1 /* $NetBSD: layer_subr.c,v 1.25 2008/01/24 17:32:55 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 1.25 2008/01/24 17:32:55 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/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() 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() 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(mp, lowervp) 132 struct mount *mp; 133 struct vnode *lowervp; 134 { 135 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 136 struct layer_node_hashhead *hd; 137 struct layer_node *a; 138 struct vnode *vp; 139 int error; 140 141 /* 142 * Find hash base, and then search the (two-way) linked 143 * list looking for a layer_node structure which is referencing 144 * the lower vnode. If found, the increment the layer_node 145 * reference count (but NOT the lower vnode's VREF counter) 146 * and return the vnode locked. 147 */ 148 hd = LAYER_NHASH(lmp, lowervp); 149 loop: 150 LIST_FOREACH(a, hd, layer_hash) { 151 if (a->layer_lowervp == lowervp && LAYERTOV(a)->v_mount == mp) { 152 vp = LAYERTOV(a); 153 mutex_enter(&vp->v_interlock); 154 /* 155 * If we find a node being cleaned out, then 156 * ignore it and continue. A thread trying to 157 * clean out the extant layer vnode needs to 158 * acquire the shared lock (i.e. the lower 159 * vnode's lock), which our caller already holds. 160 * To allow the cleaning to succeed the current 161 * thread must make progress. So, for a brief 162 * time more than one vnode in a layered file 163 * system may refer to a single vnode in the 164 * lower file system. 165 */ 166 if ((vp->v_iflag & VI_XLOCK) != 0) { 167 mutex_exit(&vp->v_interlock); 168 continue; 169 } 170 mutex_exit(&lmp->layerm_hashlock); 171 /* 172 * We must not let vget() try to lock the layer 173 * vp, since the lower vp is already locked and 174 * locking the layer vp will involve locking 175 * the lower vp (whether or not they actually 176 * share a lock). Instead, take the layer vp's 177 * lock separately afterward, but only if it 178 * does not share the lower vp's lock. 179 */ 180 error = vget(vp, LK_INTERLOCK | LK_NOWAIT); 181 if (error) { 182 kpause("layerfs", false, 1, NULL); 183 mutex_enter(&lmp->layerm_hashlock); 184 goto loop; 185 } 186 LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error); 187 return (vp); 188 } 189 } 190 return NULL; 191 } 192 193 194 /* 195 * Make a new layer_node node. 196 * Vp is the alias vnode, lowervp is the lower vnode. 197 * Maintain a reference to lowervp. 198 */ 199 int 200 layer_node_alloc(mp, lowervp, vpp) 201 struct mount *mp; 202 struct vnode *lowervp; 203 struct vnode **vpp; 204 { 205 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 206 struct layer_node_hashhead *hd; 207 struct layer_node *xp; 208 struct vnode *vp, *nvp; 209 int error; 210 extern int (**dead_vnodeop_p)(void *); 211 212 error = getnewvnode(lmp->layerm_tag, mp, lmp->layerm_vnodeop_p, &vp); 213 if (error != 0) 214 return (error); 215 vp->v_type = lowervp->v_type; 216 mutex_enter(&vp->v_interlock); 217 vp->v_iflag |= VI_LAYER; 218 mutex_exit(&vp->v_interlock); 219 220 xp = kmem_alloc(lmp->layerm_size, KM_SLEEP); 221 if (xp == NULL) { 222 ungetnewvnode(vp); 223 return ENOMEM; 224 } 225 if (vp->v_type == VBLK || vp->v_type == VCHR) { 226 spec_node_init(vp, lowervp->v_rdev); 227 } 228 229 vp->v_data = xp; 230 vp->v_vflag = (vp->v_vflag & ~VV_MPSAFE) | 231 (lowervp->v_vflag & VV_MPSAFE); 232 xp->layer_vnode = vp; 233 xp->layer_lowervp = lowervp; 234 xp->layer_flags = 0; 235 236 /* 237 * Before we insert our new node onto the hash chains, 238 * check to see if someone else has beaten us to it. 239 * (We could have slept in MALLOC.) 240 */ 241 mutex_enter(&lmp->layerm_hashlock); 242 if ((nvp = layer_node_find(mp, lowervp)) != NULL) { 243 *vpp = nvp; 244 245 /* free the substructures we've allocated. */ 246 kmem_free(xp, lmp->layerm_size); 247 if (vp->v_type == VBLK || vp->v_type == VCHR) 248 spec_node_destroy(vp); 249 250 vp->v_type = VBAD; /* node is discarded */ 251 vp->v_op = dead_vnodeop_p; /* so ops will still work */ 252 vrele(vp); /* get rid of it. */ 253 return (0); 254 } 255 256 /* 257 * Now lock the new node. We rely on the fact that we were passed 258 * a locked vnode. If the lower node is exporting a struct lock 259 * (v_vnlock != NULL) then we just set the upper v_vnlock to the 260 * lower one, and both are now locked. If the lower node is exporting 261 * NULL, then we copy that up and manually lock the upper node. 262 * 263 * LAYERFS_UPPERLOCK already has the test, so we use it after copying 264 * up the v_vnlock from below. 265 */ 266 267 vp->v_vnlock = lowervp->v_vnlock; 268 LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error); 269 KASSERT(error == 0); 270 271 /* 272 * Insert the new node into the hash. 273 * Add a reference to the lower node. 274 */ 275 276 *vpp = vp; 277 VREF(lowervp); 278 hd = LAYER_NHASH(lmp, lowervp); 279 LIST_INSERT_HEAD(hd, xp, layer_hash); 280 uvm_vnp_setsize(vp, 0); 281 mutex_exit(&lmp->layerm_hashlock); 282 return (0); 283 } 284 285 286 /* 287 * Try to find an existing layer_node vnode refering 288 * to it, otherwise make a new layer_node vnode which 289 * contains a reference to the lower vnode. 290 * 291 * >>> we assume that the lower node is already locked upon entry, so we 292 * propagate the lock state to upper node << 293 */ 294 int 295 layer_node_create(mp, lowervp, newvpp) 296 struct mount *mp; 297 struct vnode *lowervp; 298 struct vnode **newvpp; 299 { 300 struct vnode *aliasvp; 301 struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp); 302 303 mutex_enter(&lmp->layerm_hashlock); 304 aliasvp = layer_node_find(mp, lowervp); 305 if (aliasvp != NULL) { 306 /* 307 * layer_node_find has taken another reference 308 * to the alias vnode and moved the lock holding to 309 * aliasvp 310 */ 311 #ifdef LAYERFS_DIAGNOSTIC 312 if (layerfs_debug) 313 vprint("layer_node_create: exists", aliasvp); 314 #endif 315 } else { 316 int error; 317 318 mutex_exit(&lmp->layerm_hashlock); 319 320 /* 321 * Get new vnode. 322 */ 323 #ifdef LAYERFS_DIAGNOSTIC 324 if (layerfs_debug) 325 printf("layer_node_create: create new alias vnode\n"); 326 #endif 327 328 /* 329 * Make new vnode reference the layer_node. 330 */ 331 if ((error = (lmp->layerm_alloc)(mp, lowervp, &aliasvp)) != 0) 332 return error; 333 334 /* 335 * aliasvp is already VREF'd by getnewvnode() 336 */ 337 } 338 339 /* 340 * Now that we have VREF'd the upper vnode, release the reference 341 * to the lower node. The existence of the layer_node retains one 342 * reference to the lower node. 343 */ 344 vrele(lowervp); 345 346 #ifdef DIAGNOSTIC 347 if (lowervp->v_usecount < 1) { 348 /* Should never happen... */ 349 vprint("layer_node_create: alias", aliasvp); 350 vprint("layer_node_create: lower", lowervp); 351 panic("layer_node_create: lower has 0 usecount."); 352 } 353 #endif 354 355 #ifdef LAYERFS_DIAGNOSTIC 356 if (layerfs_debug) 357 vprint("layer_node_create: alias", aliasvp); 358 #endif 359 *newvpp = aliasvp; 360 return (0); 361 } 362 363 #ifdef LAYERFS_DIAGNOSTIC 364 struct vnode * 365 layer_checkvp(vp, fil, lno) 366 struct vnode *vp; 367 const char *fil; 368 int lno; 369 { 370 struct layer_node *a = VTOLAYER(vp); 371 #ifdef notyet 372 /* 373 * Can't do this check because vop_reclaim runs 374 * with a funny vop vector. 375 * 376 * WRS - no it doesnt... 377 */ 378 if (vp->v_op != layer_vnodeop_p) { 379 printf ("layer_checkvp: on non-layer-node\n"); 380 #ifdef notyet 381 while (layer_checkvp_barrier) /*WAIT*/ ; 382 #endif 383 panic("layer_checkvp"); 384 }; 385 #endif 386 if (a->layer_lowervp == NULL) { 387 /* Should never happen */ 388 int i; u_long *p; 389 printf("vp = %p, ZERO ptr\n", vp); 390 for (p = (u_long *) a, i = 0; i < 8; i++) 391 printf(" %lx", p[i]); 392 printf("\n"); 393 /* wait for debugger */ 394 panic("layer_checkvp"); 395 } 396 if (a->layer_lowervp->v_usecount < 1) { 397 int i; u_long *p; 398 printf("vp = %p, unref'ed lowervp\n", vp); 399 for (p = (u_long *) a, i = 0; i < 8; i++) 400 printf(" %lx", p[i]); 401 printf("\n"); 402 /* wait for debugger */ 403 panic ("layer with unref'ed lowervp"); 404 }; 405 #ifdef notnow 406 printf("layer %p/%d -> %p/%d [%s, %d]\n", 407 LAYERTOV(a), LAYERTOV(a)->v_usecount, 408 a->layer_lowervp, a->layer_lowervp->v_usecount, 409 fil, lno); 410 #endif 411 return a->layer_lowervp; 412 } 413 #endif 414