bsd/miscfs/bindfs/bind_subr.c

   1 /*
   2  * Copyright (c) 2019 Apple Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. Please obtain a copy of the License at
  10  * http://www.opensource.apple.com/apsl/ and read it before using this
  11  * file.
  12  *
  13  * The Original Code and all software distributed under the License are
  14  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  15  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  16  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  18  * Please see the License for the specific language governing rights and
  19  * limitations under the License.
  20  *
  21  * @APPLE_LICENSE_HEADER_END@
  22  */
  23
  24 /*-
  25  * Portions Copyright (c) 1992, 1993
  26  *  The Regents of the University of California.  All rights reserved.
  27  *
  28  * This code is derived from software donated to Berkeley by
  29  * Jan-Simon Pendry.
  30  *
  31  * Redistribution and use in source and binary forms, with or without
  32  * modification, are permitted provided that the following conditions
  33  * are met:
  34  * 1. Redistributions of source code must retain the above copyright
  35  *    notice, this list of conditions and the following disclaimer.
  36  * 2. Redistributions in binary form must reproduce the above copyright
  37  *    notice, this list of conditions and the following disclaimer in the
  38  *    documentation and/or other materials provided with the distribution.
  39  * 4. Neither the name of the University nor the names of its contributors
  40  *    may be used to endorse or promote products derived from this software
  41  *    without specific prior written permission.
  42  *
  43  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  44  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  45  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  46  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  47  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  48  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  49  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  50  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  51  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  52  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  53  * SUCH DAMAGE.
  54  *
  55  *  @(#)null_subr.c 8.7 (Berkeley) 5/14/95
  56  *
  57  * $FreeBSD$
  58  */
  59 #include <sys/param.h>
  60 #include <sys/systm.h>
  61 #include <sys/kernel.h>
  62 #include <sys/lock.h>
  63 #include <sys/malloc.h>
  64 #include <sys/mount.h>
  65 #include <sys/proc.h>
  66 #include <sys/vnode.h>
  67
  68 #include "bindfs.h"
  69
  70 /*
  71  * Null layer cache:
  72  * Each cache entry holds a reference to the lower vnode
  73  * along with a pointer to the alias vnode.  When an
  74  * entry is added the lower vnode is VREF'd.  When the
  75  * alias is removed the lower vnode is vrele'd.
  76  */
  77
  78 #define BIND_HASH_SIZE (desiredvnodes / 10)
  79
  80 /* xnu doesn't really have the functionality freebsd uses here..gonna try this
  81  * hacked hash...*/
  82 #define BIND_NHASH(vp) (&bind_node_hashtbl[((((uintptr_t)vp) >> vnsz2log) + (uintptr_t)vnode_mount(vp)) & bind_hash_mask])
  83
  84 static LIST_HEAD(bind_node_hashhead, bind_node) * bind_node_hashtbl;
  85 static LCK_GRP_DECLARE(bind_hashlck_grp, "com.apple.filesystems.bindfs");
  86 static LCK_MTX_DECLARE(bind_hashmtx, &bind_hashlck_grp);
  87 static u_long bind_hash_mask;
  88
  89 /*  xnu doesn't have hashes built into vnodes. This mimics what freebsd does
  90  *  9 is an eyeball of the log 2 size of vnode */
  91 static int vnsz2log = 9;
  92
  93 static int bind_hashins(struct mount *, struct bind_node *, struct vnode **);
  94
  95 /*
  96  * Initialise cache headers
  97  */
  98 int
  99 bindfs_init(__unused struct vfsconf * vfsp)
 100 {
 101         BINDFSDEBUG("%s\n", __FUNCTION__);
 102
 103         bind_node_hashtbl = hashinit(BIND_HASH_SIZE, M_TEMP, &bind_hash_mask);
 104         if (bind_node_hashtbl == NULL) {
 105                 goto error;
 106         }
 107
 108         BINDFSDEBUG("%s finished\n", __FUNCTION__);
 109         return 0;
 110 error:
 111         printf("BINDFS: failed to initialize globals\n");
 112         return KERN_FAILURE;
 113 }
 114
 115 int
 116 bindfs_destroy(void)
 117 {
 118         /* This gets called when the fs is uninstalled, there wasn't an exact
 119          * equivalent in vfsops */
 120         hashdestroy(bind_node_hashtbl, M_TEMP, bind_hash_mask);
 121         return 0;
 122 }
 123
 124 /*
 125  * Find the bindfs vnode mapped to lowervp. Return it in *vpp with an iocount if found.
 126  * Return 0 on success. On failure *vpp will be NULL and a non-zero error code will be returned.
 127  */
 128 int
 129 bind_hashget(struct mount * mp, struct vnode * lowervp, struct vnode ** vpp)
 130 {
 131         struct bind_node_hashhead * hd;
 132         struct bind_node * a;
 133         struct vnode * vp = NULL;
 134         uint32_t vp_vid = 0;
 135         int error = ENOENT;
 136
 137         /*
 138          * Find hash base, and then search the (two-way) linked
 139          * list looking for a bind_node structure which is referencing
 140          * the lower vnode. We only give up our reference at reclaim so
 141          * just check whether the lowervp has gotten pulled from under us
 142          */
 143         hd = BIND_NHASH(lowervp);
 144         lck_mtx_lock(&bind_hashmtx);
 145         LIST_FOREACH(a, hd, bind_hash)
 146         {
 147                 if (a->bind_lowervp == lowervp && vnode_mount(BINDTOV(a)) == mp) {
 148                         vp = BINDTOV(a);
 149                         if (a->bind_lowervid != vnode_vid(lowervp)) {
 150                                 /*lowervp has reved */
 151                                 error = EIO;
 152                                 vp = NULL;
 153                         } else {
 154                                 vp_vid = a->bind_myvid;
 155                         }
 156                         break;
 157                 }
 158         }
 159         lck_mtx_unlock(&bind_hashmtx);
 160
 161         if (vp != NULL) {
 162                 error = vnode_getwithvid(vp, vp_vid);
 163                 if (error == 0) {
 164                         *vpp = vp;
 165                 }
 166         }
 167         return error;
 168 }
 169
 170 /*
 171  * Act like bind_hashget, but add passed bind_node to hash if no existing
 172  * node found.
 173  * If we find a vnode in the hash table it is returned via vpp. If we don't
 174  * find a hit in the table, then vpp is NULL on return and xp is added to the table.
 175  * 0 is returned if a hash table hit occurs or if we insert the bind_node.
 176  * EIO is returned if we found a hash table hit but the lower vnode was recycled.
 177  */
 178 static int
 179 bind_hashins(struct mount * mp, struct bind_node * xp, struct vnode ** vpp)
 180 {
 181         struct bind_node_hashhead * hd;
 182         struct bind_node * oxp;
 183         struct vnode * ovp = NULL;
 184         uint32_t oxp_vid = 0;
 185         int error = 0;
 186
 187         hd = BIND_NHASH(xp->bind_lowervp);
 188         lck_mtx_lock(&bind_hashmtx);
 189         LIST_FOREACH(oxp, hd, bind_hash)
 190         {
 191                 if (oxp->bind_lowervp == xp->bind_lowervp && vnode_mount(BINDTOV(oxp)) == mp) {
 192                         ovp = BINDTOV(oxp);
 193                         if (oxp->bind_lowervid != vnode_vid(oxp->bind_lowervp)) {
 194                                 /*      vp doesn't exist so return null (not sure we are actually gonna catch
 195                                  *  recycle right now
 196                                  *  This is an exceptional case right now, it suggests the vnode we are
 197                                  *  trying to add has been recycled
 198                                  *  don't add it.*/
 199                                 error = EIO;
 200                                 ovp = NULL;
 201                         } else {
 202                                 oxp_vid = oxp->bind_myvid;
 203                         }
 204                         goto end;
 205                 }
 206         }
 207         /* if it wasn't in the hash map then the vnode pointed to by xp already has a
 208          * iocount so don't get another. */
 209         LIST_INSERT_HEAD(hd, xp, bind_hash);
 210         xp->bind_flags |= BIND_FLAG_HASHED;
 211 end:
 212         lck_mtx_unlock(&bind_hashmtx);
 213         if (ovp != NULL) {
 214                 /* if we found something in the hash map then grab an iocount */
 215                 error = vnode_getwithvid(ovp, oxp_vid);
 216                 if (error == 0) {
 217                         *vpp = ovp;
 218                 }
 219         }
 220         return error;
 221 }
 222
 223 /*
 224  * Remove node from hash.
 225  */
 226 void
 227 bind_hashrem(struct bind_node * xp)
 228 {
 229         if (xp->bind_flags & BIND_FLAG_HASHED) {
 230                 lck_mtx_lock(&bind_hashmtx);
 231                 LIST_REMOVE(xp, bind_hash);
 232                 lck_mtx_unlock(&bind_hashmtx);
 233         }
 234 }
 235
 236 static struct bind_node *
 237 bind_nodecreate(struct vnode * lowervp)
 238 {
 239         struct bind_node * xp;
 240
 241         MALLOC(xp, struct bind_node *, sizeof(struct bind_node), M_TEMP, M_WAITOK | M_ZERO);
 242         if (xp != NULL) {
 243                 if (lowervp) {
 244                         xp->bind_lowervp  = lowervp;
 245                         xp->bind_lowervid = vnode_vid(lowervp);
 246                 }
 247         }
 248         return xp;
 249 }
 250
 251 /* assumption is that vnode has iocount on it after vnode create */
 252 int
 253 bind_getnewvnode(
 254         struct mount * mp, struct vnode * lowervp, struct vnode * dvp, struct vnode ** vpp, struct componentname * cnp, int root)
 255 {
 256         struct vnode_fsparam vnfs_param;
 257         int error             = 0;
 258         enum vtype type       = VDIR;
 259         struct bind_node * xp = bind_nodecreate(lowervp);
 260
 261         if (xp == NULL) {
 262                 return ENOMEM;
 263         }
 264
 265         if (lowervp) {
 266                 type = vnode_vtype(lowervp);
 267         }
 268
 269         vnfs_param.vnfs_mp         = mp;
 270         vnfs_param.vnfs_vtype      = type;
 271         vnfs_param.vnfs_str        = "bindfs";
 272         vnfs_param.vnfs_dvp        = dvp;
 273         vnfs_param.vnfs_fsnode     = (void *)xp;
 274         vnfs_param.vnfs_vops       = bindfs_vnodeop_p;
 275         vnfs_param.vnfs_markroot   = root;
 276         vnfs_param.vnfs_marksystem = 0;
 277         vnfs_param.vnfs_rdev       = 0;
 278         vnfs_param.vnfs_filesize   = 0; // set this to 0 since we should only be shadowing non-regular files
 279         vnfs_param.vnfs_cnp        = cnp;
 280         vnfs_param.vnfs_flags      = VNFS_ADDFSREF;
 281
 282         error = vnode_create(VNCREATE_FLAVOR, VCREATESIZE, &vnfs_param, vpp);
 283         if (error == 0) {
 284                 xp->bind_vnode = *vpp;
 285                 xp->bind_myvid = vnode_vid(*vpp);
 286                 vnode_settag(*vpp, VT_BINDFS);
 287         } else {
 288                 FREE(xp, M_TEMP);
 289         }
 290         return error;
 291 }
 292
 293 /*
 294  * Make a new or get existing bindfs node.
 295  * Vp is the alias vnode, lowervp is the lower vnode.
 296  *
 297  * lowervp is assumed to have an iocount on it from the caller
 298  */
 299 int
 300 bind_nodeget(
 301         struct mount * mp, struct vnode * lowervp, struct vnode * dvp, struct vnode ** vpp, struct componentname * cnp, int root)
 302 {
 303         struct vnode * vp;
 304         int error;
 305
 306         /* Lookup the hash firstly. */
 307         error = bind_hashget(mp, lowervp, vpp);
 308         /* ENOENT means it wasn't found, EIO is a failure we should bail from, 0 is it
 309          * was found */
 310         if (error != ENOENT) {
 311                 /* bind_hashget checked the vid, so if we got something here its legit to
 312                  * the best of our knowledge*/
 313                 /* if we found something then there is an iocount on vpp,
 314                  *  if we didn't find something then vpp shouldn't be used by the caller */
 315                 return error;
 316         }
 317
 318         /*
 319          * We do not serialize vnode creation, instead we will check for
 320          * duplicates later, when adding new vnode to hash.
 321          */
 322         error = vnode_ref(lowervp); // take a ref on lowervp so we let the system know we care about it
 323         if (error) {
 324                 // Failed to get a reference on the lower vp so bail. Lowervp may be gone already.
 325                 return error;
 326         }
 327
 328         error = bind_getnewvnode(mp, lowervp, dvp, &vp, cnp, root);
 329
 330         if (error) {
 331                 vnode_rele(lowervp);
 332                 return error;
 333         }
 334
 335         /*
 336          * Atomically insert our new node into the hash or vget existing
 337          * if someone else has beaten us to it.
 338          */
 339         error = bind_hashins(mp, VTOBIND(vp), vpp);
 340         if (error || *vpp != NULL) {
 341                 /* recycle will call reclaim which will get rid of the internals */
 342                 vnode_recycle(vp);
 343                 vnode_put(vp);
 344                 /* if we found vpp, then bind_hashins put an iocount on it */
 345                 return error;
 346         }
 347
 348         /* vp has an iocount from bind_getnewvnode */
 349         *vpp = vp;
 350
 351         return 0;
 352 }