1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1992, 1993 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_subr.c 8.7 (Berkeley) 5/14/95 35 * 36 * $FreeBSD$ 37 */ 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/kernel.h> 42 #include <sys/lock.h> 43 #include <sys/rwlock.h> 44 #include <sys/malloc.h> 45 #include <sys/mount.h> 46 #include <sys/proc.h> 47 #include <sys/vnode.h> 48 49 #include <fs/nullfs/null.h> 50 51 /* 52 * Null layer cache: 53 * Each cache entry holds a reference to the lower vnode 54 * along with a pointer to the alias vnode. When an 55 * entry is added the lower vnode is VREF'd. When the 56 * alias is removed the lower vnode is vrele'd. 57 */ 58 59 #define NULL_NHASH(vp) (&null_node_hashtbl[vfs_hash_index(vp) & null_hash_mask]) 60 61 static LIST_HEAD(null_node_hashhead, null_node) *null_node_hashtbl; 62 static struct rwlock null_hash_lock; 63 static u_long null_hash_mask; 64 65 static MALLOC_DEFINE(M_NULLFSHASH, "nullfs_hash", "NULLFS hash table"); 66 MALLOC_DEFINE(M_NULLFSNODE, "nullfs_node", "NULLFS vnode private part"); 67 68 static struct vnode * null_hashins(struct mount *, struct null_node *); 69 70 /* 71 * Initialise cache headers 72 */ 73 int 74 nullfs_init(vfsp) 75 struct vfsconf *vfsp; 76 { 77 78 null_node_hashtbl = hashinit(desiredvnodes, M_NULLFSHASH, 79 &null_hash_mask); 80 rw_init(&null_hash_lock, "nullhs"); 81 return (0); 82 } 83 84 int 85 nullfs_uninit(vfsp) 86 struct vfsconf *vfsp; 87 { 88 89 rw_destroy(&null_hash_lock); 90 hashdestroy(null_node_hashtbl, M_NULLFSHASH, null_hash_mask); 91 return (0); 92 } 93 94 /* 95 * Return a VREF'ed alias for lower vnode if already exists, else 0. 96 * Lower vnode should be locked on entry and will be left locked on exit. 97 */ 98 struct vnode * 99 null_hashget(mp, lowervp) 100 struct mount *mp; 101 struct vnode *lowervp; 102 { 103 struct null_node_hashhead *hd; 104 struct null_node *a; 105 struct vnode *vp; 106 107 ASSERT_VOP_LOCKED(lowervp, "null_hashget"); 108 109 /* 110 * Find hash base, and then search the (two-way) linked 111 * list looking for a null_node structure which is referencing 112 * the lower vnode. If found, the increment the null_node 113 * reference count (but NOT the lower vnode's VREF counter). 114 */ 115 hd = NULL_NHASH(lowervp); 116 if (LIST_EMPTY(hd)) 117 return (NULLVP); 118 rw_rlock(&null_hash_lock); 119 LIST_FOREACH(a, hd, null_hash) { 120 if (a->null_lowervp == lowervp && NULLTOV(a)->v_mount == mp) { 121 /* 122 * Since we have the lower node locked the nullfs 123 * node can not be in the process of recycling. If 124 * it had been recycled before we grabed the lower 125 * lock it would not have been found on the hash. 126 */ 127 vp = NULLTOV(a); 128 vref(vp); 129 rw_runlock(&null_hash_lock); 130 return (vp); 131 } 132 } 133 rw_runlock(&null_hash_lock); 134 return (NULLVP); 135 } 136 137 /* 138 * Act like null_hashget, but add passed null_node to hash if no existing 139 * node found. 140 */ 141 static struct vnode * 142 null_hashins(mp, xp) 143 struct mount *mp; 144 struct null_node *xp; 145 { 146 struct null_node_hashhead *hd; 147 struct null_node *oxp; 148 struct vnode *ovp; 149 150 hd = NULL_NHASH(xp->null_lowervp); 151 rw_wlock(&null_hash_lock); 152 LIST_FOREACH(oxp, hd, null_hash) { 153 if (oxp->null_lowervp == xp->null_lowervp && 154 NULLTOV(oxp)->v_mount == mp) { 155 /* 156 * See null_hashget for a description of this 157 * operation. 158 */ 159 ovp = NULLTOV(oxp); 160 vref(ovp); 161 rw_wunlock(&null_hash_lock); 162 return (ovp); 163 } 164 } 165 LIST_INSERT_HEAD(hd, xp, null_hash); 166 rw_wunlock(&null_hash_lock); 167 return (NULLVP); 168 } 169 170 static void 171 null_destroy_proto(struct vnode *vp, void *xp) 172 { 173 174 lockmgr(&vp->v_lock, LK_EXCLUSIVE, NULL); 175 VI_LOCK(vp); 176 vp->v_data = NULL; 177 vp->v_vnlock = &vp->v_lock; 178 vp->v_op = &dead_vnodeops; 179 VI_UNLOCK(vp); 180 vgone(vp); 181 vput(vp); 182 free(xp, M_NULLFSNODE); 183 } 184 185 static void 186 null_insmntque_dtr(struct vnode *vp, void *xp) 187 { 188 189 vput(((struct null_node *)xp)->null_lowervp); 190 null_destroy_proto(vp, xp); 191 } 192 193 /* 194 * Make a new or get existing nullfs node. 195 * Vp is the alias vnode, lowervp is the lower vnode. 196 * 197 * The lowervp assumed to be locked and having "spare" reference. This routine 198 * vrele lowervp if nullfs node was taken from hash. Otherwise it "transfers" 199 * the caller's "spare" reference to created nullfs vnode. 200 */ 201 int 202 null_nodeget(mp, lowervp, vpp) 203 struct mount *mp; 204 struct vnode *lowervp; 205 struct vnode **vpp; 206 { 207 struct null_node *xp; 208 struct vnode *vp; 209 int error; 210 211 ASSERT_VOP_LOCKED(lowervp, "lowervp"); 212 VNPASS(lowervp->v_usecount > 0, lowervp); 213 214 /* Lookup the hash firstly. */ 215 *vpp = null_hashget(mp, lowervp); 216 if (*vpp != NULL) { 217 vrele(lowervp); 218 return (0); 219 } 220 221 /* 222 * The insmntque1() call below requires the exclusive lock on 223 * the nullfs vnode. Upgrade the lock now if hash failed to 224 * provide ready to use vnode. 225 */ 226 if (VOP_ISLOCKED(lowervp) != LK_EXCLUSIVE) { 227 vn_lock(lowervp, LK_UPGRADE | LK_RETRY); 228 if (VN_IS_DOOMED(lowervp)) { 229 vput(lowervp); 230 return (ENOENT); 231 } 232 } 233 234 /* 235 * We do not serialize vnode creation, instead we will check for 236 * duplicates later, when adding new vnode to hash. 237 * Note that duplicate can only appear in hash if the lowervp is 238 * locked LK_SHARED. 239 */ 240 xp = malloc(sizeof(struct null_node), M_NULLFSNODE, M_WAITOK); 241 242 error = getnewvnode("nullfs", mp, &null_vnodeops, &vp); 243 if (error) { 244 vput(lowervp); 245 free(xp, M_NULLFSNODE); 246 return (error); 247 } 248 249 xp->null_vnode = vp; 250 xp->null_lowervp = lowervp; 251 xp->null_flags = 0; 252 vp->v_type = lowervp->v_type; 253 vp->v_data = xp; 254 vp->v_vnlock = lowervp->v_vnlock; 255 error = insmntque1(vp, mp, null_insmntque_dtr, xp); 256 if (error != 0) 257 return (error); 258 if (lowervp == MOUNTTONULLMOUNT(mp)->nullm_lowerrootvp) 259 vp->v_vflag |= VV_ROOT; 260 261 /* 262 * We might miss the case where lower vnode sets VIRF_PGREAD 263 * some time after construction, which is typical case. 264 * null_open rechecks. 265 */ 266 if ((vn_irflag_read(lowervp) & VIRF_PGREAD) != 0) { 267 MPASS(lowervp->v_object != NULL); 268 if ((vn_irflag_read(vp) & VIRF_PGREAD) == 0) { 269 if (vp->v_object == NULL) 270 vp->v_object = lowervp->v_object; 271 else 272 MPASS(vp->v_object == lowervp->v_object); 273 vn_irflag_set_cond(vp, VIRF_PGREAD); 274 } else { 275 MPASS(vp->v_object != NULL); 276 } 277 } 278 279 /* 280 * Atomically insert our new node into the hash or vget existing 281 * if someone else has beaten us to it. 282 */ 283 *vpp = null_hashins(mp, xp); 284 if (*vpp != NULL) { 285 vrele(lowervp); 286 vp->v_object = NULL; /* in case VIRF_PGREAD set it */ 287 null_destroy_proto(vp, xp); 288 return (0); 289 } 290 *vpp = vp; 291 292 return (0); 293 } 294 295 /* 296 * Remove node from hash. 297 */ 298 void 299 null_hashrem(xp) 300 struct null_node *xp; 301 { 302 303 rw_wlock(&null_hash_lock); 304 LIST_REMOVE(xp, null_hash); 305 rw_wunlock(&null_hash_lock); 306 } 307 308 #ifdef DIAGNOSTIC 309 310 struct vnode * 311 null_checkvp(vp, fil, lno) 312 struct vnode *vp; 313 char *fil; 314 int lno; 315 { 316 struct null_node *a = VTONULL(vp); 317 318 #ifdef notyet 319 /* 320 * Can't do this check because vop_reclaim runs 321 * with a funny vop vector. 322 */ 323 if (vp->v_op != null_vnodeop_p) { 324 printf ("null_checkvp: on non-null-node\n"); 325 panic("null_checkvp"); 326 } 327 #endif 328 if (a->null_lowervp == NULLVP) { 329 /* Should never happen */ 330 panic("null_checkvp %p", vp); 331 } 332 VI_LOCK_FLAGS(a->null_lowervp, MTX_DUPOK); 333 if (a->null_lowervp->v_usecount < 1) 334 panic ("null with unref'ed lowervp, vp %p lvp %p", 335 vp, a->null_lowervp); 336 VI_UNLOCK(a->null_lowervp); 337 #ifdef notyet 338 printf("null %x/%d -> %x/%d [%s, %d]\n", 339 NULLTOV(a), vrefcnt(NULLTOV(a)), 340 a->null_lowervp, vrefcnt(a->null_lowervp), 341 fil, lno); 342 #endif 343 return (a->null_lowervp); 344 } 345 #endif 346