1 /* $NetBSD: layer_subr.c,v 1.29 2010/06/06 08:01:31 hannken 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 1.29 2010/06/06 08:01:31 hannken 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/kmem.h> 81 #include <sys/malloc.h> 82 83 #include <miscfs/specfs/specdev.h> 84 #include <miscfs/genfs/layer.h> 85 #include <miscfs/genfs/layer_extern.h> 86 87 #define NLAYERNODECACHE 16 88 89 #ifdef LAYERFS_DIAGNOSTIC 90 int layerfs_debug = 1; 91 #endif 92 93 /* 94 * layer cache: 95 * Each cache entry holds a reference to the lower vnode 96 * along with a pointer to the alias vnode. When an 97 * entry is added the lower vnode is VREF'd. When the 98 * alias is removed the lower vnode is vrele'd. 99 */ 100 101 /* 102 * Initialise cache headers 103 */ 104 void 105 layerfs_init(void) 106 { 107 #ifdef LAYERFS_DIAGNOSTIC 108 if (layerfs_debug) 109 printf("layerfs_init\n"); /* printed during system boot */ 110 #endif 111 } 112 113 /* 114 * Free global resources of layerfs. 115 */ 116 void 117 layerfs_done(void) 118 { 119 #ifdef LAYERFS_DIAGNOSTIC 120 if (layerfs_debug) 121 printf("layerfs_done\n"); /* printed on layerfs detach */ 122 #endif 123 } 124 125 /* 126 * Return a locked, VREF'ed alias for lower vnode if already exists, else NULL. 127 * The layermp's hashlock must be held on entry. 128 * It will be held upon return iff we return NULL. 129 */ 130 struct vnode * 131 layer_node_find(struct mount *mp, 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 mutex_enter(&vp->v_interlock); 152 /* 153 * If we find a node being cleaned out, then 154 * ignore it and continue. A thread trying to 155 * clean out the extant layer vnode needs to 156 * acquire the shared lock (i.e. the lower 157 * vnode's lock), which our caller already holds. 158 * To allow the cleaning to succeed the current 159 * thread must make progress. So, for a brief 160 * time more than one vnode in a layered file 161 * system may refer to a single vnode in the 162 * lower file system. 163 */ 164 if ((vp->v_iflag & VI_XLOCK) != 0) { 165 mutex_exit(&vp->v_interlock); 166 continue; 167 } 168 mutex_exit(&lmp->layerm_hashlock); 169 /* 170 * We must not let vget() try to lock the layer 171 * vp, since the lower vp is already locked and 172 * locking the layer vp will involve locking 173 * the lower vp. 174 */ 175 error = vget(vp, LK_INTERLOCK | LK_NOWAIT); 176 if (error) { 177 kpause("layerfs", false, 1, NULL); 178 mutex_enter(&lmp->layerm_hashlock); 179 goto loop; 180 } 181 return (vp); 182 } 183 } 184 return NULL; 185 } 186 187 188 /* 189 * Make a new layer_node node. 190 * Vp is the alias vnode, lowervp is the lower vnode. 191 * Maintain a reference to lowervp. 192 */ 193 int 194 layer_node_alloc(struct mount *mp, struct vnode *lowervp, struct vnode **vpp) 195 { 196 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 197 struct layer_node_hashhead *hd; 198 struct layer_node *xp; 199 struct vnode *vp, *nvp; 200 int error; 201 extern int (**dead_vnodeop_p)(void *); 202 203 error = getnewvnode(lmp->layerm_tag, mp, lmp->layerm_vnodeop_p, &vp); 204 if (error != 0) 205 return (error); 206 vp->v_type = lowervp->v_type; 207 mutex_enter(&vp->v_interlock); 208 vp->v_iflag |= VI_LAYER; 209 mutex_exit(&vp->v_interlock); 210 211 xp = kmem_alloc(lmp->layerm_size, KM_SLEEP); 212 if (xp == NULL) { 213 ungetnewvnode(vp); 214 return ENOMEM; 215 } 216 if (vp->v_type == VBLK || vp->v_type == VCHR) { 217 spec_node_init(vp, lowervp->v_rdev); 218 } 219 220 vp->v_data = xp; 221 vp->v_vflag = (vp->v_vflag & ~VV_MPSAFE) | 222 (lowervp->v_vflag & VV_MPSAFE); 223 xp->layer_vnode = vp; 224 xp->layer_lowervp = lowervp; 225 xp->layer_flags = 0; 226 227 /* 228 * Before we insert our new node onto the hash chains, 229 * check to see if someone else has beaten us to it. 230 * (We could have slept in MALLOC.) 231 */ 232 mutex_enter(&lmp->layerm_hashlock); 233 if ((nvp = layer_node_find(mp, lowervp)) != NULL) { 234 *vpp = nvp; 235 236 /* free the substructures we've allocated. */ 237 kmem_free(xp, lmp->layerm_size); 238 if (vp->v_type == VBLK || vp->v_type == VCHR) 239 spec_node_destroy(vp); 240 241 vp->v_type = VBAD; /* node is discarded */ 242 vp->v_op = dead_vnodeop_p; /* so ops will still work */ 243 vrele(vp); /* get rid of it. */ 244 return (0); 245 } 246 247 /* 248 * Insert the new node into the hash. 249 * Add a reference to the lower node. 250 */ 251 252 *vpp = vp; 253 vref(lowervp); 254 hd = LAYER_NHASH(lmp, lowervp); 255 LIST_INSERT_HEAD(hd, xp, layer_hash); 256 uvm_vnp_setsize(vp, 0); 257 mutex_exit(&lmp->layerm_hashlock); 258 return (0); 259 } 260 261 262 /* 263 * Try to find an existing layer_node vnode refering 264 * to it, otherwise make a new layer_node vnode which 265 * contains a reference to the lower vnode. 266 * 267 * >>> we assume that the lower node is already locked upon entry, so we 268 * propagate the lock state to upper node << 269 */ 270 int 271 layer_node_create(struct mount *mp, struct vnode *lowervp, struct vnode **newvpp) 272 { 273 struct vnode *aliasvp; 274 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 275 276 mutex_enter(&lmp->layerm_hashlock); 277 aliasvp = layer_node_find(mp, lowervp); 278 if (aliasvp != NULL) { 279 /* 280 * layer_node_find has taken another reference 281 * to the alias vnode and moved the lock holding to 282 * aliasvp 283 */ 284 #ifdef LAYERFS_DIAGNOSTIC 285 if (layerfs_debug) 286 vprint("layer_node_create: exists", aliasvp); 287 #endif 288 } else { 289 int error; 290 291 mutex_exit(&lmp->layerm_hashlock); 292 293 /* 294 * Get new vnode. 295 */ 296 #ifdef LAYERFS_DIAGNOSTIC 297 if (layerfs_debug) 298 printf("layer_node_create: create new alias vnode\n"); 299 #endif 300 301 /* 302 * Make new vnode reference the layer_node. 303 */ 304 if ((error = (lmp->layerm_alloc)(mp, lowervp, &aliasvp)) != 0) 305 return error; 306 307 /* 308 * aliasvp is already VREF'd by getnewvnode() 309 */ 310 } 311 312 /* 313 * Now that we have VREF'd the upper vnode, release the reference 314 * to the lower node. The existence of the layer_node retains one 315 * reference to the lower node. 316 */ 317 vrele(lowervp); 318 319 #ifdef DIAGNOSTIC 320 if (lowervp->v_usecount < 1) { 321 /* Should never happen... */ 322 vprint("layer_node_create: alias", aliasvp); 323 vprint("layer_node_create: lower", lowervp); 324 panic("layer_node_create: lower has 0 usecount."); 325 } 326 #endif 327 328 #ifdef LAYERFS_DIAGNOSTIC 329 if (layerfs_debug) 330 vprint("layer_node_create: alias", aliasvp); 331 #endif 332 *newvpp = aliasvp; 333 return (0); 334 } 335 336 #ifdef LAYERFS_DIAGNOSTIC 337 struct vnode * 338 layer_checkvp(struct vnode *vp, const char *fil, int lno) 339 { 340 struct layer_node *a = VTOLAYER(vp); 341 #ifdef notyet 342 /* 343 * Can't do this check because vop_reclaim runs 344 * with a funny vop vector. 345 * 346 * WRS - no it doesnt... 347 */ 348 if (vp->v_op != layer_vnodeop_p) { 349 printf ("layer_checkvp: on non-layer-node\n"); 350 #ifdef notyet 351 while (layer_checkvp_barrier) /*WAIT*/ ; 352 #endif 353 panic("layer_checkvp"); 354 }; 355 #endif 356 if (a->layer_lowervp == NULL) { 357 /* Should never happen */ 358 int i; u_long *p; 359 printf("vp = %p, ZERO ptr\n", vp); 360 for (p = (u_long *) a, i = 0; i < 8; i++) 361 printf(" %lx", p[i]); 362 printf("\n"); 363 /* wait for debugger */ 364 panic("layer_checkvp"); 365 } 366 if (a->layer_lowervp->v_usecount < 1) { 367 int i; u_long *p; 368 printf("vp = %p, unref'ed lowervp\n", vp); 369 for (p = (u_long *) a, i = 0; i < 8; i++) 370 printf(" %lx", p[i]); 371 printf("\n"); 372 /* wait for debugger */ 373 panic ("layer with unref'ed lowervp"); 374 }; 375 #ifdef notnow 376 printf("layer %p/%d -> %p/%d [%s, %d]\n", 377 LAYERTOV(a), LAYERTOV(a)->v_usecount, 378 a->layer_lowervp, a->layer_lowervp->v_usecount, 379 fil, lno); 380 #endif 381 return a->layer_lowervp; 382 } 383 #endif 384