union_vfsops.c revision 1.52.2.1
1 /* $NetBSD: union_vfsops.c,v 1.52.2.1 2008/05/18 12:35:06 yamt 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 1.52.2.1 2008/05/18 12:35:06 yamt 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 /* 104 * Mount union filesystem 105 */ 106 int 107 union_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 108 { 109 struct lwp *l = curlwp; 110 struct nameidata nd; 111 int error = 0; 112 struct union_args *args = data; 113 struct vnode *lowerrootvp = NULLVP; 114 struct vnode *upperrootvp = NULLVP; 115 struct union_mount *um = 0; 116 const char *cp; 117 char *xp; 118 int len; 119 size_t size; 120 121 if (*data_len < sizeof *args) 122 return EINVAL; 123 124 #ifdef UNION_DIAGNOSTIC 125 printf("union_mount(mp = %p)\n", mp); 126 #endif 127 128 if (mp->mnt_flag & MNT_GETARGS) { 129 um = MOUNTTOUNIONMOUNT(mp); 130 if (um == NULL) 131 return EIO; 132 args->target = NULL; 133 args->mntflags = um->um_op; 134 *data_len = sizeof *args; 135 return 0; 136 } 137 /* 138 * Update is a no-op 139 */ 140 if (mp->mnt_flag & MNT_UPDATE) { 141 /* 142 * Need to provide. 143 * 1. a way to convert between rdonly and rdwr mounts. 144 * 2. support for nfs exports. 145 */ 146 error = EOPNOTSUPP; 147 goto bad; 148 } 149 150 lowerrootvp = mp->mnt_vnodecovered; 151 VREF(lowerrootvp); 152 153 /* 154 * Find upper node. 155 */ 156 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, args->target); 157 158 if ((error = namei(&nd)) != 0) 159 goto bad; 160 161 upperrootvp = nd.ni_vp; 162 163 if (upperrootvp->v_type != VDIR) { 164 error = EINVAL; 165 goto bad; 166 } 167 168 um = (struct union_mount *) malloc(sizeof(struct union_mount), 169 M_UFSMNT, M_WAITOK); /* XXX */ 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 /* 205 * Unless the mount is readonly, ensure that the top layer 206 * supports whiteout operations 207 */ 208 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 209 error = VOP_WHITEOUT(um->um_uppervp, (struct componentname *) 0, LOOKUP); 210 if (error) 211 goto bad; 212 } 213 214 um->um_cred = l->l_cred; 215 kauth_cred_hold(um->um_cred); 216 um->um_cmode = UN_DIRMODE &~ l->l_proc->p_cwdi->cwdi_cmask; 217 218 /* 219 * Depending on what you think the MNT_LOCAL flag might mean, 220 * you may want the && to be || on the conditional below. 221 * At the moment it has been defined that the filesystem is 222 * only local if it is all local, ie the MNT_LOCAL flag implies 223 * that the entire namespace is local. If you think the MNT_LOCAL 224 * flag implies that some of the files might be stored locally 225 * then you will want to change the conditional. 226 */ 227 if (um->um_op == UNMNT_ABOVE) { 228 if (((um->um_lowervp == NULLVP) || 229 (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) && 230 (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL)) 231 mp->mnt_flag |= MNT_LOCAL; 232 } 233 234 /* 235 * Copy in the upper layer's RDONLY flag. This is for the benefit 236 * of lookup() which explicitly checks the flag, rather than asking 237 * the filesystem for it's own opinion. This means, that an update 238 * mount of the underlying filesystem to go from rdonly to rdwr 239 * will leave the unioned view as read-only. 240 */ 241 mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY); 242 243 mp->mnt_data = um; 244 vfs_getnewfsid(mp); 245 246 error = set_statvfs_info( path, UIO_USERSPACE, NULL, UIO_USERSPACE, 247 mp->mnt_op->vfs_name, mp, l); 248 if (error) 249 goto bad; 250 251 switch (um->um_op) { 252 case UNMNT_ABOVE: 253 cp = "<above>:"; 254 break; 255 case UNMNT_BELOW: 256 cp = "<below>:"; 257 break; 258 case UNMNT_REPLACE: 259 cp = ""; 260 break; 261 default: 262 cp = "<invalid>:"; 263 #ifdef DIAGNOSTIC 264 panic("union_mount: bad um_op"); 265 #endif 266 break; 267 } 268 len = strlen(cp); 269 memcpy(mp->mnt_stat.f_mntfromname, cp, len); 270 271 xp = mp->mnt_stat.f_mntfromname + len; 272 len = MNAMELEN - len; 273 274 (void) copyinstr(args->target, xp, len - 1, &size); 275 memset(xp + size, 0, len - size); 276 277 #ifdef UNION_DIAGNOSTIC 278 printf("union_mount: from %s, on %s\n", 279 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname); 280 #endif 281 282 /* Setup the readdir hook if it's not set already */ 283 if (!vn_union_readdir_hook) 284 vn_union_readdir_hook = union_readdirhook; 285 286 return (0); 287 288 bad: 289 if (um) 290 free(um, M_UFSMNT); 291 if (upperrootvp) 292 vrele(upperrootvp); 293 if (lowerrootvp) 294 vrele(lowerrootvp); 295 return (error); 296 } 297 298 /* 299 * VFS start. Nothing needed here - the start routine 300 * on the underlying filesystem(s) will have been called 301 * when that filesystem was mounted. 302 */ 303 /*ARGSUSED*/ 304 int 305 union_start(struct mount *mp, int flags) 306 { 307 308 return (0); 309 } 310 311 /* 312 * Free reference to union layer 313 */ 314 int 315 union_unmount(struct mount *mp, int mntflags) 316 { 317 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 318 int freeing; 319 int error; 320 321 #ifdef UNION_DIAGNOSTIC 322 printf("union_unmount(mp = %p)\n", mp); 323 #endif 324 325 /* 326 * Keep flushing vnodes from the mount list. 327 * This is needed because of the un_pvp held 328 * reference to the parent vnode. 329 * If more vnodes have been freed on a given pass, 330 * the try again. The loop will iterate at most 331 * (d) times, where (d) is the maximum tree depth 332 * in the filesystem. 333 */ 334 for (freeing = 0; (error = vflush(mp, NULL, 0)) != 0;) { 335 struct vnode *vp; 336 int n; 337 338 /* count #vnodes held on mount list */ 339 n = 0; 340 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 341 n++; 342 343 /* if this is unchanged then stop */ 344 if (n == freeing) 345 break; 346 347 /* otherwise try once more time */ 348 freeing = n; 349 } 350 351 /* 352 * Ok, now that we've tried doing it gently, get out the hammer. 353 */ 354 355 if (mntflags & MNT_FORCE) 356 error = vflush(mp, NULL, FORCECLOSE); 357 358 if (error) 359 return error; 360 361 /* 362 * Discard references to upper and lower target vnodes. 363 */ 364 if (um->um_lowervp) 365 vrele(um->um_lowervp); 366 vrele(um->um_uppervp); 367 kauth_cred_free(um->um_cred); 368 /* 369 * Finally, throw away the union_mount structure 370 */ 371 free(mp->mnt_data, M_UFSMNT); /* XXX */ 372 mp->mnt_data = NULL; 373 return (0); 374 } 375 376 int 377 union_root(struct mount *mp, struct vnode **vpp) 378 { 379 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 380 int error; 381 382 /* 383 * Return locked reference to root. 384 */ 385 VREF(um->um_uppervp); 386 vn_lock(um->um_uppervp, LK_EXCLUSIVE | LK_RETRY); 387 if (um->um_lowervp) 388 VREF(um->um_lowervp); 389 error = union_allocvp(vpp, mp, NULL, NULL, NULL, 390 um->um_uppervp, um->um_lowervp, 1); 391 392 if (error) { 393 vput(um->um_uppervp); 394 if (um->um_lowervp) 395 vrele(um->um_lowervp); 396 } 397 398 return (error); 399 } 400 401 int 402 union_statvfs(struct mount *mp, struct statvfs *sbp) 403 { 404 int error; 405 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 406 struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO); 407 unsigned long lbsize; 408 409 #ifdef UNION_DIAGNOSTIC 410 printf("union_statvfs(mp = %p, lvp = %p, uvp = %p)\n", mp, 411 um->um_lowervp, um->um_uppervp); 412 #endif 413 414 if (um->um_lowervp) { 415 error = VFS_STATVFS(um->um_lowervp->v_mount, sbuf); 416 if (error) 417 goto done; 418 } 419 420 /* now copy across the "interesting" information and fake the rest */ 421 lbsize = sbuf->f_bsize; 422 sbp->f_blocks = sbuf->f_blocks - sbuf->f_bfree; 423 sbp->f_files = sbuf->f_files - sbuf->f_ffree; 424 425 error = VFS_STATVFS(um->um_uppervp->v_mount, sbuf); 426 if (error) 427 goto done; 428 429 sbp->f_flag = sbuf->f_flag; 430 sbp->f_bsize = sbuf->f_bsize; 431 sbp->f_frsize = sbuf->f_frsize; 432 sbp->f_iosize = sbuf->f_iosize; 433 434 /* 435 * The "total" fields count total resources in all layers, 436 * the "free" fields count only those resources which are 437 * free in the upper layer (since only the upper layer 438 * is writable). 439 */ 440 441 if (sbuf->f_bsize != lbsize) 442 sbp->f_blocks = sbp->f_blocks * lbsize / sbuf->f_bsize; 443 sbp->f_blocks += sbuf->f_blocks; 444 sbp->f_bfree = sbuf->f_bfree; 445 sbp->f_bavail = sbuf->f_bavail; 446 sbp->f_bresvd = sbuf->f_bresvd; 447 sbp->f_files += sbuf->f_files; 448 sbp->f_ffree = sbuf->f_ffree; 449 sbp->f_favail = sbuf->f_favail; 450 sbp->f_fresvd = sbuf->f_fresvd; 451 452 copy_statvfs_info(sbp, mp); 453 done: 454 free(sbuf, M_TEMP); 455 return error; 456 } 457 458 /*ARGSUSED*/ 459 int 460 union_sync(struct mount *mp, int waitfor, 461 kauth_cred_t cred) 462 { 463 464 /* 465 * XXX - Assumes no data cached at union layer. 466 */ 467 return (0); 468 } 469 470 /*ARGSUSED*/ 471 int 472 union_vget(struct mount *mp, ino_t ino, 473 struct vnode **vpp) 474 { 475 476 return (EOPNOTSUPP); 477 } 478 479 static int 480 union_renamelock_enter(struct mount *mp) 481 { 482 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 483 484 /* Lock just the upper fs, where the action happens. */ 485 return VFS_RENAMELOCK_ENTER(um->um_uppervp->v_mount); 486 } 487 488 static void 489 union_renamelock_exit(struct mount *mp) 490 { 491 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 492 493 VFS_RENAMELOCK_EXIT(um->um_uppervp->v_mount); 494 } 495 496 SYSCTL_SETUP(sysctl_vfs_union_setup, "sysctl vfs.union subtree setup") 497 { 498 499 sysctl_createv(clog, 0, NULL, NULL, 500 CTLFLAG_PERMANENT, 501 CTLTYPE_NODE, "vfs", NULL, 502 NULL, 0, NULL, 0, 503 CTL_VFS, CTL_EOL); 504 sysctl_createv(clog, 0, NULL, NULL, 505 CTLFLAG_PERMANENT, 506 CTLTYPE_NODE, "union", 507 SYSCTL_DESCR("Union file system"), 508 NULL, 0, NULL, 0, 509 CTL_VFS, 15, CTL_EOL); 510 /* 511 * XXX the "15" above could be dynamic, thereby eliminating 512 * one more instance of the "number to vfs" mapping problem, 513 * but "15" is the order as taken from sys/mount.h 514 */ 515 } 516 517 extern const struct vnodeopv_desc union_vnodeop_opv_desc; 518 519 const struct vnodeopv_desc * const union_vnodeopv_descs[] = { 520 &union_vnodeop_opv_desc, 521 NULL, 522 }; 523 524 struct vfsops union_vfsops = { 525 MOUNT_UNION, 526 sizeof (struct union_args), 527 union_mount, 528 union_start, 529 union_unmount, 530 union_root, 531 (void *)eopnotsupp, /* vfs_quotactl */ 532 union_statvfs, 533 union_sync, 534 union_vget, 535 (void *)eopnotsupp, /* vfs_fhtovp */ 536 (void *)eopnotsupp, /* vfs_vptofh */ 537 union_init, 538 NULL, /* vfs_reinit */ 539 union_done, 540 NULL, /* vfs_mountroot */ 541 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 542 vfs_stdextattrctl, 543 (void *)eopnotsupp, /* vfs_suspendctl */ 544 union_renamelock_enter, 545 union_renamelock_exit, 546 (void *)eopnotsupp, 547 union_vnodeopv_descs, 548 0, /* vfs_refcount */ 549 { NULL, NULL }, 550 }; 551 552 static int 553 union_modcmd(modcmd_t cmd, void *arg) 554 { 555 556 switch (cmd) { 557 case MODULE_CMD_INIT: 558 return vfs_attach(&union_vfsops); 559 case MODULE_CMD_FINI: 560 return vfs_detach(&union_vfsops); 561 default: 562 return ENOTTY; 563 } 564 } 565
Indexes created Mon Oct 13 01:09:56 GMT 2025