1 /* $NetBSD: union_vfsops.c,v 1.72 2014/04/16 18:55:19 maxv 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 1.72 2014/04/16 18:55:19 maxv 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 #include <sys/module.h> 96 97 #include <fs/union/union.h> 98 99 MODULE(MODULE_CLASS_VFS, union, NULL); 100 101 VFS_PROTOS(union); 102 103 static struct sysctllog *union_sysctl_log; 104 105 /* 106 * Mount union filesystem 107 */ 108 int 109 union_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 110 { 111 struct lwp *l = curlwp; 112 int error = 0; 113 struct union_args *args = data; 114 struct vnode *lowerrootvp = NULLVP; 115 struct vnode *upperrootvp = NULLVP; 116 struct union_mount *um = 0; 117 const char *cp; 118 char *xp; 119 int len; 120 size_t size; 121 122 if (args == NULL) 123 return EINVAL; 124 if (*data_len < sizeof *args) 125 return EINVAL; 126 127 #ifdef UNION_DIAGNOSTIC 128 printf("union_mount(mp = %p)\n", mp); 129 #endif 130 131 if (mp->mnt_flag & MNT_GETARGS) { 132 um = MOUNTTOUNIONMOUNT(mp); 133 if (um == NULL) 134 return EIO; 135 args->target = NULL; 136 args->mntflags = um->um_op; 137 *data_len = sizeof *args; 138 return 0; 139 } 140 /* 141 * Update is a no-op 142 */ 143 if (mp->mnt_flag & MNT_UPDATE) { 144 /* 145 * Need to provide. 146 * 1. a way to convert between rdonly and rdwr mounts. 147 * 2. support for nfs exports. 148 */ 149 error = EOPNOTSUPP; 150 goto bad; 151 } 152 153 lowerrootvp = mp->mnt_vnodecovered; 154 vref(lowerrootvp); 155 156 /* 157 * Find upper node. 158 */ 159 error = namei_simple_user(args->target, 160 NSM_FOLLOW_NOEMULROOT, &upperrootvp); 161 if (error != 0) 162 goto bad; 163 164 if (upperrootvp->v_type != VDIR) { 165 error = EINVAL; 166 goto bad; 167 } 168 169 um = kmem_zalloc(sizeof(struct union_mount), KM_SLEEP); 170 171 /* 172 * Keep a held reference to the target vnodes. 173 * They are vrele'd in union_unmount. 174 * 175 * Depending on the _BELOW flag, the filesystems are 176 * viewed in a different order. In effect, this is the 177 * same as providing a mount under option to the mount syscall. 178 */ 179 180 um->um_op = args->mntflags & UNMNT_OPMASK; 181 switch (um->um_op) { 182 case UNMNT_ABOVE: 183 um->um_lowervp = lowerrootvp; 184 um->um_uppervp = upperrootvp; 185 break; 186 187 case UNMNT_BELOW: 188 um->um_lowervp = upperrootvp; 189 um->um_uppervp = lowerrootvp; 190 break; 191 192 case UNMNT_REPLACE: 193 vrele(lowerrootvp); 194 lowerrootvp = NULLVP; 195 um->um_uppervp = upperrootvp; 196 um->um_lowervp = lowerrootvp; 197 break; 198 199 default: 200 error = EINVAL; 201 goto bad; 202 } 203 204 mp->mnt_iflag |= IMNT_MPSAFE; 205 206 /* 207 * Unless the mount is readonly, ensure that the top layer 208 * supports whiteout operations 209 */ 210 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 211 vn_lock(um->um_uppervp, LK_EXCLUSIVE | LK_RETRY); 212 error = VOP_WHITEOUT(um->um_uppervp, 213 (struct componentname *) 0, LOOKUP); 214 VOP_UNLOCK(um->um_uppervp); 215 if (error) 216 goto bad; 217 } 218 219 um->um_cred = l->l_cred; 220 kauth_cred_hold(um->um_cred); 221 um->um_cmode = UN_DIRMODE &~ l->l_proc->p_cwdi->cwdi_cmask; 222 223 /* 224 * Depending on what you think the MNT_LOCAL flag might mean, 225 * you may want the && to be || on the conditional below. 226 * At the moment it has been defined that the filesystem is 227 * only local if it is all local, ie the MNT_LOCAL flag implies 228 * that the entire namespace is local. If you think the MNT_LOCAL 229 * flag implies that some of the files might be stored locally 230 * then you will want to change the conditional. 231 */ 232 if (um->um_op == UNMNT_ABOVE) { 233 if (((um->um_lowervp == NULLVP) || 234 (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) && 235 (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL)) 236 mp->mnt_flag |= MNT_LOCAL; 237 } 238 239 /* 240 * Copy in the upper layer's RDONLY flag. This is for the benefit 241 * of lookup() which explicitly checks the flag, rather than asking 242 * the filesystem for it's own opinion. This means, that an update 243 * mount of the underlying filesystem to go from rdonly to rdwr 244 * will leave the unioned view as read-only. 245 */ 246 mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY); 247 248 mp->mnt_data = um; 249 vfs_getnewfsid(mp); 250 251 error = set_statvfs_info( path, UIO_USERSPACE, NULL, UIO_USERSPACE, 252 mp->mnt_op->vfs_name, mp, l); 253 if (error) 254 goto bad; 255 256 switch (um->um_op) { 257 case UNMNT_ABOVE: 258 cp = "<above>:"; 259 break; 260 case UNMNT_BELOW: 261 cp = "<below>:"; 262 break; 263 case UNMNT_REPLACE: 264 cp = ""; 265 break; 266 default: 267 cp = "<invalid>:"; 268 #ifdef DIAGNOSTIC 269 panic("union_mount: bad um_op"); 270 #endif 271 break; 272 } 273 len = strlen(cp); 274 memcpy(mp->mnt_stat.f_mntfromname, cp, len); 275 276 xp = mp->mnt_stat.f_mntfromname + len; 277 len = MNAMELEN - len; 278 279 (void) copyinstr(args->target, xp, len - 1, &size); 280 memset(xp + size, 0, len - size); 281 282 #ifdef UNION_DIAGNOSTIC 283 printf("union_mount: from %s, on %s\n", 284 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname); 285 #endif 286 287 /* Setup the readdir hook if it's not set already */ 288 if (!vn_union_readdir_hook) 289 vn_union_readdir_hook = union_readdirhook; 290 291 return (0); 292 293 bad: 294 if (um) 295 kmem_free(um, sizeof(struct union_mount)); 296 if (upperrootvp) 297 vrele(upperrootvp); 298 if (lowerrootvp) 299 vrele(lowerrootvp); 300 return (error); 301 } 302 303 /* 304 * VFS start. Nothing needed here - the start routine 305 * on the underlying filesystem(s) will have been called 306 * when that filesystem was mounted. 307 */ 308 /*ARGSUSED*/ 309 int 310 union_start(struct mount *mp, int flags) 311 { 312 313 return (0); 314 } 315 316 /* 317 * Free reference to union layer 318 */ 319 int 320 union_unmount(struct mount *mp, int mntflags) 321 { 322 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 323 int freeing; 324 int error; 325 326 #ifdef UNION_DIAGNOSTIC 327 printf("union_unmount(mp = %p)\n", mp); 328 #endif 329 330 /* 331 * Keep flushing vnodes from the mount list. 332 * This is needed because of the un_pvp held 333 * reference to the parent vnode. 334 * If more vnodes have been freed on a given pass, 335 * the try again. The loop will iterate at most 336 * (d) times, where (d) is the maximum tree depth 337 * in the filesystem. 338 */ 339 for (freeing = 0; (error = vflush(mp, NULL, 0)) != 0;) { 340 struct vnode *vp; 341 int n; 342 343 /* count #vnodes held on mount list */ 344 n = 0; 345 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 346 n++; 347 348 /* if this is unchanged then stop */ 349 if (n == freeing) 350 break; 351 352 /* otherwise try once more time */ 353 freeing = n; 354 } 355 356 /* 357 * Ok, now that we've tried doing it gently, get out the hammer. 358 */ 359 360 if (mntflags & MNT_FORCE) 361 error = vflush(mp, NULL, FORCECLOSE); 362 363 if (error) 364 return error; 365 366 /* 367 * Discard references to upper and lower target vnodes. 368 */ 369 if (um->um_lowervp) 370 vrele(um->um_lowervp); 371 vrele(um->um_uppervp); 372 kauth_cred_free(um->um_cred); 373 /* 374 * Finally, throw away the union_mount structure 375 */ 376 kmem_free(um, sizeof(struct union_mount)); 377 mp->mnt_data = NULL; 378 return 0; 379 } 380 381 int 382 union_root(struct mount *mp, struct vnode **vpp) 383 { 384 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 385 int error; 386 387 /* 388 * Return locked reference to root. 389 */ 390 vref(um->um_uppervp); 391 if (um->um_lowervp) 392 vref(um->um_lowervp); 393 error = union_allocvp(vpp, mp, NULL, NULL, NULL, 394 um->um_uppervp, um->um_lowervp, 1); 395 396 if (error) { 397 vrele(um->um_uppervp); 398 if (um->um_lowervp) 399 vrele(um->um_lowervp); 400 return error; 401 } 402 403 vn_lock(*vpp, LK_EXCLUSIVE | LK_RETRY); 404 405 return 0; 406 } 407 408 int 409 union_statvfs(struct mount *mp, struct statvfs *sbp) 410 { 411 int error; 412 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 413 struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO); 414 unsigned long lbsize; 415 416 #ifdef UNION_DIAGNOSTIC 417 printf("union_statvfs(mp = %p, lvp = %p, uvp = %p)\n", mp, 418 um->um_lowervp, um->um_uppervp); 419 #endif 420 421 if (um->um_lowervp) { 422 error = VFS_STATVFS(um->um_lowervp->v_mount, sbuf); 423 if (error) 424 goto done; 425 } 426 427 /* now copy across the "interesting" information and fake the rest */ 428 lbsize = sbuf->f_bsize; 429 sbp->f_blocks = sbuf->f_blocks - sbuf->f_bfree; 430 sbp->f_files = sbuf->f_files - sbuf->f_ffree; 431 432 error = VFS_STATVFS(um->um_uppervp->v_mount, sbuf); 433 if (error) 434 goto done; 435 436 sbp->f_flag = sbuf->f_flag; 437 sbp->f_bsize = sbuf->f_bsize; 438 sbp->f_frsize = sbuf->f_frsize; 439 sbp->f_iosize = sbuf->f_iosize; 440 441 /* 442 * The "total" fields count total resources in all layers, 443 * the "free" fields count only those resources which are 444 * free in the upper layer (since only the upper layer 445 * is writable). 446 */ 447 448 if (sbuf->f_bsize != lbsize) 449 sbp->f_blocks = sbp->f_blocks * lbsize / sbuf->f_bsize; 450 sbp->f_blocks += sbuf->f_blocks; 451 sbp->f_bfree = sbuf->f_bfree; 452 sbp->f_bavail = sbuf->f_bavail; 453 sbp->f_bresvd = sbuf->f_bresvd; 454 sbp->f_files += sbuf->f_files; 455 sbp->f_ffree = sbuf->f_ffree; 456 sbp->f_favail = sbuf->f_favail; 457 sbp->f_fresvd = sbuf->f_fresvd; 458 459 copy_statvfs_info(sbp, mp); 460 done: 461 free(sbuf, M_TEMP); 462 return error; 463 } 464 465 /*ARGSUSED*/ 466 int 467 union_sync(struct mount *mp, int waitfor, 468 kauth_cred_t cred) 469 { 470 471 /* 472 * XXX - Assumes no data cached at union layer. 473 */ 474 return (0); 475 } 476 477 /*ARGSUSED*/ 478 int 479 union_vget(struct mount *mp, ino_t ino, 480 struct vnode **vpp) 481 { 482 483 return (EOPNOTSUPP); 484 } 485 486 static int 487 union_renamelock_enter(struct mount *mp) 488 { 489 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 490 491 /* Lock just the upper fs, where the action happens. */ 492 return VFS_RENAMELOCK_ENTER(um->um_uppervp->v_mount); 493 } 494 495 static void 496 union_renamelock_exit(struct mount *mp) 497 { 498 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 499 500 VFS_RENAMELOCK_EXIT(um->um_uppervp->v_mount); 501 } 502 503 extern const struct vnodeopv_desc union_vnodeop_opv_desc; 504 505 const struct vnodeopv_desc * const union_vnodeopv_descs[] = { 506 &union_vnodeop_opv_desc, 507 NULL, 508 }; 509 510 struct vfsops union_vfsops = { 511 .vfs_name = MOUNT_UNION, 512 .vfs_min_mount_data = sizeof (struct union_args), 513 .vfs_mount = union_mount, 514 .vfs_start = union_start, 515 .vfs_unmount = union_unmount, 516 .vfs_root = union_root, 517 .vfs_quotactl = (void *)eopnotsupp, 518 .vfs_statvfs = union_statvfs, 519 .vfs_sync = union_sync, 520 .vfs_vget = union_vget, 521 .vfs_fhtovp = (void *)eopnotsupp, 522 .vfs_vptofh = (void *)eopnotsupp, 523 .vfs_init = union_init, 524 .vfs_reinit = union_reinit, 525 .vfs_done = union_done, 526 .vfs_snapshot = (void *)eopnotsupp, 527 .vfs_extattrctl = vfs_stdextattrctl, 528 .vfs_suspendctl = (void *)eopnotsupp, 529 .vfs_renamelock_enter = union_renamelock_enter, 530 .vfs_renamelock_exit = union_renamelock_exit, 531 .vfs_fsync = (void *)eopnotsupp, 532 .vfs_opv_descs = union_vnodeopv_descs 533 }; 534 535 static int 536 union_modcmd(modcmd_t cmd, void *arg) 537 { 538 int error; 539 540 switch (cmd) { 541 case MODULE_CMD_INIT: 542 error = vfs_attach(&union_vfsops); 543 if (error != 0) 544 break; 545 sysctl_createv(&union_sysctl_log, 0, NULL, NULL, 546 CTLFLAG_PERMANENT, 547 CTLTYPE_NODE, "union", 548 SYSCTL_DESCR("Union file system"), 549 NULL, 0, NULL, 0, 550 CTL_VFS, 15, CTL_EOL); 551 /* 552 * XXX the "15" above could be dynamic, thereby eliminating 553 * one more instance of the "number to vfs" mapping problem, 554 * but "15" is the order as taken from sys/mount.h 555 */ 556 break; 557 case MODULE_CMD_FINI: 558 error = vfs_detach(&union_vfsops); 559 if (error != 0) 560 break; 561 sysctl_teardown(&union_sysctl_log); 562 break; 563 default: 564 error = ENOTTY; 565 break; 566 } 567 568 return (error); 569 } 570