1 /* $NetBSD: layer_subr.c,v 1.22 2007/10/10 20:42:28 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1999 National Aeronautics & Space Administration 5 * All rights reserved. 6 * 7 * This software was written by William Studenmund of the 8 * Numerical Aerospace Simulation Facility, NASA Ames Research Center. 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 National Aeronautics & Space Administration 19 * nor the names of its contributors may be used to endorse or promote 20 * products derived from this software without specific prior written 21 * permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE NATIONAL AERONAUTICS & SPACE ADMINISTRATION 24 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 25 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 26 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE ADMINISTRATION OR CONTRIB- 27 * UTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, 28 * OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 30 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 31 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 32 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 33 * POSSIBILITY OF SUCH DAMAGE. 34 */ 35 /* 36 * Copyright (c) 1992, 1993 37 * The Regents of the University of California. All rights reserved. 38 * 39 * This code is derived from software donated to Berkeley by 40 * Jan-Simon Pendry. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * from: Id: lofs_subr.c,v 1.11 1992/05/30 10:05:43 jsp Exp 67 * @(#)null_subr.c 8.7 (Berkeley) 5/14/95 68 */ 69 70 #include <sys/cdefs.h> 71 __KERNEL_RCSID(0, "$NetBSD: layer_subr.c,v 1.22 2007/10/10 20:42:28 ad Exp $"); 72 73 #include <sys/param.h> 74 #include <sys/systm.h> 75 #include <sys/proc.h> 76 #include <sys/time.h> 77 #include <sys/vnode.h> 78 #include <sys/mount.h> 79 #include <sys/namei.h> 80 #include <sys/malloc.h> 81 #include <miscfs/specfs/specdev.h> 82 #include <miscfs/genfs/layer.h> 83 #include <miscfs/genfs/layer_extern.h> 84 85 #define NLAYERNODECACHE 16 86 87 #ifdef LAYERFS_DIAGNOSTIC 88 int layerfs_debug = 1; 89 #endif 90 91 /* 92 * layer cache: 93 * Each cache entry holds a reference to the lower vnode 94 * along with a pointer to the alias vnode. When an 95 * entry is added the lower vnode is VREF'd. When the 96 * alias is removed the lower vnode is vrele'd. 97 */ 98 99 /* 100 * Initialise cache headers 101 */ 102 void 103 layerfs_init() 104 { 105 #ifdef LAYERFS_DIAGNOSTIC 106 if (layerfs_debug) 107 printf("layerfs_init\n"); /* printed during system boot */ 108 #endif 109 } 110 111 /* 112 * Free global resources of layerfs. 113 */ 114 void 115 layerfs_done() 116 { 117 #ifdef LAYERFS_DIAGNOSTIC 118 if (layerfs_debug) 119 printf("layerfs_done\n"); /* printed on layerfs detach */ 120 #endif 121 } 122 123 /* 124 * Return a locked, VREF'ed alias for lower vnode if already exists, else NULL. 125 * The layermp's hashlock must be held on entry. 126 * It will be held upon return iff we return NULL. 127 */ 128 struct vnode * 129 layer_node_find(mp, lowervp) 130 struct mount *mp; 131 struct vnode *lowervp; 132 { 133 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 134 struct layer_node_hashhead *hd; 135 struct layer_node *a; 136 struct vnode *vp; 137 int error; 138 139 /* 140 * Find hash base, and then search the (two-way) linked 141 * list looking for a layer_node structure which is referencing 142 * the lower vnode. If found, the increment the layer_node 143 * reference count (but NOT the lower vnode's VREF counter) 144 * and return the vnode locked. 145 */ 146 hd = LAYER_NHASH(lmp, lowervp); 147 loop: 148 LIST_FOREACH(a, hd, layer_hash) { 149 if (a->layer_lowervp == lowervp && LAYERTOV(a)->v_mount == mp) { 150 vp = LAYERTOV(a); 151 152 /* 153 * We must not let vget() try to lock the layer vp, 154 * since the lower vp is already locked and locking the 155 * layer vp will involve locking the lower vp (whether 156 * or not they actually share a lock). Instead, take 157 * the layer vp's lock separately afterward, but only 158 * if it does not share the lower vp's lock. 159 */ 160 mutex_exit(&lmp->layerm_hashlock); 161 error = vget(vp, 0); 162 if (error) { 163 mutex_enter(&lmp->layerm_hashlock); 164 goto loop; 165 } 166 LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error); 167 return (vp); 168 } 169 } 170 return NULL; 171 } 172 173 174 /* 175 * Make a new layer_node node. 176 * Vp is the alias vnode, lowervp is the lower vnode. 177 * Maintain a reference to lowervp. 178 */ 179 int 180 layer_node_alloc(mp, lowervp, vpp) 181 struct mount *mp; 182 struct vnode *lowervp; 183 struct vnode **vpp; 184 { 185 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 186 struct layer_node_hashhead *hd; 187 struct layer_node *xp; 188 struct vnode *vp, *nvp; 189 int error; 190 extern int (**dead_vnodeop_p)(void *); 191 192 if ((error = getnewvnode(lmp->layerm_tag, mp, lmp->layerm_vnodeop_p, 193 &vp)) != 0) 194 return (error); 195 vp->v_type = lowervp->v_type; 196 simple_lock(&vp->v_interlock); 197 vp->v_iflag |= VI_LAYER; 198 simple_unlock(&vp->v_interlock); 199 200 xp = malloc(lmp->layerm_size, M_TEMP, M_WAITOK); 201 if (vp->v_type == VBLK || vp->v_type == VCHR) { 202 MALLOC(vp->v_specinfo, struct specinfo *, 203 sizeof(struct specinfo), M_VNODE, M_WAITOK); 204 vp->v_hashchain = NULL; 205 vp->v_rdev = lowervp->v_rdev; 206 } 207 208 vp->v_data = xp; 209 xp->layer_vnode = vp; 210 xp->layer_lowervp = lowervp; 211 xp->layer_flags = 0; 212 213 /* 214 * Before we insert our new node onto the hash chains, 215 * check to see if someone else has beaten us to it. 216 * (We could have slept in MALLOC.) 217 */ 218 mutex_enter(&lmp->layerm_hashlock); 219 if ((nvp = layer_node_find(mp, lowervp)) != NULL) { 220 *vpp = nvp; 221 222 /* free the substructures we've allocated. */ 223 FREE(xp, M_TEMP); 224 if (vp->v_type == VBLK || vp->v_type == VCHR) 225 FREE(vp->v_specinfo, M_VNODE); 226 227 vp->v_type = VBAD; /* node is discarded */ 228 vp->v_op = dead_vnodeop_p; /* so ops will still work */ 229 vrele(vp); /* get rid of it. */ 230 return (0); 231 } 232 233 /* 234 * Now lock the new node. We rely on the fact that we were passed 235 * a locked vnode. If the lower node is exporting a struct lock 236 * (v_vnlock != NULL) then we just set the upper v_vnlock to the 237 * lower one, and both are now locked. If the lower node is exporting 238 * NULL, then we copy that up and manually lock the upper node. 239 * 240 * LAYERFS_UPPERLOCK already has the test, so we use it after copying 241 * up the v_vnlock from below. 242 */ 243 244 vp->v_vnlock = lowervp->v_vnlock; 245 LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error); 246 KASSERT(error == 0); 247 248 /* 249 * Insert the new node into the hash. 250 * Add a reference to the lower node. 251 */ 252 253 *vpp = vp; 254 VREF(lowervp); 255 hd = LAYER_NHASH(lmp, lowervp); 256 LIST_INSERT_HEAD(hd, xp, layer_hash); 257 uvm_vnp_setsize(vp, 0); 258 mutex_exit(&lmp->layerm_hashlock); 259 return (0); 260 } 261 262 263 /* 264 * Try to find an existing layer_node vnode refering 265 * to it, otherwise make a new layer_node vnode which 266 * contains a reference to the lower vnode. 267 * 268 * >>> we assume that the lower node is already locked upon entry, so we 269 * propagate the lock state to upper node << 270 */ 271 int 272 layer_node_create(mp, lowervp, newvpp) 273 struct mount *mp; 274 struct vnode *lowervp; 275 struct vnode **newvpp; 276 { 277 struct vnode *aliasvp; 278 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 279 280 mutex_enter(&lmp->layerm_hashlock); 281 aliasvp = layer_node_find(mp, lowervp); 282 if (aliasvp != NULL) { 283 /* 284 * layer_node_find has taken another reference 285 * to the alias vnode and moved the lock holding to 286 * aliasvp 287 */ 288 #ifdef LAYERFS_DIAGNOSTIC 289 if (layerfs_debug) 290 vprint("layer_node_create: exists", aliasvp); 291 #endif 292 } else { 293 int error; 294 295 mutex_exit(&lmp->layerm_hashlock); 296 297 /* 298 * Get new vnode. 299 */ 300 #ifdef LAYERFS_DIAGNOSTIC 301 if (layerfs_debug) 302 printf("layer_node_create: create new alias vnode\n"); 303 #endif 304 305 /* 306 * Make new vnode reference the layer_node. 307 */ 308 if ((error = (lmp->layerm_alloc)(mp, lowervp, &aliasvp)) != 0) 309 return error; 310 311 /* 312 * aliasvp is already VREF'd by getnewvnode() 313 */ 314 } 315 316 /* 317 * Now that we have VREF'd the upper vnode, release the reference 318 * to the lower node. The existence of the layer_node retains one 319 * reference to the lower node. 320 */ 321 vrele(lowervp); 322 323 #ifdef DIAGNOSTIC 324 if (lowervp->v_usecount < 1) { 325 /* Should never happen... */ 326 vprint("layer_node_create: alias", aliasvp); 327 vprint("layer_node_create: lower", lowervp); 328 panic("layer_node_create: lower has 0 usecount."); 329 } 330 #endif 331 332 #ifdef LAYERFS_DIAGNOSTIC 333 if (layerfs_debug) 334 vprint("layer_node_create: alias", aliasvp); 335 #endif 336 *newvpp = aliasvp; 337 return (0); 338 } 339 340 #ifdef LAYERFS_DIAGNOSTIC 341 struct vnode * 342 layer_checkvp(vp, fil, lno) 343 struct vnode *vp; 344 const char *fil; 345 int lno; 346 { 347 struct layer_node *a = VTOLAYER(vp); 348 #ifdef notyet 349 /* 350 * Can't do this check because vop_reclaim runs 351 * with a funny vop vector. 352 * 353 * WRS - no it doesnt... 354 */ 355 if (vp->v_op != layer_vnodeop_p) { 356 printf ("layer_checkvp: on non-layer-node\n"); 357 #ifdef notyet 358 while (layer_checkvp_barrier) /*WAIT*/ ; 359 #endif 360 panic("layer_checkvp"); 361 }; 362 #endif 363 if (a->layer_lowervp == NULL) { 364 /* Should never happen */ 365 int i; u_long *p; 366 printf("vp = %p, ZERO ptr\n", vp); 367 for (p = (u_long *) a, i = 0; i < 8; i++) 368 printf(" %lx", p[i]); 369 printf("\n"); 370 /* wait for debugger */ 371 panic("layer_checkvp"); 372 } 373 if (a->layer_lowervp->v_usecount < 1) { 374 int i; u_long *p; 375 printf("vp = %p, unref'ed lowervp\n", vp); 376 for (p = (u_long *) a, i = 0; i < 8; i++) 377 printf(" %lx", p[i]); 378 printf("\n"); 379 /* wait for debugger */ 380 panic ("layer with unref'ed lowervp"); 381 }; 382 #ifdef notnow 383 printf("layer %p/%d -> %p/%d [%s, %d]\n", 384 LAYERTOV(a), LAYERTOV(a)->v_usecount, 385 a->layer_lowervp, a->layer_lowervp->v_usecount, 386 fil, lno); 387 #endif 388 return a->layer_lowervp; 389 } 390 #endif 391