1/*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed 8 * to Berkeley by John Heidemann of the UCLA Ficus project. 9 * 10 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_init.c 8.3 (Berkeley) 1/4/94 37 */ 38 39#include <sys/cdefs.h> 40__FBSDID("$FreeBSD$"); 41 42#include <sys/param.h> 43#include <sys/systm.h> 44#include <sys/fnv_hash.h> 45#include <sys/jail.h> 46#include <sys/kernel.h> 47#include <sys/linker.h> 48#include <sys/mount.h> 49#include <sys/proc.h> 50#include <sys/sx.h> 51#include <sys/syscallsubr.h> 52#include <sys/sysctl.h> 53#include <sys/vnode.h> 54#include <sys/malloc.h> 55 56static int vfs_register(struct vfsconf *); 57static int vfs_unregister(struct vfsconf *); 58 59MALLOC_DEFINE(M_VNODE, "vnodes", "Dynamically allocated vnodes"); 60 61/* 62 * The highest defined VFS number. 63 */ 64int maxvfsconf = VFS_GENERIC + 1; 65 66/* 67 * Single-linked list of configured VFSes. 68 * New entries are added/deleted by vfs_register()/vfs_unregister() 69 */ 70struct vfsconfhead vfsconf = TAILQ_HEAD_INITIALIZER(vfsconf); 71struct sx vfsconf_sx; 72SX_SYSINIT(vfsconf, &vfsconf_sx, "vfsconf"); 73 74/* 75 * Loader.conf variable vfs.typenumhash enables setting vfc_typenum using a hash 76 * calculation on vfc_name, so that it doesn't change when file systems are 77 * loaded in a different order. This will avoid the NFS server file handles from 78 * changing for file systems that use vfc_typenum in their fsid. 79 */ 80static int vfs_typenumhash = 1; 81SYSCTL_INT(_vfs, OID_AUTO, typenumhash, CTLFLAG_RDTUN, &vfs_typenumhash, 0, 82 "Set vfc_typenum using a hash calculation on vfc_name, so that it does not" 83 "change when file systems are loaded in a different order."); 84 85/* 86 * A Zen vnode attribute structure. 87 * 88 * Initialized when the first filesystem registers by vfs_register(). 89 */ 90struct vattr va_null; 91 92/* 93 * vfs_init.c 94 * 95 * Allocate and fill in operations vectors. 96 * 97 * An undocumented feature of this approach to defining operations is that 98 * there can be multiple entries in vfs_opv_descs for the same operations 99 * vector. This allows third parties to extend the set of operations 100 * supported by another layer in a binary compatibile way. For example, 101 * assume that NFS needed to be modified to support Ficus. NFS has an entry 102 * (probably nfs_vnopdeop_decls) declaring all the operations NFS supports by 103 * default. Ficus could add another entry (ficus_nfs_vnodeop_decl_entensions) 104 * listing those new operations Ficus adds to NFS, all without modifying the 105 * NFS code. (Of couse, the OTW NFS protocol still needs to be munged, but 106 * that is a(whole)nother story.) This is a feature. 107 */ 108 109/* 110 * Routines having to do with the management of the vnode table. 111 */ 112 113static struct vfsconf * 114vfs_byname_locked(const char *name) 115{ 116 struct vfsconf *vfsp; 117 118 sx_assert(&vfsconf_sx, SA_LOCKED); 119 if (!strcmp(name, "ffs")) 120 name = "ufs"; 121 TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) { 122 if (!strcmp(name, vfsp->vfc_name)) 123 return (vfsp); 124 } 125 return (NULL); 126} 127 128struct vfsconf * 129vfs_byname(const char *name) 130{ 131 struct vfsconf *vfsp; 132 133 vfsconf_slock(); 134 vfsp = vfs_byname_locked(name); 135 vfsconf_sunlock(); 136 return (vfsp); 137} 138 139struct vfsconf * 140vfs_byname_kld(const char *fstype, struct thread *td, int *error) 141{ 142 struct vfsconf *vfsp; 143 int fileid, loaded; 144 145 vfsp = vfs_byname(fstype); 146 if (vfsp != NULL) 147 return (vfsp); 148 149 /* Try to load the respective module. */ 150 *error = kern_kldload(td, fstype, &fileid); 151 loaded = (*error == 0); 152 if (*error == EEXIST) 153 *error = 0; 154 if (*error) 155 return (NULL); 156 157 /* Look up again to see if the VFS was loaded. */ 158 vfsp = vfs_byname(fstype); 159 if (vfsp == NULL) { 160 if (loaded) 161 (void)kern_kldunload(td, fileid, LINKER_UNLOAD_FORCE); 162 *error = ENODEV; 163 return (NULL); 164 } 165 return (vfsp); 166} 167 168static int 169vfs_mount_sigdefer(struct mount *mp) 170{ 171 int prev_stops, rc; 172 173 TSRAW(curthread, TS_ENTER, "VFS_MOUNT", mp->mnt_vfc->vfc_name); 174 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 175 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_mount)(mp); 176 sigallowstop(prev_stops); 177 TSRAW(curthread, TS_EXIT, "VFS_MOUNT", mp->mnt_vfc->vfc_name); 178 return (rc); 179} 180 181static int 182vfs_unmount_sigdefer(struct mount *mp, int mntflags) 183{ 184 int prev_stops, rc; 185 186 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 187 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_unmount)(mp, mntflags); 188 sigallowstop(prev_stops); 189 return (rc); 190} 191 192static int 193vfs_root_sigdefer(struct mount *mp, int flags, struct vnode **vpp) 194{ 195 int prev_stops, rc; 196 197 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 198 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_root)(mp, flags, vpp); 199 sigallowstop(prev_stops); 200 return (rc); 201} 202 203static int 204vfs_cachedroot_sigdefer(struct mount *mp, int flags, struct vnode **vpp) 205{ 206 int prev_stops, rc; 207 208 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 209 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_cachedroot)(mp, flags, vpp); 210 sigallowstop(prev_stops); 211 return (rc); 212} 213 214static int 215vfs_quotactl_sigdefer(struct mount *mp, int cmd, uid_t uid, void *arg) 216{ 217 int prev_stops, rc; 218 219 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 220 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_quotactl)(mp, cmd, uid, arg); 221 sigallowstop(prev_stops); 222 return (rc); 223} 224 225static int 226vfs_statfs_sigdefer(struct mount *mp, struct statfs *sbp) 227{ 228 int prev_stops, rc; 229 230 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 231 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_statfs)(mp, sbp); 232 sigallowstop(prev_stops); 233 return (rc); 234} 235 236static int 237vfs_sync_sigdefer(struct mount *mp, int waitfor) 238{ 239 int prev_stops, rc; 240 241 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 242 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_sync)(mp, waitfor); 243 sigallowstop(prev_stops); 244 return (rc); 245} 246 247static int 248vfs_vget_sigdefer(struct mount *mp, ino_t ino, int flags, struct vnode **vpp) 249{ 250 int prev_stops, rc; 251 252 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 253 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_vget)(mp, ino, flags, vpp); 254 sigallowstop(prev_stops); 255 return (rc); 256} 257 258static int 259vfs_fhtovp_sigdefer(struct mount *mp, struct fid *fidp, int flags, 260 struct vnode **vpp) 261{ 262 int prev_stops, rc; 263 264 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 265 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_fhtovp)(mp, fidp, flags, vpp); 266 sigallowstop(prev_stops); 267 return (rc); 268} 269 270static int 271vfs_checkexp_sigdefer(struct mount *mp, struct sockaddr *nam, uint64_t *exflg, 272 struct ucred **credp, int *numsecflavors, int *secflavors) 273{ 274 int prev_stops, rc; 275 276 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 277 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_checkexp)(mp, nam, exflg, credp, 278 numsecflavors, secflavors); 279 sigallowstop(prev_stops); 280 return (rc); 281} 282 283static int 284vfs_extattrctl_sigdefer(struct mount *mp, int cmd, struct vnode *filename_vp, 285 int attrnamespace, const char *attrname) 286{ 287 int prev_stops, rc; 288 289 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 290 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_extattrctl)(mp, cmd, 291 filename_vp, attrnamespace, attrname); 292 sigallowstop(prev_stops); 293 return (rc); 294} 295 296static int 297vfs_sysctl_sigdefer(struct mount *mp, fsctlop_t op, struct sysctl_req *req) 298{ 299 int prev_stops, rc; 300 301 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 302 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_sysctl)(mp, op, req); 303 sigallowstop(prev_stops); 304 return (rc); 305} 306 307static void 308vfs_susp_clean_sigdefer(struct mount *mp) 309{ 310 int prev_stops; 311 312 if (*mp->mnt_vfc->vfc_vfsops_sd->vfs_susp_clean == NULL) 313 return; 314 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 315 (*mp->mnt_vfc->vfc_vfsops_sd->vfs_susp_clean)(mp); 316 sigallowstop(prev_stops); 317} 318 319static void 320vfs_reclaim_lowervp_sigdefer(struct mount *mp, struct vnode *vp) 321{ 322 int prev_stops; 323 324 if (*mp->mnt_vfc->vfc_vfsops_sd->vfs_reclaim_lowervp == NULL) 325 return; 326 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 327 (*mp->mnt_vfc->vfc_vfsops_sd->vfs_reclaim_lowervp)(mp, vp); 328 sigallowstop(prev_stops); 329} 330 331static void 332vfs_unlink_lowervp_sigdefer(struct mount *mp, struct vnode *vp) 333{ 334 int prev_stops; 335 336 if (*mp->mnt_vfc->vfc_vfsops_sd->vfs_unlink_lowervp == NULL) 337 return; 338 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 339 (*(mp)->mnt_vfc->vfc_vfsops_sd->vfs_unlink_lowervp)(mp, vp); 340 sigallowstop(prev_stops); 341} 342 343static void 344vfs_purge_sigdefer(struct mount *mp) 345{ 346 int prev_stops; 347 348 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 349 (*mp->mnt_vfc->vfc_vfsops_sd->vfs_purge)(mp); 350 sigallowstop(prev_stops); 351} 352 353static struct vfsops vfsops_sigdefer = { 354 .vfs_mount = vfs_mount_sigdefer, 355 .vfs_unmount = vfs_unmount_sigdefer, 356 .vfs_root = vfs_root_sigdefer, 357 .vfs_cachedroot = vfs_cachedroot_sigdefer, 358 .vfs_quotactl = vfs_quotactl_sigdefer, 359 .vfs_statfs = vfs_statfs_sigdefer, 360 .vfs_sync = vfs_sync_sigdefer, 361 .vfs_vget = vfs_vget_sigdefer, 362 .vfs_fhtovp = vfs_fhtovp_sigdefer, 363 .vfs_checkexp = vfs_checkexp_sigdefer, 364 .vfs_extattrctl = vfs_extattrctl_sigdefer, 365 .vfs_sysctl = vfs_sysctl_sigdefer, 366 .vfs_susp_clean = vfs_susp_clean_sigdefer, 367 .vfs_reclaim_lowervp = vfs_reclaim_lowervp_sigdefer, 368 .vfs_unlink_lowervp = vfs_unlink_lowervp_sigdefer, 369 .vfs_purge = vfs_purge_sigdefer, 370 371}; 372 373/* Register a new filesystem type in the global table */ 374static int 375vfs_register(struct vfsconf *vfc) 376{ 377 struct sysctl_oid *oidp; 378 struct vfsops *vfsops; 379 static int once; 380 struct vfsconf *tvfc; 381 uint32_t hashval; 382 int secondpass; 383 384 if (!once) { 385 vattr_null(&va_null); 386 once = 1; 387 } 388 389 if (vfc->vfc_version != VFS_VERSION) { 390 printf("ERROR: filesystem %s, unsupported ABI version %x\n", 391 vfc->vfc_name, vfc->vfc_version); 392 return (EINVAL); 393 } 394 vfsconf_lock(); 395 if (vfs_byname_locked(vfc->vfc_name) != NULL) { 396 vfsconf_unlock(); 397 return (EEXIST); 398 } 399 400 if (vfs_typenumhash != 0) { 401 /* 402 * Calculate a hash on vfc_name to use for vfc_typenum. Unless 403 * all of 1<->255 are assigned, it is limited to 8bits since 404 * that is what ZFS uses from vfc_typenum and is also the 405 * preferred range for vfs_getnewfsid(). 406 */ 407 hashval = fnv_32_str(vfc->vfc_name, FNV1_32_INIT); 408 hashval &= 0xff; 409 secondpass = 0; 410 do { 411 /* Look for and fix any collision. */ 412 TAILQ_FOREACH(tvfc, &vfsconf, vfc_list) { 413 if (hashval == tvfc->vfc_typenum) { 414 if (hashval == 255 && secondpass == 0) { 415 hashval = 1; 416 secondpass = 1; 417 } else 418 hashval++; 419 break; 420 } 421 } 422 } while (tvfc != NULL); 423 vfc->vfc_typenum = hashval; 424 if (vfc->vfc_typenum >= maxvfsconf) 425 maxvfsconf = vfc->vfc_typenum + 1; 426 } else 427 vfc->vfc_typenum = maxvfsconf++; 428 TAILQ_INSERT_TAIL(&vfsconf, vfc, vfc_list); 429 430 /* 431 * Initialise unused ``struct vfsops'' fields, to use 432 * the vfs_std*() functions. Note, we need the mount 433 * and unmount operations, at the least. The check 434 * for vfsops available is just a debugging aid. 435 */ 436 KASSERT(vfc->vfc_vfsops != NULL, 437 ("Filesystem %s has no vfsops", vfc->vfc_name)); 438 /* 439 * Check the mount and unmount operations. 440 */ 441 vfsops = vfc->vfc_vfsops; 442 KASSERT(vfsops->vfs_mount != NULL, 443 ("Filesystem %s has no mount op", vfc->vfc_name)); 444 KASSERT(vfsops->vfs_unmount != NULL, 445 ("Filesystem %s has no unmount op", vfc->vfc_name)); 446 447 if (vfsops->vfs_root == NULL) 448 /* return file system's root vnode */ 449 vfsops->vfs_root = vfs_stdroot; 450 if (vfsops->vfs_quotactl == NULL) 451 /* quota control */ 452 vfsops->vfs_quotactl = vfs_stdquotactl; 453 if (vfsops->vfs_statfs == NULL) 454 /* return file system's status */ 455 vfsops->vfs_statfs = vfs_stdstatfs; 456 if (vfsops->vfs_sync == NULL) 457 /* 458 * flush unwritten data (nosync) 459 * file systems can use vfs_stdsync 460 * explicitly by setting it in the 461 * vfsop vector. 462 */ 463 vfsops->vfs_sync = vfs_stdnosync; 464 if (vfsops->vfs_vget == NULL) 465 /* convert an inode number to a vnode */ 466 vfsops->vfs_vget = vfs_stdvget; 467 if (vfsops->vfs_fhtovp == NULL) 468 /* turn an NFS file handle into a vnode */ 469 vfsops->vfs_fhtovp = vfs_stdfhtovp; 470 if (vfsops->vfs_checkexp == NULL) 471 /* check if file system is exported */ 472 vfsops->vfs_checkexp = vfs_stdcheckexp; 473 if (vfsops->vfs_init == NULL) 474 /* file system specific initialisation */ 475 vfsops->vfs_init = vfs_stdinit; 476 if (vfsops->vfs_uninit == NULL) 477 /* file system specific uninitialisation */ 478 vfsops->vfs_uninit = vfs_stduninit; 479 if (vfsops->vfs_extattrctl == NULL) 480 /* extended attribute control */ 481 vfsops->vfs_extattrctl = vfs_stdextattrctl; 482 if (vfsops->vfs_sysctl == NULL) 483 vfsops->vfs_sysctl = vfs_stdsysctl; 484 485 if ((vfc->vfc_flags & VFCF_SBDRY) != 0) { 486 vfc->vfc_vfsops_sd = vfc->vfc_vfsops; 487 vfc->vfc_vfsops = &vfsops_sigdefer; 488 } 489 490 if (vfc->vfc_flags & VFCF_JAIL) 491 prison_add_vfs(vfc); 492 493 /* 494 * Call init function for this VFS... 495 */ 496 if ((vfc->vfc_flags & VFCF_SBDRY) != 0) 497 vfc->vfc_vfsops_sd->vfs_init(vfc); 498 else 499 vfc->vfc_vfsops->vfs_init(vfc); 500 vfsconf_unlock(); 501 502 /* 503 * If this filesystem has a sysctl node under vfs 504 * (i.e. vfs.xxfs), then change the oid number of that node to 505 * match the filesystem's type number. This allows user code 506 * which uses the type number to read sysctl variables defined 507 * by the filesystem to continue working. Since the oids are 508 * in a sorted list, we need to make sure the order is 509 * preserved by re-registering the oid after modifying its 510 * number. 511 */ 512 sysctl_wlock(); 513 SLIST_FOREACH(oidp, SYSCTL_CHILDREN(&sysctl___vfs), oid_link) { 514 if (strcmp(oidp->oid_name, vfc->vfc_name) == 0) { 515 sysctl_unregister_oid(oidp); 516 oidp->oid_number = vfc->vfc_typenum; 517 sysctl_register_oid(oidp); 518 break; 519 } 520 } 521 sysctl_wunlock(); 522 523 return (0); 524} 525 526/* Remove registration of a filesystem type */ 527static int 528vfs_unregister(struct vfsconf *vfc) 529{ 530 struct vfsconf *vfsp; 531 int error, maxtypenum; 532 533 vfsconf_lock(); 534 vfsp = vfs_byname_locked(vfc->vfc_name); 535 if (vfsp == NULL) { 536 vfsconf_unlock(); 537 return (EINVAL); 538 } 539 if (vfsp->vfc_refcount != 0) { 540 vfsconf_unlock(); 541 return (EBUSY); 542 } 543 error = 0; 544 if ((vfc->vfc_flags & VFCF_SBDRY) != 0) { 545 if (vfc->vfc_vfsops_sd->vfs_uninit != NULL) 546 error = vfc->vfc_vfsops_sd->vfs_uninit(vfsp); 547 } else { 548 if (vfc->vfc_vfsops->vfs_uninit != NULL) { 549 error = vfc->vfc_vfsops->vfs_uninit(vfsp); 550 } 551 if (error != 0) { 552 vfsconf_unlock(); 553 return (error); 554 } 555 } 556 TAILQ_REMOVE(&vfsconf, vfsp, vfc_list); 557 maxtypenum = VFS_GENERIC; 558 TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) 559 if (maxtypenum < vfsp->vfc_typenum) 560 maxtypenum = vfsp->vfc_typenum; 561 maxvfsconf = maxtypenum + 1; 562 vfsconf_unlock(); 563 return (0); 564} 565 566/* 567 * Standard kernel module handling code for filesystem modules. 568 * Referenced from VFS_SET(). 569 */ 570int 571vfs_modevent(module_t mod, int type, void *data) 572{ 573 struct vfsconf *vfc; 574 int error = 0; 575 576 vfc = (struct vfsconf *)data; 577 578 switch (type) { 579 case MOD_LOAD: 580 if (vfc) 581 error = vfs_register(vfc); 582 break; 583 584 case MOD_UNLOAD: 585 if (vfc) 586 error = vfs_unregister(vfc); 587 break; 588 default: 589 error = EOPNOTSUPP; 590 break; 591 } 592 return (error); 593} 594