bsd/miscfs/union/union_vfsops.c

   1 /*
   2  * Copyright (c) 2000-2006 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28 /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
  29 /*
  30  * Copyright (c) 1994, 1995 The Regents of the University of California.
  31  * Copyright (c) 1994, 1995 Jan-Simon Pendry.
  32  * All rights reserved.
  33  *
  34  * This code is derived from software donated to Berkeley by
  35  * Jan-Simon Pendry.
  36  *
  37  * Redistribution and use in source and binary forms, with or without
  38  * modification, are permitted provided that the following conditions
  39  * are met:
  40  * 1. Redistributions of source code must retain the above copyright
  41  *    notice, this list of conditions and the following disclaimer.
  42  * 2. Redistributions in binary form must reproduce the above copyright
  43  *    notice, this list of conditions and the following disclaimer in the
  44  *    documentation and/or other materials provided with the distribution.
  45  * 3. All advertising materials mentioning features or use of this software
  46  *    must display the following acknowledgement:
  47  *      This product includes software developed by the University of
  48  *      California, Berkeley and its contributors.
  49  * 4. Neither the name of the University nor the names of its contributors
  50  *    may be used to endorse or promote products derived from this software
  51  *    without specific prior written permission.
  52  *
  53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  63  * SUCH DAMAGE.
  64  *
  65  *      @(#)union_vfsops.c      8.20 (Berkeley) 5/20/95
  66  */
  67
  68 /*
  69  * Union Layer
  70  */
  71
  72 #include <sys/param.h>
  73 #include <sys/systm.h>
  74 #include <sys/time.h>
  75 #include <sys/types.h>
  76 #include <sys/proc_internal.h>
  77 #include <sys/kauth.h>
  78 #include <sys/vnode_internal.h>
  79 #include <sys/mount_internal.h>
  80 #include <sys/namei.h>
  81 #include <sys/malloc.h>
  82 #include <sys/filedesc.h>
  83 #include <sys/queue.h>
  84 #include <miscfs/union/union.h>
  85
  86 static  int union_itercallback(vnode_t, void *);
  87 static int union_root(mount_t, vnode_t *, vfs_context_t);
  88
  89 /*
  90  * Mount union filesystem
  91  */
  92 static int
  93 union_mount(mount_t mp, __unused vnode_t devvp, user_addr_t data, vfs_context_t context)
  94 {
  95         proc_t p = vfs_context_proc(context);
  96         int error = 0;
  97         struct user_union_args args;
  98         struct vnode *lowerrootvp = NULLVP;
  99         struct vnode *upperrootvp = NULLVP;
 100         struct union_mount *um = NULL;
 101         kauth_cred_t cred = NOCRED;
 102         const char *cp = NULL;
 103         char *vcp;
 104         int len;
 105         u_int size;
 106         struct nameidata nd;
 107
 108 #ifdef UNION_DIAGNOSTIC
 109         printf("union_mount(mp = %x)\n", mp);
 110 #endif
 111
 112         /*
 113          * Update is a no-op
 114          */
 115         if (mp->mnt_flag & MNT_UPDATE) {
 116                 /*
 117                  * Need to provide.
 118                  * 1. a way to convert between rdonly and rdwr mounts.
 119                  * 2. support for nfs exports.
 120                  */
 121                 error = ENOTSUP;
 122                 goto bad;
 123         }
 124
 125         /*
 126          * Get argument
 127          */
 128         if (vfs_context_is64bit(context)) {
 129                 error = copyin(data, (caddr_t)&args, sizeof(args));
 130         }
 131         else {
 132                 struct union_args temp;
 133                 error = copyin(data, (caddr_t)&temp, sizeof (temp));
 134                 args.target = CAST_USER_ADDR_T(temp.target);
 135                 args.mntflags = temp.mntflags;
 136         }
 137         if (error)
 138                 goto bad;
 139
 140         lowerrootvp = mp->mnt_vnodecovered;
 141         vnode_get(lowerrootvp);
 142
 143         /*
 144          * Find upper node.
 145          */
 146         NDINIT(&nd, LOOKUP, FOLLOW|WANTPARENT,
 147                (IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32),
 148                args.target, context);
 149
 150         if ((error = namei(&nd)))
 151                 goto bad;
 152
 153         nameidone(&nd);
 154         upperrootvp = nd.ni_vp;
 155         vnode_put(nd.ni_dvp);
 156         nd.ni_dvp = NULL;
 157
 158         if (upperrootvp->v_type != VDIR) {
 159                 error = EINVAL;
 160                 goto bad;
 161         }
 162
 163         MALLOC(um, struct union_mount *, sizeof(struct union_mount),
 164                                 M_UFSMNT, M_WAITOK);
 165
 166         /*
 167          * Keep a held reference to the target vnodes.
 168          * They are vnode_put'd in union_unmount.
 169          *
 170          * Depending on the _BELOW flag, the filesystems are
 171          * viewed in a different order.  In effect, this is the
 172          * same as providing a mount under option to the mount syscall.
 173          */
 174
 175         um->um_op = args.mntflags & UNMNT_OPMASK;
 176         switch (um->um_op) {
 177         case UNMNT_ABOVE:
 178                 um->um_lowervp = lowerrootvp;
 179                 um->um_uppervp = upperrootvp;
 180                 break;
 181
 182         case UNMNT_BELOW:
 183                 um->um_lowervp = upperrootvp;
 184                 um->um_uppervp = lowerrootvp;
 185                 break;
 186
 187         case UNMNT_REPLACE:
 188                 vnode_put(lowerrootvp);
 189                 lowerrootvp = NULLVP;
 190                 um->um_uppervp = upperrootvp;
 191                 um->um_lowervp = lowerrootvp;
 192                 break;
 193
 194 #ifdef FAULTFS
 195         case UNMNT_FAULTIN:
 196                 um->um_lowervp = upperrootvp;
 197                 um->um_uppervp = lowerrootvp;
 198                 break;
 199 #endif
 200
 201         default:
 202                 error = EINVAL;
 203                 goto bad;
 204         }
 205
 206         if (um->um_lowervp != NULLVP)
 207                 um->um_lowervid = vnode_vid(um->um_lowervp);
 208         if (um->um_uppervp != NULLVP)
 209                 um->um_uppervid = vnode_vid(um->um_uppervp);
 210         /*
 211          * Unless the mount is readonly, ensure that the top layer
 212          * supports whiteout operations
 213          */
 214 #ifdef FAULTFS
 215         if ((um->um_op != UNMNT_FAULTIN) && (mp->mnt_flag & MNT_RDONLY) == 0)
 216 #else
 217         if ((mp->mnt_flag & MNT_RDONLY) == 0)
 218 #endif
 219         {
 220                 error = VNOP_WHITEOUT(um->um_uppervp, (struct componentname *) 0,
 221                                       LOOKUP, context);
 222                 if (error)
 223                         goto bad;
 224         }
 225
 226         um->um_cred = kauth_cred_get_with_ref();
 227         um->um_cmode = UN_DIRMODE &~ p->p_fd->fd_cmask;
 228
 229         /*
 230          * Depending on what you think the MNT_LOCAL flag might mean,
 231          * you may want the && to be || on the conditional below.
 232          * At the moment it has been defined that the filesystem is
 233          * only local if it is all local, ie the MNT_LOCAL flag implies
 234          * that the entire namespace is local.  If you think the MNT_LOCAL
 235          * flag implies that some of the files might be stored locally
 236          * then you will want to change the conditional.
 237          */
 238         if (um->um_op == UNMNT_ABOVE) {
 239                 if (((um->um_lowervp == NULLVP) ||
 240                      (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) &&
 241                     (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
 242                         mp->mnt_flag |= MNT_LOCAL;
 243         }
 244
 245         /*
 246          * Copy in the upper layer's RDONLY flag.  This is for the benefit
 247          * of lookup() which explicitly checks the flag, rather than asking
 248          * the filesystem for it's own opinion.  This means, that an update
 249          * mount of the underlying filesystem to go from rdonly to rdwr
 250          * will leave the unioned view as read-only.
 251          */
 252         mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY);
 253
 254         mp->mnt_data = (qaddr_t) um;
 255         vfs_getnewfsid(mp);
 256
 257
 258         switch (um->um_op) {
 259         case UNMNT_ABOVE:
 260                 cp = "<above>:";
 261                 break;
 262         case UNMNT_BELOW:
 263                 cp = "<below>:";
 264                 break;
 265         case UNMNT_REPLACE:
 266                 cp = "";
 267                 break;
 268 #ifdef FAULTFS
 269         case UNMNT_FAULTIN:
 270                 cp = "/FaultingFS/";
 271                 break;
 272 #endif
 273         }
 274         len = strlen(cp);
 275         bcopy(cp, mp->mnt_vfsstat.f_mntfromname, len);
 276
 277         vcp = mp->mnt_vfsstat.f_mntfromname + len;
 278         len = MNAMELEN - len;
 279
 280         (void) copyinstr(args.target, vcp, len - 1, (size_t *)&size);
 281         bzero(vcp + size, len - size);
 282
 283         /* mark the filesystem thred safe */
 284          mp->mnt_vtable->vfc_threadsafe = TRUE;
 285
 286 #ifdef UNION_DIAGNOSTIC
 287         printf("union_mount: from %s, on %s\n",
 288                 mp->mnt_vfsstat.f_mntfromname, mp->mnt_vfsstat.f_mntonname);
 289 #endif
 290         return (0);
 291
 292 bad:
 293         if (um)
 294                 _FREE(um, M_UFSMNT);
 295         if (IS_VALID_CRED(cred))
 296                 kauth_cred_unref(&cred);
 297         if (upperrootvp)
 298                 vnode_put(upperrootvp);
 299         if (lowerrootvp)
 300                 vnode_put(lowerrootvp);
 301         return (error);
 302 }
 303
 304 /*
 305  * VFS start.  Nothing needed here - the start routine
 306  * on the underlying filesystem(s) will have been called
 307  * when that filesystem was mounted.
 308  */
 309 static int
 310 union_start(__unused struct mount *mp, __unused int flags, __unused vfs_context_t context)
 311 {
 312
 313         return (0);
 314 }
 315
 316 static int
 317 union_itercallback(__unused vnode_t vp, void *args)
 318 {
 319         int  num = *(int *)args;
 320
 321         *(int *)args = num + 1;
 322         return(VNODE_RETURNED);
 323 }
 324
 325
 326
 327 /*
 328  * Free reference to union layer
 329  */
 330 static int
 331 union_unmount(mount_t mp, int mntflags, vfs_context_t context)
 332 {
 333         struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
 334         struct vnode *um_rootvp;
 335         int error;
 336         int freeing;
 337         int flags = 0;
 338
 339 #ifdef UNION_DIAGNOSTIC
 340         printf("union_unmount(mp = %x)\n", mp);
 341 #endif
 342
 343         if (mntflags & MNT_FORCE)
 344                 flags |= FORCECLOSE;
 345
 346         if ((error = union_root(mp, &um_rootvp, context)))
 347                 return (error);
 348
 349         /*
 350          * Keep flushing vnodes from the mount list.
 351          * This is needed because of the un_pvp held
 352          * reference to the parent vnode.
 353          * If more vnodes have been freed on a given pass,
 354          * the try again.  The loop will iterate at most
 355          * (d) times, where (d) is the maximum tree depth
 356          * in the filesystem.
 357          */
 358         for (freeing = 0; vflush(mp, um_rootvp, flags) != 0;) {
 359                 int n = 0;
 360
 361                 vnode_iterate(mp, VNODE_NOLOCK_INTERNAL, union_itercallback, &n);
 362
 363                 /* if this is unchanged then stop */
 364                 if (n == freeing)
 365                         break;
 366
 367                 /* otherwise try once more time */
 368                 freeing = n;
 369         }
 370
 371         /* At this point the root vnode should have a single reference */
 372         if (vnode_isinuse(um_rootvp, 0)) {
 373                 vnode_put(um_rootvp);
 374                 return (EBUSY);
 375         }
 376
 377 #ifdef UNION_DIAGNOSTIC
 378         vprint("union root", um_rootvp);
 379 #endif
 380         /*
 381          * Discard references to upper and lower target vnodes.
 382          */
 383         if (um->um_lowervp)
 384                 vnode_put(um->um_lowervp);
 385         vnode_put(um->um_uppervp);
 386         if (IS_VALID_CRED(um->um_cred)) {
 387                 kauth_cred_unref(&um->um_cred);
 388         }
 389         /*
 390          * Release reference on underlying root vnode
 391          */
 392         vnode_put(um_rootvp);
 393         /*
 394          * And blow it away for future re-use
 395          */
 396         vnode_reclaim(um_rootvp);
 397         /*
 398          * Finally, throw away the union_mount structure
 399          */
 400         _FREE(mp->mnt_data, M_UFSMNT);  /* XXX */
 401         mp->mnt_data = NULL;
 402         return (0);
 403 }
 404
 405 static int
 406 union_root(mount_t mp, vnode_t *vpp, __unused vfs_context_t context)
 407 {
 408         struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
 409         int error;
 410
 411         /*
 412          * Return locked reference to root.
 413          */
 414         vnode_get(um->um_uppervp);
 415         if (um->um_lowervp)
 416                 vnode_get(um->um_lowervp);
 417
 418         union_lock();
 419         error = union_allocvp(vpp, mp,
 420                               (struct vnode *) 0,
 421                               (struct vnode *) 0,
 422                               (struct componentname *) 0,
 423                               um->um_uppervp,
 424                               um->um_lowervp,
 425                               1);
 426         union_unlock();
 427
 428         if (error) {
 429                 vnode_put(um->um_uppervp);
 430                 if (um->um_lowervp)
 431                         vnode_put(um->um_lowervp);
 432         }
 433
 434         return (error);
 435 }
 436
 437 static int
 438 union_vfs_getattr(mount_t mp, struct vfs_attr *fsap, vfs_context_t context)
 439 {
 440         int error;
 441         struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
 442         struct vfs_attr attr;
 443         uint32_t lbsize = 0;
 444
 445 #ifdef UNION_DIAGNOSTIC
 446         printf("union_vfs_getattr(mp = %x, lvp = %x, uvp = %x)\n", mp,
 447                         um->um_lowervp,
 448                         um->um_uppervp);
 449 #endif
 450
 451         /* Get values from lower file system (if any) */
 452         if (um->um_lowervp) {
 453                 VFSATTR_INIT(&attr);
 454                 VFSATTR_WANTED(&attr, f_bsize);
 455                 VFSATTR_WANTED(&attr, f_blocks);
 456                 VFSATTR_WANTED(&attr, f_bused);
 457                 VFSATTR_WANTED(&attr, f_files);
 458                 error = vfs_getattr(um->um_lowervp->v_mount, &attr, context);
 459                 if (error)
 460                         return (error);
 461
 462                 /* now copy across the "interesting" information and fake the rest */
 463                 if (VFSATTR_IS_SUPPORTED(&attr, f_bsize))
 464                         lbsize = attr.f_bsize;
 465                 else
 466                         lbsize = um->um_lowervp->v_mount->mnt_devblocksize;
 467                 fsap->f_blocks = VFSATTR_IS_SUPPORTED(&attr, f_blocks) ? attr.f_blocks : 0;
 468                 fsap->f_bused  = VFSATTR_IS_SUPPORTED(&attr, f_bused)  ? attr.f_bused  : 0;
 469                 fsap->f_files  = VFSATTR_IS_SUPPORTED(&attr, f_files)  ? attr.f_files  : 0;
 470         } else {
 471                 fsap->f_blocks = 0;
 472                 fsap->f_bused = 0;
 473                 fsap->f_files = 0;
 474         }
 475
 476         VFSATTR_INIT(&attr);
 477         VFSATTR_WANTED(&attr, f_bsize);
 478         VFSATTR_WANTED(&attr, f_blocks);
 479         VFSATTR_WANTED(&attr, f_bfree);
 480         VFSATTR_WANTED(&attr, f_bavail);
 481         VFSATTR_WANTED(&attr, f_files);
 482         VFSATTR_WANTED(&attr, f_ffree);
 483         error = vfs_getattr(um->um_uppervp->v_mount, &attr, context);
 484         if (error)
 485                 return (error);
 486
 487         if (VFSATTR_IS_SUPPORTED(&attr, f_bsize)) {
 488                 fsap->f_bsize = attr.f_bsize;
 489                 VFSATTR_SET_SUPPORTED(fsap, f_bsize);
 490         }
 491         if (VFSATTR_IS_SUPPORTED(&attr, f_iosize)) {
 492                 fsap->f_iosize = attr.f_iosize;
 493                 VFSATTR_SET_SUPPORTED(fsap, f_iosize);
 494         }
 495
 496         /*
 497          * if the lower and upper blocksizes differ, then frig the
 498          * block counts so that the sizes reported by df make some
 499          * kind of sense.  none of this makes sense though.
 500          */
 501         if (VFSATTR_IS_SUPPORTED(&attr, f_bsize))
 502                 fsap->f_bsize = attr.f_bsize;
 503         else
 504                 fsap->f_bsize =  um->um_uppervp->v_mount->mnt_devblocksize;
 505         VFSATTR_RETURN(fsap, f_bsize, attr.f_bsize);
 506         if (fsap->f_bsize != lbsize)
 507                 fsap->f_blocks = fsap->f_blocks * lbsize / attr.f_bsize;
 508
 509         /*
 510          * The "total" fields count total resources in all layers,
 511          * the "free" fields count only those resources which are
 512          * free in the upper layer (since only the upper layer
 513          * is writeable).
 514          */
 515         if (VFSATTR_IS_SUPPORTED(&attr, f_blocks))
 516                 fsap->f_blocks += attr.f_blocks;
 517         if (VFSATTR_IS_SUPPORTED(&attr, f_bfree))
 518                 fsap->f_bfree = attr.f_bfree;
 519         if (VFSATTR_IS_SUPPORTED(&attr, f_bavail))
 520                 fsap->f_bavail = attr.f_bavail;
 521         if (VFSATTR_IS_SUPPORTED(&attr, f_bused))
 522                 fsap->f_bused += attr.f_bused;
 523         if (VFSATTR_IS_SUPPORTED(&attr, f_files))
 524                 fsap->f_files += attr.f_files;
 525         if (VFSATTR_IS_SUPPORTED(&attr, f_ffree))
 526                 fsap->f_ffree = attr.f_ffree;
 527
 528         VFSATTR_SET_SUPPORTED(fsap, f_bsize);
 529         VFSATTR_SET_SUPPORTED(fsap, f_blocks);
 530         VFSATTR_SET_SUPPORTED(fsap, f_bfree);
 531         VFSATTR_SET_SUPPORTED(fsap, f_bavail);
 532         VFSATTR_SET_SUPPORTED(fsap, f_bused);
 533         VFSATTR_SET_SUPPORTED(fsap, f_files);
 534         VFSATTR_SET_SUPPORTED(fsap, f_ffree);
 535
 536         return (0);
 537 }
 538
 539 /*
 540  * XXX - Assumes no data cached at union layer.
 541  */
 542 #define union_sync (int (*) (mount_t, int, vfs_context_t))nullop
 543
 544 #define union_fhtovp (int (*) (mount_t, int, unsigned char *, vnode_t *, vfs_context_t))eopnotsupp
 545 #define union_sysctl (int (*) (int *, u_int, user_addr_t, size_t *, user_addr_t, size_t, vfs_context_t))eopnotsupp
 546 #define union_vget (int (*) (mount_t, ino64_t, vnode_t *, vfs_context_t))eopnotsupp
 547 #define union_vptofh (int (*) (vnode_t, int *, unsigned char *, vfs_context_t))eopnotsupp
 548
 549 struct vfsops union_vfsops = {
 550         union_mount,
 551         union_start,
 552         union_unmount,
 553         union_root,
 554         NULL,                   /* quotactl */
 555         union_vfs_getattr,
 556         union_sync,
 557         union_vget,
 558         union_fhtovp,
 559         union_vptofh,
 560         union_init,
 561         union_sysctl,
 562         NULL,
 563         {NULL}
 564 };
 565
 566