1/*- 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed 6 * to Berkeley by John Heidemann of the UCLA Ficus project. 7 * 8 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project 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 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)vfs_init.c 8.3 (Berkeley) 1/4/94 35 */ 36 37#include <sys/cdefs.h> 38__FBSDID("$FreeBSD$"); 39 40#include <sys/param.h> 41#include <sys/systm.h> 42#include <sys/fnv_hash.h> 43#include <sys/kernel.h> 44#include <sys/linker.h> 45#include <sys/mount.h> 46#include <sys/proc.h> 47#include <sys/sx.h> 48#include <sys/syscallsubr.h> 49#include <sys/sysctl.h> 50#include <sys/vnode.h> 51#include <sys/malloc.h> 52 53static int vfs_register(struct vfsconf *); 54static int vfs_unregister(struct vfsconf *); 55 56MALLOC_DEFINE(M_VNODE, "vnodes", "Dynamically allocated vnodes"); 57 58/* 59 * The highest defined VFS number. 60 */ 61int maxvfsconf = VFS_GENERIC + 1; 62 63/* 64 * Single-linked list of configured VFSes. 65 * New entries are added/deleted by vfs_register()/vfs_unregister() 66 */ 67struct vfsconfhead vfsconf = TAILQ_HEAD_INITIALIZER(vfsconf); 68struct sx vfsconf_sx; 69SX_SYSINIT(vfsconf, &vfsconf_sx, "vfsconf"); 70 71/* 72 * Loader.conf variable vfs.typenumhash enables setting vfc_typenum using a hash 73 * calculation on vfc_name, so that it doesn't change when file systems are 74 * loaded in a different order. This will avoid the NFS server file handles from 75 * changing for file systems that use vfc_typenum in their fsid. 76 */ 77static int vfs_typenumhash = 1; 78TUNABLE_INT("vfs.typenumhash", &vfs_typenumhash); 79SYSCTL_INT(_vfs, OID_AUTO, typenumhash, CTLFLAG_RDTUN, &vfs_typenumhash, 0, 80 "Set vfc_typenum using a hash calculation on vfc_name, so that it does not" 81 "change when file systems are loaded in a different order."); 82 83/* 84 * A Zen vnode attribute structure. 85 * 86 * Initialized when the first filesystem registers by vfs_register(). 87 */ 88struct vattr va_null; 89 90/* 91 * vfs_init.c 92 * 93 * Allocate and fill in operations vectors. 94 * 95 * An undocumented feature of this approach to defining operations is that 96 * there can be multiple entries in vfs_opv_descs for the same operations 97 * vector. This allows third parties to extend the set of operations 98 * supported by another layer in a binary compatibile way. For example, 99 * assume that NFS needed to be modified to support Ficus. NFS has an entry 100 * (probably nfs_vnopdeop_decls) declaring all the operations NFS supports by 101 * default. Ficus could add another entry (ficus_nfs_vnodeop_decl_entensions) 102 * listing those new operations Ficus adds to NFS, all without modifying the 103 * NFS code. (Of couse, the OTW NFS protocol still needs to be munged, but 104 * that is a(whole)nother story.) This is a feature. 105 */ 106 107/* 108 * Routines having to do with the management of the vnode table. 109 */ 110 111static struct vfsconf * 112vfs_byname_locked(const char *name) 113{ 114 struct vfsconf *vfsp; 115 116 sx_assert(&vfsconf_sx, SA_LOCKED); 117 if (!strcmp(name, "ffs")) 118 name = "ufs"; 119 TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) { 120 if (!strcmp(name, vfsp->vfc_name)) 121 return (vfsp); 122 } 123 return (NULL); 124} 125 126struct vfsconf * 127vfs_byname(const char *name) 128{ 129 struct vfsconf *vfsp; 130 131 vfsconf_slock(); 132 vfsp = vfs_byname_locked(name); 133 vfsconf_sunlock(); 134 return (vfsp); 135} 136 137struct vfsconf * 138vfs_byname_kld(const char *fstype, struct thread *td, int *error) 139{ 140 struct vfsconf *vfsp; 141 int fileid, loaded; 142 143 vfsp = vfs_byname(fstype); 144 if (vfsp != NULL) 145 return (vfsp); 146 147 /* Try to load the respective module. */ 148 *error = kern_kldload(td, fstype, &fileid); 149 loaded = (*error == 0); 150 if (*error == EEXIST) 151 *error = 0; 152 if (*error) 153 return (NULL); 154 155 /* Look up again to see if the VFS was loaded. */ 156 vfsp = vfs_byname(fstype); 157 if (vfsp == NULL) { 158 if (loaded) 159 (void)kern_kldunload(td, fileid, LINKER_UNLOAD_FORCE); 160 *error = ENODEV; 161 return (NULL); 162 } 163 return (vfsp); 164} 165 166 167/* Register a new filesystem type in the global table */ 168static int 169vfs_register(struct vfsconf *vfc) 170{ 171 struct sysctl_oid *oidp; 172 struct vfsops *vfsops; 173 static int once; 174 struct vfsconf *tvfc; 175 uint32_t hashval; 176 int secondpass; 177 178 if (!once) { 179 vattr_null(&va_null); 180 once = 1; 181 } 182 183 if (vfc->vfc_version != VFS_VERSION) { 184 printf("ERROR: filesystem %s, unsupported ABI version %x\n", 185 vfc->vfc_name, vfc->vfc_version); 186 return (EINVAL); 187 } 188 vfsconf_lock(); 189 if (vfs_byname_locked(vfc->vfc_name) != NULL) { 190 vfsconf_unlock(); 191 return (EEXIST); 192 } 193 194 if (vfs_typenumhash != 0) { 195 /* 196 * Calculate a hash on vfc_name to use for vfc_typenum. Unless 197 * all of 1<->255 are assigned, it is limited to 8bits since 198 * that is what ZFS uses from vfc_typenum and is also the 199 * preferred range for vfs_getnewfsid(). 200 */ 201 hashval = fnv_32_str(vfc->vfc_name, FNV1_32_INIT); 202 hashval &= 0xff; 203 secondpass = 0; 204 do { 205 /* Look for and fix any collision. */ 206 TAILQ_FOREACH(tvfc, &vfsconf, vfc_list) { 207 if (hashval == tvfc->vfc_typenum) { 208 if (hashval == 255 && secondpass == 0) { 209 hashval = 1; 210 secondpass = 1; 211 } else 212 hashval++; 213 break; 214 } 215 } 216 } while (tvfc != NULL); 217 vfc->vfc_typenum = hashval; 218 if (vfc->vfc_typenum >= maxvfsconf) 219 maxvfsconf = vfc->vfc_typenum + 1; 220 } else 221 vfc->vfc_typenum = maxvfsconf++; 222 TAILQ_INSERT_TAIL(&vfsconf, vfc, vfc_list); 223 224 /* 225 * Initialise unused ``struct vfsops'' fields, to use 226 * the vfs_std*() functions. Note, we need the mount 227 * and unmount operations, at the least. The check 228 * for vfsops available is just a debugging aid. 229 */ 230 KASSERT(vfc->vfc_vfsops != NULL, 231 ("Filesystem %s has no vfsops", vfc->vfc_name)); 232 /* 233 * Check the mount and unmount operations. 234 */ 235 vfsops = vfc->vfc_vfsops; 236 KASSERT(vfsops->vfs_mount != NULL, 237 ("Filesystem %s has no mount op", vfc->vfc_name)); 238 KASSERT(vfsops->vfs_unmount != NULL, 239 ("Filesystem %s has no unmount op", vfc->vfc_name)); 240 241 if (vfsops->vfs_root == NULL) 242 /* return file system's root vnode */ 243 vfsops->vfs_root = vfs_stdroot; 244 if (vfsops->vfs_quotactl == NULL) 245 /* quota control */ 246 vfsops->vfs_quotactl = vfs_stdquotactl; 247 if (vfsops->vfs_statfs == NULL) 248 /* return file system's status */ 249 vfsops->vfs_statfs = vfs_stdstatfs; 250 if (vfsops->vfs_sync == NULL) 251 /* 252 * flush unwritten data (nosync) 253 * file systems can use vfs_stdsync 254 * explicitly by setting it in the 255 * vfsop vector. 256 */ 257 vfsops->vfs_sync = vfs_stdnosync; 258 if (vfsops->vfs_vget == NULL) 259 /* convert an inode number to a vnode */ 260 vfsops->vfs_vget = vfs_stdvget; 261 if (vfsops->vfs_fhtovp == NULL) 262 /* turn an NFS file handle into a vnode */ 263 vfsops->vfs_fhtovp = vfs_stdfhtovp; 264 if (vfsops->vfs_checkexp == NULL) 265 /* check if file system is exported */ 266 vfsops->vfs_checkexp = vfs_stdcheckexp; 267 if (vfsops->vfs_init == NULL) 268 /* file system specific initialisation */ 269 vfsops->vfs_init = vfs_stdinit; 270 if (vfsops->vfs_uninit == NULL) 271 /* file system specific uninitialisation */ 272 vfsops->vfs_uninit = vfs_stduninit; 273 if (vfsops->vfs_extattrctl == NULL) 274 /* extended attribute control */ 275 vfsops->vfs_extattrctl = vfs_stdextattrctl; 276 if (vfsops->vfs_sysctl == NULL) 277 vfsops->vfs_sysctl = vfs_stdsysctl; 278 279 /* 280 * Call init function for this VFS... 281 */ 282 (*(vfc->vfc_vfsops->vfs_init))(vfc); 283 vfsconf_unlock(); 284 285 /* 286 * If this filesystem has a sysctl node under vfs 287 * (i.e. vfs.xxfs), then change the oid number of that node to 288 * match the filesystem's type number. This allows user code 289 * which uses the type number to read sysctl variables defined 290 * by the filesystem to continue working. Since the oids are 291 * in a sorted list, we need to make sure the order is 292 * preserved by re-registering the oid after modifying its 293 * number. 294 */ 295 sysctl_lock(); 296 SLIST_FOREACH(oidp, &sysctl__vfs_children, oid_link) { 297 if (strcmp(oidp->oid_name, vfc->vfc_name) == 0) { 298 sysctl_unregister_oid(oidp); 299 oidp->oid_number = vfc->vfc_typenum; 300 sysctl_register_oid(oidp); 301 break; 302 } 303 } 304 sysctl_unlock(); 305 306 return (0); 307} 308 309 310/* Remove registration of a filesystem type */ 311static int 312vfs_unregister(struct vfsconf *vfc) 313{ 314 struct vfsconf *vfsp; 315 int error, i, maxtypenum; 316 317 i = vfc->vfc_typenum; 318 319 vfsconf_lock(); 320 vfsp = vfs_byname_locked(vfc->vfc_name); 321 if (vfsp == NULL) { 322 vfsconf_unlock(); 323 return (EINVAL); 324 } 325 if (vfsp->vfc_refcount != 0) { 326 vfsconf_unlock(); 327 return (EBUSY); 328 } 329 if (vfc->vfc_vfsops->vfs_uninit != NULL) { 330 error = (*vfc->vfc_vfsops->vfs_uninit)(vfsp); 331 if (error != 0) { 332 vfsconf_unlock(); 333 return (error); 334 } 335 } 336 TAILQ_REMOVE(&vfsconf, vfsp, vfc_list); 337 maxtypenum = VFS_GENERIC; 338 TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) 339 if (maxtypenum < vfsp->vfc_typenum) 340 maxtypenum = vfsp->vfc_typenum; 341 maxvfsconf = maxtypenum + 1; 342 vfsconf_unlock(); 343 return (0); 344} 345 346/* 347 * Standard kernel module handling code for filesystem modules. 348 * Referenced from VFS_SET(). 349 */ 350int 351vfs_modevent(module_t mod, int type, void *data) 352{ 353 struct vfsconf *vfc; 354 int error = 0; 355 356 vfc = (struct vfsconf *)data; 357 358 switch (type) { 359 case MOD_LOAD: 360 if (vfc) 361 error = vfs_register(vfc); 362 break; 363 364 case MOD_UNLOAD: 365 if (vfc) 366 error = vfs_unregister(vfc); 367 break; 368 default: 369 error = EOPNOTSUPP; 370 break; 371 } 372 return (error); 373} 374