct componentname *a_cnp, struct vattr *a_vap)
 849  */
 850 static int
 851 ext2_mkdir(struct vop_old_mkdir_args *ap)
 852 {
 853         struct vnode *dvp = ap->a_dvp;
 854         struct vattr *vap = ap->a_vap;
 855         struct componentname *cnp = ap->a_cnp;
 856         struct inode *ip, *dp;
 857         struct vnode *tvp;
 858         struct dirtemplate dirtemplate, *dtp;
 859         int error, dmode;
 860
 861         dp = VTOI(dvp);
 862         if ((nlink_t)dp->i_nlink >= LINK_MAX) {
 863                 error = EMLINK;
 864                 goto out;
 865         }
 866         dmode = vap->va_mode & 0777;
 867         dmode |= IFDIR;
 868         /*
 869          * Must simulate part of ext2_makeinode here to acquire the inode,
 870          * but not have it entered in the parent directory. The entry is
 871          * made later after writing "." and ".." entries.
 872          */
 873         error = EXT2_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
 874         if (error)
 875                 goto out;
 876         ip = VTOI(tvp);
 877         ip->i_gid = dp->i_gid;
 878 #ifdef SUIDDIR
 879         {
 880 #ifdef QUOTA
 881                 struct ucred ucred, *ucp;
 882                 ucp = cnp->cn_cred;
 883 #endif
 884                 /*
 885                  * if we are hacking owners here, (only do this where told to)
 886                  * and we are not giving it TOO root, (would subvert quotas)
 887                  * then go ahead and give it to the other user.
 888                  * The new directory also inherits the SUID bit.
 889                  * If user's UID and dir UID are the same,
 890                  * 'give it away' so that the SUID is still forced on.
 891                  */
 892                 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 893                    (dp->i_mode & ISUID) && dp->i_uid) {
 894                         dmode |= ISUID;
 895                         ip->i_uid = dp->i_uid;
 896 #ifdef QUOTA
 897                         if (dp->i_uid != cnp->cn_cred->cr_uid) {
 898                                 /*
 899                                  * make sure the correct user gets charged
 900                                  * for the space.
 901                                  * Make a dummy credential for the victim.
 902                                  * XXX This seems to never be accessed out of
 903                                  * our context so a stack variable is ok.
 904                                  */
 905                                 ucred.cr_ref = 1;
 906                                 ucred.cr_uid = ip->i_uid;
 907                                 ucred.cr_ngroups = 1;
 908                                 ucred.cr_groups[0] = dp->i_gid;
 909                                 ucp = &ucred;
 910                         }
 911 #endif
 912                 } else {
 913                         ip->i_uid = cnp->cn_cred->cr_uid;
 914                 }
 915 #ifdef QUOTA
 916                 if ((error = ext2_getinoquota(ip)) ||
 917                 (error = ext2_chkiq(ip, 1, ucp, 0))) {
 918                         EXT2_VFREE(tvp, ip->i_number, dmode);
 919                         vput(tvp);
 920                         return (error);
 921                 }
 922 #endif
 923         }
 924 #else
 925         ip->i_uid = cnp->cn_cred->cr_uid;
 926 #ifdef QUOTA
 927         if ((error = ext2_getinoquota(ip)) ||
 928             (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
 929                 EXT2_VFREE(tvp, ip->i_number, dmode);
 930                 vput(tvp);
 931                 return (error);
 932         }
 933 #endif
 934 #endif
 935         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 936         ip->i_mode = dmode;
 937         tvp->v_type = VDIR;     /* Rest init'd in getnewvnode(). */
 938         ip->i_nlink = 2;
 939         if (cnp->cn_flags & CNP_ISWHITEOUT)
 940                 ip->i_flags |= UF_OPAQUE;
 941         error = EXT2_UPDATE(tvp, 1);
 942
 943         /*
 944          * The vnode must have a VM object in order to issue buffer cache
 945          * ops on it.
 946          */
 947         vinitvmio(tvp, 0);
 948
 949         /*
 950          * Bump link count in parent directory
 951          * to reflect work done below.  Should
 952          * be done before reference is created
 953          * so reparation is possible if we crash.
 954          */
 955         dp->i_nlink++;
 956         dp->i_flag |= IN_CHANGE;
 957         error = EXT2_UPDATE(dvp, 1);
 958         if (error)
 959                 goto bad;
 960
 961         /* Initialize directory with "." and ".." from static template. */
 962         if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs->s_es,
 963             EXT2_FEATURE_INCOMPAT_FILETYPE))
 964                 dtp = &ext2_mastertemplate;
 965         else
 966                 dtp = &ext2_omastertemplate;
 967         dirtemplate = *dtp;
 968         dirtemplate.dot_ino = ip->i_number;
 969         dirtemplate.dotdot_ino = dp->i_number;
 970         /* note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE
 971          * so let's just redefine it - for this function only
 972          */
 973 #undef  DIRBLKSIZ
 974 #define DIRBLKSIZ  VTOI(dvp)->i_e2fs->s_blocksize
 975         dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
 976         error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
 977                         sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
 978                         IO_NODELOCKED|IO_SYNC, cnp->cn_cred, (int *)0);
 979         if (error) {
 980                 dp->i_nlink--;
 981                 dp->i_flag |= IN_CHANGE;
 982                 goto bad;
 983         }
 984         if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
 985                 panic("ext2_mkdir: blksize"); /* XXX should grow with balloc() */
 986         else {
 987                 ip->i_size = DIRBLKSIZ;
 988                 ip->i_flag |= IN_CHANGE;
 989         }
 990
 991         /* Directory set up, now install its entry in the parent directory. */
 992         error = ext2_direnter(ip, dvp, cnp);
 993         if (error) {
 994                 dp->i_nlink--;
 995                 dp->i_flag |= IN_CHANGE;
 996         }
 997 bad:
 998         /*
 999          * No need to do an explicit VOP_TRUNCATE here, vrele will do this
1000          * for us because we set the link count to 0.
1001          */
1002         if (error) {
1003                 ip->i_nlink = 0;
1004                 ip->i_flag |= IN_CHANGE;
1005                 vput(tvp);
1006         } else
1007                 *ap->a_vpp = tvp;
1008 out:
1009         return (error);
1010 #undef  DIRBLKSIZ
1011 #define DIRBLKSIZ  DEV_BSIZE
1012 }
1013
1014 /*
1015  * Rmdir system call.
1016  *
1017  * ext2_rmdir(struct vnode *a_dvp, struct vnode *a_vp,
1018  *            struct componentname *a_cnp)
1019  */
1020 static int
1021 ext2_rmdir(struct vop_old_rmdir_args *ap)
1022 {
1023         struct vnode *vp = ap->a_vp;
1024         struct vnode *dvp = ap->a_dvp;
1025         struct componentname *cnp = ap->a_cnp;
1026         struct inode *ip, *dp;
1027         int error;
1028
1029         ip = VTOI(vp);
1030         dp = VTOI(dvp);
1031
1032         /*
1033          * Verify the directory is empty (and valid).
1034          * (Rmdir ".." won't be valid since
1035          *  ".." will contain a reference to
1036          *  the current directory and thus be
1037          *  non-empty.)
1038          */
1039         error = 0;
1040         if (ip->i_nlink != 2 || !ext2_dirempty(ip, dp->i_number, cnp->cn_cred)) {
1041                 error = ENOTEMPTY;
1042                 goto out;
1043         }
1044         if ((dp->i_flags & APPEND)
1045             || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1046                 error = EPERM;
1047                 goto out;
1048         }
1049         /*
1050          * Delete reference to directory before purging
1051          * inode.  If we crash in between, the directory
1052          * will be reattached to lost+found,
1053          */
1054         error = ext2_dirremove(dvp, cnp);
1055         if (error)
1056                 goto out;
1057         dp->i_nlink--;
1058         dp->i_flag |= IN_CHANGE;
1059         vn_unlock(dvp);
1060         /*
1061          * Truncate inode.  The only stuff left
1062          * in the directory is "." and "..".  The
1063          * "." reference is inconsequential since
1064          * we're quashing it.  The ".." reference
1065          * has already been adjusted above.  We've
1066          * removed the "." reference and the reference
1067          * in the parent directory, but there may be
1068          * other hard links so decrement by 2 and
1069          * worry about them later.
1070          */
1071         ip->i_nlink -= 2;
1072         error = EXT2_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred);
1073         vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
1074 out:
1075         return (error);
1076 }
1077
1078 /*
1079  * symlink -- make a symbolic link
1080  *
1081  * ext2_symlink(struct vnode *a_dvp, struct vnode **a_vpp,
1082  *              struct componentname *a_cnp, struct vattr *a_vap,
1083  *              char *a_target)
1084  */
1085 static int
1086 ext2_symlink(struct vop_old_symlink_args *ap)
1087 {
1088         struct vnode *vp, **vpp = ap->a_vpp;
1089         struct inode *ip;
1090         int len, error;
1091
1092         error = ext2_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
1093             vpp, ap->a_cnp);
1094         if (error)
1095                 return (error);
1096         vp = *vpp;
1097         len = strlen(ap->a_target);
1098         if (len < vp->v_mount->mnt_maxsymlinklen) {
1099                 ip = VTOI(vp);
1100                 bcopy(ap->a_target, (char *)ip->i_shortlink, len);
1101                 ip->i_size = len;
1102                 ip->i_flag |= IN_CHANGE | IN_UPDATE;
1103         } else {
1104                 /*
1105                  * Make sure we have a VM object in order to use
1106                  * the buffer cache.
1107                  */
1108                 if (vp->v_object == NULL)
1109                         vinitvmio(vp, 0);
1110
1111                 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
1112                                 UIO_SYSSPACE, IO_NODELOCKED,
1113                                 ap->a_cnp->cn_cred, (int *)0);
1114
1115                 if (error)
1116                         vput(vp);
1117         }
1118
1119         return (error);
1120 }
1121
1122 /*
1123  * Allocate a new inode.
1124  */
1125 static int
1126 ext2_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
1127                struct componentname *cnp)
1128 {
1129         struct inode *ip, *pdir;
1130         struct vnode *tvp;
1131         int error;
1132
1133         pdir = VTOI(dvp);
1134         *vpp = NULL;
1135         if ((mode & IFMT) == 0)
1136                 mode |= IFREG;
1137
1138         error = EXT2_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
1139         if (error) {
1140                 return (error);
1141         }
1142         ip = VTOI(tvp);
1143         ip->i_gid = pdir->i_gid;
1144 #ifdef SUIDDIR
1145         {
1146 #ifdef QUOTA
1147                 struct ucred ucred, *ucp;
1148                 ucp = cnp->cn_cred;
1149 #endif
1150                 /*
1151                  * if we are
1152                  * not the owner of the directory,
1153                  * and we are hacking owners here, (only do this where told to)
1154                  * and we are not giving it TOO root, (would subvert quotas)
1155                  * then go ahead and give it to the other user.
1156                  * Note that this drops off the execute bits for security.
1157                  */
1158                 if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
1159                      (pdir->i_mode & ISUID) &&
1160                      (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
1161                         ip->i_uid = pdir->i_uid;
1162                         mode &= ~07111;
1163 #ifdef QUOTA
1164                         /*
1165                          * make sure the correct user gets charged
1166                          * for the space.
1167                          * Quickly knock up a dummy credential for the victim.
1168                          * XXX This seems to never be accessed out of our
1169                          * context so a stack variable is ok.
1170                          */
1171                         ucred.cr_ref = 1;
1172                         ucred.cr_uid = ip->i_uid;
1173                         ucred.cr_ngroups = 1;
1174                         ucred.cr_groups[0] = pdir->i_gid;
1175                         ucp = &ucred;
1176 #endif
1177                 } else {
1178                         ip->i_uid = cnp->cn_cred->cr_uid;
1179                 }
1180
1181 #ifdef QUOTA
1182                 if ((error = ext2_getinoquota(ip)) ||
1183                 (error = ext2_chkiq(ip, 1, ucp, 0))) {
1184                         EXT2_VFREE(tvp, ip->i_number, mode);
1185                         vput(tvp);
1186                         return (error);
1187                 }
1188 #endif
1189         }
1190 #else
1191         ip->i_uid = cnp->cn_cred->cr_uid;
1192 #ifdef QUOTA
1193         if ((error = ext2_getinoquota(ip)) ||
1194             (error = ext2_chkiq(ip, 1, cnp->cn_cred, 0))) {
1195                 EXT2_VFREE(tvp, ip->i_number, mode);
1196                 vput(tvp);
1197                 return (error);
1198         }
1199 #endif
1200 #endif
1201         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1202         ip->i_mode = mode;
1203         tvp->v_type = IFTOVT(mode);     /* Rest init'd in getnewvnode(). */
1204         ip->i_nlink = 1;
1205         if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
1206             priv_check_cred(cnp->cn_cred, PRIV_ROOT, PRISON_ROOT))
1207                 ip->i_mode &= ~ISGID;
1208
1209         if (cnp->cn_flags & CNP_ISWHITEOUT)
1210                 ip->i_flags |= UF_OPAQUE;
1211
1212         /*
1213          * Regular files and directories need VM objects.  Softlinks do
1214          * not (not immediately anyway).
1215          */
1216         if (tvp->v_type == VREG || tvp->v_type == VDIR)
1217                 vinitvmio(tvp, 0);
1218
1219         /*
1220          * Make sure inode goes to disk before directory entry.
1221          */
1222         error = EXT2_UPDATE(tvp, 1);
1223         if (error)
1224                 goto bad;
1225         error = ext2_direnter(ip, dvp, cnp);
1226         if (error)
1227                 goto bad;
1228
1229         *vpp = tvp;
1230         return (0);
1231
1232 bad:
1233         /*
1234          * Write error occurred trying to update the inode
1235          * or the directory so must deallocate the inode.
1236          */
1237         ip->i_nlink = 0;
1238         ip->i_flag |= IN_CHANGE;
1239         vput(tvp);
1240         return (error);
1241 }
1242
1243 /*
1244  * get page routine
1245  *
1246  * XXX By default, wimp out... note that a_offset is ignored (and always
1247  * XXX has been).
1248  */
1249 static int
1250 ext2_getpages(struct vop_getpages_args *ap)
1251 {
1252         return (vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
1253                 ap->a_reqpage));
1254 }
1255
1256 void
1257 ext2_itimes(struct vnode *vp)
1258 {
1259         struct inode *ip;
1260         struct timespec ts;
1261
1262         ip = VTOI(vp);
1263         if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
1264                 return;
1265         if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
1266                 ip->i_flag |= IN_LAZYMOD;
1267         else
1268                 ip->i_flag |= IN_MODIFIED;
1269         if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
1270                 vfs_timestamp(&ts);
1271                 if (ip->i_flag & IN_ACCESS) {
1272                         ip->i_atime = ts.tv_sec;
1273                         ip->i_atimensec = ts.tv_nsec;
1274                 }
1275                 if (ip->i_flag & IN_UPDATE) {
1276                         ip->i_mtime = ts.tv_sec;
1277                         ip->i_mtimensec = ts.tv_nsec;
1278                         ip->i_modrev++;
1279                 }
1280                 if (ip->i_flag & IN_CHANGE) {
1281                         ip->i_ctime = ts.tv_sec;
1282                         ip->i_ctimensec = ts.tv_nsec;
1283                 }
1284         }
1285         ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
1286 }
1287
1288 /*
1289  * Open called.
1290  *
1291  * Nothing to do.
1292  *
1293  * ext2_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
1294  *           struct file *a_fp)
1295  */
1296 /* ARGSUSED */
1297 static
1298 int
1299 ext2_open(struct vop_open_args *ap)
1300 {
1301         struct vnode *vp = ap->a_vp;
1302
1303         /*
1304          * Files marked append-only must be opened for appending.
1305          */
1306         if ((VTOI(vp)->i_flags & APPEND) &&
1307             (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE) {
1308                 return (EPERM);
1309         }
1310         return (vop_stdopen(ap));
1311 }
1312
1313 /*
1314  * Close called.
1315  *
1316  * Update the times on the inode.
1317  *
1318  * ext2_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
1319  */
1320 /* ARGSUSED */
1321 static
1322 int
1323 ext2_close(struct vop_close_args *ap)
1324 {
1325         struct vnode *vp = ap->a_vp;
1326
1327         if (vp->v_sysref.refcnt > 1)
1328                 ext2_itimes(vp);
1329         return (vop_stdclose(ap));
1330 }
1331
1332 /*
1333  * ext2_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred)
1334  */
1335 static
1336 int
1337 ext2_access(struct vop_access_args *ap)
1338 {
1339         struct vnode *vp = ap->a_vp;
1340         struct inode *ip = VTOI(vp);
1341         struct ucred *cred = ap->a_cred;
1342         mode_t mask, mode = ap->a_mode;
1343         gid_t *gp;
1344         int i;
1345 #ifdef QUOTA
1346         int error;
1347 #endif
1348
1349         /*
1350          * Disallow write attempts on read-only filesystems;
1351          * unless the file is a socket, fifo, or a block or
1352          * character device resident on the filesystem.
1353          */
1354         if (mode & VWRITE) {
1355                 switch (vp->v_type) {
1356                 case VDIR:
1357                 case VLNK:
1358                 case VREG:
1359                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
1360                                 return (EROFS);
1361 #ifdef QUOTA
1362                         if ((error = ext2_getinoquota(ip)) != 0)
1363                                 return (error);
1364 #endif
1365                         break;
1366                 default:
1367                         break;
1368                 }
1369         }
1370
1371         /* If immutable bit set, nobody gets to write it. */
1372         if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
1373                 return (EPERM);
1374
1375         /* Otherwise, user id 0 always gets access. */
1376         if (cred->cr_uid == 0)
1377                 return (0);
1378
1379         mask = 0;
1380
1381         /* Otherwise, check the owner. */
1382         if (cred->cr_uid == ip->i_uid) {
1383                 if (mode & VEXEC)
1384                         mask |= S_IXUSR;
1385                 if (mode & VREAD)
1386                         mask |= S_IRUSR;
1387                 if (mode & VWRITE)
1388                         mask |= S_IWUSR;
1389                 return ((ip->i_mode & mask) == mask ? 0 : EACCES);
1390         }
1391
1392         /* Otherwise, check the groups. */
1393         for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++)
1394                 if (ip->i_gid == *gp) {
1395                         if (mode & VEXEC)
1396                                 mask |= S_IXGRP;
1397                         if (mode & VREAD)
1398                                 mask |= S_IRGRP;
1399                         if (mode & VWRITE)
1400                                 mask |= S_IWGRP;
1401                         return ((ip->i_mode & mask) == mask ? 0 : EACCES);
1402                 }
1403
1404         /* Otherwise, check everyone else. */
1405         if (mode & VEXEC)
1406                 mask |= S_IXOTH;
1407         if (mode & VREAD)
1408                 mask |= S_IROTH;
1409         if (mode & VWRITE)
1410                 mask |= S_IWOTH;
1411         return ((ip->i_mode & mask) == mask ? 0 : EACCES);
1412 }
1413
1414 /*
1415  * ext2_getattr(struct vnode *a_vp, struct vattr *a_vap)
1416  */
1417 /* ARGSUSED */
1418 static
1419 int
1420 ext2_getattr(struct vop_getattr_args *ap)
1421 {
1422         struct vnode *vp = ap->a_vp;
1423         struct inode *ip = VTOI(vp);
1424         struct vattr *vap = ap->a_vap;
1425
1426         ext2_itimes(vp);
1427         /*
1428          * Copy from inode table
1429          */
1430         vap->va_fsid = dev2udev(ip->i_dev);
1431         vap->va_fileid = ip->i_number;
1432         vap->va_mode = ip->i_mode & ~IFMT;
1433         vap->va_nlink = VFSTOEXT2(vp->v_mount)->um_i_effnlink_valid ?
1434             ip->i_effnlink : ip->i_nlink;
1435         vap->va_uid = ip->i_uid;
1436         vap->va_gid = ip->i_gid;
1437         vap->va_rmajor = umajor(ip->i_rdev);
1438         vap->va_rminor = uminor(ip->i_rdev);
1439         vap->va_size = ip->i_din.di_size;
1440         vap->va_atime.tv_sec = ip->i_atime;
1441         vap->va_atime.tv_nsec = ip->i_atimensec;
1442         vap->va_mtime.tv_sec = ip->i_mtime;
1443         vap->va_mtime.tv_nsec = ip->i_mtimensec;
1444         vap->va_ctime.tv_sec = ip->i_ctime;
1445         vap->va_ctime.tv_nsec = ip->i_ctimensec;
1446         vap->va_flags = ip->i_flags;
1447         vap->va_gen = ip->i_gen;
1448         vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
1449         vap->va_bytes = dbtob((u_quad_t)ip->i_blocks);
1450         vap->va_type = IFTOVT(ip->i_mode);
1451         vap->va_filerev = ip->i_modrev;
1452         return (0);
1453 }
1454
1455 /*
1456  * Set attribute vnode op. called from several syscalls
1457  *
1458  * ext2_setattr(struct vnode *a_vp, struct vattr *a_vap, struct ucred *a_cred)
1459  */
1460 static
1461 int
1462 ext2_setattr(struct vop_setattr_args *ap)
1463 {
1464         struct vattr *vap = ap->a_vap;
1465         struct vnode *vp = ap->a_vp;
1466         struct inode *ip = VTOI(vp);
1467         struct ucred *cred = ap->a_cred;
1468         int error;
1469
1470         /*
1471          * Check for unsettable attributes.
1472          */
1473         if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
1474             (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
1475             (vap->va_blocksize != VNOVAL) || (vap->va_rmajor != VNOVAL) ||
1476             ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
1477                 return (EINVAL);
1478         }
1479         if (vap->va_flags != VNOVAL) {
1480                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1481                         return (EROFS);
1482                 if (cred->cr_uid != ip->i_uid &&
1483                     (error = priv_check_cred(cred, PRIV_ROOT, PRISON_ROOT)))
1484                         return (error);
1485                 /*
1486                  * Note that a root chflags becomes a user chflags when
1487                  * we are jailed, unless the jail.chflags_allowed sysctl
1488                  * is set.
1489                  */
1490                 if (cred->cr_uid == 0 &&
1491                     (!jailed(cred) || jail_chflags_allowed)) {
1492                         if ((ip->i_flags
1493                             & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) &&
1494                             securelevel > 0)
1495                                 return (EPERM);
1496                         ip->i_flags = vap->va_flags;
1497                 } else {
1498                         if (ip->i_flags
1499                             & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
1500                             (vap->va_flags & UF_SETTABLE) != vap->va_flags)
1501                                 return (EPERM);
1502                         ip->i_flags &= SF_SETTABLE;
1503                         ip->i_flags |= (vap->va_flags & UF_SETTABLE);
1504                 }
1505                 ip->i_flag |= IN_CHANGE;
1506                 if (vap->va_flags & (IMMUTABLE | APPEND))
1507                         return (0);
1508         }
1509         if (ip->i_flags & (IMMUTABLE | APPEND))
1510                 return (EPERM);
1511         /*
1512          * Go through the fields and update iff not VNOVAL.
1513          */
1514         if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
1515                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1516                         return (EROFS);
1517                 if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred)) != 0)
1518                         return (error);
1519         }
1520         if (vap->va_size != VNOVAL) {
1521                 /*
1522                  * Disallow write attempts on read-only filesystems;
1523                  * unless the file is a socket, fifo, or a block or
1524                  * character device resident on the filesystem.
1525                  */
1526                 switch (vp->v_type) {
1527                 case VDIR:
1528                         return (EISDIR);
1529                 case VLNK:
1530                 case VREG:
1531                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
1532                                 return (EROFS);
1533                         break;
1534                 default:
1535                         break;
1536                 }
1537                 if ((error = EXT2_TRUNCATE(vp, vap->va_size, 0, cred)) != 0)
1538                         return (error);
1539         }
1540         ip = VTOI(vp);
1541         if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
1542                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1543                         return (EROFS);
1544                 if (cred->cr_uid != ip->i_uid &&
1545                     (error = priv_check_cred(cred, PRIV_ROOT, PRISON_ROOT)) &&
1546                     ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
1547                     (error = VOP_ACCESS(vp, VWRITE, cred))))
1548                         return (error);
1549                 if (vap->va_atime.tv_sec != VNOVAL)
1550                         ip->i_flag |= IN_ACCESS;
1551                 if (vap->va_mtime.tv_sec != VNOVAL)
1552                         ip->i_flag |= IN_CHANGE | IN_UPDATE;
1553                 ext2_itimes(vp);
1554                 if (vap->va_atime.tv_sec != VNOVAL) {
1555                         ip->i_atime = vap->va_atime.tv_sec;
1556                         ip->i_atimensec = vap->va_atime.tv_nsec;
1557                 }
1558                 if (vap->va_mtime.tv_sec != VNOVAL) {
1559                         ip->i_mtime = vap->va_mtime.tv_sec;
1560                         ip->i_mtimensec = vap->va_mtime.tv_nsec;
1561                 }
1562                 error = EXT2_UPDATE(vp, 0);
1563                 if (error)
1564                         return (error);
1565         }
1566         error = 0;
1567         if (vap->va_mode != (mode_t)VNOVAL) {
1568                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
1569                         return (EROFS);
1570                 error = ext2_chmod(vp, (int)vap->va_mode, cred);
1571         }
1572         VN_KNOTE(vp, NOTE_ATTRIB);
1573         return (error);
1574 }
1575
1576 /*
1577  * Change the mode on a file.
1578  * Inode must be locked before calling.
1579  */
1580 static int
1581 ext2_chmod(struct vnode *vp, int mode, struct ucred *cred)
1582 {
1583         struct inode *ip = VTOI(vp);
1584         int error;
1585
1586         if (cred->cr_uid != ip->i_uid) {
1587             error = priv_check_cred(cred, PRIV_ROOT, PRISON_ROOT);
1588             if (error)
1589                 return (error);
1590         }
1591         if (cred->cr_uid) {
1592                 if (vp->v_type != VDIR && (mode & S_ISTXT))
1593                         return (EFTYPE);
1594                 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
1595                         return (EPERM);
1596         }
1597         ip->i_mode &= ~ALLPERMS;
1598         ip->i_mode |= (mode & ALLPERMS);
1599         ip->i_flag |= IN_CHANGE;
1600         return (0);
1601 }
1602
1603 /*
1604  * Perform chown operation on inode ip;
1605  * inode must be locked prior to call.
1606  */
1607 static int
1608 ext2_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred)
1609 {
1610         struct inode *ip = VTOI(vp);
1611         uid_t ouid;
1612         gid_t ogid;
1613         int error = 0;
1614 #ifdef QUOTA
1615         int i;
1616         long change;
1617 #endif
1618
1619         if (uid == (uid_t)VNOVAL)
1620                 uid = ip->i_uid;
1621         if (gid == (gid_t)VNOVAL)
1622                 gid = ip->i_gid;
1623         /*
1624          * If we don't own the file, are trying to change the owner
1625          * of the file, or are not a member of the target group,
1626          * the caller must be superuser or the call fails.
1627          */
1628         if ((cred->cr_uid != ip->i_uid || uid != ip->i_uid ||
1629             (gid != ip->i_gid && !(cred->cr_gid == gid ||
1630             groupmember((gid_t)gid, cred)))) &&
1631             (error = priv_check_cred(cred, PRIV_ROOT, PRISON_ROOT)))
1632                 return (error);
1633         ogid = ip->i_gid;
1634         ouid = ip->i_uid;
1635 #ifdef QUOTA
1636         if ((error = ext2_getinoquota(ip)) != 0)
1637                 return (error);
1638         if (ouid == uid) {
1639                 ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1640                 ip->i_dquot[USRQUOTA] = NODQUOT;
1641         }
1642         if (ogid == gid) {
1643                 ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1644                 ip->i_dquot[GRPQUOTA] = NODQUOT;
1645         }
1646         change = ip->i_blocks;
1647         (void) ext2_chkdq(ip, -change, cred, CHOWN);
1648         (void) ext2_chkiq(ip, -1, cred, CHOWN);
1649         for (i = 0; i < MAXQUOTAS; i++) {
1650                 ext2_dqrele(vp, ip->i_dquot[i]);
1651                 ip->i_dquot[i] = NODQUOT;
1652         }
1653 #endif
1654         ip->i_gid = gid;
1655         ip->i_uid = uid;
1656 #ifdef QUOTA
1657         if ((error = ext2_getinoquota(ip)) == 0) {
1658                 if (ouid == uid) {
1659                         ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1660                         ip->i_dquot[USRQUOTA] = NODQUOT;
1661                 }
1662                 if (ogid == gid) {
1663                         ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1664                         ip->i_dquot[GRPQUOTA] = NODQUOT;
1665                 }
1666                 if ((error = ext2_chkdq(ip, change, cred, CHOWN)) == 0) {
1667                         if ((error = ext2_chkiq(ip, 1, cred, CHOWN)) == 0)
1668                                 goto good;
1669                         else
1670                                 (void)ext2_chkdq(ip, -change, cred, CHOWN|FORCE);
1671                 }
1672                 for (i = 0; i < MAXQUOTAS; i++) {
1673                         ext2_dqrele(vp, ip->i_dquot[i]);
1674                         ip->i_dquot[i] = NODQUOT;
1675                 }
1676         }
1677         ip->i_gid = ogid;
1678         ip->i_uid = ouid;
1679         if (ext2_getinoquota(ip) == 0) {
1680                 if (ouid == uid) {
1681                         ext2_dqrele(vp, ip->i_dquot[USRQUOTA]);
1682                         ip->i_dquot[USRQUOTA] = NODQUOT;
1683                 }
1684                 if (ogid == gid) {
1685                         ext2_dqrele(vp, ip->i_dquot[GRPQUOTA]);
1686                         ip->i_dquot[GRPQUOTA] = NODQUOT;
1687                 }
1688                 (void) ext2_chkdq(ip, change, cred, FORCE|CHOWN);
1689                 (void) ext2_chkiq(ip, 1, cred, FORCE|CHOWN);
1690                 (void) ext2_getinoquota(ip);
1691         }
1692         return (error);
1693 good:
1694         if (ext2_getinoquota(ip))
1695                 panic("ext2_chown: lost quota");
1696 #endif /* QUOTA */
1697         ip->i_flag |= IN_CHANGE;
1698         if (cred->cr_uid != 0 && (ouid != uid || ogid != gid))
1699                 ip->i_mode &= ~(ISUID | ISGID);
1700         return (0);
1701 }
1702
1703 /*
1704  * Mmap a file
1705  *
1706  * NB Currently unsupported.
1707  *
1708  * ext2_mmap(struct vnode *a_vp, int a_fflags, struct ucred *a_cred)
1709  */
1710 /* ARGSUSED */
1711 static
1712 int
1713 ext2_mmap(struct vop_mmap_args *ap)
1714 {
1715         return (EINVAL);
1716 }
1717
1718 /*
1719  * whiteout vnode call
1720  *
1721  * ext2_whiteout(struct vnode *a_dvp, struct componentname *a_cnp, int a_flags)
1722  */
1723 static
1724 int
1725 ext2_whiteout(struct vop_old_whiteout_args *ap)
1726 {
1727         return (EOPNOTSUPP);
1728 }
1729
1730 /*
1731  * Return target name of a symbolic link
1732  *
1733  * ext2_readlink(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred)
1734  */
1735 static
1736 int
1737 ext2_readlink(struct vop_readlink_args *ap)
1738 {
1739         struct vnode *vp = ap->a_vp;
1740         struct inode *ip = VTOI(vp);
1741         int isize;
1742
1743         isize = ip->i_size;
1744         if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
1745             (ip->i_din.di_blocks == 0)) {   /* XXX - for old fastlink support */
1746                 uiomove((char *)ip->i_shortlink, isize, ap->a_uio);
1747                 return (0);
1748         }
1749         return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1750 }
1751
1752 /*
1753  * Calculate the logical to physical mapping if not done already,
1754  * then call the device strategy routine.
1755  *
1756  * In order to be able to swap to a file, the VOP_BMAP operation may not
1757  * deadlock on memory.  See ext2_bmap() for details.
1758  *
1759  * ext2_strategy(struct vnode *a_vp, struct bio *a_bio)
1760  */
1761 static
1762 int
1763 ext2_strategy(struct vop_strategy_args *ap)
1764 {
1765         struct bio *bio = ap->a_bio;
1766         struct bio *nbio;
1767         struct buf *bp = bio->bio_buf;
1768         struct vnode *vp = ap->a_vp;
1769         struct inode *ip;
1770         int error;
1771
1772         ip = VTOI(vp);
1773         if (vp->v_type == VBLK || vp->v_type == VCHR)
1774                 panic("ext2_strategy: spec");
1775         nbio = push_bio(bio);
1776         if (nbio->bio_offset == NOOFFSET) {
1777                 error = VOP_BMAP(vp, bio->bio_offset, &nbio->bio_offset,
1778                                  NULL, NULL, bp->b_cmd);
1779                 if (error) {
1780                         bp->b_error = error;
1781                         bp->b_flags |= B_ERROR;
1782                         /* I/O was never started on nbio, must biodone(bio) */
1783                         biodone(bio);
1784                         return (error);
1785                 }
1786                 if (nbio->bio_offset == NOOFFSET)
1787                         vfs_bio_clrbuf(bp);
1788         }
1789         if (nbio->bio_offset == NOOFFSET) {
1790                 /* I/O was never started on nbio, must biodone(bio) */
1791                 biodone(bio);
1792                 return (0);
1793         }
1794         vn_strategy(ip->i_devvp, nbio);
1795         return (0);
1796 }
1797
1798 /*
1799  * Print out the contents of an inode.
1800  *
1801  * ext2_print(struct vnode *a_vp)
1802  */
1803 static
1804 int
1805 ext2_print(struct vop_print_args *ap)
1806 {
1807         struct vnode *vp = ap->a_vp;
1808         struct inode *ip = VTOI(vp);
1809
1810         kprintf("tag VT_EXT2FS, ino %lu, on dev %s (%d, %d)",
1811             (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev),
1812             minor(ip->i_dev));
1813         if (vp->v_type == VFIFO)
1814                 fifo_printinfo(vp);
1815         lockmgr_printinfo(&vp->v_lock);
1816         kprintf("\n");
1817         return (0);
1818 }
1819
1820 /*
1821  * Read wrapper for special devices.
1822  *
1823  * ext2spec_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1824  *              struct ucred *a_cred)
1825  */
1826 static
1827 int
1828 ext2spec_read(struct vop_read_args *ap)
1829 {
1830         int error, resid;
1831         struct inode *ip;
1832         struct uio *uio;
1833
1834         uio = ap->a_uio;
1835         resid = uio->uio_resid;
1836         error = VOCALL(&spec_vnode_vops, &ap->a_head);
1837         /*
1838          * The inode may have been revoked during the call, so it must not
1839          * be accessed blindly here or in the other wrapper functions.
1840          */
1841         ip = VTOI(ap->a_vp);
1842         if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1843                 ip->i_flag |= IN_ACCESS;
1844         return (error);
1845 }
1846
1847 /*
1848  * Write wrapper for special devices.
1849  *
1850  * ext2spec_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1851  *               struct ucred *a_cred)
1852  */
1853 static
1854 int
1855 ext2spec_write(struct vop_write_args *ap)
1856 {
1857         int error, resid;
1858         struct inode *ip;
1859         struct uio *uio;
1860
1861         uio = ap->a_uio;
1862         resid = uio->uio_resid;
1863         error = VOCALL(&spec_vnode_vops, &ap->a_head);
1864         ip = VTOI(ap->a_vp);
1865         if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1866                 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1867         return (error);
1868 }
1869
1870 /*
1871  * Close wrapper for special devices.
1872  *
1873  * Update the times on the inode then do device close.
1874  *
1875  * ext2spec_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
1876  */
1877 static
1878 int
1879 ext2spec_close(struct vop_close_args *ap)
1880 {
1881         struct vnode *vp = ap->a_vp;
1882
1883         if (vp->v_sysref.refcnt > 1)
1884                 ext2_itimes(vp);
1885         return (VOCALL(&spec_vnode_vops, &ap->a_head));
1886 }
1887
1888 /*
1889  * Read wrapper for fifos.
1890  *
1891  * ext2fifo_read(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1892  *              struct ucred *a_cred)
1893  */
1894 static
1895 int
1896 ext2fifo_read(struct vop_read_args *ap)
1897 {
1898         int error, resid;
1899         struct inode *ip;
1900         struct uio *uio;
1901
1902         uio = ap->a_uio;
1903         resid = uio->uio_resid;
1904         error = VOCALL(&fifo_vnode_vops, &ap->a_head);
1905         ip = VTOI(ap->a_vp);
1906         if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
1907             (uio->uio_resid != resid || (error == 0 && resid != 0)))
1908                 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
1909         return (error);
1910 }
1911
1912 /*
1913  * Write wrapper for fifos.
1914  *
1915  * ext2fifo_write(struct vnode *a_vp, struct uio *a_uio, int a_ioflag,
1916  *               struct ucred *a_cred)
1917  */
1918 static
1919 int
1920 ext2fifo_write(struct vop_write_args *ap)
1921 {
1922         int error, resid;
1923         struct inode *ip;
1924         struct uio *uio;
1925
1926         uio = ap->a_uio;
1927         resid = uio->uio_resid;
1928         error = VOCALL(&fifo_vnode_vops, &ap->a_head);
1929         ip = VTOI(ap->a_vp);
1930         if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
1931                 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
1932         return (error);
1933 }
1934
1935 /*
1936  * Close wrapper for fifos.
1937  *
1938  * Update the times on the inode then do device close.
1939  *
1940  * ext2fifo_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred)
1941  */
1942 static
1943 int
1944 ext2fifo_close(struct vop_close_args *ap)
1945 {
1946         struct vnode *vp = ap->a_vp;
1947
1948         if (vp->v_sysref.refcnt > 1)
1949                 ext2_itimes(vp);
1950         return (VOCALL(&fifo_vnode_vops, &ap->a_head));
1951 }
1952
1953 /*
1954  * Kqfilter wrapper for fifos.
1955  *
1956  * Fall through to ext2 kqfilter routines if needed
1957  */
1958 static
1959 int
1960 ext2fifo_kqfilter(struct vop_kqfilter_args *ap)
1961 {
1962         int error;
1963
1964         error = VOCALL(&fifo_vnode_vops, &ap->a_head);
1965         if (error)
1966                 error = ext2_kqfilter(ap);
1967         return (error);
1968 }
1969
1970 /*
1971  * Return POSIX pathconf information applicable to ext2 filesystems.
1972  *
1973  * ext2_pathconf(struct vnode *a_vp, int a_name, int *a_retval)
1974  */
1975 static
1976 int
1977 ext2_pathconf(struct vop_pathconf_args *ap)
1978 {
1979         switch (ap->a_name) {
1980         case _PC_LINK_MAX:
1981                 *ap->a_retval = LINK_MAX;
1982                 return (0);
1983         case _PC_NAME_MAX:
1984                 *ap->a_retval = NAME_MAX;
1985                 return (0);
1986         case _PC_PATH_MAX:
1987                 *ap->a_retval = PATH_MAX;
1988                 return (0);
1989         case _PC_PIPE_BUF:
1990                 *ap->a_retval = PIPE_BUF;
1991                 return (0);
1992         case _PC_CHOWN_RESTRICTED:
1993                 *ap->a_retval = 1;
1994                 return (0);
1995         case _PC_NO_TRUNC:
1996                 *ap->a_retval = 1;
1997                 return (0);
1998         default:
1999                 return (EINVAL);
2000         }
2001         /* NOTREACHED */
2002 }
2003
2004 /*
2005  * Advisory record locking support
2006  *
2007  * ext2_advlock(struct vnode *a_vp, caddr_t a_id, int a_op, struct flock *a_fl,
2008  *             int a_flags)
2009  */
2010 static
2011 int
2012 ext2_advlock(struct vop_advlock_args *ap)
2013 {
2014         struct inode *ip = VTOI(ap->a_vp);
2015
2016         return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
2017 }
2018
2019 /*
2020  * Initialize the vnode associated with a new inode, handle aliased
2021  * vnodes.
2022  */
2023 int
2024 ext2_vinit(struct mount *mntp, struct vnode **vpp)
2025 {
2026         struct inode *ip;
2027         struct vnode *vp;
2028         struct timeval tv;
2029
2030         vp = *vpp;
2031         ip = VTOI(vp);
2032
2033         switch(vp->v_type = IFTOVT(ip->i_mode)) {
2034         case VCHR:
2035         case VBLK:
2036                 vp->v_ops = &mntp->mnt_vn_spec_ops;
2037                 addaliasu(vp, umajor(ip->i_rdev), uminor(ip->i_rdev));
2038                 break;
2039         case VFIFO:
2040                 vp->v_ops = &mntp->mnt_vn_fifo_ops;
2041                 break;
2042         case VDIR:
2043         case VREG:
2044                 vinitvmio(vp, ip->i_size);
2045                 break;
2046         case VLNK:
2047                 if ((ip->i_size >= vp->v_mount->mnt_maxsymlinklen) &&
2048                     ip->i_din.di_blocks != 0
2049                 ) {
2050                         vinitvmio(vp, ip->i_size);
2051                 }
2052                 break;
2053         default:
2054                 break;
2055
2056         }
2057
2058         if (ip->i_number == ROOTINO)
2059                 vp->v_flag |= VROOT;
2060         /*
2061          * Initialize modrev times
2062          */
2063         getmicrouptime(&tv);
2064         SETHIGH(ip->i_modrev, tv.tv_sec);
2065         SETLOW(ip->i_modrev, tv.tv_usec * 4294);
2066         *vpp = vp;
2067         return (0);
2068 }
2069
2070 static struct filterops ext2read_filtops =
2071         { 1, NULL, filt_ext2detach, filt_ext2read };
2072 static struct filterops ext2write_filtops =
2073         { 1, NULL, filt_ext2detach, filt_ext2write };
2074 static struct filterops ext2vnode_filtops =
2075         { 1, NULL, filt_ext2detach, filt_ext2vnode };
2076
2077 /*
2078  * ext2_kqfilter(struct vnode *a_vp, struct knote *a_kn)
2079  */
2080 static int
2081 ext2_kqfilter(struct vop_kqfilter_args *ap)
2082 {
2083         struct vnode *vp = ap->a_vp;
2084         struct knote *kn = ap->a_kn;
2085         lwkt_tokref ilock;
2086
2087         switch (kn->kn_filter) {
2088         case EVFILT_READ:
2089                 kn->kn_fop = &ext2read_filtops;
2090                 break;
2091         case EVFILT_WRITE:
2092                 kn->kn_fop = &ext2write_filtops;
2093                 break;
2094         case EVFILT_VNODE:
2095                 kn->kn_fop = &ext2vnode_filtops;
2096                 break;
2097         default:
2098                 return (1);
2099         }
2100
2101         kn->kn_hook = (caddr_t)vp;
2102
2103         lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2104         SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
2105         lwkt_reltoken(&ilock);
2106
2107         return (0);
2108 }
2109
2110 static void
2111 filt_ext2detach(struct knote *kn)
2112 {
2113         struct vnode *vp = (struct vnode *)kn->kn_hook;
2114         lwkt_tokref ilock;
2115
2116         lwkt_gettoken(&ilock, &vp->v_pollinfo.vpi_token);
2117         SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
2118             kn, knote, kn_selnext);
2119         lwkt_reltoken(&ilock);
2120 }
2121
2122 /*ARGSUSED*/
2123 static int
2124 filt_ext2read(struct knote *kn, long hint)
2125 {
2126         struct vnode *vp = (struct vnode *)kn->kn_hook;
2127         struct inode *ip = VTOI(vp);
2128
2129         /*
2130          * filesystem is gone, so set the EOF flag and schedule
2131          * the knote for deletion.
2132          */
2133         if (hint == NOTE_REVOKE) {
2134                 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2135                 return (1);
2136         }
2137
2138         kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
2139         return (kn->kn_data != 0);
2140 }
2141
2142 /*ARGSUSED*/
2143 static int
2144 filt_ext2write(struct knote *kn, long hint)
2145 {
2146         /*
2147          * filesystem is gone, so set the EOF flag and schedule
2148          * the knote for deletion.
2149          */
2150         if (hint == NOTE_REVOKE)
2151                 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
2152
2153         kn->kn_data = 0;
2154         return (1);
2155 }
2156
2157 static int
2158 filt_ext2vnode(struct knote *kn, long hint)
2159 {
2160         if (kn->kn_sfflags & hint)
2161                 kn->kn_fflags |= hint;
2162         if (hint == NOTE_REVOKE) {
2163                 kn->kn_flags |= EV_EOF;
2164                 return (1);
2165         }
2166         return (kn->kn_fflags != 0);
2167 }
2168
2169 struct vop_ops ext2_vnode_vops = {
2170         .vop_default =          vop_defaultop,
2171         .vop_fsync =            ext2_fsync,
2172         .vop_read =             ext2_read,
2173         .vop_reallocblks =      ext2_reallocblks,
2174         .vop_write =            ext2_write,
2175         .vop_access =           ext2_access,
2176         .vop_advlock =          ext2_advlock,
2177         .vop_bmap =             ext2_bmap,
2178         .vop_old_lookup =       ext2_lookup,
2179         .vop_close =            ext2_close,
2180         .vop_old_create =       ext2_create,
2181         .vop_getattr =          ext2_getattr,
2182         .vop_inactive =         ext2_inactive,
2183         .vop_old_link =         ext2_link,
2184         .vop_old_mkdir =        ext2_mkdir,
2185         .vop_old_mknod =        ext2_mknod,
2186         .vop_mmap =             ext2_mmap,
2187         .vop_open =             ext2_open,
2188         .vop_pathconf =         ext2_pathconf,
2189         .vop_poll =             vop_stdpoll,
2190         .vop_kqfilter =         ext2_kqfilter,
2191         .vop_print =            ext2_print,
2192         .vop_readdir =          ext2_readdir,
2193         .vop_readlink =         ext2_readlink,
2194         .vop_reclaim =          ext2_reclaim,
2195         .vop_old_remove =       ext2_remove,
2196         .vop_old_rename =       ext2_rename,
2197         .vop_old_rmdir =        ext2_rmdir,
2198         .vop_setattr =          ext2_setattr,
2199         .vop_strategy =         ext2_strategy,
2200         .vop_old_symlink =      ext2_symlink,
2201         .vop_old_whiteout =     ext2_whiteout,
2202         .vop_getpages =         ext2_getpages,
2203         .vop_putpages =         vop_stdputpages
2204 };
2205
2206 struct vop_ops ext2_spec_vops = {
2207         .vop_default =          ext2_vnoperatespec,
2208         .vop_fsync =            ext2_fsync,
2209         .vop_access =           ext2_access,
2210         .vop_close =            ext2spec_close,
2211         .vop_getattr =          ext2_getattr,
2212         .vop_inactive =         ext2_inactive,
2213         .vop_print =            ext2_print,
2214         .vop_read =             ext2spec_read,
2215         .vop_reclaim =          ext2_reclaim,
2216         .vop_setattr =          ext2_setattr,
2217         .vop_write =            ext2spec_write
2218 };
2219
2220 struct vop_ops ext2_fifo_vops = {
2221         .vop_default =          ext2_vnoperatefifo,
2222         .vop_fsync =            ext2_fsync,
2223         .vop_access =           ext2_access,
2224         .vop_close =            ext2fifo_close,
2225         .vop_getattr =          ext2_getattr,
2226         .vop_inactive =         ext2_inactive,
2227         .vop_kqfilter =         ext2fifo_kqfilter,
2228         .vop_print =            ext2_print,
2229         .vop_read =             ext2fifo_read,
2230         .vop_reclaim =          ext2_reclaim,
2231         .vop_setattr =          ext2_setattr,
2232         .vop_write =            ext2fifo_write
2233 };
2234
2235 VNODEOP_SET(ext2_vnode_vops);
2236 VNODEOP_SET(ext2_spec_vops);
2237 VNODEOP_SET(ext2_fifo_vops);
2238
2239 /*
2240  * ext2_vnoperate()
2241  */
2242 int
2243 ext2_vnoperate(struct vop_generic_args *ap)
2244 {
2245         return (VOCALL(&ext2_vnode_vops, ap));
2246 }
2247
2248 /*
2249  * ext2_vnoperatefifo()
2250  */
2251 int
2252 ext2_vnoperatefifo(struct vop_generic_args *ap)
2253 {
2254         return (VOCALL(&ext2_fifo_vops, ap));
2255 }
2256
2257 /*
2258  * ext2_vnoperatespec()
2259  */
2260 int
2261 ext2_vnoperatespec(struct vop_generic_args *ap)
2262 {
2263         return (VOCALL(&ext2_spec_vops, ap));
2264 }
2265