1 /* $NetBSD: layer_subr.c,v 1.18 2005/12/11 12:24:50 christos 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 1.18 2005/12/11 12:24:50 christos 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 0. 125 */ 126 struct vnode * 127 layer_node_find(mp, lowervp) 128 struct mount *mp; 129 struct vnode *lowervp; 130 { 131 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 132 struct layer_node_hashhead *hd; 133 struct layer_node *a; 134 struct vnode *vp; 135 136 /* 137 * Find hash base, and then search the (two-way) linked 138 * list looking for a layer_node structure which is referencing 139 * the lower vnode. If found, the increment the layer_node 140 * reference count (but NOT the lower vnode's VREF counter) 141 * and return the vnode locked. 142 */ 143 hd = LAYER_NHASH(lmp, lowervp); 144 loop: 145 simple_lock(&lmp->layerm_hashlock); 146 for (a = hd->lh_first; a != 0; a = a->layer_hash.le_next) { 147 if (a->layer_lowervp == lowervp && LAYERTOV(a)->v_mount == mp) { 148 vp = LAYERTOV(a); 149 simple_unlock(&lmp->layerm_hashlock); 150 /* 151 * We must be careful here as the fact the lower 152 * vnode is locked will imply vp is locked unless 153 * someone has decided to start vclean'ing either 154 * vp or lowervp. 155 * 156 * So we try for an exclusive, recursive lock 157 * on the upper vnode. If it fails, vcleaning 158 * is in progress (so when we try again, we'll 159 * fail). If it succeeds, we now have double 160 * locked the bottom node. So we do an explicit 161 * VOP_UNLOCK on it to keep the counts right. Note 162 * that we will end up with the upper node and 163 * the lower node locked once. 164 */ 165 if (vget(vp, LK_EXCLUSIVE | LK_CANRECURSE)) { 166 printf ("layer_node_find: vget failed.\n"); 167 goto loop; 168 }; 169 VOP_UNLOCK(lowervp, 0); 170 return (vp); 171 } 172 } 173 174 simple_unlock(&lmp->layerm_hashlock); 175 return NULL; 176 } 177 178 179 /* 180 * Make a new layer_node node. 181 * Vp is the alias vnode, lowervp is the lower vnode. 182 * Maintain a reference to lowervp. 183 */ 184 int 185 layer_node_alloc(mp, lowervp, vpp) 186 struct mount *mp; 187 struct vnode *lowervp; 188 struct vnode **vpp; 189 { 190 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 191 struct layer_node_hashhead *hd; 192 struct layer_node *xp; 193 struct vnode *vp, *nvp; 194 int error; 195 extern int (**dead_vnodeop_p)(void *); 196 197 if ((error = getnewvnode(lmp->layerm_tag, mp, lmp->layerm_vnodeop_p, 198 &vp)) != 0) 199 return (error); 200 vp->v_type = lowervp->v_type; 201 vp->v_flag |= VLAYER; 202 203 xp = malloc(lmp->layerm_size, M_TEMP, M_WAITOK); 204 if (vp->v_type == VBLK || vp->v_type == VCHR) { 205 MALLOC(vp->v_specinfo, struct specinfo *, 206 sizeof(struct specinfo), M_VNODE, M_WAITOK); 207 vp->v_hashchain = NULL; 208 vp->v_rdev = lowervp->v_rdev; 209 } 210 211 vp->v_data = xp; 212 xp->layer_vnode = vp; 213 xp->layer_lowervp = lowervp; 214 xp->layer_flags = 0; 215 /* 216 * Before we insert our new node onto the hash chains, 217 * check to see if someone else has beaten us to it. 218 * (We could have slept in MALLOC.) 219 */ 220 if ((nvp = layer_node_find(mp, lowervp)) != NULL) { 221 *vpp = nvp; 222 223 /* free the substructures we've allocated. */ 224 FREE(xp, M_TEMP); 225 if (vp->v_type == VBLK || vp->v_type == VCHR) 226 FREE(vp->v_specinfo, M_VNODE); 227 228 vp->v_type = VBAD; /* node is discarded */ 229 vp->v_op = dead_vnodeop_p; /* so ops will still work */ 230 vrele(vp); /* get rid of it. */ 231 return (0); 232 } 233 234 simple_lock(&lmp->layerm_hashlock); 235 236 /* 237 * Now lock the new node. We rely on the fact that we were passed 238 * a locked vnode. If the lower node is exporting a struct lock 239 * (v_vnlock != NULL) then we just set the upper v_vnlock to the 240 * lower one, and both are now locked. If the lower node is exporting 241 * NULL, then we copy that up and manually lock the upper node. 242 * 243 * LAYERFS_UPPERLOCK already has the test, so we use it after copying 244 * up the v_vnlock from below. 245 */ 246 247 vp->v_vnlock = lowervp->v_vnlock; 248 LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error); 249 250 if (error) { 251 /* 252 * How did we get a locking error? The node just came off 253 * of the free list, and we're the only routine which 254 * knows it's there... 255 */ 256 vp->v_vnlock = &vp->v_lock; 257 *vpp = NULL; 258 259 /* free the substructures we've allocated. */ 260 FREE(xp, M_TEMP); 261 if (vp->v_type == VBLK || vp->v_type == VCHR) 262 FREE(vp->v_specinfo, M_VNODE); 263 264 vp->v_type = VBAD; /* node is discarded */ 265 vp->v_op = dead_vnodeop_p; /* so ops will still work */ 266 vrele(vp); /* get rid of it. */ 267 return (error); 268 } 269 /* 270 * NetBSD used to do an inlined checkalias here. We do not, as 271 * we never flag device nodes as being aliased. The lowervp 272 * node will, when appropriate, be flaged as an alias. 273 */ 274 275 *vpp = vp; 276 VREF(lowervp); /* Take into account reference held in layer_node */ 277 hd = LAYER_NHASH(lmp, lowervp); 278 LIST_INSERT_HEAD(hd, xp, layer_hash); 279 uvm_vnp_setsize(vp, 0); 280 simple_unlock(&lmp->layerm_hashlock); 281 return (0); 282 } 283 284 285 /* 286 * Try to find an existing layer_node vnode refering 287 * to it, otherwise make a new layer_node vnode which 288 * contains a reference to the lower vnode. 289 * 290 * >>> we assume that the lower node is already locked upon entry, so we 291 * propagate the lock state to upper node << 292 */ 293 int 294 layer_node_create(mp, lowervp, newvpp) 295 struct mount *mp; 296 struct vnode *lowervp; 297 struct vnode **newvpp; 298 { 299 struct vnode *aliasvp; 300 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 301 302 if ((aliasvp = layer_node_find(mp, lowervp)) != NULL) { 303 /* 304 * layer_node_find has taken another reference 305 * to the alias vnode and moved the lock holding to 306 * aliasvp 307 */ 308 #ifdef LAYERFS_DIAGNOSTIC 309 if (layerfs_debug) 310 vprint("layer_node_create: exists", aliasvp); 311 #endif 312 } else { 313 int error; 314 315 /* 316 * Get new vnode. 317 */ 318 #ifdef LAYERFS_DIAGNOSTIC 319 if (layerfs_debug) 320 printf("layer_node_create: create new alias vnode\n"); 321 #endif 322 323 /* 324 * Make new vnode reference the layer_node. 325 */ 326 if ((error = (lmp->layerm_alloc)(mp, lowervp, &aliasvp)) != 0) 327 return error; 328 329 /* 330 * aliasvp is already VREF'd by getnewvnode() 331 */ 332 } 333 334 /* 335 * Now that we have VREF'd the upper vnode, release the reference 336 * to the lower node. The existance of the layer_node retains one 337 * reference to the lower node. 338 */ 339 vrele(lowervp); 340 341 #ifdef DIAGNOSTIC 342 if (lowervp->v_usecount < 1) { 343 /* Should never happen... */ 344 vprint("layer_node_create: alias", aliasvp); 345 vprint("layer_node_create: lower", lowervp); 346 panic("layer_node_create: lower has 0 usecount."); 347 }; 348 #endif 349 350 #ifdef LAYERFS_DIAGNOSTIC 351 if (layerfs_debug) 352 vprint("layer_node_create: alias", aliasvp); 353 #endif 354 *newvpp = aliasvp; 355 return (0); 356 } 357 358 #ifdef LAYERFS_DIAGNOSTIC 359 struct vnode * 360 layer_checkvp(vp, fil, lno) 361 struct vnode *vp; 362 const char *fil; 363 int lno; 364 { 365 struct layer_node *a = VTOLAYER(vp); 366 #ifdef notyet 367 /* 368 * Can't do this check because vop_reclaim runs 369 * with a funny vop vector. 370 * 371 * WRS - no it doesnt... 372 */ 373 if (vp->v_op != layer_vnodeop_p) { 374 printf ("layer_checkvp: on non-layer-node\n"); 375 #ifdef notyet 376 while (layer_checkvp_barrier) /*WAIT*/ ; 377 #endif 378 panic("layer_checkvp"); 379 }; 380 #endif 381 if (a->layer_lowervp == NULL) { 382 /* Should never happen */ 383 int i; u_long *p; 384 printf("vp = %p, ZERO ptr\n", vp); 385 for (p = (u_long *) a, i = 0; i < 8; i++) 386 printf(" %lx", p[i]); 387 printf("\n"); 388 /* wait for debugger */ 389 panic("layer_checkvp"); 390 } 391 if (a->layer_lowervp->v_usecount < 1) { 392 int i; u_long *p; 393 printf("vp = %p, unref'ed lowervp\n", vp); 394 for (p = (u_long *) a, i = 0; i < 8; i++) 395 printf(" %lx", p[i]); 396 printf("\n"); 397 /* wait for debugger */ 398 panic ("layer with unref'ed lowervp"); 399 }; 400 #ifdef notnow 401 printf("layer %p/%d -> %p/%d [%s, %d]\n", 402 LAYERTOV(a), LAYERTOV(a)->v_usecount, 403 a->layer_lowervp, a->layer_lowervp->v_usecount, 404 fil, lno); 405 #endif 406 return a->layer_lowervp; 407 } 408 #endif 409