union_vfsops.c revision 1.64.2.1
1 /* $NetBSD: union_vfsops.c,v 1.64.2.1 2012/04/17 00:08:21 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 1.64.2.1 2012/04/17 00:08:21 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 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 (*data_len < sizeof *args) 123 return EINVAL; 124 125 #ifdef UNION_DIAGNOSTIC 126 printf("union_mount(mp = %p)\n", mp); 127 #endif 128 129 if (mp->mnt_flag & MNT_GETARGS) { 130 um = MOUNTTOUNIONMOUNT(mp); 131 if (um == NULL) 132 return EIO; 133 args->target = NULL; 134 args->mntflags = um->um_op; 135 *data_len = sizeof *args; 136 return 0; 137 } 138 /* 139 * Update is a no-op 140 */ 141 if (mp->mnt_flag & MNT_UPDATE) { 142 /* 143 * Need to provide. 144 * 1. a way to convert between rdonly and rdwr mounts. 145 * 2. support for nfs exports. 146 */ 147 error = EOPNOTSUPP; 148 goto bad; 149 } 150 151 lowerrootvp = mp->mnt_vnodecovered; 152 vref(lowerrootvp); 153 154 /* 155 * Find upper node. 156 */ 157 error = namei_simple_user(args->target, 158 NSM_FOLLOW_NOEMULROOT, &upperrootvp); 159 if (error != 0) 160 goto bad; 161 162 if (upperrootvp->v_type != VDIR) { 163 error = EINVAL; 164 goto bad; 165 } 166 167 um = (struct union_mount *) malloc(sizeof(struct union_mount), 168 M_UFSMNT, M_WAITOK); /* XXX */ 169 170 /* 171 * Keep a held reference to the target vnodes. 172 * They are vrele'd in union_unmount. 173 * 174 * Depending on the _BELOW flag, the filesystems are 175 * viewed in a different order. In effect, this is the 176 * same as providing a mount under option to the mount syscall. 177 */ 178 179 um->um_op = args->mntflags & UNMNT_OPMASK; 180 switch (um->um_op) { 181 case UNMNT_ABOVE: 182 um->um_lowervp = lowerrootvp; 183 um->um_uppervp = upperrootvp; 184 break; 185 186 case UNMNT_BELOW: 187 um->um_lowervp = upperrootvp; 188 um->um_uppervp = lowerrootvp; 189 break; 190 191 case UNMNT_REPLACE: 192 vrele(lowerrootvp); 193 lowerrootvp = NULLVP; 194 um->um_uppervp = upperrootvp; 195 um->um_lowervp = lowerrootvp; 196 break; 197 198 default: 199 error = EINVAL; 200 goto bad; 201 } 202 203 mp->mnt_iflag |= IMNT_MPSAFE; 204 205 /* 206 * Unless the mount is readonly, ensure that the top layer 207 * supports whiteout operations 208 */ 209 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 210 vn_lock(um->um_uppervp, LK_EXCLUSIVE | LK_RETRY); 211 error = VOP_WHITEOUT(um->um_uppervp, 212 (struct componentname *) 0, LOOKUP); 213 VOP_UNLOCK(um->um_uppervp); 214 if (error) 215 goto bad; 216 } 217 218 um->um_cred = l->l_cred; 219 kauth_cred_hold(um->um_cred); 220 um->um_cmode = UN_DIRMODE &~ l->l_proc->p_cwdi->cwdi_cmask; 221 222 /* 223 * Depending on what you think the MNT_LOCAL flag might mean, 224 * you may want the && to be || on the conditional below. 225 * At the moment it has been defined that the filesystem is 226 * only local if it is all local, ie the MNT_LOCAL flag implies 227 * that the entire namespace is local. If you think the MNT_LOCAL 228 * flag implies that some of the files might be stored locally 229 * then you will want to change the conditional. 230 */ 231 if (um->um_op == UNMNT_ABOVE) { 232 if (((um->um_lowervp == NULLVP) || 233 (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) && 234 (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL)) 235 mp->mnt_flag |= MNT_LOCAL; 236 } 237 238 /* 239 * Copy in the upper layer's RDONLY flag. This is for the benefit 240 * of lookup() which explicitly checks the flag, rather than asking 241 * the filesystem for it's own opinion. This means, that an update 242 * mount of the underlying filesystem to go from rdonly to rdwr 243 * will leave the unioned view as read-only. 244 */ 245 mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY); 246 247 mp->mnt_data = um; 248 vfs_getnewfsid(mp); 249 250 error = set_statvfs_info( path, UIO_USERSPACE, NULL, UIO_USERSPACE, 251 mp->mnt_op->vfs_name, mp, l); 252 if (error) 253 goto bad; 254 255 switch (um->um_op) { 256 case UNMNT_ABOVE: 257 cp = "<above>:"; 258 break; 259 case UNMNT_BELOW: 260 cp = "<below>:"; 261 break; 262 case UNMNT_REPLACE: 263 cp = ""; 264 break; 265 default: 266 cp = "<invalid>:"; 267 #ifdef DIAGNOSTIC 268 panic("union_mount: bad um_op"); 269 #endif 270 break; 271 } 272 len = strlen(cp); 273 memcpy(mp->mnt_stat.f_mntfromname, cp, len); 274 275 xp = mp->mnt_stat.f_mntfromname + len; 276 len = MNAMELEN - len; 277 278 (void) copyinstr(args->target, xp, len - 1, &size); 279 memset(xp + size, 0, len - size); 280 281 #ifdef UNION_DIAGNOSTIC 282 printf("union_mount: from %s, on %s\n", 283 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname); 284 #endif 285 286 /* Setup the readdir hook if it's not set already */ 287 if (!vn_union_readdir_hook) 288 vn_union_readdir_hook = union_readdirhook; 289 290 return (0); 291 292 bad: 293 if (um) 294 free(um, M_UFSMNT); 295 if (upperrootvp) 296 vrele(upperrootvp); 297 if (lowerrootvp) 298 vrele(lowerrootvp); 299 return (error); 300 } 301 302 /* 303 * VFS start. Nothing needed here - the start routine 304 * on the underlying filesystem(s) will have been called 305 * when that filesystem was mounted. 306 */ 307 /*ARGSUSED*/ 308 int 309 union_start(struct mount *mp, int flags) 310 { 311 312 return (0); 313 } 314 315 /* 316 * Free reference to union layer 317 */ 318 int 319 union_unmount(struct mount *mp, int mntflags) 320 { 321 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 322 int freeing; 323 int error; 324 325 #ifdef UNION_DIAGNOSTIC 326 printf("union_unmount(mp = %p)\n", mp); 327 #endif 328 329 /* 330 * Keep flushing vnodes from the mount list. 331 * This is needed because of the un_pvp held 332 * reference to the parent vnode. 333 * If more vnodes have been freed on a given pass, 334 * the try again. The loop will iterate at most 335 * (d) times, where (d) is the maximum tree depth 336 * in the filesystem. 337 */ 338 for (freeing = 0; (error = vflush(mp, NULL, 0)) != 0;) { 339 struct vnode *vp; 340 int n; 341 342 /* count #vnodes held on mount list */ 343 n = 0; 344 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) 345 n++; 346 347 /* if this is unchanged then stop */ 348 if (n == freeing) 349 break; 350 351 /* otherwise try once more time */ 352 freeing = n; 353 } 354 355 /* 356 * Ok, now that we've tried doing it gently, get out the hammer. 357 */ 358 359 if (mntflags & MNT_FORCE) 360 error = vflush(mp, NULL, FORCECLOSE); 361 362 if (error) 363 return error; 364 365 /* 366 * Discard references to upper and lower target vnodes. 367 */ 368 if (um->um_lowervp) 369 vrele(um->um_lowervp); 370 vrele(um->um_uppervp); 371 kauth_cred_free(um->um_cred); 372 /* 373 * Finally, throw away the union_mount structure 374 */ 375 free(mp->mnt_data, M_UFSMNT); /* XXX */ 376 mp->mnt_data = NULL; 377 return (0); 378 } 379 380 int 381 union_root(struct mount *mp, struct vnode **vpp) 382 { 383 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 384 int error; 385 386 /* 387 * Return locked reference to root. 388 */ 389 vref(um->um_uppervp); 390 vn_lock(um->um_uppervp, LK_EXCLUSIVE | LK_RETRY); 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 vput(um->um_uppervp); 398 if (um->um_lowervp) 399 vrele(um->um_lowervp); 400 } 401 402 return (error); 403 } 404 405 int 406 union_statvfs(struct mount *mp, struct statvfs *sbp) 407 { 408 int error; 409 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 410 struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO); 411 unsigned long lbsize; 412 413 #ifdef UNION_DIAGNOSTIC 414 printf("union_statvfs(mp = %p, lvp = %p, uvp = %p)\n", mp, 415 um->um_lowervp, um->um_uppervp); 416 #endif 417 418 if (um->um_lowervp) { 419 error = VFS_STATVFS(um->um_lowervp->v_mount, sbuf); 420 if (error) 421 goto done; 422 } 423 424 /* now copy across the "interesting" information and fake the rest */ 425 lbsize = sbuf->f_bsize; 426 sbp->f_blocks = sbuf->f_blocks - sbuf->f_bfree; 427 sbp->f_files = sbuf->f_files - sbuf->f_ffree; 428 429 error = VFS_STATVFS(um->um_uppervp->v_mount, sbuf); 430 if (error) 431 goto done; 432 433 sbp->f_flag = sbuf->f_flag; 434 sbp->f_bsize = sbuf->f_bsize; 435 sbp->f_frsize = sbuf->f_frsize; 436 sbp->f_iosize = sbuf->f_iosize; 437 438 /* 439 * The "total" fields count total resources in all layers, 440 * the "free" fields count only those resources which are 441 * free in the upper layer (since only the upper layer 442 * is writable). 443 */ 444 445 if (sbuf->f_bsize != lbsize) 446 sbp->f_blocks = sbp->f_blocks * lbsize / sbuf->f_bsize; 447 sbp->f_blocks += sbuf->f_blocks; 448 sbp->f_bfree = sbuf->f_bfree; 449 sbp->f_bavail = sbuf->f_bavail; 450 sbp->f_bresvd = sbuf->f_bresvd; 451 sbp->f_files += sbuf->f_files; 452 sbp->f_ffree = sbuf->f_ffree; 453 sbp->f_favail = sbuf->f_favail; 454 sbp->f_fresvd = sbuf->f_fresvd; 455 456 copy_statvfs_info(sbp, mp); 457 done: 458 free(sbuf, M_TEMP); 459 return error; 460 } 461 462 /*ARGSUSED*/ 463 int 464 union_sync(struct mount *mp, int waitfor, 465 kauth_cred_t cred) 466 { 467 468 /* 469 * XXX - Assumes no data cached at union layer. 470 */ 471 return (0); 472 } 473 474 /*ARGSUSED*/ 475 int 476 union_vget(struct mount *mp, ino_t ino, 477 struct vnode **vpp) 478 { 479 480 return (EOPNOTSUPP); 481 } 482 483 static int 484 union_renamelock_enter(struct mount *mp) 485 { 486 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 487 488 /* Lock just the upper fs, where the action happens. */ 489 return VFS_RENAMELOCK_ENTER(um->um_uppervp->v_mount); 490 } 491 492 static void 493 union_renamelock_exit(struct mount *mp) 494 { 495 struct union_mount *um = MOUNTTOUNIONMOUNT(mp); 496 497 VFS_RENAMELOCK_EXIT(um->um_uppervp->v_mount); 498 } 499 500 extern const struct vnodeopv_desc union_vnodeop_opv_desc; 501 502 const struct vnodeopv_desc * const union_vnodeopv_descs[] = { 503 &union_vnodeop_opv_desc, 504 NULL, 505 }; 506 507 struct vfsops union_vfsops = { 508 MOUNT_UNION, 509 sizeof (struct union_args), 510 union_mount, 511 union_start, 512 union_unmount, 513 union_root, 514 (void *)eopnotsupp, /* vfs_quotactl */ 515 union_statvfs, 516 union_sync, 517 union_vget, 518 (void *)eopnotsupp, /* vfs_fhtovp */ 519 (void *)eopnotsupp, /* vfs_vptofh */ 520 union_init, 521 union_reinit, 522 union_done, 523 NULL, /* vfs_mountroot */ 524 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 525 vfs_stdextattrctl, 526 (void *)eopnotsupp, /* vfs_suspendctl */ 527 union_renamelock_enter, 528 union_renamelock_exit, 529 (void *)eopnotsupp, 530 union_vnodeopv_descs, 531 0, /* vfs_refcount */ 532 { NULL, NULL }, 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, "vfs", NULL, 548 NULL, 0, NULL, 0, 549 CTL_VFS, CTL_EOL); 550 sysctl_createv(&union_sysctl_log, 0, NULL, NULL, 551 CTLFLAG_PERMANENT, 552 CTLTYPE_NODE, "union", 553 SYSCTL_DESCR("Union file system"), 554 NULL, 0, NULL, 0, 555 CTL_VFS, 15, CTL_EOL); 556 /* 557 * XXX the "15" above could be dynamic, thereby eliminating 558 * one more instance of the "number to vfs" mapping problem, 559 * but "15" is the order as taken from sys/mount.h 560 */ 561 break; 562 case MODULE_CMD_FINI: 563 error = vfs_detach(&union_vfsops); 564 if (error != 0) 565 break; 566 sysctl_teardown(&union_sysctl_log); 567 break; 568 default: 569 error = ENOTTY; 570 break; 571 } 572 573 return (error); 574 } 575
Indexes created Mon Oct 13 01:09:56 GMT 2025