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 56 static int vfs_register(struct vfsconf *); 57 static int vfs_unregister(struct vfsconf *); 58 59 MALLOC_DEFINE(M_VNODE, "vnodes", "Dynamically allocated vnodes"); 60 61 /* 62 * The highest defined VFS number. 63 */ 64 int 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 */ 70 struct vfsconfhead vfsconf = TAILQ_HEAD_INITIALIZER(vfsconf); 71 struct sx vfsconf_sx; 72 SX_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 */ 80 static int vfs_typenumhash = 1; 81 SYSCTL_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 */ 90 struct 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 */ 112 113 static struct vfsconf * 114 vfs_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 } 127 128 struct vfsconf * 129 vfs_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 } 138 139 struct vfsconf * 140 vfs_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 } 167 168 static int 169 vfs_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 } 180 181 static int 182 vfs_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 } 191 192 static int 193 vfs_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 } 202 203 static int 204 vfs_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 } 213 214 static int 215 vfs_quotactl_sigdefer(struct mount *mp, int cmd, uid_t uid, void *arg, 216 bool *mp_busy) 217 { 218 int prev_stops, rc; 219 220 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 221 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_quotactl)(mp, cmd, uid, arg, 222 mp_busy); 223 sigallowstop(prev_stops); 224 return (rc); 225 } 226 227 static int 228 vfs_statfs_sigdefer(struct mount *mp, struct statfs *sbp) 229 { 230 int prev_stops, rc; 231 232 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 233 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_statfs)(mp, sbp); 234 sigallowstop(prev_stops); 235 return (rc); 236 } 237 238 static int 239 vfs_sync_sigdefer(struct mount *mp, int waitfor) 240 { 241 int prev_stops, rc; 242 243 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 244 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_sync)(mp, waitfor); 245 sigallowstop(prev_stops); 246 return (rc); 247 } 248 249 static int 250 vfs_vget_sigdefer(struct mount *mp, ino_t ino, int flags, struct vnode **vpp) 251 { 252 int prev_stops, rc; 253 254 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 255 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_vget)(mp, ino, flags, vpp); 256 sigallowstop(prev_stops); 257 return (rc); 258 } 259 260 static int 261 vfs_fhtovp_sigdefer(struct mount *mp, struct fid *fidp, int flags, 262 struct vnode **vpp) 263 { 264 int prev_stops, rc; 265 266 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 267 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_fhtovp)(mp, fidp, flags, vpp); 268 sigallowstop(prev_stops); 269 return (rc); 270 } 271 272 static int 273 vfs_checkexp_sigdefer(struct mount *mp, struct sockaddr *nam, uint64_t *exflg, 274 struct ucred **credp, int *numsecflavors, int *secflavors) 275 { 276 int prev_stops, rc; 277 278 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 279 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_checkexp)(mp, nam, exflg, credp, 280 numsecflavors, secflavors); 281 sigallowstop(prev_stops); 282 return (rc); 283 } 284 285 static int 286 vfs_extattrctl_sigdefer(struct mount *mp, int cmd, struct vnode *filename_vp, 287 int attrnamespace, const char *attrname) 288 { 289 int prev_stops, rc; 290 291 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 292 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_extattrctl)(mp, cmd, 293 filename_vp, attrnamespace, attrname); 294 sigallowstop(prev_stops); 295 return (rc); 296 } 297 298 static int 299 vfs_sysctl_sigdefer(struct mount *mp, fsctlop_t op, struct sysctl_req *req) 300 { 301 int prev_stops, rc; 302 303 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 304 rc = (*mp->mnt_vfc->vfc_vfsops_sd->vfs_sysctl)(mp, op, req); 305 sigallowstop(prev_stops); 306 return (rc); 307 } 308 309 static void 310 vfs_susp_clean_sigdefer(struct mount *mp) 311 { 312 int prev_stops; 313 314 if (*mp->mnt_vfc->vfc_vfsops_sd->vfs_susp_clean == NULL) 315 return; 316 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 317 (*mp->mnt_vfc->vfc_vfsops_sd->vfs_susp_clean)(mp); 318 sigallowstop(prev_stops); 319 } 320 321 static void 322 vfs_reclaim_lowervp_sigdefer(struct mount *mp, struct vnode *vp) 323 { 324 int prev_stops; 325 326 if (*mp->mnt_vfc->vfc_vfsops_sd->vfs_reclaim_lowervp == NULL) 327 return; 328 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 329 (*mp->mnt_vfc->vfc_vfsops_sd->vfs_reclaim_lowervp)(mp, vp); 330 sigallowstop(prev_stops); 331 } 332 333 static void 334 vfs_unlink_lowervp_sigdefer(struct mount *mp, struct vnode *vp) 335 { 336 int prev_stops; 337 338 if (*mp->mnt_vfc->vfc_vfsops_sd->vfs_unlink_lowervp == NULL) 339 return; 340 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 341 (*(mp)->mnt_vfc->vfc_vfsops_sd->vfs_unlink_lowervp)(mp, vp); 342 sigallowstop(prev_stops); 343 } 344 345 static void 346 vfs_purge_sigdefer(struct mount *mp) 347 { 348 int prev_stops; 349 350 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT); 351 (*mp->mnt_vfc->vfc_vfsops_sd->vfs_purge)(mp); 352 sigallowstop(prev_stops); 353 } 354 355 static struct vfsops vfsops_sigdefer = { 356 .vfs_mount = vfs_mount_sigdefer, 357 .vfs_unmount = vfs_unmount_sigdefer, 358 .vfs_root = vfs_root_sigdefer, 359 .vfs_cachedroot = vfs_cachedroot_sigdefer, 360 .vfs_quotactl = vfs_quotactl_sigdefer, 361 .vfs_statfs = vfs_statfs_sigdefer, 362 .vfs_sync = vfs_sync_sigdefer, 363 .vfs_vget = vfs_vget_sigdefer, 364 .vfs_fhtovp = vfs_fhtovp_sigdefer, 365 .vfs_checkexp = vfs_checkexp_sigdefer, 366 .vfs_extattrctl = vfs_extattrctl_sigdefer, 367 .vfs_sysctl = vfs_sysctl_sigdefer, 368 .vfs_susp_clean = vfs_susp_clean_sigdefer, 369 .vfs_reclaim_lowervp = vfs_reclaim_lowervp_sigdefer, 370 .vfs_unlink_lowervp = vfs_unlink_lowervp_sigdefer, 371 .vfs_purge = vfs_purge_sigdefer, 372 373 }; 374 375 /* Register a new filesystem type in the global table */ 376 static int 377 vfs_register(struct vfsconf *vfc) 378 { 379 struct sysctl_oid *oidp; 380 struct vfsops *vfsops; 381 static int once; 382 struct vfsconf *tvfc; 383 uint32_t hashval; 384 int secondpass; 385 386 if (!once) { 387 vattr_null(&va_null); 388 once = 1; 389 } 390 391 if (vfc->vfc_version != VFS_VERSION) { 392 printf("ERROR: filesystem %s, unsupported ABI version %x\n", 393 vfc->vfc_name, vfc->vfc_version); 394 return (EINVAL); 395 } 396 vfsconf_lock(); 397 if (vfs_byname_locked(vfc->vfc_name) != NULL) { 398 vfsconf_unlock(); 399 return (EEXIST); 400 } 401 402 if (vfs_typenumhash != 0) { 403 /* 404 * Calculate a hash on vfc_name to use for vfc_typenum. Unless 405 * all of 1<->255 are assigned, it is limited to 8bits since 406 * that is what ZFS uses from vfc_typenum and is also the 407 * preferred range for vfs_getnewfsid(). 408 */ 409 hashval = fnv_32_str(vfc->vfc_name, FNV1_32_INIT); 410 hashval &= 0xff; 411 secondpass = 0; 412 do { 413 /* Look for and fix any collision. */ 414 TAILQ_FOREACH(tvfc, &vfsconf, vfc_list) { 415 if (hashval == tvfc->vfc_typenum) { 416 if (hashval == 255 && secondpass == 0) { 417 hashval = 1; 418 secondpass = 1; 419 } else 420 hashval++; 421 break; 422 } 423 } 424 } while (tvfc != NULL); 425 vfc->vfc_typenum = hashval; 426 if (vfc->vfc_typenum >= maxvfsconf) 427 maxvfsconf = vfc->vfc_typenum + 1; 428 } else 429 vfc->vfc_typenum = maxvfsconf++; 430 TAILQ_INSERT_TAIL(&vfsconf, vfc, vfc_list); 431 432 /* 433 * Initialise unused ``struct vfsops'' fields, to use 434 * the vfs_std*() functions. Note, we need the mount 435 * and unmount operations, at the least. The check 436 * for vfsops available is just a debugging aid. 437 */ 438 KASSERT(vfc->vfc_vfsops != NULL, 439 ("Filesystem %s has no vfsops", vfc->vfc_name)); 440 /* 441 * Check the mount and unmount operations. 442 */ 443 vfsops = vfc->vfc_vfsops; 444 KASSERT(vfsops->vfs_mount != NULL, 445 ("Filesystem %s has no mount op", vfc->vfc_name)); 446 KASSERT(vfsops->vfs_unmount != NULL, 447 ("Filesystem %s has no unmount op", vfc->vfc_name)); 448 449 if (vfsops->vfs_root == NULL) 450 /* return file system's root vnode */ 451 vfsops->vfs_root = vfs_stdroot; 452 if (vfsops->vfs_quotactl == NULL) 453 /* quota control */ 454 vfsops->vfs_quotactl = vfs_stdquotactl; 455 if (vfsops->vfs_statfs == NULL) 456 /* return file system's status */ 457 vfsops->vfs_statfs = vfs_stdstatfs; 458 if (vfsops->vfs_sync == NULL) 459 /* 460 * flush unwritten data (nosync) 461 * file systems can use vfs_stdsync 462 * explicitly by setting it in the 463 * vfsop vector. 464 */ 465 vfsops->vfs_sync = vfs_stdnosync; 466 if (vfsops->vfs_vget == NULL) 467 /* convert an inode number to a vnode */ 468 vfsops->vfs_vget = vfs_stdvget; 469 if (vfsops->vfs_fhtovp == NULL) 470 /* turn an NFS file handle into a vnode */ 471 vfsops->vfs_fhtovp = vfs_stdfhtovp; 472 if (vfsops->vfs_checkexp == NULL) 473 /* check if file system is exported */ 474 vfsops->vfs_checkexp = vfs_stdcheckexp; 475 if (vfsops->vfs_init == NULL) 476 /* file system specific initialisation */ 477 vfsops->vfs_init = vfs_stdinit; 478 if (vfsops->vfs_uninit == NULL) 479 /* file system specific uninitialisation */ 480 vfsops->vfs_uninit = vfs_stduninit; 481 if (vfsops->vfs_extattrctl == NULL) 482 /* extended attribute control */ 483 vfsops->vfs_extattrctl = vfs_stdextattrctl; 484 if (vfsops->vfs_sysctl == NULL) 485 vfsops->vfs_sysctl = vfs_stdsysctl; 486 487 if ((vfc->vfc_flags & VFCF_SBDRY) != 0) { 488 vfc->vfc_vfsops_sd = vfc->vfc_vfsops; 489 vfc->vfc_vfsops = &vfsops_sigdefer; 490 } 491 492 if (vfc->vfc_flags & VFCF_JAIL) 493 prison_add_vfs(vfc); 494 495 /* 496 * Call init function for this VFS... 497 */ 498 if ((vfc->vfc_flags & VFCF_SBDRY) != 0) 499 vfc->vfc_vfsops_sd->vfs_init(vfc); 500 else 501 vfc->vfc_vfsops->vfs_init(vfc); 502 vfsconf_unlock(); 503 504 /* 505 * If this filesystem has a sysctl node under vfs 506 * (i.e. vfs.xxfs), then change the oid number of that node to 507 * match the filesystem's type number. This allows user code 508 * which uses the type number to read sysctl variables defined 509 * by the filesystem to continue working. Since the oids are 510 * in a sorted list, we need to make sure the order is 511 * preserved by re-registering the oid after modifying its 512 * number. 513 */ 514 sysctl_wlock(); 515 SLIST_FOREACH(oidp, SYSCTL_CHILDREN(&sysctl___vfs), oid_link) { 516 if (strcmp(oidp->oid_name, vfc->vfc_name) == 0) { 517 sysctl_unregister_oid(oidp); 518 oidp->oid_number = vfc->vfc_typenum; 519 sysctl_register_oid(oidp); 520 break; 521 } 522 } 523 sysctl_wunlock(); 524 525 return (0); 526 } 527 528 /* Remove registration of a filesystem type */ 529 static int 530 vfs_unregister(struct vfsconf *vfc) 531 { 532 struct vfsconf *vfsp; 533 int error, maxtypenum; 534 535 vfsconf_lock(); 536 vfsp = vfs_byname_locked(vfc->vfc_name); 537 if (vfsp == NULL) { 538 vfsconf_unlock(); 539 return (EINVAL); 540 } 541 if (vfsp->vfc_refcount != 0) { 542 vfsconf_unlock(); 543 return (EBUSY); 544 } 545 error = 0; 546 if ((vfc->vfc_flags & VFCF_SBDRY) != 0) { 547 if (vfc->vfc_vfsops_sd->vfs_uninit != NULL) 548 error = vfc->vfc_vfsops_sd->vfs_uninit(vfsp); 549 } else { 550 if (vfc->vfc_vfsops->vfs_uninit != NULL) { 551 error = vfc->vfc_vfsops->vfs_uninit(vfsp); 552 } 553 if (error != 0) { 554 vfsconf_unlock(); 555 return (error); 556 } 557 } 558 TAILQ_REMOVE(&vfsconf, vfsp, vfc_list); 559 maxtypenum = VFS_GENERIC; 560 TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) 561 if (maxtypenum < vfsp->vfc_typenum) 562 maxtypenum = vfsp->vfc_typenum; 563 maxvfsconf = maxtypenum + 1; 564 vfsconf_unlock(); 565 return (0); 566 } 567 568 /* 569 * Standard kernel module handling code for filesystem modules. 570 * Referenced from VFS_SET(). 571 */ 572 int 573 vfs_modevent(module_t mod, int type, void *data) 574 { 575 struct vfsconf *vfc; 576 int error = 0; 577 578 vfc = (struct vfsconf *)data; 579 580 switch (type) { 581 case MOD_LOAD: 582 if (vfc) 583 error = vfs_register(vfc); 584 break; 585 586 case MOD_UNLOAD: 587 if (vfc) 588 error = vfs_unregister(vfc); 589 break; 590 default: 591 error = EOPNOTSUPP; 592 break; 593 } 594 return (error); 595 } 596