bsd/miscfs/union/union_vfsops.c

   1 /*
   2  * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * The contents of this file constitute Original Code as defined in and
   7  * are subject to the Apple Public Source License Version 1.1 (the
   8  * "License").  You may not use this file except in compliance with the
   9  * License.  Please obtain a copy of the License at
  10  * http://www.apple.com/publicsource and read it before using this file.
  11  *
  12  * This Original Code and all software distributed under the License are
  13  * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  14  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  15  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
  17  * License for the specific language governing rights and limitations
  18  * under the License.
  19  *
  20  * @APPLE_LICENSE_HEADER_END@
  21  */
  22 /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
  23 /*
  24  * Copyright (c) 1994, 1995 The Regents of the University of California.
  25  * Copyright (c) 1994, 1995 Jan-Simon Pendry.
  26  * All rights reserved.
  27  *
  28  * This code is derived from software donated to Berkeley by
  29  * Jan-Simon Pendry.
  30  *
  31  * Redistribution and use in source and binary forms, with or without
  32  * modification, are permitted provided that the following conditions
  33  * are met:
  34  * 1. Redistributions of source code must retain the above copyright
  35  *    notice, this list of conditions and the following disclaimer.
  36  * 2. Redistributions in binary form must reproduce the above copyright
  37  *    notice, this list of conditions and the following disclaimer in the
  38  *    documentation and/or other materials provided with the distribution.
  39  * 3. All advertising materials mentioning features or use of this software
  40  *    must display the following acknowledgement:
  41  *      This product includes software developed by the University of
  42  *      California, Berkeley and its contributors.
  43  * 4. Neither the name of the University nor the names of its contributors
  44  *    may be used to endorse or promote products derived from this software
  45  *    without specific prior written permission.
  46  *
  47  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  48  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  49  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  50  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  51  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  52  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  53  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  54  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  55  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  56  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  57  * SUCH DAMAGE.
  58  *
  59  *      @(#)union_vfsops.c      8.20 (Berkeley) 5/20/95
  60  */
  61
  62 /*
  63  * Union Layer
  64  */
  65
  66 #include <sys/param.h>
  67 #include <sys/systm.h>
  68 #include <sys/time.h>
  69 #include <sys/types.h>
  70 #include <sys/proc_internal.h>
  71 #include <sys/kauth.h>
  72 #include <sys/vnode_internal.h>
  73 #include <sys/mount_internal.h>
  74 #include <sys/namei.h>
  75 #include <sys/malloc.h>
  76 #include <sys/filedesc.h>
  77 #include <sys/queue.h>
  78 #include <miscfs/union/union.h>
  79
  80 static  int union_itercallback(__unused vnode_t, void *);
  81
  82 /*
  83  * Mount union filesystem
  84  */
  85 int
  86 union_mount(mount_t mp, __unused vnode_t devvp, user_addr_t data, vfs_context_t context)
  87 {
  88         proc_t p = vfs_context_proc(context);
  89         int error = 0;
  90         struct user_union_args args;
  91         struct vnode *lowerrootvp = NULLVP;
  92         struct vnode *upperrootvp = NULLVP;
  93         struct union_mount *um = 0;
  94         kauth_cred_t cred = NOCRED;
  95         char *cp;
  96         int len;
  97         u_int size;
  98         struct nameidata nd;
  99
 100 #ifdef UNION_DIAGNOSTIC
 101         printf("union_mount(mp = %x)\n", mp);
 102 #endif
 103
 104         /*
 105          * Update is a no-op
 106          */
 107         if (mp->mnt_flag & MNT_UPDATE) {
 108                 /*
 109                  * Need to provide.
 110                  * 1. a way to convert between rdonly and rdwr mounts.
 111                  * 2. support for nfs exports.
 112                  */
 113                 error = ENOTSUP;
 114                 goto bad;
 115         }
 116
 117         /*
 118          * Get argument
 119          */
 120         if (vfs_context_is64bit(context)) {
 121                 error = copyin(data, (caddr_t)&args, sizeof(args));
 122         }
 123         else {
 124                 struct union_args temp;
 125                 error = copyin(data, (caddr_t)&temp, sizeof (temp));
 126                 args.target = CAST_USER_ADDR_T(temp.target);
 127                 args.mntflags = temp.mntflags;
 128         }
 129         if (error)
 130                 goto bad;
 131
 132         lowerrootvp = mp->mnt_vnodecovered;
 133         vnode_get(lowerrootvp);
 134
 135         /*
 136          * Find upper node.
 137          */
 138         NDINIT(&nd, LOOKUP, FOLLOW|WANTPARENT,
 139                (IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32),
 140                args.target, context);
 141
 142         if ((error = namei(&nd)))
 143                 goto bad;
 144
 145         nameidone(&nd);
 146         upperrootvp = nd.ni_vp;
 147         vnode_put(nd.ni_dvp);
 148         nd.ni_dvp = NULL;
 149
 150         if (upperrootvp->v_type != VDIR) {
 151                 error = EINVAL;
 152                 goto bad;
 153         }
 154
 155 //      um = (struct union_mount *) malloc(sizeof(struct union_mount),
 156 //                              M_UFSMNT, M_WAITOK);    /* XXX */
 157         MALLOC(um, struct union_mount *, sizeof(struct union_mount),
 158                                 M_UFSMNT, M_WAITOK);
 159
 160         /*
 161          * Keep a held reference to the target vnodes.
 162          * They are vnode_put'd in union_unmount.
 163          *
 164          * Depending on the _BELOW flag, the filesystems are
 165          * viewed in a different order.  In effect, this is the
 166          * same as providing a mount under option to the mount syscall.
 167          */
 168
 169         um->um_op = args.mntflags & UNMNT_OPMASK;
 170         switch (um->um_op) {
 171         case UNMNT_ABOVE:
 172                 um->um_lowervp = lowerrootvp;
 173                 um->um_uppervp = upperrootvp;
 174                 break;
 175
 176         case UNMNT_BELOW:
 177                 um->um_lowervp = upperrootvp;
 178                 um->um_uppervp = lowerrootvp;
 179                 break;
 180
 181         case UNMNT_REPLACE:
 182                 vnode_put(lowerrootvp);
 183                 lowerrootvp = NULLVP;
 184                 um->um_uppervp = upperrootvp;
 185                 um->um_lowervp = lowerrootvp;
 186                 break;
 187
 188         default:
 189                 error = EINVAL;
 190                 goto bad;
 191         }
 192
 193         /*
 194          * Unless the mount is readonly, ensure that the top layer
 195          * supports whiteout operations
 196          */
 197         if ((mp->mnt_flag & MNT_RDONLY) == 0) {
 198                 error = VNOP_WHITEOUT(um->um_uppervp, (struct componentname *) 0,
 199                                       LOOKUP, context);
 200                 if (error)
 201                         goto bad;
 202         }
 203
 204         um->um_cred = kauth_cred_get_with_ref();
 205         um->um_cmode = UN_DIRMODE &~ p->p_fd->fd_cmask;
 206
 207         /*
 208          * Depending on what you think the MNT_LOCAL flag might mean,
 209          * you may want the && to be || on the conditional below.
 210          * At the moment it has been defined that the filesystem is
 211          * only local if it is all local, ie the MNT_LOCAL flag implies
 212          * that the entire namespace is local.  If you think the MNT_LOCAL
 213          * flag implies that some of the files might be stored locally
 214          * then you will want to change the conditional.
 215          */
 216         if (um->um_op == UNMNT_ABOVE) {
 217                 if (((um->um_lowervp == NULLVP) ||
 218                      (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) &&
 219                     (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
 220                         mp->mnt_flag |= MNT_LOCAL;
 221         }
 222
 223         /*
 224          * Copy in the upper layer's RDONLY flag.  This is for the benefit
 225          * of lookup() which explicitly checks the flag, rather than asking
 226          * the filesystem for it's own opinion.  This means, that an update
 227          * mount of the underlying filesystem to go from rdonly to rdwr
 228          * will leave the unioned view as read-only.
 229          */
 230         mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY);
 231
 232         mp->mnt_data = (qaddr_t) um;
 233         vfs_getnewfsid(mp);
 234
 235
 236         switch (um->um_op) {
 237         case UNMNT_ABOVE:
 238                 cp = "<above>:";
 239                 break;
 240         case UNMNT_BELOW:
 241                 cp = "<below>:";
 242                 break;
 243         case UNMNT_REPLACE:
 244                 cp = "";
 245                 break;
 246         }
 247         len = strlen(cp);
 248         bcopy(cp, mp->mnt_vfsstat.f_mntfromname, len);
 249
 250         cp = mp->mnt_vfsstat.f_mntfromname + len;
 251         len = MNAMELEN - len;
 252
 253         (void) copyinstr(args.target, cp, len - 1, (size_t *)&size);
 254         bzero(cp + size, len - size);
 255
 256 #ifdef UNION_DIAGNOSTIC
 257         printf("union_mount: from %s, on %s\n",
 258                 mp->mnt_vfsstat.f_mntfromname, mp->mnt_vfsstat.f_mntonname);
 259 #endif
 260         return (0);
 261
 262 bad:
 263         if (um)
 264                 _FREE(um, M_UFSMNT);
 265         if (cred != NOCRED)
 266                 kauth_cred_rele(cred);
 267         if (upperrootvp)
 268                 vnode_put(upperrootvp);
 269         if (lowerrootvp)
 270                 vnode_put(lowerrootvp);
 271         return (error);
 272 }
 273
 274 /*
 275  * VFS start.  Nothing needed here - the start routine
 276  * on the underlying filesystem(s) will have been called
 277  * when that filesystem was mounted.
 278  */
 279 int
 280 union_start(__unused struct mount *mp, __unused int flags, __unused vfs_context_t context)
 281 {
 282
 283         return (0);
 284 }
 285
 286 static int
 287 union_itercallback(__unused vnode_t vp, void *args)
 288 {
 289         int  num = *(int *)args;
 290
 291         *(int *)args = num + 1;
 292         return(VNODE_RETURNED);
 293 }
 294
 295
 296
 297 /*
 298  * Free reference to union layer
 299  */
 300 int
 301 union_unmount(mount_t mp, int mntflags, __unused vfs_context_t context)
 302 {
 303         struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
 304         struct vnode *um_rootvp;
 305         int error;
 306         int freeing;
 307         int flags = 0;
 308         kauth_cred_t cred;
 309
 310 #ifdef UNION_DIAGNOSTIC
 311         printf("union_unmount(mp = %x)\n", mp);
 312 #endif
 313
 314         if (mntflags & MNT_FORCE)
 315                 flags |= FORCECLOSE;
 316
 317         if ((error = union_root(mp, &um_rootvp)))
 318                 return (error);
 319
 320         /*
 321          * Keep flushing vnodes from the mount list.
 322          * This is needed because of the un_pvp held
 323          * reference to the parent vnode.
 324          * If more vnodes have been freed on a given pass,
 325          * the try again.  The loop will iterate at most
 326          * (d) times, where (d) is the maximum tree depth
 327          * in the filesystem.
 328          */
 329         for (freeing = 0; vflush(mp, um_rootvp, flags) != 0;) {
 330                 int n = 0;
 331
 332                 vnode_iterate(mp, VNODE_NOLOCK_INTERNAL, union_itercallback, &n);
 333
 334                 /* if this is unchanged then stop */
 335                 if (n == freeing)
 336                         break;
 337
 338                 /* otherwise try once more time */
 339                 freeing = n;
 340         }
 341
 342         /* At this point the root vnode should have a single reference */
 343         if (vnode_isinuse(um_rootvp, 0)) {
 344                 vnode_put(um_rootvp);
 345                 return (EBUSY);
 346         }
 347
 348 #ifdef UNION_DIAGNOSTIC
 349         vprint("union root", um_rootvp);
 350 #endif
 351         /*
 352          * Discard references to upper and lower target vnodes.
 353          */
 354         if (um->um_lowervp)
 355                 vnode_put(um->um_lowervp);
 356         vnode_put(um->um_uppervp);
 357         cred = um->um_cred;
 358         if (cred != NOCRED) {
 359                 um->um_cred = NOCRED;
 360                 kauth_cred_rele(cred);
 361         }
 362         /*
 363          * Release reference on underlying root vnode
 364          */
 365         vnode_put(um_rootvp);
 366         /*
 367          * And blow it away for future re-use
 368          */
 369         vnode_reclaim(um_rootvp);
 370         /*
 371          * Finally, throw away the union_mount structure
 372          */
 373         _FREE(mp->mnt_data, M_UFSMNT);  /* XXX */
 374         mp->mnt_data = 0;
 375         return (0);
 376 }
 377
 378 int
 379 union_root(mount_t mp, vnode_t *vpp, __unused vfs_context_t context)
 380 {
 381         struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
 382         int error;
 383
 384         /*
 385          * Return locked reference to root.
 386          */
 387         vnode_get(um->um_uppervp);
 388         if (um->um_lowervp)
 389                 vnode_get(um->um_lowervp);
 390         error = union_allocvp(vpp, mp,
 391                               (struct vnode *) 0,
 392                               (struct vnode *) 0,
 393                               (struct componentname *) 0,
 394                               um->um_uppervp,
 395                               um->um_lowervp,
 396                               1);
 397
 398         if (error) {
 399                 vnode_put(um->um_uppervp);
 400                 if (um->um_lowervp)
 401                         vnode_put(um->um_lowervp);
 402         }
 403
 404         return (error);
 405 }
 406
 407 static int
 408 union_vfs_getattr(mount_t mp, struct vfs_attr *fsap, vfs_context_t context)
 409 {
 410         int error;
 411         struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
 412         struct vfs_attr attr;
 413         uint32_t lbsize = 0;
 414
 415 #ifdef UNION_DIAGNOSTIC
 416         printf("union_vfs_getattr(mp = %x, lvp = %x, uvp = %x)\n", mp,
 417                         um->um_lowervp,
 418                         um->um_uppervp);
 419 #endif
 420
 421         /* Get values from lower file system (if any) */
 422         if (um->um_lowervp) {
 423                 VFSATTR_INIT(&attr);
 424                 VFSATTR_WANTED(&attr, f_bsize);
 425                 VFSATTR_WANTED(&attr, f_blocks);
 426                 VFSATTR_WANTED(&attr, f_bused);
 427                 VFSATTR_WANTED(&attr, f_files);
 428                 error = vfs_getattr(um->um_lowervp->v_mount, &attr, context);
 429                 if (error)
 430                         return (error);
 431
 432                 /* now copy across the "interesting" information and fake the rest */
 433                 if (VFSATTR_IS_SUPPORTED(&attr, f_bsize))
 434                         lbsize = attr.f_bsize;
 435                 else
 436                         lbsize = um->um_lowervp->v_mount->mnt_devblocksize;
 437                 fsap->f_blocks = VFSATTR_IS_SUPPORTED(&attr, f_blocks) ? attr.f_blocks : 0;
 438                 fsap->f_bused  = VFSATTR_IS_SUPPORTED(&attr, f_bused)  ? attr.f_bused  : 0;
 439                 fsap->f_files  = VFSATTR_IS_SUPPORTED(&attr, f_files)  ? attr.f_files  : 0;
 440         } else {
 441                 fsap->f_blocks = 0;
 442                 fsap->f_bused = 0;
 443                 fsap->f_files = 0;
 444         }
 445
 446         VFSATTR_INIT(&attr);
 447         VFSATTR_WANTED(&attr, f_bsize);
 448         VFSATTR_WANTED(&attr, f_blocks);
 449         VFSATTR_WANTED(&attr, f_bfree);
 450         VFSATTR_WANTED(&attr, f_bavail);
 451         VFSATTR_WANTED(&attr, f_files);
 452         VFSATTR_WANTED(&attr, f_ffree);
 453         error = vfs_getattr(um->um_uppervp->v_mount, &attr, context);
 454         if (error)
 455                 return (error);
 456
 457         if (VFSATTR_IS_SUPPORTED(&attr, f_bsize)) {
 458                 fsap->f_bsize = attr.f_bsize;
 459                 VFSATTR_SET_SUPPORTED(fsap, f_bsize);
 460         }
 461         if (VFSATTR_IS_SUPPORTED(&attr, f_iosize)) {
 462                 fsap->f_iosize = attr.f_iosize;
 463                 VFSATTR_SET_SUPPORTED(fsap, f_iosize);
 464         }
 465
 466         /*
 467          * if the lower and upper blocksizes differ, then frig the
 468          * block counts so that the sizes reported by df make some
 469          * kind of sense.  none of this makes sense though.
 470          */
 471         if (VFSATTR_IS_SUPPORTED(&attr, f_bsize))
 472                 fsap->f_bsize = attr.f_bsize;
 473         else
 474                 fsap->f_bsize =  um->um_uppervp->v_mount->mnt_devblocksize;
 475         VFSATTR_RETURN(fsap, f_bsize, attr.f_bsize);
 476         if (fsap->f_bsize != lbsize)
 477                 fsap->f_blocks = fsap->f_blocks * lbsize / attr.f_bsize;
 478
 479         /*
 480          * The "total" fields count total resources in all layers,
 481          * the "free" fields count only those resources which are
 482          * free in the upper layer (since only the upper layer
 483          * is writeable).
 484          */
 485         if (VFSATTR_IS_SUPPORTED(&attr, f_blocks))
 486                 fsap->f_blocks += attr.f_blocks;
 487         if (VFSATTR_IS_SUPPORTED(&attr, f_bfree))
 488                 fsap->f_bfree = attr.f_bfree;
 489         if (VFSATTR_IS_SUPPORTED(&attr, f_bavail))
 490                 fsap->f_bavail = attr.f_bavail;
 491         if (VFSATTR_IS_SUPPORTED(&attr, f_bused))
 492                 fsap->f_bused += attr.f_bused;
 493         if (VFSATTR_IS_SUPPORTED(&attr, f_files))
 494                 fsap->f_files += attr.f_files;
 495         if (VFSATTR_IS_SUPPORTED(&attr, f_ffree))
 496                 fsap->f_ffree = attr.f_ffree;
 497
 498         VFSATTR_SET_SUPPORTED(fsap, f_bsize);
 499         VFSATTR_SET_SUPPORTED(fsap, f_blocks);
 500         VFSATTR_SET_SUPPORTED(fsap, f_bfree);
 501         VFSATTR_SET_SUPPORTED(fsap, f_bavail);
 502         VFSATTR_SET_SUPPORTED(fsap, f_bused);
 503         VFSATTR_SET_SUPPORTED(fsap, f_files);
 504         VFSATTR_SET_SUPPORTED(fsap, f_ffree);
 505
 506         return (0);
 507 }
 508
 509 /*
 510  * XXX - Assumes no data cached at union layer.
 511  */
 512 #define union_sync (int (*) (mount_t, int, ucred_t, vfs_context_t))nullop
 513
 514 #define union_fhtovp (int (*) (mount_t, int, unsigned char *, vnode_t *, vfs_context_t))eopnotsupp
 515 int union_init (struct vfsconf *);
 516 #define union_sysctl (int (*) (int *, u_int, user_addr_t, size_t *, user_addr_t, size_t, vfs_context_t))eopnotsupp
 517 #define union_vget (int (*) (mount_t, ino64_t, vnode_t *, vfs_context_t))eopnotsupp
 518 #define union_vptofh (int (*) (vnode_t, int *, unsigned char *, vfs_context_t))eopnotsupp
 519
 520 struct vfsops union_vfsops = {
 521         union_mount,
 522         union_start,
 523         union_unmount,
 524         union_root,
 525         NULL,                   /* quotactl */
 526         union_vfs_getattr,
 527         union_sync,
 528         union_vget,
 529         union_fhtovp,
 530         union_vptofh,
 531         union_init,
 532         union_sysctl
 533 };