union_vfsops.c revision 1.57.6.2.2.1
1 /* $NetBSD: union_vfsops.c,v 1.57.6.2.2.1 2014/04/28 16:05:36 sborrill 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 1.57.6.2.2.1 2014/04/28 16:05:36 sborrill 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 static const char *warn_user = 105 "WARNING: the union file system is experimental\n" 106 "WARNING: it can cause crashes and file system corruption\n"; 107 108 /* 109 * Mount union filesystem 110 */ 111 int 112 union_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 113 { 114 struct lwp *l = curlwp; 115 struct nameidata nd; 116 int error = 0; 117 struct union_args *args = data; 118 struct vnode *lowerrootvp = NULLVP; 119 struct vnode *upperrootvp = NULLVP; 120 struct union_mount *um = 0; 121 const char *cp; 122 char *xp; 123 int len; 124 size_t size; 125 126 if (args == NULL) 127 return EINVAL; 128 if (*data_len < sizeof *args) 129 return EINVAL; 130 131 #ifdef UNION_DIAGNOSTIC 132 printf("union_mount(mp = %p)\n", mp); 133 #endif 134 135 if (mp->mnt_flag & MNT_GETARGS) { 136 um = MOUNTTOUNIONMOUNT(mp); 137 if (um == NULL) 138 return EIO; 139 args->target = NULL; 140 args->mntflags = um->um_op; 141 *data_len = sizeof *args; 142 return 0; 143 } 144 /* 145 * Update is a no-op 146 */ 147 if (mp->mnt_flag & MNT_UPDATE) { 148 /* 149 * Need to provide. 150 * 1. a way to convert between rdonly and rdwr mounts. 151 * 2. support for nfs exports. 152 */ 153 error = EOPNOTSUPP; 154 goto bad; 155 } 156 157 if (warn_user != NULL) { 158 printf("%s", warn_user); 159 warn_user = NULL; 160 } 161 162 lowerrootvp = mp->mnt_vnodecovered; 163 VREF(lowerrootvp); 164 165 /* 166 * Find upper node. 167 */ 168 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, args->target); 169 170 if ((error = namei(&nd)) != 0) 171 goto bad; 172 173 upperrootvp = nd.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 vfs_getnewfsid(mp); 257 258 error = set_statvfs_info( path, UIO_USERSPACE, NULL, UIO_USERSPACE, 259 mp->mnt_op->vfs_name, mp, l); 260 if (error) 261 goto bad; 262 263 switch (um->um_op) { 264 case UNMNT_ABOVE: 265 cp = "<above>:"; 266 break; 267 case UNMNT_BELOW: 268 cp = "<below>:"; 269 break; 270 case UNMNT_REPLACE: 271 cp = ""; 272 break; 273 default: 274 cp = "<invalid>:"; 275 #ifdef DIAGNOSTIC 276 panic("union_mount: bad um_op"); 277 #endif 278 break; 279 } 280 len = strlen(cp); 281 memcpy(mp->mnt_stat.f_mntfromname, cp, len); 282 283 xp = mp->mnt_stat.f_mntfromname + len; 284 len = MNAMELEN - len; 285 286 (void) copyinstr(args->target, xp, len - 1, &size); 287 memset(xp + size, 0, len - size); 288 289 #ifdef UNION_DIAGNOSTIC 290 printf("union_mount: from %s, on %s\n", 291 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname); 292 #endif 293 294 /* Setup the readdir hook if it's not set already */ 295 if (!vn_union_readdir_hook) 296 vn_union_readdir_hook = union_readdirhook; 297 298 return (0); 299 300 bad: 301 if (um) 302 free(um, M_UFSMNT); 303 if (upperrootvp) 304 vrele(upperrootvp); 305 if (lowerrootvp) 306 vrele(lowerrootvp); 307 return (error); 308 } 309 310 /* 311 * VFS start. Nothing needed here - the start routine 312 * on the underlying filesystem(s) will have been called 313 * when that filesystem was mounted. 314 */ 315 /*ARGSUSED*/ 316 int 317 union_start(struct mount *mp, int flags) 318 { 319 320 return (0); 321 } 322 323 /* 324 * Free reference to union layer 325 */ 326 int 327 union_unmount(struct mount *mp, int mntflags) 328 { 329 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 330 int freeing; 331 int error; 332 333 #ifdef UNION_DIAGNOSTIC 334 printf("union_unmount(mp = %p)\n", mp); 335 #endif 336 337 /* 338 * Keep flushing vnodes from the mount list. 339 * This is needed because of the un_pvp held 340 * reference to the parent vnode. 341 * If more vnodes have been freed on a given pass, 342 * the try again. The loop will iterate at most 343 * (d) times, where (d) is the maximum tree depth 344 * in the filesystem. 345 */ 346 for (freeing = 0; (error = vflush(mp, NULL, 0)) != 0;) { 347 struct vnode *vp; 348 int n; 349 350 /* count #vnodes held on mount list */ 351 n = 0; 352 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 353 n++; 354 355 /* if this is unchanged then stop */ 356 if (n == freeing) 357 break; 358 359 /* otherwise try once more time */ 360 freeing = n; 361 } 362 363 /* 364 * Ok, now that we've tried doing it gently, get out the hammer. 365 */ 366 367 if (mntflags & MNT_FORCE) 368 error = vflush(mp, NULL, FORCECLOSE); 369 370 if (error) 371 return error; 372 373 /* 374 * Discard references to upper and lower target vnodes. 375 */ 376 if (um->um_lowervp) 377 vrele(um->um_lowervp); 378 vrele(um->um_uppervp); 379 kauth_cred_free(um->um_cred); 380 /* 381 * Finally, throw away the union_mount structure 382 */ 383 free(mp->mnt_data, M_UFSMNT); /* XXX */ 384 mp->mnt_data = NULL; 385 return (0); 386 } 387 388 int 389 union_root(struct mount *mp, struct vnode **vpp) 390 { 391 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 392 int error; 393 394 /* 395 * Return locked reference to root. 396 */ 397 VREF(um->um_uppervp); 398 vn_lock(um->um_uppervp, LK_EXCLUSIVE | LK_RETRY); 399 if (um->um_lowervp) 400 VREF(um->um_lowervp); 401 error = union_allocvp(vpp, mp, NULL, NULL, NULL, 402 um->um_uppervp, um->um_lowervp, 1); 403 404 if (error) { 405 vput(um->um_uppervp); 406 if (um->um_lowervp) 407 vrele(um->um_lowervp); 408 } 409 410 return (error); 411 } 412 413 int 414 union_statvfs(struct mount *mp, struct statvfs *sbp) 415 { 416 int error; 417 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 418 struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO); 419 unsigned long lbsize; 420 421 #ifdef UNION_DIAGNOSTIC 422 printf("union_statvfs(mp = %p, lvp = %p, uvp = %p)\n", mp, 423 um->um_lowervp, um->um_uppervp); 424 #endif 425 426 if (um->um_lowervp) { 427 error = VFS_STATVFS(um->um_lowervp->v_mount, sbuf); 428 if (error) 429 goto done; 430 } 431 432 /* now copy across the "interesting" information and fake the rest */ 433 lbsize = sbuf->f_bsize; 434 sbp->f_blocks = sbuf->f_blocks - sbuf->f_bfree; 435 sbp->f_files = sbuf->f_files - sbuf->f_ffree; 436 437 error = VFS_STATVFS(um->um_uppervp->v_mount, sbuf); 438 if (error) 439 goto done; 440 441 sbp->f_flag = sbuf->f_flag; 442 sbp->f_bsize = sbuf->f_bsize; 443 sbp->f_frsize = sbuf->f_frsize; 444 sbp->f_iosize = sbuf->f_iosize; 445 446 /* 447 * The "total" fields count total resources in all layers, 448 * the "free" fields count only those resources which are 449 * free in the upper layer (since only the upper layer 450 * is writable). 451 */ 452 453 if (sbuf->f_bsize != lbsize) 454 sbp->f_blocks = sbp->f_blocks * lbsize / sbuf->f_bsize; 455 sbp->f_blocks += sbuf->f_blocks; 456 sbp->f_bfree = sbuf->f_bfree; 457 sbp->f_bavail = sbuf->f_bavail; 458 sbp->f_bresvd = sbuf->f_bresvd; 459 sbp->f_files += sbuf->f_files; 460 sbp->f_ffree = sbuf->f_ffree; 461 sbp->f_favail = sbuf->f_favail; 462 sbp->f_fresvd = sbuf->f_fresvd; 463 464 copy_statvfs_info(sbp, mp); 465 done: 466 free(sbuf, M_TEMP); 467 return error; 468 } 469 470 /*ARGSUSED*/ 471 int 472 union_sync(struct mount *mp, int waitfor, 473 kauth_cred_t cred) 474 { 475 476 /* 477 * XXX - Assumes no data cached at union layer. 478 */ 479 return (0); 480 } 481 482 /*ARGSUSED*/ 483 int 484 union_vget(struct mount *mp, ino_t ino, 485 struct vnode **vpp) 486 { 487 488 return (EOPNOTSUPP); 489 } 490 491 static int 492 union_renamelock_enter(struct mount *mp) 493 { 494 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 495 496 /* Lock just the upper fs, where the action happens. */ 497 return VFS_RENAMELOCK_ENTER(um->um_uppervp->v_mount); 498 } 499 500 static void 501 union_renamelock_exit(struct mount *mp) 502 { 503 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 504 505 VFS_RENAMELOCK_EXIT(um->um_uppervp->v_mount); 506 } 507 508 extern const struct vnodeopv_desc union_vnodeop_opv_desc; 509 510 const struct vnodeopv_desc * const union_vnodeopv_descs[] = { 511 &union_vnodeop_opv_desc, 512 NULL, 513 }; 514 515 struct vfsops union_vfsops = { 516 MOUNT_UNION, 517 sizeof (struct union_args), 518 union_mount, 519 union_start, 520 union_unmount, 521 union_root, 522 (void *)eopnotsupp, /* vfs_quotactl */ 523 union_statvfs, 524 union_sync, 525 union_vget, 526 (void *)eopnotsupp, /* vfs_fhtovp */ 527 (void *)eopnotsupp, /* vfs_vptofh */ 528 union_init, 529 NULL, /* vfs_reinit */ 530 union_done, 531 NULL, /* vfs_mountroot */ 532 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 533 vfs_stdextattrctl, 534 (void *)eopnotsupp, /* vfs_suspendctl */ 535 union_renamelock_enter, 536 union_renamelock_exit, 537 (void *)eopnotsupp, 538 union_vnodeopv_descs, 539 0, /* vfs_refcount */ 540 { NULL, NULL }, 541 }; 542 543 static int 544 union_modcmd(modcmd_t cmd, void *arg) 545 { 546 int error; 547 548 switch (cmd) { 549 case MODULE_CMD_INIT: 550 error = vfs_attach(&union_vfsops); 551 if (error != 0) 552 break; 553 sysctl_createv(&union_sysctl_log, 0, NULL, NULL, 554 CTLFLAG_PERMANENT, 555 CTLTYPE_NODE, "vfs", NULL, 556 NULL, 0, NULL, 0, 557 CTL_VFS, CTL_EOL); 558 sysctl_createv(&union_sysctl_log, 0, NULL, NULL, 559 CTLFLAG_PERMANENT, 560 CTLTYPE_NODE, "union", 561 SYSCTL_DESCR("Union file system"), 562 NULL, 0, NULL, 0, 563 CTL_VFS, 15, CTL_EOL); 564 /* 565 * XXX the "15" above could be dynamic, thereby eliminating 566 * one more instance of the "number to vfs" mapping problem, 567 * but "15" is the order as taken from sys/mount.h 568 */ 569 break; 570 case MODULE_CMD_FINI: 571 error = vfs_detach(&union_vfsops); 572 if (error != 0) 573 break; 574 sysctl_teardown(&union_sysctl_log); 575 break; 576 default: 577 error = ENOTTY; 578 break; 579 } 580 581 return (error); 582 } 583
Indexes created Wed Oct 15 16:09:53 GMT 2025