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 int error; 111 112 ASSERT_VOP_LOCKED(lowervp, "null_hashget"); 113 114 /* 115 * Find hash base, and then search the (two-way) linked 116 * list looking for a null_node structure which is referencing 117 * the lower vnode. If found, the increment the null_node 118 * reference count (but NOT the lower vnode's VREF counter). 119 */ 120 hd = NULL_NHASH(lowervp); 121 mtx_lock(&null_hashmtx); 122 LIST_FOREACH(a, hd, null_hash) { 123 if (a->null_lowervp == lowervp && NULLTOV(a)->v_mount == mp) { 124 vp = NULLTOV(a); 125 VI_LOCK(vp); 126 mtx_unlock(&null_hashmtx); 127 /* 128 * We need to clear the OWEINACT flag here as this 129 * may lead vget() to try to lock our vnode which 130 * is already locked via lowervp. 131 */ 132 vp->v_iflag &= ~VI_OWEINACT; 133 error = vget(vp, LK_INTERLOCK, td); 134 /* 135 * Since we have the lower node locked the nullfs 136 * node can not be in the process of recycling. If 137 * it had been recycled before we grabed the lower 138 * lock it would not have been found on the hash. 139 */ 140 if (error) 141 panic("null_hashget: vget error %d", error); 142 return (vp); 143 } 144 } 145 mtx_unlock(&null_hashmtx); 146 return (NULLVP); 147 } 148 149 /* 150 * Act like null_hashget, but add passed null_node to hash if no existing 151 * node found. 152 */ 153 static struct vnode * 154 null_hashins(mp, xp) 155 struct mount *mp; 156 struct null_node *xp; 157 { 158 struct thread *td = curthread; /* XXX */ 159 struct null_node_hashhead *hd; 160 struct null_node *oxp; 161 struct vnode *ovp; 162 int error; 163 164 hd = NULL_NHASH(xp->null_lowervp); 165 mtx_lock(&null_hashmtx); 166 LIST_FOREACH(oxp, hd, null_hash) { 167 if (oxp->null_lowervp == xp->null_lowervp && 168 NULLTOV(oxp)->v_mount == mp) { 169 /* 170 * See null_hashget for a description of this 171 * operation. 172 */ 173 ovp = NULLTOV(oxp); 174 VI_LOCK(ovp); 175 mtx_unlock(&null_hashmtx); 176 ovp->v_iflag &= ~VI_OWEINACT; 177 error = vget(ovp, LK_INTERLOCK, td); 178 if (error) 179 panic("null_hashins: vget error %d", error); 180 return (ovp); 181 } 182 } 183 LIST_INSERT_HEAD(hd, xp, null_hash); 184 mtx_unlock(&null_hashmtx); 185 return (NULLVP); 186 } 187 188 static void 189 null_insmntque_dtr(struct vnode *vp, void *xp) 190 { 191 vp->v_data = NULL; 192 vp->v_vnlock = &vp->v_lock; 193 FREE(xp, M_NULLFSNODE); 194 vp->v_op = &dead_vnodeops; 195 (void) vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, curthread); 196 vgone(vp); 197 vput(vp); 198 } 199 200 /* 201 * Make a new or get existing nullfs node. 202 * Vp is the alias vnode, lowervp is the lower vnode. 203 * 204 * The lowervp assumed to be locked and having "spare" reference. This routine 205 * vrele lowervp if nullfs node was taken from hash. Otherwise it "transfers" 206 * the caller's "spare" reference to created nullfs vnode. 207 */ 208 int 209 null_nodeget(mp, lowervp, vpp) 210 struct mount *mp; 211 struct vnode *lowervp; 212 struct vnode **vpp; 213 { 214 struct null_node *xp; 215 struct vnode *vp; 216 int error; 217 218 /* Lookup the hash firstly */ 219 *vpp = null_hashget(mp, lowervp); 220 if (*vpp != NULL) { 221 vrele(lowervp); 222 return (0); 223 } 224 225 /* 226 * We do not serialize vnode creation, instead we will check for 227 * duplicates later, when adding new vnode to hash. 228 * 229 * Note that duplicate can only appear in hash if the lowervp is 230 * locked LK_SHARED. 231 */ 232 233 /* 234 * Do the MALLOC before the getnewvnode since doing so afterward 235 * might cause a bogus v_data pointer to get dereferenced 236 * elsewhere if MALLOC should block. 237 */ 238 MALLOC(xp, struct null_node *, sizeof(struct null_node), 239 M_NULLFSNODE, M_WAITOK); 240 241 error = getnewvnode("null", mp, &null_vnodeops, &vp); 242 if (error) { 243 FREE(xp, M_NULLFSNODE); 244 return (error); 245 } 246 247 xp->null_vnode = vp; 248 xp->null_lowervp = lowervp; 249 vp->v_type = lowervp->v_type; 250 vp->v_data = xp; 251 vp->v_vnlock = lowervp->v_vnlock; 252 if (vp->v_vnlock == NULL) 253 panic("null_nodeget: Passed a NULL vnlock.\n"); 254 error = insmntque1(vp, mp, null_insmntque_dtr, xp); 255 if (error != 0) 256 return (error); 257 /* 258 * Atomically insert our new node into the hash or vget existing 259 * if someone else has beaten us to it. 260 */ 261 *vpp = null_hashins(mp, xp); 262 if (*vpp != NULL) { 263 vrele(lowervp); 264 vp->v_vnlock = &vp->v_lock; 265 xp->null_lowervp = NULL; 266 vrele(vp); 267 return (0); 268 } 269 *vpp = vp; 270 271 return (0); 272 } 273 274 /* 275 * Remove node from hash. 276 */ 277 void 278 null_hashrem(xp) 279 struct null_node *xp; 280 { 281 282 mtx_lock(&null_hashmtx); 283 LIST_REMOVE(xp, null_hash); 284 mtx_unlock(&null_hashmtx); 285 } 286 287 #ifdef DIAGNOSTIC 288 289 #ifdef KDB 290 #define null_checkvp_barrier 1 291 #else 292 #define null_checkvp_barrier 0 293 #endif 294 295 struct vnode * 296 null_checkvp(vp, fil, lno) 297 struct vnode *vp; 298 char *fil; 299 int lno; 300 { 301 int interlock = 0; 302 struct null_node *a = VTONULL(vp); 303 #ifdef notyet 304 /* 305 * Can't do this check because vop_reclaim runs 306 * with a funny vop vector. 307 */ 308 if (vp->v_op != null_vnodeop_p) { 309 printf ("null_checkvp: on non-null-node\n"); 310 while (null_checkvp_barrier) /*WAIT*/ ; 311 panic("null_checkvp"); 312 }; 313 #endif 314 if (a->null_lowervp == NULLVP) { 315 /* Should never happen */ 316 int i; u_long *p; 317 printf("vp = %p, ZERO ptr\n", (void *)vp); 318 for (p = (u_long *) a, i = 0; i < 8; i++) 319 printf(" %lx", p[i]); 320 printf("\n"); 321 /* wait for debugger */ 322 while (null_checkvp_barrier) /*WAIT*/ ; 323 panic("null_checkvp"); 324 } 325 if (mtx_owned(VI_MTX(vp)) != 0) { 326 VI_UNLOCK(vp); 327 interlock = 1; 328 } 329 if (vrefcnt(a->null_lowervp) < 1) { 330 int i; u_long *p; 331 printf("vp = %p, unref'ed lowervp\n", (void *)vp); 332 for (p = (u_long *) a, i = 0; i < 8; i++) 333 printf(" %lx", p[i]); 334 printf("\n"); 335 /* wait for debugger */ 336 while (null_checkvp_barrier) /*WAIT*/ ; 337 panic ("null with unref'ed lowervp"); 338 }; 339 if (interlock != 0) 340 VI_LOCK(vp); 341 #ifdef notyet 342 printf("null %x/%d -> %x/%d [%s, %d]\n", 343 NULLTOV(a), vrefcnt(NULLTOV(a)), 344 a->null_lowervp, vrefcnt(a->null_lowervp), 345 fil, lno); 346 #endif 347 return a->null_lowervp; 348 } 349 #endif 350