1 /* $NetBSD: union_vfsops.c,v 1.38 2006/10/12 01:32:14 christos 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 1.38 2006/10/12 01:32:14 christos 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_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(struct mount *mp __unused, int flags __unused, 319 struct lwp *l __unused) 320 { 321 322 return (0); 323 } 324 325 /* 326 * Free reference to union layer 327 */ 328 int 329 union_unmount(struct mount *mp, int mntflags, struct lwp *l __unused) 330 { 331 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 332 int freeing; 333 int error; 334 335 #ifdef UNION_DIAGNOSTIC 336 printf("union_unmount(mp = %p)\n", mp); 337 #endif 338 339 /* 340 * Keep flushing vnodes from the mount list. 341 * This is needed because of the un_pvp held 342 * reference to the parent vnode. 343 * If more vnodes have been freed on a given pass, 344 * the try again. The loop will iterate at most 345 * (d) times, where (d) is the maximum tree depth 346 * in the filesystem. 347 */ 348 for (freeing = 0; (error = vflush(mp, NULL, 0)) != 0;) { 349 struct vnode *vp; 350 int n; 351 352 /* count #vnodes held on mount list */ 353 for (n = 0, vp = mp->mnt_vnodelist.lh_first; 354 vp != NULLVP; 355 vp = vp->v_mntvnodes.le_next) 356 n++; 357 358 /* if this is unchanged then stop */ 359 if (n == freeing) 360 break; 361 362 /* otherwise try once more time */ 363 freeing = n; 364 } 365 366 /* 367 * Ok, now that we've tried doing it gently, get out the hammer. 368 */ 369 370 if (mntflags & MNT_FORCE) 371 error = vflush(mp, NULL, FORCECLOSE); 372 373 if (error) 374 return error; 375 376 /* 377 * Discard references to upper and lower target vnodes. 378 */ 379 if (um->um_lowervp) 380 vrele(um->um_lowervp); 381 vrele(um->um_uppervp); 382 kauth_cred_free(um->um_cred); 383 /* 384 * Finally, throw away the union_mount structure 385 */ 386 free(mp->mnt_data, M_UFSMNT); /* XXX */ 387 mp->mnt_data = 0; 388 return (0); 389 } 390 391 int 392 union_root(mp, vpp) 393 struct mount *mp; 394 struct vnode **vpp; 395 { 396 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 397 int error; 398 int loselock; 399 400 /* 401 * Return locked reference to root. 402 */ 403 VREF(um->um_uppervp); 404 if ((um->um_op == UNMNT_BELOW) && 405 VOP_ISLOCKED(um->um_uppervp)) { 406 loselock = 1; 407 } else { 408 vn_lock(um->um_uppervp, LK_EXCLUSIVE | LK_RETRY); 409 loselock = 0; 410 } 411 if (um->um_lowervp) 412 VREF(um->um_lowervp); 413 error = union_allocvp(vpp, mp, 414 (struct vnode *) 0, 415 (struct vnode *) 0, 416 (struct componentname *) 0, 417 um->um_uppervp, 418 um->um_lowervp, 419 1); 420 421 if (error) { 422 if (!loselock) 423 VOP_UNLOCK(um->um_uppervp, 0); 424 vrele(um->um_uppervp); 425 if (um->um_lowervp) 426 vrele(um->um_lowervp); 427 } else { 428 if (loselock) 429 VTOUNION(*vpp)->un_flags &= ~UN_ULOCK; 430 } 431 432 return (error); 433 } 434 435 /*ARGSUSED*/ 436 int 437 union_quotactl(struct mount *mp __unused, int cmd __unused, uid_t uid __unused, 438 void *arg __unused, struct lwp *l __unused) 439 { 440 441 return (EOPNOTSUPP); 442 } 443 444 int 445 union_statvfs(mp, sbp, l) 446 struct mount *mp; 447 struct statvfs *sbp; 448 struct lwp *l; 449 { 450 int error; 451 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 452 struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO); 453 unsigned long lbsize; 454 455 #ifdef UNION_DIAGNOSTIC 456 printf("union_statvfs(mp = %p, lvp = %p, uvp = %p)\n", mp, 457 um->um_lowervp, um->um_uppervp); 458 #endif 459 460 if (um->um_lowervp) { 461 error = VFS_STATVFS(um->um_lowervp->v_mount, sbuf, l); 462 if (error) 463 goto done; 464 } 465 466 /* now copy across the "interesting" information and fake the rest */ 467 lbsize = sbuf->f_bsize; 468 sbp->f_blocks = sbuf->f_blocks - sbuf->f_bfree; 469 sbp->f_files = sbuf->f_files - sbuf->f_ffree; 470 471 error = VFS_STATVFS(um->um_uppervp->v_mount, sbuf, l); 472 if (error) 473 goto done; 474 475 sbp->f_flag = sbuf->f_flag; 476 sbp->f_bsize = sbuf->f_bsize; 477 sbp->f_frsize = sbuf->f_frsize; 478 sbp->f_iosize = sbuf->f_iosize; 479 480 /* 481 * The "total" fields count total resources in all layers, 482 * the "free" fields count only those resources which are 483 * free in the upper layer (since only the upper layer 484 * is writable). 485 */ 486 487 if (sbuf->f_bsize != lbsize) 488 sbp->f_blocks = sbp->f_blocks * lbsize / sbuf->f_bsize; 489 sbp->f_blocks += sbuf->f_blocks; 490 sbp->f_bfree = sbuf->f_bfree; 491 sbp->f_bavail = sbuf->f_bavail; 492 sbp->f_bresvd = sbuf->f_bresvd; 493 sbp->f_files += sbuf->f_files; 494 sbp->f_ffree = sbuf->f_ffree; 495 sbp->f_favail = sbuf->f_favail; 496 sbp->f_fresvd = sbuf->f_fresvd; 497 498 copy_statvfs_info(sbp, mp); 499 done: 500 free(sbuf, M_TEMP); 501 return error; 502 } 503 504 /*ARGSUSED*/ 505 int 506 union_sync(struct mount *mp __unused, int waitfor __unused, 507 kauth_cred_t cred __unused, struct lwp *l __unused) 508 { 509 510 /* 511 * XXX - Assumes no data cached at union layer. 512 */ 513 return (0); 514 } 515 516 /*ARGSUSED*/ 517 int 518 union_vget(struct mount *mp __unused, ino_t ino __unused, 519 struct vnode **vpp __unused) 520 { 521 522 return (EOPNOTSUPP); 523 } 524 525 SYSCTL_SETUP(sysctl_vfs_union_setup, "sysctl vfs.union subtree setup") 526 { 527 528 sysctl_createv(clog, 0, NULL, NULL, 529 CTLFLAG_PERMANENT, 530 CTLTYPE_NODE, "vfs", NULL, 531 NULL, 0, NULL, 0, 532 CTL_VFS, CTL_EOL); 533 sysctl_createv(clog, 0, NULL, NULL, 534 CTLFLAG_PERMANENT, 535 CTLTYPE_NODE, "union", 536 SYSCTL_DESCR("Union file system"), 537 NULL, 0, NULL, 0, 538 CTL_VFS, 15, CTL_EOL); 539 /* 540 * XXX the "15" above could be dynamic, thereby eliminating 541 * one more instance of the "number to vfs" mapping problem, 542 * but "15" is the order as taken from sys/mount.h 543 */ 544 } 545 546 extern const struct vnodeopv_desc union_vnodeop_opv_desc; 547 548 const struct vnodeopv_desc * const union_vnodeopv_descs[] = { 549 &union_vnodeop_opv_desc, 550 NULL, 551 }; 552 553 struct vfsops union_vfsops = { 554 MOUNT_UNION, 555 union_mount, 556 union_start, 557 union_unmount, 558 union_root, 559 union_quotactl, 560 union_statvfs, 561 union_sync, 562 union_vget, 563 NULL, /* vfs_fhtovp */ 564 NULL, /* vfs_vptofh */ 565 union_init, 566 NULL, /* vfs_reinit */ 567 union_done, 568 NULL, /* vfs_mountroot */ 569 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 570 vfs_stdextattrctl, 571 union_vnodeopv_descs, 572 0, /* vfs_refcount */ 573 { NULL, NULL }, 574 }; 575 VFS_ATTACH(union_vfsops); 576