vfs_init.c revision 1.60
1 /* $NetBSD: vfs_init.c,v 1.60 2023/04/29 10:07:13 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 1.60 2023/04/29 10:07:13 riastradh Exp $"); 71 72 #include <sys/param.h> 73 #include <sys/types.h> 74 75 #include <sys/buf.h> 76 #include <sys/dirhash.h> 77 #include <sys/errno.h> 78 #include <sys/kauth.h> 79 #include <sys/kmem.h> 80 #include <sys/module.h> 81 #include <sys/mount.h> 82 #include <sys/namei.h> 83 #include <sys/stat.h> 84 #include <sys/sysctl.h> 85 #include <sys/systm.h> 86 #include <sys/time.h> 87 #include <sys/ucred.h> 88 #include <sys/vnode.h> 89 90 #include <miscfs/deadfs/deadfs.h> 91 #include <miscfs/fifofs/fifo.h> 92 #include <miscfs/specfs/specdev.h> 93 94 /* 95 * Sigh, such primitive tools are these... 96 */ 97 #if 0 98 #define DODEBUG(A) A 99 #else 100 #define DODEBUG(A) 101 #endif 102 103 pool_cache_t pnbuf_cache; 104 105 /* 106 * These vnodeopv_descs are listed here because they are not 107 * associated with any particular file system, and thus cannot 108 * be initialized by vfs_attach(). 109 */ 110 const struct vnodeopv_desc * const vfs_special_vnodeopv_descs[] = { 111 &dead_vnodeop_opv_desc, 112 &fifo_vnodeop_opv_desc, 113 &spec_vnodeop_opv_desc, 114 NULL, 115 }; 116 117 struct vfs_list_head vfs_list = /* vfs list */ 118 LIST_HEAD_INITIALIZER(vfs_list); 119 120 static kauth_listener_t mount_listener; 121 122 /* 123 * This code doesn't work if the defn is **vnodop_defns with cc. 124 * The problem is because of the compiler sometimes putting in an 125 * extra level of indirection for arrays. It's an interesting 126 * "feature" of C. 127 */ 128 typedef int (*PFI)(void *); 129 130 /* 131 * A miscellaneous routine. 132 * A generic "default" routine that just returns an error. 133 */ 134 /*ARGSUSED*/ 135 int 136 vn_default_error(void *v) 137 { 138 139 return (EOPNOTSUPP); 140 } 141 142 static struct sysctllog *vfs_sysctllog; 143 144 /* 145 * Top level filesystem related information gathering. 146 */ 147 static void 148 sysctl_vfs_setup(void) 149 { 150 151 sysctl_createv(&vfs_sysctllog, 0, NULL, NULL, 152 CTLFLAG_PERMANENT, 153 CTLTYPE_NODE, "generic", 154 SYSCTL_DESCR("Non-specific vfs related information"), 155 NULL, 0, NULL, 0, 156 CTL_VFS, VFS_GENERIC, CTL_EOL); 157 sysctl_createv(&vfs_sysctllog, 0, NULL, NULL, 158 CTLFLAG_PERMANENT, 159 CTLTYPE_STRING, "fstypes", 160 SYSCTL_DESCR("List of file systems present"), 161 sysctl_vfs_generic_fstypes, 0, NULL, 0, 162 CTL_VFS, VFS_GENERIC, CTL_CREATE, CTL_EOL); 163 sysctl_createv(&vfs_sysctllog, 0, NULL, NULL, 164 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 165 CTLTYPE_INT, "magiclinks", 166 SYSCTL_DESCR("Whether \"magic\" symlinks are expanded"), 167 NULL, 0, &vfs_magiclinks, 0, 168 CTL_VFS, VFS_GENERIC, VFS_MAGICLINKS, CTL_EOL); 169 sysctl_createv(&vfs_sysctllog, 0, NULL, NULL, 170 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 171 CTLTYPE_INT, "timestamp_precision", 172 SYSCTL_DESCR("File timestamp precision"), 173 NULL, 0, &vfs_timestamp_precision, 0, 174 CTL_VFS, VFS_GENERIC, VFS_TIMESTAMP_PRECISION, 175 CTL_EOL); 176 } 177 178 179 /* 180 * vfs_init.c 181 * 182 * Allocate and fill in operations vectors. 183 * 184 * An undocumented feature of this approach to defining operations is that 185 * there can be multiple entries in vfs_opv_descs for the same operations 186 * vector. This allows third parties to extend the set of operations 187 * supported by another layer in a binary compatibile way. For example, 188 * assume that NFS needed to be modified to support Ficus. NFS has an entry 189 * (probably nfs_vnopdeop_decls) declaring all the operations NFS supports by 190 * default. Ficus could add another entry (ficus_nfs_vnodeop_decl_entensions) 191 * listing those new operations Ficus adds to NFS, all without modifying the 192 * NFS code. (Of couse, the OTW NFS protocol still needs to be munged, but 193 * that is a(whole)nother story.) This is a feature. 194 */ 195 196 /* 197 * Init the vector, if it needs it. 198 * Also handle backwards compatibility. 199 */ 200 static void 201 vfs_opv_init_explicit(const struct vnodeopv_desc *vfs_opv_desc) 202 { 203 int (**opv_desc_vector)(void *); 204 const struct vnodeopv_entry_desc *opve_descp; 205 206 opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p); 207 208 for (opve_descp = vfs_opv_desc->opv_desc_ops; 209 opve_descp->opve_op; 210 opve_descp++) { 211 /* 212 * Sanity check: is this operation listed 213 * in the list of operations? We check this 214 * by seeing if its offset is zero. Since 215 * the default routine should always be listed 216 * first, it should be the only one with a zero 217 * offset. Any other operation with a zero 218 * offset is probably not listed in 219 * vfs_op_descs, and so is probably an error. 220 * 221 * A panic here means the layer programmer 222 * has committed the all-too common bug 223 * of adding a new operation to the layer's 224 * list of vnode operations but 225 * not adding the operation to the system-wide 226 * list of supported operations. 227 */ 228 if (opve_descp->opve_op->vdesc_offset == 0 && 229 opve_descp->opve_op->vdesc_offset != VOFFSET(vop_default)) { 230 printf("operation %s not listed in %s.\n", 231 opve_descp->opve_op->vdesc_name, "vfs_op_descs"); 232 panic ("vfs_opv_init: bad operation"); 233 } 234 235 /* 236 * Fill in this entry. 237 */ 238 opv_desc_vector[opve_descp->opve_op->vdesc_offset] = 239 opve_descp->opve_impl; 240 } 241 } 242 243 static void 244 vfs_opv_init_default(const struct vnodeopv_desc *vfs_opv_desc) 245 { 246 int j; 247 int (**opv_desc_vector)(void *); 248 249 opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p); 250 251 /* 252 * Force every operations vector to have a default routine. 253 */ 254 if (opv_desc_vector[VOFFSET(vop_default)] == NULL) 255 panic("vfs_opv_init: operation vector without default routine."); 256 257 for (j = 0; j < VNODE_OPS_COUNT; j++) 258 if (opv_desc_vector[j] == NULL) 259 opv_desc_vector[j] = 260 opv_desc_vector[VOFFSET(vop_default)]; 261 } 262 263 void 264 vfs_opv_init(const struct vnodeopv_desc * const *vopvdpp) 265 { 266 int (**opv_desc_vector)(void *); 267 int i; 268 269 /* 270 * Allocate the vectors. 271 */ 272 for (i = 0; vopvdpp[i] != NULL; i++) { 273 opv_desc_vector = 274 kmem_alloc(VNODE_OPS_COUNT * sizeof(PFI), KM_SLEEP); 275 memset(opv_desc_vector, 0, VNODE_OPS_COUNT * sizeof(PFI)); 276 *(vopvdpp[i]->opv_desc_vector_p) = opv_desc_vector; 277 DODEBUG(printf("vector at %p allocated\n", 278 opv_desc_vector_p)); 279 } 280 281 /* 282 * ...and fill them in. 283 */ 284 for (i = 0; vopvdpp[i] != NULL; i++) 285 vfs_opv_init_explicit(vopvdpp[i]); 286 287 /* 288 * Finally, go back and replace unfilled routines 289 * with their default. 290 */ 291 for (i = 0; vopvdpp[i] != NULL; i++) 292 vfs_opv_init_default(vopvdpp[i]); 293 } 294 295 void 296 vfs_opv_free(const struct vnodeopv_desc * const *vopvdpp) 297 { 298 int i; 299 300 /* 301 * Free the vectors allocated in vfs_opv_init(). 302 */ 303 for (i = 0; vopvdpp[i] != NULL; i++) { 304 kmem_free(*(vopvdpp[i]->opv_desc_vector_p), 305 VNODE_OPS_COUNT * sizeof(PFI)); 306 *(vopvdpp[i]->opv_desc_vector_p) = NULL; 307 } 308 } 309 310 #ifdef DEBUG 311 static void 312 vfs_op_check(void) 313 { 314 int i; 315 316 DODEBUG(printf("Vnode_interface_init.\n")); 317 318 /* 319 * Check offset of each op. 320 */ 321 for (i = 0; vfs_op_descs[i]; i++) { 322 if (vfs_op_descs[i]->vdesc_offset != i) 323 panic("vfs_op_check: vfs_op_desc[] offset mismatch"); 324 } 325 326 if (i != VNODE_OPS_COUNT) { 327 panic("vfs_op_check: vnode ops count mismatch (%d != %d)", 328 i, VNODE_OPS_COUNT); 329 } 330 331 DODEBUG(printf ("vfs_opv_numops=%d\n", VNODE_OPS_COUNT)); 332 } 333 #endif /* DEBUG */ 334 335 /* 336 * Common routine to check if an unprivileged mount is allowed. 337 * 338 * We export just this part (i.e., without the access control) so that if a 339 * secmodel wants to implement finer grained user mounts it can do so without 340 * copying too much code. More elaborate policies (i.e., specific users allowed 341 * to also create devices and/or introduce set-id binaries, or export 342 * file-systems) will require a different implementation. 343 * 344 * This routine is intended to be called from listener context, and as such 345 * does not take credentials as an argument. 346 */ 347 int 348 usermount_common_policy(struct mount *mp, u_long flags) 349 { 350 351 /* No exporting if unprivileged. */ 352 if (flags & MNT_EXPORTED) 353 return EPERM; 354 355 /* Must have 'nosuid' and 'nodev'. */ 356 if ((flags & MNT_NODEV) == 0 || (flags & MNT_NOSUID) == 0) 357 return EPERM; 358 359 /* Retain 'noexec'. */ 360 if ((mp->mnt_flag & MNT_NOEXEC) && (flags & MNT_NOEXEC) == 0) 361 return EPERM; 362 363 return 0; 364 } 365 366 static int 367 mount_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 368 void *arg0, void *arg1, void *arg2, void *arg3) 369 { 370 int result; 371 enum kauth_system_req req; 372 373 result = KAUTH_RESULT_DEFER; 374 req = (enum kauth_system_req)(uintptr_t)(uintptr_t)arg0; 375 376 if (action != KAUTH_SYSTEM_MOUNT) 377 return result; 378 379 if (req == KAUTH_REQ_SYSTEM_MOUNT_GET) 380 result = KAUTH_RESULT_ALLOW; 381 else if (req == KAUTH_REQ_SYSTEM_MOUNT_DEVICE) { 382 vnode_t *devvp = arg2; 383 accmode_t accmode = (accmode_t)(unsigned long)arg3; 384 int error; 385 386 error = VOP_ACCESS(devvp, accmode, cred); 387 if (!error) 388 result = KAUTH_RESULT_ALLOW; 389 } 390 391 return result; 392 } 393 394 /* 395 * Initialize the vnode structures and initialize each file system type. 396 */ 397 void 398 vfsinit(void) 399 { 400 401 /* 402 * Attach sysctl nodes 403 */ 404 sysctl_vfs_setup(); 405 406 /* 407 * Initialize the namei pathname buffer pool and cache. 408 */ 409 pnbuf_cache = pool_cache_init(MAXPATHLEN, 0, 0, 0, "pnbufpl", 410 NULL, IPL_NONE, NULL, NULL, NULL); 411 KASSERT(pnbuf_cache != NULL); 412 413 /* 414 * Initialize the vnode table 415 */ 416 vntblinit(); 417 418 /* 419 * Initialize the vnode name cache 420 */ 421 nchinit(); 422 423 #ifdef DEBUG 424 /* 425 * Check the list of vnode operations. 426 */ 427 vfs_op_check(); 428 #endif 429 430 /* 431 * Initialize the special vnode operations. 432 */ 433 vfs_opv_init(vfs_special_vnodeopv_descs); 434 435 /* 436 * Initialise generic dirhash. 437 */ 438 dirhash_init(); 439 440 /* 441 * Initialise VFS hooks. 442 */ 443 vfs_hooks_init(); 444 445 mount_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM, 446 mount_listener_cb, NULL); 447 448 /* 449 * Establish each file system which was statically 450 * included in the kernel. 451 */ 452 module_init_class(MODULE_CLASS_VFS); 453 454 /* 455 * Initialize EVFILT_FS for kqueue. 456 */ 457 vfs_evfilt_fs_init(); 458 } 459 460 /* 461 * Drop a reference to a file system type. 462 */ 463 void 464 vfs_delref(struct vfsops *vfs) 465 { 466 467 mutex_enter(&vfs_list_lock); 468 vfs->vfs_refcount--; 469 mutex_exit(&vfs_list_lock); 470 } 471 472 /* 473 * Establish a file system and initialize it. 474 */ 475 int 476 vfs_attach(struct vfsops *vfs) 477 { 478 struct vfsops *v; 479 int error = 0; 480 481 mutex_enter(&vfs_list_lock); 482 483 /* 484 * Make sure this file system doesn't already exist. 485 */ 486 LIST_FOREACH(v, &vfs_list, vfs_list) { 487 if (strcmp(vfs->vfs_name, v->vfs_name) == 0) { 488 error = EEXIST; 489 goto out; 490 } 491 } 492 493 /* 494 * Initialize the vnode operations for this file system. 495 */ 496 vfs_opv_init(vfs->vfs_opv_descs); 497 498 /* 499 * Now initialize the file system itself. 500 */ 501 (*vfs->vfs_init)(); 502 503 /* 504 * ...and link it into the kernel's list. 505 */ 506 LIST_INSERT_HEAD(&vfs_list, vfs, vfs_list); 507 508 /* 509 * Sanity: make sure the reference count is 0. 510 */ 511 vfs->vfs_refcount = 0; 512 out: 513 mutex_exit(&vfs_list_lock); 514 return (error); 515 } 516 517 /* 518 * Remove a file system from the kernel. 519 */ 520 int 521 vfs_detach(struct vfsops *vfs) 522 { 523 struct vfsops *v; 524 int error = 0; 525 526 mutex_enter(&vfs_list_lock); 527 528 /* 529 * Make sure no one is using the filesystem. 530 */ 531 if (vfs->vfs_refcount != 0) { 532 error = EBUSY; 533 goto out; 534 } 535 536 /* 537 * ...and remove it from the kernel's list. 538 */ 539 LIST_FOREACH(v, &vfs_list, vfs_list) { 540 if (v == vfs) { 541 LIST_REMOVE(v, vfs_list); 542 break; 543 } 544 } 545 546 if (v == NULL) { 547 error = ESRCH; 548 goto out; 549 } 550 551 /* 552 * Now run the file system-specific cleanups. 553 */ 554 (*vfs->vfs_done)(); 555 556 /* 557 * Free the vnode operations vector. 558 */ 559 vfs_opv_free(vfs->vfs_opv_descs); 560 out: 561 mutex_exit(&vfs_list_lock); 562 return (error); 563 } 564 565 void 566 vfs_reinit(void) 567 { 568 struct vfsops *vfs; 569 570 mutex_enter(&vfs_list_lock); 571 LIST_FOREACH(vfs, &vfs_list, vfs_list) { 572 if (vfs->vfs_reinit) { 573 vfs->vfs_refcount++; 574 mutex_exit(&vfs_list_lock); 575 (*vfs->vfs_reinit)(); 576 mutex_enter(&vfs_list_lock); 577 vfs->vfs_refcount--; 578 } 579 } 580 mutex_exit(&vfs_list_lock); 581 } 582
Indexes created Sat Sep 20 22:09:52 GMT 2025