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_mtx_t bind_hashmtx;
  86 static lck_attr_t * bind_hashlck_attr;
  87 static lck_grp_t * bind_hashlck_grp;
  88 static lck_grp_attr_t * bind_hashlck_grp_attr;
  89 static u_long bind_hash_mask;
  90
  91 /*  xnu doesn't have hashes built into vnodes. This mimics what freebsd does
  92  *  9 is an eyeball of the log 2 size of vnode */
  93 static int vnsz2log = 9;
  94
  95 static int bind_hashins(struct mount *, struct bind_node *, struct vnode **);
  96
  97 int
  98 bindfs_init_lck(lck_mtx_t * lck)
  99 {
 100         int error = 1;
 101         if (lck && bind_hashlck_grp && bind_hashlck_attr) {
 102                 lck_mtx_init(lck, bind_hashlck_grp, bind_hashlck_attr);
 103                 error = 0;
 104         }
 105         return error;
 106 }
 107
 108 int
 109 bindfs_destroy_lck(lck_mtx_t * lck)
 110 {
 111         int error = 1;
 112         if (lck && bind_hashlck_grp) {
 113                 lck_mtx_destroy(lck, bind_hashlck_grp);
 114                 error = 0;
 115         }
 116         return error;
 117 }
 118
 119 /*
 120  * Initialise cache headers
 121  */
 122 int
 123 bindfs_init(__unused struct vfsconf * vfsp)
 124 {
 125         BINDFSDEBUG("%s\n", __FUNCTION__);
 126
 127         /* assuming for now that this happens immediately and by default after fs
 128          * installation */
 129         bind_hashlck_grp_attr = lck_grp_attr_alloc_init();
 130         if (bind_hashlck_grp_attr == NULL) {
 131                 goto error;
 132         }
 133         bind_hashlck_grp = lck_grp_alloc_init("com.apple.filesystems.bindfs", bind_hashlck_grp_attr);
 134         if (bind_hashlck_grp == NULL) {
 135                 goto error;
 136         }
 137         bind_hashlck_attr = lck_attr_alloc_init();
 138         if (bind_hashlck_attr == NULL) {
 139                 goto error;
 140         }
 141
 142         bind_node_hashtbl = hashinit(BIND_HASH_SIZE, M_TEMP, &bind_hash_mask);
 143         if (bind_node_hashtbl == NULL) {
 144                 goto error;
 145         }
 146         lck_mtx_init(&bind_hashmtx, bind_hashlck_grp, bind_hashlck_attr);
 147
 148         BINDFSDEBUG("%s finished\n", __FUNCTION__);
 149         return 0;
 150 error:
 151         printf("BINDFS: failed to initialize globals\n");
 152         if (bind_hashlck_grp_attr) {
 153                 lck_grp_attr_free(bind_hashlck_grp_attr);
 154                 bind_hashlck_grp_attr = NULL;
 155         }
 156         if (bind_hashlck_grp) {
 157                 lck_grp_free(bind_hashlck_grp);
 158                 bind_hashlck_grp = NULL;
 159         }
 160         if (bind_hashlck_attr) {
 161                 lck_attr_free(bind_hashlck_attr);
 162                 bind_hashlck_attr = NULL;
 163         }
 164         return KERN_FAILURE;
 165 }
 166
 167 int
 168 bindfs_destroy(void)
 169 {
 170         /* This gets called when the fs is uninstalled, there wasn't an exact
 171          * equivalent in vfsops */
 172         lck_mtx_destroy(&bind_hashmtx, bind_hashlck_grp);
 173         hashdestroy(bind_node_hashtbl, M_TEMP, bind_hash_mask);
 174         if (bind_hashlck_grp_attr) {
 175                 lck_grp_attr_free(bind_hashlck_grp_attr);
 176                 bind_hashlck_grp_attr = NULL;
 177         }
 178         if (bind_hashlck_grp) {
 179                 lck_grp_free(bind_hashlck_grp);
 180                 bind_hashlck_grp = NULL;
 181         }
 182         if (bind_hashlck_attr) {
 183                 lck_attr_free(bind_hashlck_attr);
 184                 bind_hashlck_attr = NULL;
 185         }
 186         return 0;
 187 }
 188
 189 /*
 190  * Find the bindfs vnode mapped to lowervp. Return it in *vpp with an iocount if found.
 191  * Return 0 on success. On failure *vpp will be NULL and a non-zero error code will be returned.
 192  */
 193 int
 194 bind_hashget(struct mount * mp, struct vnode * lowervp, struct vnode ** vpp)
 195 {
 196         struct bind_node_hashhead * hd;
 197         struct bind_node * a;
 198         struct vnode * vp = NULL;
 199         int error = ENOENT;
 200
 201         /*
 202          * Find hash base, and then search the (two-way) linked
 203          * list looking for a bind_node structure which is referencing
 204          * the lower vnode. We only give up our reference at reclaim so
 205          * just check whether the lowervp has gotten pulled from under us
 206          */
 207         hd = BIND_NHASH(lowervp);
 208         lck_mtx_lock(&bind_hashmtx);
 209         LIST_FOREACH(a, hd, bind_hash)
 210         {
 211                 if (a->bind_lowervp == lowervp && vnode_mount(BINDTOV(a)) == mp) {
 212                         vp = BINDTOV(a);
 213                         if (a->bind_lowervid != vnode_vid(lowervp)) {
 214                                 /*lowervp has reved */
 215                                 error = EIO;
 216                                 vp = NULL;
 217                         }
 218                         break;
 219                 }
 220         }
 221         lck_mtx_unlock(&bind_hashmtx);
 222
 223         if (vp != NULL) {
 224                 error = vnode_getwithvid(vp, a->bind_myvid);
 225                 if (error == 0) {
 226                         *vpp = vp;
 227                 }
 228         }
 229         return error;
 230 }
 231
 232 /*
 233  * Act like bind_hashget, but add passed bind_node to hash if no existing
 234  * node found.
 235  * If we find a vnode in the hash table it is returned via vpp. If we don't
 236  * find a hit in the table, then vpp is NULL on return and xp is added to the table.
 237  * 0 is returned if a hash table hit occurs or if we insert the bind_node.
 238  * EIO is returned if we found a hash table hit but the lower vnode was recycled.
 239  */
 240 static int
 241 bind_hashins(struct mount * mp, struct bind_node * xp, struct vnode ** vpp)
 242 {
 243         struct bind_node_hashhead * hd;
 244         struct bind_node * oxp;
 245         struct vnode * ovp = NULL;
 246         int error = 0;
 247
 248         hd = BIND_NHASH(xp->bind_lowervp);
 249         lck_mtx_lock(&bind_hashmtx);
 250         LIST_FOREACH(oxp, hd, bind_hash)
 251         {
 252                 if (oxp->bind_lowervp == xp->bind_lowervp && vnode_mount(BINDTOV(oxp)) == mp) {
 253                         ovp = BINDTOV(oxp);
 254                         if (oxp->bind_lowervid != vnode_vid(oxp->bind_lowervp)) {
 255                                 /*      vp doesn't exist so return null (not sure we are actually gonna catch
 256                                  *  recycle right now
 257                                  *  This is an exceptional case right now, it suggests the vnode we are
 258                                  *  trying to add has been recycled
 259                                  *  don't add it.*/
 260                                 error = EIO;
 261                                 ovp = NULL;
 262                         }
 263                         goto end;
 264                 }
 265         }
 266         /* if it wasn't in the hash map then the vnode pointed to by xp already has a
 267          * iocount so don't get another. */
 268         LIST_INSERT_HEAD(hd, xp, bind_hash);
 269         xp->bind_flags |= BIND_FLAG_HASHED;
 270 end:
 271         lck_mtx_unlock(&bind_hashmtx);
 272         if (ovp != NULL) {
 273                 /* if we found something in the hash map then grab an iocount */
 274                 error = vnode_getwithvid(ovp, oxp->bind_myvid);
 275                 if (error == 0) {
 276                         *vpp = ovp;
 277                 }
 278         }
 279         return error;
 280 }
 281
 282 /*
 283  * Remove node from hash.
 284  */
 285 void
 286 bind_hashrem(struct bind_node * xp)
 287 {
 288         if (xp->bind_flags & BIND_FLAG_HASHED) {
 289                 lck_mtx_lock(&bind_hashmtx);
 290                 LIST_REMOVE(xp, bind_hash);
 291                 lck_mtx_unlock(&bind_hashmtx);
 292         }
 293 }
 294
 295 static struct bind_node *
 296 bind_nodecreate(struct vnode * lowervp)
 297 {
 298         struct bind_node * xp;
 299
 300         MALLOC(xp, struct bind_node *, sizeof(struct bind_node), M_TEMP, M_WAITOK | M_ZERO);
 301         if (xp != NULL) {
 302                 if (lowervp) {
 303                         xp->bind_lowervp  = lowervp;
 304                         xp->bind_lowervid = vnode_vid(lowervp);
 305                 }
 306         }
 307         return xp;
 308 }
 309
 310 /* assumption is that vnode has iocount on it after vnode create */
 311 int
 312 bind_getnewvnode(
 313         struct mount * mp, struct vnode * lowervp, struct vnode * dvp, struct vnode ** vpp, struct componentname * cnp, int root)
 314 {
 315         struct vnode_fsparam vnfs_param;
 316         int error             = 0;
 317         enum vtype type       = VDIR;
 318         struct bind_node * xp = bind_nodecreate(lowervp);
 319
 320         if (xp == NULL) {
 321                 return ENOMEM;
 322         }
 323
 324         if (lowervp) {
 325                 type = vnode_vtype(lowervp);
 326         }
 327
 328         vnfs_param.vnfs_mp         = mp;
 329         vnfs_param.vnfs_vtype      = type;
 330         vnfs_param.vnfs_str        = "bindfs";
 331         vnfs_param.vnfs_dvp        = dvp;
 332         vnfs_param.vnfs_fsnode     = (void *)xp;
 333         vnfs_param.vnfs_vops       = bindfs_vnodeop_p;
 334         vnfs_param.vnfs_markroot   = root;
 335         vnfs_param.vnfs_marksystem = 0;
 336         vnfs_param.vnfs_rdev       = 0;
 337         vnfs_param.vnfs_filesize   = 0; // set this to 0 since we should only be shadowing non-regular files
 338         vnfs_param.vnfs_cnp        = cnp;
 339         vnfs_param.vnfs_flags      = VNFS_ADDFSREF;
 340
 341         error = vnode_create(VNCREATE_FLAVOR, VCREATESIZE, &vnfs_param, vpp);
 342         if (error == 0) {
 343                 xp->bind_vnode = *vpp;
 344                 xp->bind_myvid = vnode_vid(*vpp);
 345                 vnode_settag(*vpp, VT_BINDFS);
 346         } else {
 347                 FREE(xp, M_TEMP);
 348         }
 349         return error;
 350 }
 351
 352 /*
 353  * Make a new or get existing bindfs node.
 354  * Vp is the alias vnode, lowervp is the lower vnode.
 355  *
 356  * lowervp is assumed to have an iocount on it from the caller
 357  */
 358 int
 359 bind_nodeget(
 360         struct mount * mp, struct vnode * lowervp, struct vnode * dvp, struct vnode ** vpp, struct componentname * cnp, int root)
 361 {
 362         struct vnode * vp;
 363         int error;
 364
 365         /* Lookup the hash firstly. */
 366         error = bind_hashget(mp, lowervp, vpp);
 367         /* ENOENT means it wasn't found, EIO is a failure we should bail from, 0 is it
 368          * was found */
 369         if (error != ENOENT) {
 370                 /* bind_hashget checked the vid, so if we got something here its legit to
 371                  * the best of our knowledge*/
 372                 /* if we found something then there is an iocount on vpp,
 373                  *  if we didn't find something then vpp shouldn't be used by the caller */
 374                 return error;
 375         }
 376
 377         /*
 378          * We do not serialize vnode creation, instead we will check for
 379          * duplicates later, when adding new vnode to hash.
 380          */
 381         error = vnode_ref(lowervp); // take a ref on lowervp so we let the system know we care about it
 382         if (error) {
 383                 // Failed to get a reference on the lower vp so bail. Lowervp may be gone already.
 384                 return error;
 385         }
 386
 387         error = bind_getnewvnode(mp, lowervp, dvp, &vp, cnp, root);
 388
 389         if (error) {
 390                 vnode_rele(lowervp);
 391                 return error;
 392         }
 393
 394         /*
 395          * Atomically insert our new node into the hash or vget existing
 396          * if someone else has beaten us to it.
 397          */
 398         error = bind_hashins(mp, VTOBIND(vp), vpp);
 399         if (error || *vpp != NULL) {
 400                 /* recycle will call reclaim which will get rid of the internals */
 401                 vnode_recycle(vp);
 402                 vnode_put(vp);
 403                 /* if we found vpp, then bind_hashins put an iocount on it */
 404                 return error;
 405         }
 406
 407         /* vp has an iocount from bind_getnewvnode */
 408         *vpp = vp;
 409
 410         return 0;
 411 }