1139804Simp/*- 21541Srgrimes * Copyright (c) 1989, 1993 31541Srgrimes * The Regents of the University of California. All rights reserved. 41541Srgrimes * 51541Srgrimes * This code is derived from software contributed 61541Srgrimes * to Berkeley by John Heidemann of the UCLA Ficus project. 71541Srgrimes * 81541Srgrimes * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project 91541Srgrimes * 101541Srgrimes * Redistribution and use in source and binary forms, with or without 111541Srgrimes * modification, are permitted provided that the following conditions 121541Srgrimes * are met: 131541Srgrimes * 1. Redistributions of source code must retain the above copyright 141541Srgrimes * notice, this list of conditions and the following disclaimer. 151541Srgrimes * 2. Redistributions in binary form must reproduce the above copyright 161541Srgrimes * notice, this list of conditions and the following disclaimer in the 171541Srgrimes * documentation and/or other materials provided with the distribution. 181541Srgrimes * 4. Neither the name of the University nor the names of its contributors 191541Srgrimes * may be used to endorse or promote products derived from this software 201541Srgrimes * without specific prior written permission. 211541Srgrimes * 221541Srgrimes * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 231541Srgrimes * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 241541Srgrimes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 251541Srgrimes * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 261541Srgrimes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 271541Srgrimes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 281541Srgrimes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 291541Srgrimes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 301541Srgrimes * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 311541Srgrimes * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 321541Srgrimes * SUCH DAMAGE. 331541Srgrimes * 341541Srgrimes * @(#)vfs_init.c 8.3 (Berkeley) 1/4/94 351541Srgrimes */ 361541Srgrimes 37116182Sobrien#include <sys/cdefs.h> 38116182Sobrien__FBSDID("$FreeBSD$"); 391541Srgrimes 401541Srgrimes#include <sys/param.h> 412112Swollman#include <sys/systm.h> 42225537Srmacklem#include <sys/fnv_hash.h> 432946Swollman#include <sys/kernel.h> 44138350Sphk#include <sys/linker.h> 451541Srgrimes#include <sys/mount.h> 46138350Sphk#include <sys/proc.h> 47269457Skib#include <sys/sx.h> 48159590Sjhb#include <sys/syscallsubr.h> 4938869Sbde#include <sys/sysctl.h> 501541Srgrimes#include <sys/vnode.h> 511541Srgrimes#include <sys/malloc.h> 521541Srgrimes 53141634Sphkstatic int vfs_register(struct vfsconf *); 54141634Sphkstatic int vfs_unregister(struct vfsconf *); 5512577Sbde 5630354SphkMALLOC_DEFINE(M_VNODE, "vnodes", "Dynamically allocated vnodes"); 5730354Sphk 5810358Sjulian/* 5969664Speter * The highest defined VFS number. 6029653Sdyson */ 6169664Speterint maxvfsconf = VFS_GENERIC + 1; 6291690Seivind 6391690Seivind/* 6491690Seivind * Single-linked list of configured VFSes. 6591690Seivind * New entries are added/deleted by vfs_register()/vfs_unregister() 6691690Seivind */ 67132710Sphkstruct vfsconfhead vfsconf = TAILQ_HEAD_INITIALIZER(vfsconf); 68269457Skibstruct sx vfsconf_sx; 69269457SkibSX_SYSINIT(vfsconf, &vfsconf_sx, "vfsconf"); 7052780Smsmith 7152780Smsmith/* 72225537Srmacklem * Loader.conf variable vfs.typenumhash enables setting vfc_typenum using a hash 73225537Srmacklem * calculation on vfc_name, so that it doesn't change when file systems are 74225537Srmacklem * loaded in a different order. This will avoid the NFS server file handles from 75225537Srmacklem * changing for file systems that use vfc_typenum in their fsid. 76225537Srmacklem */ 77225537Srmacklemstatic int vfs_typenumhash = 1; 78225537SrmacklemSYSCTL_INT(_vfs, OID_AUTO, typenumhash, CTLFLAG_RDTUN, &vfs_typenumhash, 0, 79225537Srmacklem "Set vfc_typenum using a hash calculation on vfc_name, so that it does not" 80225537Srmacklem "change when file systems are loaded in a different order."); 81225537Srmacklem 82225537Srmacklem/* 83135279Sphk * A Zen vnode attribute structure. 84135279Sphk * 85135279Sphk * Initialized when the first filesystem registers by vfs_register(). 86135279Sphk */ 87135279Sphkstruct vattr va_null; 88135279Sphk 89135279Sphk/* 901541Srgrimes * vfs_init.c 911541Srgrimes * 921541Srgrimes * Allocate and fill in operations vectors. 931541Srgrimes * 941541Srgrimes * An undocumented feature of this approach to defining operations is that 951541Srgrimes * there can be multiple entries in vfs_opv_descs for the same operations 961541Srgrimes * vector. This allows third parties to extend the set of operations 971541Srgrimes * supported by another layer in a binary compatibile way. For example, 981541Srgrimes * assume that NFS needed to be modified to support Ficus. NFS has an entry 991541Srgrimes * (probably nfs_vnopdeop_decls) declaring all the operations NFS supports by 1001541Srgrimes * default. Ficus could add another entry (ficus_nfs_vnodeop_decl_entensions) 1011541Srgrimes * listing those new operations Ficus adds to NFS, all without modifying the 1021541Srgrimes * NFS code. (Of couse, the OTW NFS protocol still needs to be munged, but 1031541Srgrimes * that is a(whole)nother story.) This is a feature. 1041541Srgrimes */ 10541056Speter 106138290Sphk/* 107138290Sphk * Routines having to do with the management of the vnode table. 108138290Sphk */ 10941056Speter 110269457Skibstatic struct vfsconf * 111269457Skibvfs_byname_locked(const char *name) 112132710Sphk{ 113132710Sphk struct vfsconf *vfsp; 114132710Sphk 115269457Skib sx_assert(&vfsconf_sx, SA_LOCKED); 116138497Sphk if (!strcmp(name, "ffs")) 117138497Sphk name = "ufs"; 118269457Skib TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) { 119132710Sphk if (!strcmp(name, vfsp->vfc_name)) 120132710Sphk return (vfsp); 121269457Skib } 122132710Sphk return (NULL); 123132710Sphk} 124132710Sphk 125138350Sphkstruct vfsconf * 126269457Skibvfs_byname(const char *name) 127269457Skib{ 128269457Skib struct vfsconf *vfsp; 129269457Skib 130269457Skib vfsconf_slock(); 131269457Skib vfsp = vfs_byname_locked(name); 132269457Skib vfsconf_sunlock(); 133269457Skib return (vfsp); 134269457Skib} 135269457Skib 136269457Skibstruct vfsconf * 137138350Sphkvfs_byname_kld(const char *fstype, struct thread *td, int *error) 138138350Sphk{ 139138350Sphk struct vfsconf *vfsp; 140247071Sjamie int fileid, loaded; 141138350Sphk 142138350Sphk vfsp = vfs_byname(fstype); 143138350Sphk if (vfsp != NULL) 144138350Sphk return (vfsp); 145138350Sphk 146159956Sjhb /* Try to load the respective module. */ 147159590Sjhb *error = kern_kldload(td, fstype, &fileid); 148247071Sjamie loaded = (*error == 0); 149247071Sjamie if (*error == EEXIST) 150247071Sjamie *error = 0; 151138350Sphk if (*error) 152138350Sphk return (NULL); 153159590Sjhb 154138350Sphk /* Look up again to see if the VFS was loaded. */ 155138350Sphk vfsp = vfs_byname(fstype); 156138350Sphk if (vfsp == NULL) { 157247071Sjamie if (loaded) 158247071Sjamie (void)kern_kldunload(td, fileid, LINKER_UNLOAD_FORCE); 159138350Sphk *error = ENODEV; 160138350Sphk return (NULL); 161138350Sphk } 162138350Sphk return (vfsp); 163138350Sphk} 164138350Sphk 165138350Sphk 16696755Strhodes/* Register a new filesystem type in the global table */ 167141634Sphkstatic int 16841056Spetervfs_register(struct vfsconf *vfc) 16940435Speter{ 17044549Sdfr struct sysctl_oid *oidp; 171116271Sphk struct vfsops *vfsops; 172135279Sphk static int once; 173225537Srmacklem struct vfsconf *tvfc; 174225537Srmacklem uint32_t hashval; 175225537Srmacklem int secondpass; 176135279Sphk 177135279Sphk if (!once) { 178135279Sphk vattr_null(&va_null); 179135279Sphk once = 1; 180135279Sphk } 181116271Sphk 182132902Sphk if (vfc->vfc_version != VFS_VERSION) { 183132902Sphk printf("ERROR: filesystem %s, unsupported ABI version %x\n", 184132902Sphk vfc->vfc_name, vfc->vfc_version); 185132902Sphk return (EINVAL); 186132902Sphk } 187269457Skib vfsconf_lock(); 188269457Skib if (vfs_byname_locked(vfc->vfc_name) != NULL) { 189269457Skib vfsconf_unlock(); 190241896Skib return (EEXIST); 191269457Skib } 19240435Speter 193225537Srmacklem if (vfs_typenumhash != 0) { 194225537Srmacklem /* 195225537Srmacklem * Calculate a hash on vfc_name to use for vfc_typenum. Unless 196225537Srmacklem * all of 1<->255 are assigned, it is limited to 8bits since 197225537Srmacklem * that is what ZFS uses from vfc_typenum and is also the 198225537Srmacklem * preferred range for vfs_getnewfsid(). 199225537Srmacklem */ 200225537Srmacklem hashval = fnv_32_str(vfc->vfc_name, FNV1_32_INIT); 201225537Srmacklem hashval &= 0xff; 202225537Srmacklem secondpass = 0; 203225537Srmacklem do { 204225537Srmacklem /* Look for and fix any collision. */ 205225537Srmacklem TAILQ_FOREACH(tvfc, &vfsconf, vfc_list) { 206225537Srmacklem if (hashval == tvfc->vfc_typenum) { 207225537Srmacklem if (hashval == 255 && secondpass == 0) { 208225537Srmacklem hashval = 1; 209225537Srmacklem secondpass = 1; 210225537Srmacklem } else 211225537Srmacklem hashval++; 212225537Srmacklem break; 213225537Srmacklem } 214225537Srmacklem } 215225537Srmacklem } while (tvfc != NULL); 216225537Srmacklem vfc->vfc_typenum = hashval; 217225537Srmacklem if (vfc->vfc_typenum >= maxvfsconf) 218225537Srmacklem maxvfsconf = vfc->vfc_typenum + 1; 219225537Srmacklem } else 220225537Srmacklem vfc->vfc_typenum = maxvfsconf++; 221132710Sphk TAILQ_INSERT_TAIL(&vfsconf, vfc, vfc_list); 22240435Speter 22340435Speter /* 224116271Sphk * Initialise unused ``struct vfsops'' fields, to use 225116271Sphk * the vfs_std*() functions. Note, we need the mount 226116271Sphk * and unmount operations, at the least. The check 227116271Sphk * for vfsops available is just a debugging aid. 228116271Sphk */ 229116271Sphk KASSERT(vfc->vfc_vfsops != NULL, 230116271Sphk ("Filesystem %s has no vfsops", vfc->vfc_name)); 231116271Sphk /* 232116271Sphk * Check the mount and unmount operations. 233116271Sphk */ 234116271Sphk vfsops = vfc->vfc_vfsops; 235138509Sphk KASSERT(vfsops->vfs_mount != NULL, 236138509Sphk ("Filesystem %s has no mount op", vfc->vfc_name)); 237116271Sphk KASSERT(vfsops->vfs_unmount != NULL, 238116271Sphk ("Filesystem %s has no unmount op", vfc->vfc_name)); 239116271Sphk 240116271Sphk if (vfsops->vfs_root == NULL) 241116271Sphk /* return file system's root vnode */ 242116271Sphk vfsops->vfs_root = vfs_stdroot; 243116271Sphk if (vfsops->vfs_quotactl == NULL) 244116271Sphk /* quota control */ 245116271Sphk vfsops->vfs_quotactl = vfs_stdquotactl; 246116271Sphk if (vfsops->vfs_statfs == NULL) 247116271Sphk /* return file system's status */ 248116271Sphk vfsops->vfs_statfs = vfs_stdstatfs; 249116271Sphk if (vfsops->vfs_sync == NULL) 250116271Sphk /* 251116271Sphk * flush unwritten data (nosync) 252116271Sphk * file systems can use vfs_stdsync 253116271Sphk * explicitly by setting it in the 254116271Sphk * vfsop vector. 255116271Sphk */ 256116271Sphk vfsops->vfs_sync = vfs_stdnosync; 257116271Sphk if (vfsops->vfs_vget == NULL) 258116271Sphk /* convert an inode number to a vnode */ 259116271Sphk vfsops->vfs_vget = vfs_stdvget; 260116271Sphk if (vfsops->vfs_fhtovp == NULL) 261116271Sphk /* turn an NFS file handle into a vnode */ 262116271Sphk vfsops->vfs_fhtovp = vfs_stdfhtovp; 263116271Sphk if (vfsops->vfs_checkexp == NULL) 264116271Sphk /* check if file system is exported */ 265116271Sphk vfsops->vfs_checkexp = vfs_stdcheckexp; 266116271Sphk if (vfsops->vfs_init == NULL) 267116271Sphk /* file system specific initialisation */ 268116271Sphk vfsops->vfs_init = vfs_stdinit; 269116271Sphk if (vfsops->vfs_uninit == NULL) 270116271Sphk /* file system specific uninitialisation */ 271116271Sphk vfsops->vfs_uninit = vfs_stduninit; 272116271Sphk if (vfsops->vfs_extattrctl == NULL) 273116271Sphk /* extended attribute control */ 274116271Sphk vfsops->vfs_extattrctl = vfs_stdextattrctl; 275131733Salfred if (vfsops->vfs_sysctl == NULL) 276131733Salfred vfsops->vfs_sysctl = vfs_stdsysctl; 277116271Sphk 278116271Sphk /* 27940435Speter * Call init function for this VFS... 28040435Speter */ 28140435Speter (*(vfc->vfc_vfsops->vfs_init))(vfc); 282269457Skib vfsconf_unlock(); 28340435Speter 284269457Skib /* 285269457Skib * If this filesystem has a sysctl node under vfs 286269457Skib * (i.e. vfs.xxfs), then change the oid number of that node to 287269457Skib * match the filesystem's type number. This allows user code 288269457Skib * which uses the type number to read sysctl variables defined 289269457Skib * by the filesystem to continue working. Since the oids are 290269457Skib * in a sorted list, we need to make sure the order is 291269457Skib * preserved by re-registering the oid after modifying its 292269457Skib * number. 293269457Skib */ 294287835Smjg sysctl_wlock(); 295269457Skib SLIST_FOREACH(oidp, SYSCTL_CHILDREN(&sysctl___vfs), oid_link) { 296269457Skib if (strcmp(oidp->oid_name, vfc->vfc_name) == 0) { 297269457Skib sysctl_unregister_oid(oidp); 298269457Skib oidp->oid_number = vfc->vfc_typenum; 299269457Skib sysctl_register_oid(oidp); 300269457Skib break; 301269457Skib } 302269457Skib } 303287835Smjg sysctl_wunlock(); 304269457Skib 305269457Skib return (0); 3061541Srgrimes} 3072946Swollman 30840435Speter 30996755Strhodes/* Remove registration of a filesystem type */ 310141634Sphkstatic int 31141056Spetervfs_unregister(struct vfsconf *vfc) 31240435Speter{ 313132710Sphk struct vfsconf *vfsp; 314283735Skib int error, maxtypenum; 31540435Speter 316269457Skib vfsconf_lock(); 317269457Skib vfsp = vfs_byname_locked(vfc->vfc_name); 318269457Skib if (vfsp == NULL) { 319269457Skib vfsconf_unlock(); 320269457Skib return (EINVAL); 321269457Skib } 322269457Skib if (vfsp->vfc_refcount != 0) { 323269457Skib vfsconf_unlock(); 324269457Skib return (EBUSY); 325269457Skib } 32640435Speter if (vfc->vfc_vfsops->vfs_uninit != NULL) { 32740435Speter error = (*vfc->vfc_vfsops->vfs_uninit)(vfsp); 328269457Skib if (error != 0) { 329269457Skib vfsconf_unlock(); 33040435Speter return (error); 331269457Skib } 33240435Speter } 333132710Sphk TAILQ_REMOVE(&vfsconf, vfsp, vfc_list); 33440435Speter maxtypenum = VFS_GENERIC; 335132710Sphk TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) 33640435Speter if (maxtypenum < vfsp->vfc_typenum) 33740435Speter maxtypenum = vfsp->vfc_typenum; 33840435Speter maxvfsconf = maxtypenum + 1; 339269457Skib vfsconf_unlock(); 340269457Skib return (0); 34140435Speter} 34241056Speter 34391690Seivind/* 34496755Strhodes * Standard kernel module handling code for filesystem modules. 34591690Seivind * Referenced from VFS_SET(). 34691690Seivind */ 34741056Speterint 34841170Sbdevfs_modevent(module_t mod, int type, void *data) 34941056Speter{ 35041056Speter struct vfsconf *vfc; 35141056Speter int error = 0; 35241056Speter 35341056Speter vfc = (struct vfsconf *)data; 35441056Speter 35541056Speter switch (type) { 35641056Speter case MOD_LOAD: 35741056Speter if (vfc) 35841056Speter error = vfs_register(vfc); 35941056Speter break; 36041056Speter 36141056Speter case MOD_UNLOAD: 36241056Speter if (vfc) 36341056Speter error = vfs_unregister(vfc); 36441056Speter break; 365132199Sphk default: 366132199Sphk error = EOPNOTSUPP; 36741056Speter break; 36841056Speter } 36941056Speter return (error); 37041056Speter} 371