1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1992, 1993, 1995 5 * The Regents of the University of California. 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 * @(#)null_vfsops.c 8.2 (Berkeley) 1/21/94 35 * 36 * @(#)lofs_vfsops.c 1.2 (Berkeley) 6/18/92 37 * $FreeBSD$ 38 */ 39 40 /* 41 * Null Layer 42 * (See null_vnops.c for a description of what this does.) 43 */ 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/fcntl.h> 48 #include <sys/kernel.h> 49 #include <sys/lock.h> 50 #include <sys/malloc.h> 51 #include <sys/mount.h> 52 #include <sys/namei.h> 53 #include <sys/proc.h> 54 #include <sys/vnode.h> 55 #include <sys/jail.h> 56 57 #include <fs/nullfs/null.h> 58 59 static MALLOC_DEFINE(M_NULLFSMNT, "nullfs_mount", "NULLFS mount structure"); 60 61 static vfs_fhtovp_t nullfs_fhtovp; 62 static vfs_mount_t nullfs_mount; 63 static vfs_quotactl_t nullfs_quotactl; 64 static vfs_root_t nullfs_root; 65 static vfs_sync_t nullfs_sync; 66 static vfs_statfs_t nullfs_statfs; 67 static vfs_unmount_t nullfs_unmount; 68 static vfs_vget_t nullfs_vget; 69 static vfs_extattrctl_t nullfs_extattrctl; 70 71 /* 72 * Mount null layer 73 */ 74 static int 75 nullfs_mount(struct mount *mp) 76 { 77 struct vnode *lowerrootvp; 78 struct vnode *nullm_rootvp; 79 struct null_mount *xmp; 80 struct null_node *nn; 81 struct nameidata nd, *ndp; 82 char *target; 83 int error, len; 84 bool isvnunlocked; 85 86 NULLFSDEBUG("nullfs_mount(mp = %p)\n", (void *)mp); 87 88 if (mp->mnt_flag & MNT_ROOTFS) 89 return (EOPNOTSUPP); 90 91 /* 92 * Update is a no-op 93 */ 94 if (mp->mnt_flag & MNT_UPDATE) { 95 /* 96 * Only support update mounts for NFS export. 97 */ 98 if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0)) 99 return (0); 100 else 101 return (EOPNOTSUPP); 102 } 103 104 /* 105 * Get argument 106 */ 107 error = vfs_getopt(mp->mnt_optnew, "from", (void **)&target, &len); 108 if (error != 0) 109 error = vfs_getopt(mp->mnt_optnew, "target", (void **)&target, &len); 110 if (error || target[len - 1] != '\0') 111 return (EINVAL); 112 113 /* 114 * Unlock lower node to avoid possible deadlock. 115 */ 116 if (mp->mnt_vnodecovered->v_op == &null_vnodeops && 117 VOP_ISLOCKED(mp->mnt_vnodecovered) == LK_EXCLUSIVE) { 118 VOP_UNLOCK(mp->mnt_vnodecovered); 119 isvnunlocked = true; 120 } else { 121 isvnunlocked = false; 122 } 123 124 /* 125 * Find lower node 126 */ 127 ndp = &nd; 128 NDINIT(ndp, LOOKUP, FOLLOW|LOCKLEAF, UIO_SYSSPACE, target); 129 error = namei(ndp); 130 131 /* 132 * Re-lock vnode. 133 * XXXKIB This is deadlock-prone as well. 134 */ 135 if (isvnunlocked) 136 vn_lock(mp->mnt_vnodecovered, LK_EXCLUSIVE | LK_RETRY); 137 138 if (error) 139 return (error); 140 NDFREE_PNBUF(ndp); 141 142 /* 143 * Sanity check on lower vnode 144 */ 145 lowerrootvp = ndp->ni_vp; 146 147 /* 148 * Check multi null mount to avoid `lock against myself' panic. 149 */ 150 if (mp->mnt_vnodecovered->v_op == &null_vnodeops) { 151 nn = VTONULL(mp->mnt_vnodecovered); 152 if (nn == NULL || lowerrootvp == nn->null_lowervp) { 153 NULLFSDEBUG("nullfs_mount: multi null mount?\n"); 154 vput(lowerrootvp); 155 return (EDEADLK); 156 } 157 } 158 159 /* 160 * Lower vnode must be the same type as the covered vnode - we 161 * don't allow mounting directories to files or vice versa. 162 */ 163 if ((lowerrootvp->v_type != VDIR && lowerrootvp->v_type != VREG) || 164 lowerrootvp->v_type != mp->mnt_vnodecovered->v_type) { 165 NULLFSDEBUG("nullfs_mount: target must be same type as fspath"); 166 vput(lowerrootvp); 167 return (EINVAL); 168 } 169 170 xmp = (struct null_mount *) malloc(sizeof(struct null_mount), 171 M_NULLFSMNT, M_WAITOK | M_ZERO); 172 173 /* 174 * Save pointer to underlying FS and the reference to the 175 * lower root vnode. 176 */ 177 xmp->nullm_vfs = vfs_register_upper_from_vp(lowerrootvp, mp, 178 &xmp->upper_node); 179 if (xmp->nullm_vfs == NULL) { 180 vput(lowerrootvp); 181 free(xmp, M_NULLFSMNT); 182 return (ENOENT); 183 } 184 vref(lowerrootvp); 185 xmp->nullm_lowerrootvp = lowerrootvp; 186 mp->mnt_data = xmp; 187 188 /* 189 * Make sure the node alias worked. 190 */ 191 error = null_nodeget(mp, lowerrootvp, &nullm_rootvp); 192 if (error != 0) { 193 vfs_unregister_upper(xmp->nullm_vfs, &xmp->upper_node); 194 vrele(lowerrootvp); 195 free(xmp, M_NULLFSMNT); 196 return (error); 197 } 198 199 if (NULLVPTOLOWERVP(nullm_rootvp)->v_mount->mnt_flag & MNT_LOCAL) { 200 MNT_ILOCK(mp); 201 mp->mnt_flag |= MNT_LOCAL; 202 MNT_IUNLOCK(mp); 203 } 204 205 xmp->nullm_flags |= NULLM_CACHE; 206 if (vfs_getopt(mp->mnt_optnew, "nocache", NULL, NULL) == 0 || 207 (xmp->nullm_vfs->mnt_kern_flag & MNTK_NULL_NOCACHE) != 0) 208 xmp->nullm_flags &= ~NULLM_CACHE; 209 210 if ((xmp->nullm_flags & NULLM_CACHE) != 0) { 211 vfs_register_for_notification(xmp->nullm_vfs, mp, 212 &xmp->notify_node); 213 } 214 215 if (lowerrootvp == mp->mnt_vnodecovered) { 216 vn_lock(lowerrootvp, LK_EXCLUSIVE | LK_RETRY | LK_CANRECURSE); 217 lowerrootvp->v_vflag |= VV_CROSSLOCK; 218 VOP_UNLOCK(lowerrootvp); 219 } 220 221 MNT_ILOCK(mp); 222 if ((xmp->nullm_flags & NULLM_CACHE) != 0) { 223 mp->mnt_kern_flag |= lowerrootvp->v_mount->mnt_kern_flag & 224 (MNTK_SHARED_WRITES | MNTK_LOOKUP_SHARED | 225 MNTK_EXTENDED_SHARED); 226 } 227 mp->mnt_kern_flag |= MNTK_NOMSYNC | MNTK_UNLOCKED_INSMNTQUE; 228 mp->mnt_kern_flag |= lowerrootvp->v_mount->mnt_kern_flag & 229 (MNTK_USES_BCACHE | MNTK_NO_IOPF | MNTK_UNMAPPED_BUFS); 230 MNT_IUNLOCK(mp); 231 vfs_getnewfsid(mp); 232 vfs_mountedfrom(mp, target); 233 vput(nullm_rootvp); 234 235 NULLFSDEBUG("nullfs_mount: lower %s, alias at %s\n", 236 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname); 237 return (0); 238 } 239 240 /* 241 * Free reference to null layer 242 */ 243 static int 244 nullfs_unmount(mp, mntflags) 245 struct mount *mp; 246 int mntflags; 247 { 248 struct null_mount *mntdata; 249 int error, flags; 250 251 NULLFSDEBUG("nullfs_unmount: mp = %p\n", (void *)mp); 252 253 if (mntflags & MNT_FORCE) 254 flags = FORCECLOSE; 255 else 256 flags = 0; 257 258 for (;;) { 259 /* There is 1 extra root vnode reference (nullm_rootvp). */ 260 error = vflush(mp, 0, flags, curthread); 261 if (error) 262 return (error); 263 MNT_ILOCK(mp); 264 if (mp->mnt_nvnodelistsize == 0) { 265 MNT_IUNLOCK(mp); 266 break; 267 } 268 MNT_IUNLOCK(mp); 269 if ((mntflags & MNT_FORCE) == 0) 270 return (EBUSY); 271 } 272 273 /* 274 * Finally, throw away the null_mount structure 275 */ 276 mntdata = mp->mnt_data; 277 if ((mntdata->nullm_flags & NULLM_CACHE) != 0) { 278 vfs_unregister_for_notification(mntdata->nullm_vfs, 279 &mntdata->notify_node); 280 } 281 if (mntdata->nullm_lowerrootvp == mp->mnt_vnodecovered) { 282 vn_lock(mp->mnt_vnodecovered, LK_EXCLUSIVE | LK_RETRY | LK_CANRECURSE); 283 mp->mnt_vnodecovered->v_vflag &= ~VV_CROSSLOCK; 284 VOP_UNLOCK(mp->mnt_vnodecovered); 285 } 286 vfs_unregister_upper(mntdata->nullm_vfs, &mntdata->upper_node); 287 vrele(mntdata->nullm_lowerrootvp); 288 mp->mnt_data = NULL; 289 free(mntdata, M_NULLFSMNT); 290 return (0); 291 } 292 293 static int 294 nullfs_root(mp, flags, vpp) 295 struct mount *mp; 296 int flags; 297 struct vnode **vpp; 298 { 299 struct vnode *vp; 300 struct null_mount *mntdata; 301 int error; 302 303 mntdata = MOUNTTONULLMOUNT(mp); 304 NULLFSDEBUG("nullfs_root(mp = %p, vp = %p)\n", mp, 305 mntdata->nullm_lowerrootvp); 306 307 error = vget(mntdata->nullm_lowerrootvp, flags); 308 if (error == 0) { 309 error = null_nodeget(mp, mntdata->nullm_lowerrootvp, &vp); 310 if (error == 0) { 311 *vpp = vp; 312 } 313 } 314 return (error); 315 } 316 317 static int 318 nullfs_quotactl(mp, cmd, uid, arg, mp_busy) 319 struct mount *mp; 320 int cmd; 321 uid_t uid; 322 void *arg; 323 bool *mp_busy; 324 { 325 struct mount *lowermp; 326 struct null_mount *mntdata; 327 int error; 328 bool unbusy; 329 330 mntdata = MOUNTTONULLMOUNT(mp); 331 lowermp = atomic_load_ptr(&mntdata->nullm_vfs); 332 KASSERT(*mp_busy == true, ("upper mount not busy")); 333 /* 334 * See comment in sys_quotactl() for an explanation of why the 335 * lower mount needs to be busied by the caller of VFS_QUOTACTL() 336 * but may be unbusied by the implementation. We must unbusy 337 * the upper mount for the same reason; otherwise a namei lookup 338 * issued by the VFS_QUOTACTL() implementation could traverse the 339 * upper mount and deadlock. 340 */ 341 vfs_unbusy(mp); 342 *mp_busy = false; 343 unbusy = true; 344 error = vfs_busy(lowermp, 0); 345 if (error == 0) 346 error = VFS_QUOTACTL(lowermp, cmd, uid, arg, &unbusy); 347 if (unbusy) 348 vfs_unbusy(lowermp); 349 350 return (error); 351 } 352 353 static int 354 nullfs_statfs(mp, sbp) 355 struct mount *mp; 356 struct statfs *sbp; 357 { 358 int error; 359 struct statfs *mstat; 360 361 NULLFSDEBUG("nullfs_statfs(mp = %p, vp = %p->%p)\n", (void *)mp, 362 (void *)MOUNTTONULLMOUNT(mp)->nullm_rootvp, 363 (void *)NULLVPTOLOWERVP(MOUNTTONULLMOUNT(mp)->nullm_rootvp)); 364 365 mstat = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK | M_ZERO); 366 367 error = VFS_STATFS(MOUNTTONULLMOUNT(mp)->nullm_vfs, mstat); 368 if (error) { 369 free(mstat, M_STATFS); 370 return (error); 371 } 372 373 /* now copy across the "interesting" information and fake the rest */ 374 sbp->f_type = mstat->f_type; 375 sbp->f_flags = (sbp->f_flags & (MNT_RDONLY | MNT_NOEXEC | MNT_NOSUID | 376 MNT_UNION | MNT_NOSYMFOLLOW | MNT_AUTOMOUNTED)) | 377 (mstat->f_flags & ~(MNT_ROOTFS | MNT_AUTOMOUNTED)); 378 sbp->f_bsize = mstat->f_bsize; 379 sbp->f_iosize = mstat->f_iosize; 380 sbp->f_blocks = mstat->f_blocks; 381 sbp->f_bfree = mstat->f_bfree; 382 sbp->f_bavail = mstat->f_bavail; 383 sbp->f_files = mstat->f_files; 384 sbp->f_ffree = mstat->f_ffree; 385 386 free(mstat, M_STATFS); 387 return (0); 388 } 389 390 static int 391 nullfs_sync(mp, waitfor) 392 struct mount *mp; 393 int waitfor; 394 { 395 /* 396 * XXX - Assumes no data cached at null layer. 397 */ 398 return (0); 399 } 400 401 static int 402 nullfs_vget(mp, ino, flags, vpp) 403 struct mount *mp; 404 ino_t ino; 405 int flags; 406 struct vnode **vpp; 407 { 408 int error; 409 410 KASSERT((flags & LK_TYPE_MASK) != 0, 411 ("nullfs_vget: no lock requested")); 412 413 error = VFS_VGET(MOUNTTONULLMOUNT(mp)->nullm_vfs, ino, flags, vpp); 414 if (error != 0) 415 return (error); 416 return (null_nodeget(mp, *vpp, vpp)); 417 } 418 419 static int 420 nullfs_fhtovp(mp, fidp, flags, vpp) 421 struct mount *mp; 422 struct fid *fidp; 423 int flags; 424 struct vnode **vpp; 425 { 426 int error; 427 428 error = VFS_FHTOVP(MOUNTTONULLMOUNT(mp)->nullm_vfs, fidp, flags, 429 vpp); 430 if (error != 0) 431 return (error); 432 return (null_nodeget(mp, *vpp, vpp)); 433 } 434 435 static int 436 nullfs_extattrctl(mp, cmd, filename_vp, namespace, attrname) 437 struct mount *mp; 438 int cmd; 439 struct vnode *filename_vp; 440 int namespace; 441 const char *attrname; 442 { 443 444 return (VFS_EXTATTRCTL(MOUNTTONULLMOUNT(mp)->nullm_vfs, cmd, 445 filename_vp, namespace, attrname)); 446 } 447 448 static void 449 nullfs_reclaim_lowervp(struct mount *mp, struct vnode *lowervp) 450 { 451 struct vnode *vp; 452 453 vp = null_hashget(mp, lowervp); 454 if (vp == NULL) 455 return; 456 VTONULL(vp)->null_flags |= NULLV_NOUNLOCK; 457 vgone(vp); 458 vput(vp); 459 } 460 461 static void 462 nullfs_unlink_lowervp(struct mount *mp, struct vnode *lowervp) 463 { 464 struct vnode *vp; 465 struct null_node *xp; 466 467 vp = null_hashget(mp, lowervp); 468 if (vp == NULL) 469 return; 470 xp = VTONULL(vp); 471 xp->null_flags |= NULLV_DROP | NULLV_NOUNLOCK; 472 vhold(vp); 473 vunref(vp); 474 475 if (vp->v_usecount == 0) { 476 /* 477 * If vunref() dropped the last use reference on the 478 * nullfs vnode, it must be reclaimed, and its lock 479 * was split from the lower vnode lock. Need to do 480 * extra unlock before allowing the final vdrop() to 481 * free the vnode. 482 */ 483 KASSERT(VN_IS_DOOMED(vp), 484 ("not reclaimed nullfs vnode %p", vp)); 485 VOP_UNLOCK(vp); 486 } else { 487 /* 488 * Otherwise, the nullfs vnode still shares the lock 489 * with the lower vnode, and must not be unlocked. 490 * Also clear the NULLV_NOUNLOCK, the flag is not 491 * relevant for future reclamations. 492 */ 493 ASSERT_VOP_ELOCKED(vp, "unlink_lowervp"); 494 KASSERT(!VN_IS_DOOMED(vp), 495 ("reclaimed nullfs vnode %p", vp)); 496 xp->null_flags &= ~NULLV_NOUNLOCK; 497 } 498 vdrop(vp); 499 } 500 501 static struct vfsops null_vfsops = { 502 .vfs_extattrctl = nullfs_extattrctl, 503 .vfs_fhtovp = nullfs_fhtovp, 504 .vfs_init = nullfs_init, 505 .vfs_mount = nullfs_mount, 506 .vfs_quotactl = nullfs_quotactl, 507 .vfs_root = nullfs_root, 508 .vfs_statfs = nullfs_statfs, 509 .vfs_sync = nullfs_sync, 510 .vfs_uninit = nullfs_uninit, 511 .vfs_unmount = nullfs_unmount, 512 .vfs_vget = nullfs_vget, 513 .vfs_reclaim_lowervp = nullfs_reclaim_lowervp, 514 .vfs_unlink_lowervp = nullfs_unlink_lowervp, 515 }; 516 517 VFS_SET(null_vfsops, nullfs, VFCF_LOOPBACK | VFCF_JAIL | VFCF_FILEMOUNT); 518