1 /*- 2 * Copyright (c) 1992, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software donated to Berkeley by 6 * Jan-Simon Pendry. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 4. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)null_subr.c 8.7 (Berkeley) 5/14/95 33 * 34 * $FreeBSD$ 35 */ 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/kernel.h> 40 #include <sys/lock.h> 41 #include <sys/mutex.h> 42 #include <sys/malloc.h> 43 #include <sys/mount.h> 44 #include <sys/proc.h> 45 #include <sys/vnode.h> 46 47 #include <fs/nullfs/null.h> 48 49 #define LOG2_SIZEVNODE 8 /* log2(sizeof struct vnode) */ 50 #define NNULLNODECACHE 16 51 52 /* 53 * Null layer cache: 54 * Each cache entry holds a reference to the lower vnode 55 * along with a pointer to the alias vnode. When an 56 * entry is added the lower vnode is VREF'd. When the 57 * alias is removed the lower vnode is vrele'd. 58 */ 59 60 #define NULL_NHASH(vp) \ 61 (&null_node_hashtbl[(((uintptr_t)vp)>>LOG2_SIZEVNODE) & null_node_hash]) 62 63 static LIST_HEAD(null_node_hashhead, null_node) *null_node_hashtbl; 64 static u_long null_node_hash; 65 struct mtx null_hashmtx; 66 67 static MALLOC_DEFINE(M_NULLFSHASH, "nullfs_hash", "NULLFS hash table"); 68 MALLOC_DEFINE(M_NULLFSNODE, "nullfs_node", "NULLFS vnode private part"); 69 70 static struct vnode * null_hashget(struct mount *, struct vnode *); 71 static struct vnode * null_hashins(struct mount *, struct null_node *); 72 73 /* 74 * Initialise cache headers 75 */ 76 int 77 nullfs_init(vfsp) 78 struct vfsconf *vfsp; 79 { 80 81 NULLFSDEBUG("nullfs_init\n"); /* printed during system boot */ 82 null_node_hashtbl = hashinit(NNULLNODECACHE, M_NULLFSHASH, &null_node_hash); 83 mtx_init(&null_hashmtx, "nullhs", NULL, MTX_DEF); 84 return (0); 85 } 86 87 int 88 nullfs_uninit(vfsp) 89 struct vfsconf *vfsp; 90 { 91 92 mtx_destroy(&null_hashmtx); 93 free(null_node_hashtbl, M_NULLFSHASH); 94 return (0); 95 } 96 97 /* 98 * Return a VREF'ed alias for lower vnode if already exists, else 0. 99 * Lower vnode should be locked on entry and will be left locked on exit. 100 */ 101 static struct vnode * 102 null_hashget(mp, lowervp) 103 struct mount *mp; 104 struct vnode *lowervp; 105 { 106 struct null_node_hashhead *hd; 107 struct null_node *a; 108 struct vnode *vp; 109 110 ASSERT_VOP_LOCKED(lowervp, "null_hashget"); 111 112 /* 113 * Find hash base, and then search the (two-way) linked 114 * list looking for a null_node structure which is referencing 115 * the lower vnode. If found, the increment the null_node 116 * reference count (but NOT the lower vnode's VREF counter). 117 */ 118 hd = NULL_NHASH(lowervp); 119 mtx_lock(&null_hashmtx); 120 LIST_FOREACH(a, hd, null_hash) { 121 if (a->null_lowervp == lowervp && NULLTOV(a)->v_mount == mp) { 122 /* 123 * Since we have the lower node locked the nullfs 124 * node can not be in the process of recycling. If 125 * it had been recycled before we grabed the lower 126 * lock it would not have been found on the hash. 127 */ 128 vp = NULLTOV(a); 129 vref(vp); 130 mtx_unlock(&null_hashmtx); 131 return (vp); 132 } 133 } 134 mtx_unlock(&null_hashmtx); 135 return (NULLVP); 136 } 137 138 /* 139 * Act like null_hashget, but add passed null_node to hash if no existing 140 * node found. 141 */ 142 static struct vnode * 143 null_hashins(mp, xp) 144 struct mount *mp; 145 struct null_node *xp; 146 { 147 struct null_node_hashhead *hd; 148 struct null_node *oxp; 149 struct vnode *ovp; 150 151 hd = NULL_NHASH(xp->null_lowervp); 152 mtx_lock(&null_hashmtx); 153 LIST_FOREACH(oxp, hd, null_hash) { 154 if (oxp->null_lowervp == xp->null_lowervp && 155 NULLTOV(oxp)->v_mount == mp) { 156 /* 157 * See null_hashget for a description of this 158 * operation. 159 */ 160 ovp = NULLTOV(oxp); 161 vref(ovp); 162 mtx_unlock(&null_hashmtx); 163 return (ovp); 164 } 165 } 166 LIST_INSERT_HEAD(hd, xp, null_hash); 167 mtx_unlock(&null_hashmtx); 168 return (NULLVP); 169 } 170 171 static void 172 null_insmntque_dtr(struct vnode *vp, void *xp) 173 { 174 vp->v_data = NULL; 175 vp->v_vnlock = &vp->v_lock; 176 FREE(xp, M_NULLFSNODE); 177 vp->v_op = &dead_vnodeops; 178 (void) vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 179 vgone(vp); 180 vput(vp); 181 } 182 183 /* 184 * Make a new or get existing nullfs node. 185 * Vp is the alias vnode, lowervp is the lower vnode. 186 * 187 * The lowervp assumed to be locked and having "spare" reference. This routine 188 * vrele lowervp if nullfs node was taken from hash. Otherwise it "transfers" 189 * the caller's "spare" reference to created nullfs vnode. 190 */ 191 int 192 null_nodeget(mp, lowervp, vpp) 193 struct mount *mp; 194 struct vnode *lowervp; 195 struct vnode **vpp; 196 { 197 struct null_node *xp; 198 struct vnode *vp; 199 int error; 200 201 /* Lookup the hash firstly */ 202 *vpp = null_hashget(mp, lowervp); 203 if (*vpp != NULL) { 204 vrele(lowervp); 205 return (0); 206 } 207 208 /* 209 * We do not serialize vnode creation, instead we will check for 210 * duplicates later, when adding new vnode to hash. 211 * 212 * Note that duplicate can only appear in hash if the lowervp is 213 * locked LK_SHARED. 214 */ 215 216 /* 217 * Do the MALLOC before the getnewvnode since doing so afterward 218 * might cause a bogus v_data pointer to get dereferenced 219 * elsewhere if MALLOC should block. 220 */ 221 MALLOC(xp, struct null_node *, sizeof(struct null_node), 222 M_NULLFSNODE, M_WAITOK); 223 224 error = getnewvnode("null", mp, &null_vnodeops, &vp); 225 if (error) { 226 FREE(xp, M_NULLFSNODE); 227 return (error); 228 } 229 230 xp->null_vnode = vp; 231 xp->null_lowervp = lowervp; 232 vp->v_type = lowervp->v_type; 233 vp->v_data = xp; 234 vp->v_vnlock = lowervp->v_vnlock; 235 if (vp->v_vnlock == NULL) 236 panic("null_nodeget: Passed a NULL vnlock.\n"); 237 error = insmntque1(vp, mp, null_insmntque_dtr, xp); 238 if (error != 0) 239 return (error); 240 /* 241 * Atomically insert our new node into the hash or vget existing 242 * if someone else has beaten us to it. 243 */ 244 *vpp = null_hashins(mp, xp); 245 if (*vpp != NULL) { 246 vrele(lowervp); 247 vp->v_vnlock = &vp->v_lock; 248 xp->null_lowervp = NULL; 249 vrele(vp); 250 return (0); 251 } 252 *vpp = vp; 253 254 return (0); 255 } 256 257 /* 258 * Remove node from hash. 259 */ 260 void 261 null_hashrem(xp) 262 struct null_node *xp; 263 { 264 265 mtx_lock(&null_hashmtx); 266 LIST_REMOVE(xp, null_hash); 267 mtx_unlock(&null_hashmtx); 268 } 269 270 #ifdef DIAGNOSTIC 271 272 #ifdef KDB 273 #define null_checkvp_barrier 1 274 #else 275 #define null_checkvp_barrier 0 276 #endif 277 278 struct vnode * 279 null_checkvp(vp, fil, lno) 280 struct vnode *vp; 281 char *fil; 282 int lno; 283 { 284 int interlock = 0; 285 struct null_node *a = VTONULL(vp); 286 #ifdef notyet 287 /* 288 * Can't do this check because vop_reclaim runs 289 * with a funny vop vector. 290 */ 291 if (vp->v_op != null_vnodeop_p) { 292 printf ("null_checkvp: on non-null-node\n"); 293 while (null_checkvp_barrier) /*WAIT*/ ; 294 panic("null_checkvp"); 295 }; 296 #endif 297 if (a->null_lowervp == NULLVP) { 298 /* Should never happen */ 299 int i; u_long *p; 300 printf("vp = %p, ZERO ptr\n", (void *)vp); 301 for (p = (u_long *) a, i = 0; i < 8; i++) 302 printf(" %lx", p[i]); 303 printf("\n"); 304 /* wait for debugger */ 305 while (null_checkvp_barrier) /*WAIT*/ ; 306 panic("null_checkvp"); 307 } 308 if (mtx_owned(VI_MTX(vp)) != 0) { 309 VI_UNLOCK(vp); 310 interlock = 1; 311 } 312 if (vrefcnt(a->null_lowervp) < 1) { 313 int i; u_long *p; 314 printf("vp = %p, unref'ed lowervp\n", (void *)vp); 315 for (p = (u_long *) a, i = 0; i < 8; i++) 316 printf(" %lx", p[i]); 317 printf("\n"); 318 /* wait for debugger */ 319 while (null_checkvp_barrier) /*WAIT*/ ; 320 panic ("null with unref'ed lowervp"); 321 }; 322 if (interlock != 0) 323 VI_LOCK(vp); 324 #ifdef notyet 325 printf("null %x/%d -> %x/%d [%s, %d]\n", 326 NULLTOV(a), vrefcnt(NULLTOV(a)), 327 a->null_lowervp, vrefcnt(a->null_lowervp), 328 fil, lno); 329 #endif 330 return a->null_lowervp; 331 } 332 #endif 333