1 /* $NetBSD: union_vfsops.c,v 1.35 2006/05/14 21:31:52 elad Exp $ */ 2 3 /* 4 * Copyright (c) 1994 The Regents of the University of California. 5 * All rights reserved. 6 * 7 * This code is derived from software donated to Berkeley by 8 * Jan-Simon Pendry. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)union_vfsops.c 8.20 (Berkeley) 5/20/95 35 */ 36 37 /* 38 * Copyright (c) 1994 Jan-Simon Pendry. 39 * All rights reserved. 40 * 41 * This code is derived from software donated to Berkeley by 42 * Jan-Simon Pendry. 43 * 44 * Redistribution and use in source and binary forms, with or without 45 * modification, are permitted provided that the following conditions 46 * are met: 47 * 1. Redistributions of source code must retain the above copyright 48 * notice, this list of conditions and the following disclaimer. 49 * 2. Redistributions in binary form must reproduce the above copyright 50 * notice, this list of conditions and the following disclaimer in the 51 * documentation and/or other materials provided with the distribution. 52 * 3. All advertising materials mentioning features or use of this software 53 * must display the following acknowledgement: 54 * This product includes software developed by the University of 55 * California, Berkeley and its contributors. 56 * 4. Neither the name of the University nor the names of its contributors 57 * may be used to endorse or promote products derived from this software 58 * without specific prior written permission. 59 * 60 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 61 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 62 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 63 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 64 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 65 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 66 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 67 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 68 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 69 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 70 * SUCH DAMAGE. 71 * 72 * @(#)union_vfsops.c 8.20 (Berkeley) 5/20/95 73 */ 74 75 /* 76 * Union Layer 77 */ 78 79 #include <sys/cdefs.h> 80 __KERNEL_RCSID(0, "$NetBSD: union_vfsops.c,v 1.35 2006/05/14 21:31:52 elad Exp $"); 81 82 #include <sys/param.h> 83 #include <sys/systm.h> 84 #include <sys/sysctl.h> 85 #include <sys/time.h> 86 #include <sys/proc.h> 87 #include <sys/vnode.h> 88 #include <sys/mount.h> 89 #include <sys/namei.h> 90 #include <sys/malloc.h> 91 #include <sys/filedesc.h> 92 #include <sys/queue.h> 93 #include <sys/stat.h> 94 #include <sys/kauth.h> 95 96 #include <fs/union/union.h> 97 98 int union_mount(struct mount *, const char *, void *, struct nameidata *, 99 struct lwp *); 100 int union_start(struct mount *, int, struct lwp *); 101 int union_unmount(struct mount *, int, struct lwp *); 102 int union_root(struct mount *, struct vnode **); 103 int union_quotactl(struct mount *, int, uid_t, void *, struct lwp *); 104 int union_statvfs(struct mount *, struct statvfs *, struct lwp *); 105 int union_sync(struct mount *, int, kauth_cred_t, struct lwp *); 106 int union_vget(struct mount *, ino_t, struct vnode **); 107 108 /* 109 * Mount union filesystem 110 */ 111 int 112 union_mount(mp, path, data, ndp, l) 113 struct mount *mp; 114 const char *path; 115 void *data; 116 struct nameidata *ndp; 117 struct lwp *l; 118 { 119 int error = 0; 120 struct union_args args; 121 struct vnode *lowerrootvp = NULLVP; 122 struct vnode *upperrootvp = NULLVP; 123 struct union_mount *um = 0; 124 const char *cp; 125 char *xp; 126 int len; 127 size_t size; 128 129 #ifdef UNION_DIAGNOSTIC 130 printf("union_mount(mp = %p)\n", mp); 131 #endif 132 133 if (mp->mnt_flag & MNT_GETARGS) { 134 um = MOUNTTOUNIONMOUNT(mp); 135 if (um == NULL) 136 return EIO; 137 args.target = NULL; 138 args.mntflags = um->um_op; 139 return copyout(&args, data, sizeof(args)); 140 } 141 /* 142 * Update is a no-op 143 */ 144 if (mp->mnt_flag & MNT_UPDATE) { 145 /* 146 * Need to provide. 147 * 1. a way to convert between rdonly and rdwr mounts. 148 * 2. support for nfs exports. 149 */ 150 error = EOPNOTSUPP; 151 goto bad; 152 } 153 154 /* 155 * Get argument 156 */ 157 error = copyin(data, &args, sizeof(struct union_args)); 158 if (error) 159 goto bad; 160 161 lowerrootvp = mp->mnt_vnodecovered; 162 VREF(lowerrootvp); 163 164 /* 165 * Find upper node. 166 */ 167 NDINIT(ndp, LOOKUP, FOLLOW, 168 UIO_USERSPACE, args.target, l); 169 170 if ((error = namei(ndp)) != 0) 171 goto bad; 172 173 upperrootvp = ndp->ni_vp; 174 175 if (upperrootvp->v_type != VDIR) { 176 error = EINVAL; 177 goto bad; 178 } 179 180 um = (struct union_mount *) malloc(sizeof(struct union_mount), 181 M_UFSMNT, M_WAITOK); /* XXX */ 182 183 /* 184 * Keep a held reference to the target vnodes. 185 * They are vrele'd in union_unmount. 186 * 187 * Depending on the _BELOW flag, the filesystems are 188 * viewed in a different order. In effect, this is the 189 * same as providing a mount under option to the mount syscall. 190 */ 191 192 um->um_op = args.mntflags & UNMNT_OPMASK; 193 switch (um->um_op) { 194 case UNMNT_ABOVE: 195 um->um_lowervp = lowerrootvp; 196 um->um_uppervp = upperrootvp; 197 break; 198 199 case UNMNT_BELOW: 200 um->um_lowervp = upperrootvp; 201 um->um_uppervp = lowerrootvp; 202 break; 203 204 case UNMNT_REPLACE: 205 vrele(lowerrootvp); 206 lowerrootvp = NULLVP; 207 um->um_uppervp = upperrootvp; 208 um->um_lowervp = lowerrootvp; 209 break; 210 211 default: 212 error = EINVAL; 213 goto bad; 214 } 215 216 /* 217 * Unless the mount is readonly, ensure that the top layer 218 * supports whiteout operations 219 */ 220 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 221 error = VOP_WHITEOUT(um->um_uppervp, (struct componentname *) 0, LOOKUP); 222 if (error) 223 goto bad; 224 } 225 226 um->um_cred = l->l_proc->p_cred; 227 kauth_cred_hold(um->um_cred); 228 um->um_cmode = UN_DIRMODE &~ l->l_proc->p_cwdi->cwdi_cmask; 229 230 /* 231 * Depending on what you think the MNT_LOCAL flag might mean, 232 * you may want the && to be || on the conditional below. 233 * At the moment it has been defined that the filesystem is 234 * only local if it is all local, ie the MNT_LOCAL flag implies 235 * that the entire namespace is local. If you think the MNT_LOCAL 236 * flag implies that some of the files might be stored locally 237 * then you will want to change the conditional. 238 */ 239 if (um->um_op == UNMNT_ABOVE) { 240 if (((um->um_lowervp == NULLVP) || 241 (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) && 242 (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL)) 243 mp->mnt_flag |= MNT_LOCAL; 244 } 245 246 /* 247 * Copy in the upper layer's RDONLY flag. This is for the benefit 248 * of lookup() which explicitly checks the flag, rather than asking 249 * the filesystem for it's own opinion. This means, that an update 250 * mount of the underlying filesystem to go from rdonly to rdwr 251 * will leave the unioned view as read-only. 252 */ 253 mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY); 254 255 mp->mnt_data = um; 256 mp->mnt_leaf = um->um_uppervp->v_mount->mnt_leaf; 257 vfs_getnewfsid(mp); 258 259 error = set_statvfs_info( path, UIO_USERSPACE, NULL, UIO_USERSPACE, 260 mp, l); 261 if (error) 262 goto bad; 263 264 switch (um->um_op) { 265 case UNMNT_ABOVE: 266 cp = "<above>:"; 267 break; 268 case UNMNT_BELOW: 269 cp = "<below>:"; 270 break; 271 case UNMNT_REPLACE: 272 cp = ""; 273 break; 274 default: 275 cp = "<invalid>:"; 276 #ifdef DIAGNOSTIC 277 panic("union_mount: bad um_op"); 278 #endif 279 break; 280 } 281 len = strlen(cp); 282 memcpy(mp->mnt_stat.f_mntfromname, cp, len); 283 284 xp = mp->mnt_stat.f_mntfromname + len; 285 len = MNAMELEN - len; 286 287 (void) copyinstr(args.target, xp, len - 1, &size); 288 memset(xp + size, 0, len - size); 289 290 #ifdef UNION_DIAGNOSTIC 291 printf("union_mount: from %s, on %s\n", 292 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname); 293 #endif 294 295 /* Setup the readdir hook if it's not set already */ 296 if (!vn_union_readdir_hook) 297 vn_union_readdir_hook = union_readdirhook; 298 299 return (0); 300 301 bad: 302 if (um) 303 free(um, M_UFSMNT); 304 if (upperrootvp) 305 vrele(upperrootvp); 306 if (lowerrootvp) 307 vrele(lowerrootvp); 308 return (error); 309 } 310 311 /* 312 * VFS start. Nothing needed here - the start routine 313 * on the underlying filesystem(s) will have been called 314 * when that filesystem was mounted. 315 */ 316 /*ARGSUSED*/ 317 int 318 union_start(mp, flags, l) 319 struct mount *mp; 320 int flags; 321 struct lwp *l; 322 { 323 324 return (0); 325 } 326 327 /* 328 * Free reference to union layer 329 */ 330 int 331 union_unmount(mp, mntflags, l) 332 struct mount *mp; 333 int mntflags; 334 struct lwp *l; 335 { 336 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 337 int freeing; 338 int error; 339 340 #ifdef UNION_DIAGNOSTIC 341 printf("union_unmount(mp = %p)\n", mp); 342 #endif 343 344 /* 345 * Keep flushing vnodes from the mount list. 346 * This is needed because of the un_pvp held 347 * reference to the parent vnode. 348 * If more vnodes have been freed on a given pass, 349 * the try again. The loop will iterate at most 350 * (d) times, where (d) is the maximum tree depth 351 * in the filesystem. 352 */ 353 for (freeing = 0; (error = vflush(mp, NULL, 0)) != 0;) { 354 struct vnode *vp; 355 int n; 356 357 /* count #vnodes held on mount list */ 358 for (n = 0, vp = mp->mnt_vnodelist.lh_first; 359 vp != NULLVP; 360 vp = vp->v_mntvnodes.le_next) 361 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 /* 372 * Ok, now that we've tried doing it gently, get out the hammer. 373 */ 374 375 if (mntflags & MNT_FORCE) 376 error = vflush(mp, NULL, FORCECLOSE); 377 378 if (error) 379 return error; 380 381 /* 382 * Discard references to upper and lower target vnodes. 383 */ 384 if (um->um_lowervp) 385 vrele(um->um_lowervp); 386 vrele(um->um_uppervp); 387 kauth_cred_free(um->um_cred); 388 /* 389 * Finally, throw away the union_mount structure 390 */ 391 free(mp->mnt_data, M_UFSMNT); /* XXX */ 392 mp->mnt_data = 0; 393 return (0); 394 } 395 396 int 397 union_root(mp, vpp) 398 struct mount *mp; 399 struct vnode **vpp; 400 { 401 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 402 int error; 403 int loselock; 404 405 /* 406 * Return locked reference to root. 407 */ 408 VREF(um->um_uppervp); 409 if ((um->um_op == UNMNT_BELOW) && 410 VOP_ISLOCKED(um->um_uppervp)) { 411 loselock = 1; 412 } else { 413 vn_lock(um->um_uppervp, LK_EXCLUSIVE | LK_RETRY); 414 loselock = 0; 415 } 416 if (um->um_lowervp) 417 VREF(um->um_lowervp); 418 error = union_allocvp(vpp, mp, 419 (struct vnode *) 0, 420 (struct vnode *) 0, 421 (struct componentname *) 0, 422 um->um_uppervp, 423 um->um_lowervp, 424 1); 425 426 if (error) { 427 if (!loselock) 428 VOP_UNLOCK(um->um_uppervp, 0); 429 vrele(um->um_uppervp); 430 if (um->um_lowervp) 431 vrele(um->um_lowervp); 432 } else { 433 if (loselock) 434 VTOUNION(*vpp)->un_flags &= ~UN_ULOCK; 435 } 436 437 return (error); 438 } 439 440 /*ARGSUSED*/ 441 int 442 union_quotactl(mp, cmd, uid, arg, l) 443 struct mount *mp; 444 int cmd; 445 uid_t uid; 446 void *arg; 447 struct lwp *l; 448 { 449 450 return (EOPNOTSUPP); 451 } 452 453 int 454 union_statvfs(mp, sbp, l) 455 struct mount *mp; 456 struct statvfs *sbp; 457 struct lwp *l; 458 { 459 int error; 460 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 461 struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO); 462 unsigned long lbsize; 463 464 #ifdef UNION_DIAGNOSTIC 465 printf("union_statvfs(mp = %p, lvp = %p, uvp = %p)\n", mp, 466 um->um_lowervp, um->um_uppervp); 467 #endif 468 469 if (um->um_lowervp) { 470 error = VFS_STATVFS(um->um_lowervp->v_mount, sbuf, l); 471 if (error) 472 goto done; 473 } 474 475 /* now copy across the "interesting" information and fake the rest */ 476 lbsize = sbuf->f_bsize; 477 sbp->f_blocks = sbuf->f_blocks - sbuf->f_bfree; 478 sbp->f_files = sbuf->f_files - sbuf->f_ffree; 479 480 error = VFS_STATVFS(um->um_uppervp->v_mount, sbuf, l); 481 if (error) 482 goto done; 483 484 sbp->f_flag = sbuf->f_flag; 485 sbp->f_bsize = sbuf->f_bsize; 486 sbp->f_frsize = sbuf->f_frsize; 487 sbp->f_iosize = sbuf->f_iosize; 488 489 /* 490 * The "total" fields count total resources in all layers, 491 * the "free" fields count only those resources which are 492 * free in the upper layer (since only the upper layer 493 * is writable). 494 */ 495 496 if (sbuf->f_bsize != lbsize) 497 sbp->f_blocks = sbp->f_blocks * lbsize / sbuf->f_bsize; 498 sbp->f_blocks += sbuf->f_blocks; 499 sbp->f_bfree = sbuf->f_bfree; 500 sbp->f_bavail = sbuf->f_bavail; 501 sbp->f_bresvd = sbuf->f_bresvd; 502 sbp->f_files += sbuf->f_files; 503 sbp->f_ffree = sbuf->f_ffree; 504 sbp->f_favail = sbuf->f_favail; 505 sbp->f_fresvd = sbuf->f_fresvd; 506 507 copy_statvfs_info(sbp, mp); 508 done: 509 free(sbuf, M_TEMP); 510 return error; 511 } 512 513 /*ARGSUSED*/ 514 int 515 union_sync(mp, waitfor, cred, l) 516 struct mount *mp; 517 int waitfor; 518 kauth_cred_t cred; 519 struct lwp *l; 520 { 521 522 /* 523 * XXX - Assumes no data cached at union layer. 524 */ 525 return (0); 526 } 527 528 /*ARGSUSED*/ 529 int 530 union_vget(mp, ino, vpp) 531 struct mount *mp; 532 ino_t ino; 533 struct vnode **vpp; 534 { 535 536 return (EOPNOTSUPP); 537 } 538 539 SYSCTL_SETUP(sysctl_vfs_union_setup, "sysctl vfs.union subtree setup") 540 { 541 542 sysctl_createv(clog, 0, NULL, NULL, 543 CTLFLAG_PERMANENT, 544 CTLTYPE_NODE, "vfs", NULL, 545 NULL, 0, NULL, 0, 546 CTL_VFS, CTL_EOL); 547 sysctl_createv(clog, 0, NULL, NULL, 548 CTLFLAG_PERMANENT, 549 CTLTYPE_NODE, "union", 550 SYSCTL_DESCR("Union file system"), 551 NULL, 0, NULL, 0, 552 CTL_VFS, 15, CTL_EOL); 553 /* 554 * XXX the "15" above could be dynamic, thereby eliminating 555 * one more instance of the "number to vfs" mapping problem, 556 * but "15" is the order as taken from sys/mount.h 557 */ 558 } 559 560 extern const struct vnodeopv_desc union_vnodeop_opv_desc; 561 562 const struct vnodeopv_desc * const union_vnodeopv_descs[] = { 563 &union_vnodeop_opv_desc, 564 NULL, 565 }; 566 567 struct vfsops union_vfsops = { 568 MOUNT_UNION, 569 union_mount, 570 union_start, 571 union_unmount, 572 union_root, 573 union_quotactl, 574 union_statvfs, 575 union_sync, 576 union_vget, 577 NULL, /* vfs_fhtovp */ 578 NULL, /* vfs_vptofh */ 579 union_init, 580 NULL, /* vfs_reinit */ 581 union_done, 582 NULL, /* vfs_mountroot */ 583 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 584 vfs_stdextattrctl, 585 union_vnodeopv_descs, 586 }; 587 VFS_ATTACH(union_vfsops); 588