sys/miscfs/genfs/layer_subr.c

   1 /*      $NetBSD: layer_subr.c,v 1.27 2009/03/15 17:22:38 cegger 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 1.27 2009/03/15 17:22:38 cegger 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 (whether or not they actually
 174                          * share a lock).  Instead, take the layer vp's
 175                          * lock separately afterward, but only if it
 176                          * does not share the lower vp's lock.
 177                          */
 178                         error = vget(vp, LK_INTERLOCK | LK_NOWAIT);
 179                         if (error) {
 180                                 kpause("layerfs", false, 1, NULL);
 181                                 mutex_enter(&lmp->layerm_hashlock);
 182                                 goto loop;
 183                         }
 184                         LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error);
 185                         return (vp);
 186                 }
 187         }
 188         return NULL;
 189 }
 190
 191
 192 /*
 193  * Make a new layer_node node.
 194  * Vp is the alias vnode, lowervp is the lower vnode.
 195  * Maintain a reference to lowervp.
 196  */
 197 int
 198 layer_node_alloc(struct mount *mp, struct vnode *lowervp, struct vnode **vpp)
 199 {
 200         struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
 201         struct layer_node_hashhead *hd;
 202         struct layer_node *xp;
 203         struct vnode *vp, *nvp;
 204         int error;
 205         extern int (**dead_vnodeop_p)(void *);
 206
 207         error = getnewvnode(lmp->layerm_tag, mp, lmp->layerm_vnodeop_p, &vp);
 208         if (error != 0)
 209                 return (error);
 210         vp->v_type = lowervp->v_type;
 211         mutex_enter(&vp->v_interlock);
 212         vp->v_iflag |= VI_LAYER;
 213         mutex_exit(&vp->v_interlock);
 214
 215         xp = kmem_alloc(lmp->layerm_size, KM_SLEEP);
 216         if (xp == NULL) {
 217                 ungetnewvnode(vp);
 218                 return ENOMEM;
 219         }
 220         if (vp->v_type == VBLK || vp->v_type == VCHR) {
 221                 spec_node_init(vp, lowervp->v_rdev);
 222         }
 223
 224         vp->v_data = xp;
 225         vp->v_vflag = (vp->v_vflag & ~VV_MPSAFE) |
 226             (lowervp->v_vflag & VV_MPSAFE);
 227         xp->layer_vnode = vp;
 228         xp->layer_lowervp = lowervp;
 229         xp->layer_flags = 0;
 230
 231         /*
 232          * Before we insert our new node onto the hash chains,
 233          * check to see if someone else has beaten us to it.
 234          * (We could have slept in MALLOC.)
 235          */
 236         mutex_enter(&lmp->layerm_hashlock);
 237         if ((nvp = layer_node_find(mp, lowervp)) != NULL) {
 238                 *vpp = nvp;
 239
 240                 /* free the substructures we've allocated. */
 241                 kmem_free(xp, lmp->layerm_size);
 242                 if (vp->v_type == VBLK || vp->v_type == VCHR)
 243                         spec_node_destroy(vp);
 244
 245                 vp->v_type = VBAD;              /* node is discarded */
 246                 vp->v_op = dead_vnodeop_p;      /* so ops will still work */
 247                 vrele(vp);                      /* get rid of it. */
 248                 return (0);
 249         }
 250
 251         /*
 252          * Now lock the new node. We rely on the fact that we were passed
 253          * a locked vnode. If the lower node is exporting a struct lock
 254          * (v_vnlock != NULL) then we just set the upper v_vnlock to the
 255          * lower one, and both are now locked. If the lower node is exporting
 256          * NULL, then we copy that up and manually lock the upper node.
 257          *
 258          * LAYERFS_UPPERLOCK already has the test, so we use it after copying
 259          * up the v_vnlock from below.
 260          */
 261
 262         vp->v_vnlock = lowervp->v_vnlock;
 263         LAYERFS_UPPERLOCK(vp, LK_EXCLUSIVE, error);
 264         KASSERT(error == 0);
 265
 266         /*
 267          * Insert the new node into the hash.
 268          * Add a reference to the lower node.
 269          */
 270
 271         *vpp = vp;
 272         vref(lowervp);
 273         hd = LAYER_NHASH(lmp, lowervp);
 274         LIST_INSERT_HEAD(hd, xp, layer_hash);
 275         uvm_vnp_setsize(vp, 0);
 276         mutex_exit(&lmp->layerm_hashlock);
 277         return (0);
 278 }
 279
 280
 281 /*
 282  * Try to find an existing layer_node vnode refering
 283  * to it, otherwise make a new layer_node vnode which
 284  * contains a reference to the lower vnode.
 285  *
 286  * >>> we assume that the lower node is already locked upon entry, so we
 287  * propagate the lock state to upper node <<
 288  */
 289 int
 290 layer_node_create(struct mount *mp, struct vnode *lowervp, struct vnode **newvpp)
 291 {
 292         struct vnode *aliasvp;
 293         struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
 294
 295         mutex_enter(&lmp->layerm_hashlock);
 296         aliasvp = layer_node_find(mp, lowervp);
 297         if (aliasvp != NULL) {
 298                 /*
 299                  * layer_node_find has taken another reference
 300                  * to the alias vnode and moved the lock holding to
 301                  * aliasvp
 302                  */
 303 #ifdef LAYERFS_DIAGNOSTIC
 304                 if (layerfs_debug)
 305                         vprint("layer_node_create: exists", aliasvp);
 306 #endif
 307         } else {
 308                 int error;
 309
 310                 mutex_exit(&lmp->layerm_hashlock);
 311
 312                 /*
 313                  * Get new vnode.
 314                  */
 315 #ifdef LAYERFS_DIAGNOSTIC
 316                 if (layerfs_debug)
 317                         printf("layer_node_create: create new alias vnode\n");
 318 #endif
 319
 320                 /*
 321                  * Make new vnode reference the layer_node.
 322                  */
 323                 if ((error = (lmp->layerm_alloc)(mp, lowervp, &aliasvp)) != 0)
 324                         return error;
 325
 326                 /*
 327                  * aliasvp is already VREF'd by getnewvnode()
 328                  */
 329         }
 330
 331         /*
 332          * Now that we have VREF'd the upper vnode, release the reference
 333          * to the lower node. The existence of the layer_node retains one
 334          * reference to the lower node.
 335          */
 336         vrele(lowervp);
 337
 338 #ifdef DIAGNOSTIC
 339         if (lowervp->v_usecount < 1) {
 340                 /* Should never happen... */
 341                 vprint("layer_node_create: alias", aliasvp);
 342                 vprint("layer_node_create: lower", lowervp);
 343                 panic("layer_node_create: lower has 0 usecount.");
 344         }
 345 #endif
 346
 347 #ifdef LAYERFS_DIAGNOSTIC
 348         if (layerfs_debug)
 349                 vprint("layer_node_create: alias", aliasvp);
 350 #endif
 351         *newvpp = aliasvp;
 352         return (0);
 353 }
 354
 355 #ifdef LAYERFS_DIAGNOSTIC
 356 struct vnode *
 357 layer_checkvp(struct vnode *vp, const char *fil, int lno)
 358 {
 359         struct layer_node *a = VTOLAYER(vp);
 360 #ifdef notyet
 361         /*
 362          * Can't do this check because vop_reclaim runs
 363          * with a funny vop vector.
 364          *
 365          * WRS - no it doesnt...
 366          */
 367         if (vp->v_op != layer_vnodeop_p) {
 368                 printf ("layer_checkvp: on non-layer-node\n");
 369 #ifdef notyet
 370                 while (layer_checkvp_barrier) /*WAIT*/ ;
 371 #endif
 372                 panic("layer_checkvp");
 373         };
 374 #endif
 375         if (a->layer_lowervp == NULL) {
 376                 /* Should never happen */
 377                 int i; u_long *p;
 378                 printf("vp = %p, ZERO ptr\n", vp);
 379                 for (p = (u_long *) a, i = 0; i < 8; i++)
 380                         printf(" %lx", p[i]);
 381                 printf("\n");
 382                 /* wait for debugger */
 383                 panic("layer_checkvp");
 384         }
 385         if (a->layer_lowervp->v_usecount < 1) {
 386                 int i; u_long *p;
 387                 printf("vp = %p, unref'ed lowervp\n", vp);
 388                 for (p = (u_long *) a, i = 0; i < 8; i++)
 389                         printf(" %lx", p[i]);
 390                 printf("\n");
 391                 /* wait for debugger */
 392                 panic ("layer with unref'ed lowervp");
 393         };
 394 #ifdef notnow
 395         printf("layer %p/%d -> %p/%d [%s, %d]\n",
 396                 LAYERTOV(a), LAYERTOV(a)->v_usecount,
 397                 a->layer_lowervp, a->layer_lowervp->v_usecount,
 398                 fil, lno);
 399 #endif
 400         return a->layer_lowervp;
 401 }
 402 #endif