1 /* $NetBSD: vfs_init.c,v 1.59 2022/11/18 00:10:03 riastradh Exp $ */ 2 3 /*- 4 * Copyright (c) 1998, 2000, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1989, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * This code is derived from software contributed 38 * to Berkeley by John Heidemann of the UCLA Ficus project. 39 * 40 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)vfs_init.c 8.5 (Berkeley) 5/11/95 67 */ 68 69 #include <sys/cdefs.h> 70 __KERNEL_RCSID(0, "$NetBSD: vfs_init.c,v 1.59 2022/11/18 00:10:03 riastradh Exp $"); 71 72 #include <sys/param.h> 73 #include <sys/mount.h> 74 #include <sys/time.h> 75 #include <sys/vnode.h> 76 #include <sys/stat.h> 77 #include <sys/namei.h> 78 #include <sys/ucred.h> 79 #include <sys/buf.h> 80 #include <sys/errno.h> 81 #include <sys/kmem.h> 82 #include <sys/systm.h> 83 #include <sys/module.h> 84 #include <sys/dirhash.h> 85 #include <sys/sysctl.h> 86 #include <sys/kauth.h> 87 88 #include <miscfs/deadfs/deadfs.h> 89 #include <miscfs/fifofs/fifo.h> 90 #include <miscfs/specfs/specdev.h> 91 92 /* 93 * Sigh, such primitive tools are these... 94 */ 95 #if 0 96 #define DODEBUG(A) A 97 #else 98 #define DODEBUG(A) 99 #endif 100 101 pool_cache_t pnbuf_cache; 102 103 /* 104 * These vnodeopv_descs are listed here because they are not 105 * associated with any particular file system, and thus cannot 106 * be initialized by vfs_attach(). 107 */ 108 const struct vnodeopv_desc * const vfs_special_vnodeopv_descs[] = { 109 &dead_vnodeop_opv_desc, 110 &fifo_vnodeop_opv_desc, 111 &spec_vnodeop_opv_desc, 112 NULL, 113 }; 114 115 struct vfs_list_head vfs_list = /* vfs list */ 116 LIST_HEAD_INITIALIZER(vfs_list); 117 118 static kauth_listener_t mount_listener; 119 120 /* 121 * This code doesn't work if the defn is **vnodop_defns with cc. 122 * The problem is because of the compiler sometimes putting in an 123 * extra level of indirection for arrays. It's an interesting 124 * "feature" of C. 125 */ 126 typedef int (*PFI)(void *); 127 128 /* 129 * A miscellaneous routine. 130 * A generic "default" routine that just returns an error. 131 */ 132 /*ARGSUSED*/ 133 int 134 vn_default_error(void *v) 135 { 136 137 return (EOPNOTSUPP); 138 } 139 140 static struct sysctllog *vfs_sysctllog; 141 142 /* 143 * Top level filesystem related information gathering. 144 */ 145 static void 146 sysctl_vfs_setup(void) 147 { 148 149 sysctl_createv(&vfs_sysctllog, 0, NULL, NULL, 150 CTLFLAG_PERMANENT, 151 CTLTYPE_NODE, "generic", 152 SYSCTL_DESCR("Non-specific vfs related information"), 153 NULL, 0, NULL, 0, 154 CTL_VFS, VFS_GENERIC, CTL_EOL); 155 sysctl_createv(&vfs_sysctllog, 0, NULL, NULL, 156 CTLFLAG_PERMANENT, 157 CTLTYPE_STRING, "fstypes", 158 SYSCTL_DESCR("List of file systems present"), 159 sysctl_vfs_generic_fstypes, 0, NULL, 0, 160 CTL_VFS, VFS_GENERIC, CTL_CREATE, CTL_EOL); 161 sysctl_createv(&vfs_sysctllog, 0, NULL, NULL, 162 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 163 CTLTYPE_INT, "magiclinks", 164 SYSCTL_DESCR("Whether \"magic\" symlinks are expanded"), 165 NULL, 0, &vfs_magiclinks, 0, 166 CTL_VFS, VFS_GENERIC, VFS_MAGICLINKS, CTL_EOL); 167 sysctl_createv(&vfs_sysctllog, 0, NULL, NULL, 168 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 169 CTLTYPE_INT, "timestamp_precision", 170 SYSCTL_DESCR("File timestamp precision"), 171 NULL, 0, &vfs_timestamp_precision, 0, 172 CTL_VFS, VFS_GENERIC, VFS_TIMESTAMP_PRECISION, 173 CTL_EOL); 174 } 175 176 177 /* 178 * vfs_init.c 179 * 180 * Allocate and fill in operations vectors. 181 * 182 * An undocumented feature of this approach to defining operations is that 183 * there can be multiple entries in vfs_opv_descs for the same operations 184 * vector. This allows third parties to extend the set of operations 185 * supported by another layer in a binary compatibile way. For example, 186 * assume that NFS needed to be modified to support Ficus. NFS has an entry 187 * (probably nfs_vnopdeop_decls) declaring all the operations NFS supports by 188 * default. Ficus could add another entry (ficus_nfs_vnodeop_decl_entensions) 189 * listing those new operations Ficus adds to NFS, all without modifying the 190 * NFS code. (Of couse, the OTW NFS protocol still needs to be munged, but 191 * that is a(whole)nother story.) This is a feature. 192 */ 193 194 /* 195 * Init the vector, if it needs it. 196 * Also handle backwards compatibility. 197 */ 198 static void 199 vfs_opv_init_explicit(const struct vnodeopv_desc *vfs_opv_desc) 200 { 201 int (**opv_desc_vector)(void *); 202 const struct vnodeopv_entry_desc *opve_descp; 203 204 opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p); 205 206 for (opve_descp = vfs_opv_desc->opv_desc_ops; 207 opve_descp->opve_op; 208 opve_descp++) { 209 /* 210 * Sanity check: is this operation listed 211 * in the list of operations? We check this 212 * by seeing if its offset is zero. Since 213 * the default routine should always be listed 214 * first, it should be the only one with a zero 215 * offset. Any other operation with a zero 216 * offset is probably not listed in 217 * vfs_op_descs, and so is probably an error. 218 * 219 * A panic here means the layer programmer 220 * has committed the all-too common bug 221 * of adding a new operation to the layer's 222 * list of vnode operations but 223 * not adding the operation to the system-wide 224 * list of supported operations. 225 */ 226 if (opve_descp->opve_op->vdesc_offset == 0 && 227 opve_descp->opve_op->vdesc_offset != VOFFSET(vop_default)) { 228 printf("operation %s not listed in %s.\n", 229 opve_descp->opve_op->vdesc_name, "vfs_op_descs"); 230 panic ("vfs_opv_init: bad operation"); 231 } 232 233 /* 234 * Fill in this entry. 235 */ 236 opv_desc_vector[opve_descp->opve_op->vdesc_offset] = 237 opve_descp->opve_impl; 238 } 239 } 240 241 static void 242 vfs_opv_init_default(const struct vnodeopv_desc *vfs_opv_desc) 243 { 244 int j; 245 int (**opv_desc_vector)(void *); 246 247 opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p); 248 249 /* 250 * Force every operations vector to have a default routine. 251 */ 252 if (opv_desc_vector[VOFFSET(vop_default)] == NULL) 253 panic("vfs_opv_init: operation vector without default routine."); 254 255 for (j = 0; j < VNODE_OPS_COUNT; j++) 256 if (opv_desc_vector[j] == NULL) 257 opv_desc_vector[j] = 258 opv_desc_vector[VOFFSET(vop_default)]; 259 } 260 261 void 262 vfs_opv_init(const struct vnodeopv_desc * const *vopvdpp) 263 { 264 int (**opv_desc_vector)(void *); 265 int i; 266 267 /* 268 * Allocate the vectors. 269 */ 270 for (i = 0; vopvdpp[i] != NULL; i++) { 271 opv_desc_vector = 272 kmem_alloc(VNODE_OPS_COUNT * sizeof(PFI), KM_SLEEP); 273 memset(opv_desc_vector, 0, VNODE_OPS_COUNT * sizeof(PFI)); 274 *(vopvdpp[i]->opv_desc_vector_p) = opv_desc_vector; 275 DODEBUG(printf("vector at %p allocated\n", 276 opv_desc_vector_p)); 277 } 278 279 /* 280 * ...and fill them in. 281 */ 282 for (i = 0; vopvdpp[i] != NULL; i++) 283 vfs_opv_init_explicit(vopvdpp[i]); 284 285 /* 286 * Finally, go back and replace unfilled routines 287 * with their default. 288 */ 289 for (i = 0; vopvdpp[i] != NULL; i++) 290 vfs_opv_init_default(vopvdpp[i]); 291 } 292 293 void 294 vfs_opv_free(const struct vnodeopv_desc * const *vopvdpp) 295 { 296 int i; 297 298 /* 299 * Free the vectors allocated in vfs_opv_init(). 300 */ 301 for (i = 0; vopvdpp[i] != NULL; i++) { 302 kmem_free(*(vopvdpp[i]->opv_desc_vector_p), 303 VNODE_OPS_COUNT * sizeof(PFI)); 304 *(vopvdpp[i]->opv_desc_vector_p) = NULL; 305 } 306 } 307 308 #ifdef DEBUG 309 static void 310 vfs_op_check(void) 311 { 312 int i; 313 314 DODEBUG(printf("Vnode_interface_init.\n")); 315 316 /* 317 * Check offset of each op. 318 */ 319 for (i = 0; vfs_op_descs[i]; i++) { 320 if (vfs_op_descs[i]->vdesc_offset != i) 321 panic("vfs_op_check: vfs_op_desc[] offset mismatch"); 322 } 323 324 if (i != VNODE_OPS_COUNT) { 325 panic("vfs_op_check: vnode ops count mismatch (%d != %d)", 326 i, VNODE_OPS_COUNT); 327 } 328 329 DODEBUG(printf ("vfs_opv_numops=%d\n", VNODE_OPS_COUNT)); 330 } 331 #endif /* DEBUG */ 332 333 /* 334 * Common routine to check if an unprivileged mount is allowed. 335 * 336 * We export just this part (i.e., without the access control) so that if a 337 * secmodel wants to implement finer grained user mounts it can do so without 338 * copying too much code. More elaborate policies (i.e., specific users allowed 339 * to also create devices and/or introduce set-id binaries, or export 340 * file-systems) will require a different implementation. 341 * 342 * This routine is intended to be called from listener context, and as such 343 * does not take credentials as an argument. 344 */ 345 int 346 usermount_common_policy(struct mount *mp, u_long flags) 347 { 348 349 /* No exporting if unprivileged. */ 350 if (flags & MNT_EXPORTED) 351 return EPERM; 352 353 /* Must have 'nosuid' and 'nodev'. */ 354 if ((flags & MNT_NODEV) == 0 || (flags & MNT_NOSUID) == 0) 355 return EPERM; 356 357 /* Retain 'noexec'. */ 358 if ((mp->mnt_flag & MNT_NOEXEC) && (flags & MNT_NOEXEC) == 0) 359 return EPERM; 360 361 return 0; 362 } 363 364 static int 365 mount_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 366 void *arg0, void *arg1, void *arg2, void *arg3) 367 { 368 int result; 369 enum kauth_system_req req; 370 371 result = KAUTH_RESULT_DEFER; 372 req = (enum kauth_system_req)(uintptr_t)(uintptr_t)arg0; 373 374 if (action != KAUTH_SYSTEM_MOUNT) 375 return result; 376 377 if (req == KAUTH_REQ_SYSTEM_MOUNT_GET) 378 result = KAUTH_RESULT_ALLOW; 379 else if (req == KAUTH_REQ_SYSTEM_MOUNT_DEVICE) { 380 vnode_t *devvp = arg2; 381 accmode_t accmode = (accmode_t)(unsigned long)arg3; 382 int error; 383 384 error = VOP_ACCESS(devvp, accmode, cred); 385 if (!error) 386 result = KAUTH_RESULT_ALLOW; 387 } 388 389 return result; 390 } 391 392 /* 393 * Initialize the vnode structures and initialize each file system type. 394 */ 395 void 396 vfsinit(void) 397 { 398 399 /* 400 * Attach sysctl nodes 401 */ 402 sysctl_vfs_setup(); 403 404 /* 405 * Initialize the namei pathname buffer pool and cache. 406 */ 407 pnbuf_cache = pool_cache_init(MAXPATHLEN, 0, 0, 0, "pnbufpl", 408 NULL, IPL_NONE, NULL, NULL, NULL); 409 KASSERT(pnbuf_cache != NULL); 410 411 /* 412 * Initialize the vnode table 413 */ 414 vntblinit(); 415 416 /* 417 * Initialize the vnode name cache 418 */ 419 nchinit(); 420 421 #ifdef DEBUG 422 /* 423 * Check the list of vnode operations. 424 */ 425 vfs_op_check(); 426 #endif 427 428 /* 429 * Initialize the special vnode operations. 430 */ 431 vfs_opv_init(vfs_special_vnodeopv_descs); 432 433 /* 434 * Initialise generic dirhash. 435 */ 436 dirhash_init(); 437 438 /* 439 * Initialise VFS hooks. 440 */ 441 vfs_hooks_init(); 442 443 mount_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM, 444 mount_listener_cb, NULL); 445 446 /* 447 * Establish each file system which was statically 448 * included in the kernel. 449 */ 450 module_init_class(MODULE_CLASS_VFS); 451 452 /* 453 * Initialize EVFILT_FS for kqueue. 454 */ 455 vfs_evfilt_fs_init(); 456 } 457 458 /* 459 * Drop a reference to a file system type. 460 */ 461 void 462 vfs_delref(struct vfsops *vfs) 463 { 464 465 mutex_enter(&vfs_list_lock); 466 vfs->vfs_refcount--; 467 mutex_exit(&vfs_list_lock); 468 } 469 470 /* 471 * Establish a file system and initialize it. 472 */ 473 int 474 vfs_attach(struct vfsops *vfs) 475 { 476 struct vfsops *v; 477 int error = 0; 478 479 mutex_enter(&vfs_list_lock); 480 481 /* 482 * Make sure this file system doesn't already exist. 483 */ 484 LIST_FOREACH(v, &vfs_list, vfs_list) { 485 if (strcmp(vfs->vfs_name, v->vfs_name) == 0) { 486 error = EEXIST; 487 goto out; 488 } 489 } 490 491 /* 492 * Initialize the vnode operations for this file system. 493 */ 494 vfs_opv_init(vfs->vfs_opv_descs); 495 496 /* 497 * Now initialize the file system itself. 498 */ 499 (*vfs->vfs_init)(); 500 501 /* 502 * ...and link it into the kernel's list. 503 */ 504 LIST_INSERT_HEAD(&vfs_list, vfs, vfs_list); 505 506 /* 507 * Sanity: make sure the reference count is 0. 508 */ 509 vfs->vfs_refcount = 0; 510 out: 511 mutex_exit(&vfs_list_lock); 512 return (error); 513 } 514 515 /* 516 * Remove a file system from the kernel. 517 */ 518 int 519 vfs_detach(struct vfsops *vfs) 520 { 521 struct vfsops *v; 522 int error = 0; 523 524 mutex_enter(&vfs_list_lock); 525 526 /* 527 * Make sure no one is using the filesystem. 528 */ 529 if (vfs->vfs_refcount != 0) { 530 error = EBUSY; 531 goto out; 532 } 533 534 /* 535 * ...and remove it from the kernel's list. 536 */ 537 LIST_FOREACH(v, &vfs_list, vfs_list) { 538 if (v == vfs) { 539 LIST_REMOVE(v, vfs_list); 540 break; 541 } 542 } 543 544 if (v == NULL) { 545 error = ESRCH; 546 goto out; 547 } 548 549 /* 550 * Now run the file system-specific cleanups. 551 */ 552 (*vfs->vfs_done)(); 553 554 /* 555 * Free the vnode operations vector. 556 */ 557 vfs_opv_free(vfs->vfs_opv_descs); 558 out: 559 mutex_exit(&vfs_list_lock); 560 return (error); 561 } 562 563 void 564 vfs_reinit(void) 565 { 566 struct vfsops *vfs; 567 568 mutex_enter(&vfs_list_lock); 569 LIST_FOREACH(vfs, &vfs_list, vfs_list) { 570 if (vfs->vfs_reinit) { 571 vfs->vfs_refcount++; 572 mutex_exit(&vfs_list_lock); 573 (*vfs->vfs_reinit)(); 574 mutex_enter(&vfs_list_lock); 575 vfs->vfs_refcount--; 576 } 577 } 578 mutex_exit(&vfs_list_lock); 579 } 580