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