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 thread *td = curthread; /* XXX */ 107 struct null_node_hashhead *hd; 108 struct null_node *a; 109 struct vnode *vp; 110 111 /* 112 * Find hash base, and then search the (two-way) linked 113 * list looking for a null_node structure which is referencing 114 * the lower vnode. If found, the increment the null_node 115 * reference count (but NOT the lower vnode's VREF counter). 116 */ 117 hd = NULL_NHASH(lowervp); 118 loop: 119 mtx_lock(&null_hashmtx); 120 LIST_FOREACH(a, hd, null_hash) { 121 if (a->null_lowervp == lowervp && NULLTOV(a)->v_mount == mp) { 122 vp = NULLTOV(a); 123 VI_LOCK(vp); 124 /* 125 * If the nullfs node is being recycled we have 126 * to wait until it finishes prior to scanning 127 * again. 128 */ 129 mtx_unlock(&null_hashmtx); 130 if ((vp->v_iflag & VI_DOOMED) != 0) { 131 /* Wait for recycling to finish. */ 132 VOP_LOCK(vp, LK_EXCLUSIVE|LK_INTERLOCK, td); 133 VOP_UNLOCK(vp, 0, td); 134 goto loop; 135 } 136 /* 137 * We need to clear the OWEINACT flag here as this 138 * may lead vget() to try to lock our vnode which 139 * is already locked via lowervp. 140 */ 141 vp->v_iflag &= ~VI_OWEINACT; 142 vget(vp, LK_INTERLOCK, td); 143 return (vp); 144 } 145 } 146 mtx_unlock(&null_hashmtx); 147 return (NULLVP); 148 } 149 150 /* 151 * Act like null_hashget, but add passed null_node to hash if no existing 152 * node found. 153 */ 154 static struct vnode * 155 null_hashins(mp, xp) 156 struct mount *mp; 157 struct null_node *xp; 158 { 159 struct thread *td = curthread; /* XXX */ 160 struct null_node_hashhead *hd; 161 struct null_node *oxp; 162 struct vnode *ovp; 163 164 hd = NULL_NHASH(xp->null_lowervp); 165 loop: 166 mtx_lock(&null_hashmtx); 167 LIST_FOREACH(oxp, hd, null_hash) { 168 if (oxp->null_lowervp == xp->null_lowervp && 169 NULLTOV(oxp)->v_mount == mp) { 170 ovp = NULLTOV(oxp); 171 VI_LOCK(ovp); 172 /* 173 * If the nullfs node is being recycled we have 174 * to wait until it finishes prior to scanning 175 * again. 176 */ 177 mtx_unlock(&null_hashmtx); 178 if ((ovp->v_iflag & VI_DOOMED) != 0) { 179 VOP_LOCK(ovp, LK_EXCLUSIVE|LK_INTERLOCK, td); 180 VOP_UNLOCK(ovp, 0, td); 181 goto loop; 182 } 183 ovp->v_iflag &= ~VI_OWEINACT; /* see hashget comment */ 184 vget(ovp, LK_INTERLOCK, td); 185 return (ovp); 186 } 187 } 188 LIST_INSERT_HEAD(hd, xp, null_hash); 189 mtx_unlock(&null_hashmtx); 190 return (NULLVP); 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 /* Lookup the hash firstly */ 212 *vpp = null_hashget(mp, lowervp); 213 if (*vpp != NULL) { 214 vrele(lowervp); 215 return (0); 216 } 217 218 /* 219 * We do not serialize vnode creation, instead we will check for 220 * duplicates later, when adding new vnode to hash. 221 * 222 * Note that duplicate can only appear in hash if the lowervp is 223 * locked LK_SHARED. 224 */ 225 226 /* 227 * Do the MALLOC before the getnewvnode since doing so afterward 228 * might cause a bogus v_data pointer to get dereferenced 229 * elsewhere if MALLOC should block. 230 */ 231 MALLOC(xp, struct null_node *, sizeof(struct null_node), 232 M_NULLFSNODE, M_WAITOK); 233 234 error = getnewvnode("null", mp, &null_vnodeops, &vp); 235 if (error) { 236 FREE(xp, M_NULLFSNODE); 237 return (error); 238 } 239 240 xp->null_vnode = vp; 241 xp->null_lowervp = lowervp; 242 vp->v_type = lowervp->v_type; 243 vp->v_data = xp; 244 vp->v_vnlock = lowervp->v_vnlock; 245 if (vp->v_vnlock == NULL) 246 panic("null_nodeget: Passed a NULL vnlock.\n"); 247 /* 248 * Atomically insert our new node into the hash or vget existing 249 * if someone else has beaten us to it. 250 */ 251 *vpp = null_hashins(mp, xp); 252 if (*vpp != NULL) { 253 vrele(lowervp); 254 vp->v_vnlock = &vp->v_lock; 255 xp->null_lowervp = NULL; 256 vrele(vp); 257 return (0); 258 } 259 *vpp = vp; 260 261 return (0); 262 } 263 264 /* 265 * Remove node from hash. 266 */ 267 void 268 null_hashrem(xp) 269 struct null_node *xp; 270 { 271 272 mtx_lock(&null_hashmtx); 273 LIST_REMOVE(xp, null_hash); 274 mtx_unlock(&null_hashmtx); 275 } 276 277 #ifdef DIAGNOSTIC 278 279 #ifdef KDB 280 #define null_checkvp_barrier 1 281 #else 282 #define null_checkvp_barrier 0 283 #endif 284 285 struct vnode * 286 null_checkvp(vp, fil, lno) 287 struct vnode *vp; 288 char *fil; 289 int lno; 290 { 291 struct null_node *a = VTONULL(vp); 292 #ifdef notyet 293 /* 294 * Can't do this check because vop_reclaim runs 295 * with a funny vop vector. 296 */ 297 if (vp->v_op != null_vnodeop_p) { 298 printf ("null_checkvp: on non-null-node\n"); 299 while (null_checkvp_barrier) /*WAIT*/ ; 300 panic("null_checkvp"); 301 }; 302 #endif 303 if (a->null_lowervp == NULLVP) { 304 /* Should never happen */ 305 int i; u_long *p; 306 printf("vp = %p, ZERO ptr\n", (void *)vp); 307 for (p = (u_long *) a, i = 0; i < 8; i++) 308 printf(" %lx", p[i]); 309 printf("\n"); 310 /* wait for debugger */ 311 while (null_checkvp_barrier) /*WAIT*/ ; 312 panic("null_checkvp"); 313 } 314 if (vrefcnt(a->null_lowervp) < 1) { 315 int i; u_long *p; 316 printf("vp = %p, unref'ed lowervp\n", (void *)vp); 317 for (p = (u_long *) a, i = 0; i < 8; i++) 318 printf(" %lx", p[i]); 319 printf("\n"); 320 /* wait for debugger */ 321 while (null_checkvp_barrier) /*WAIT*/ ; 322 panic ("null with unref'ed lowervp"); 323 }; 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