xref: /netbsd-src/sys/kern/vfs_init.c (revision da9817918ec7e88db2912a2882967c7570a83f47)
1 /*	$NetBSD: vfs_init.c,v 1.44 2009/05/03 21:25:44 elad 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 1.44 2009/05/03 21:25:44 elad 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 
87 /*
88  * Sigh, such primitive tools are these...
89  */
90 #if 0
91 #define DODEBUG(A) A
92 #else
93 #define DODEBUG(A)
94 #endif
95 
96 /*
97  * The global list of vnode operations.
98  */
99 extern const struct vnodeop_desc * const vfs_op_descs[];
100 
101 /*
102  * These vnodeopv_descs are listed here because they are not
103  * associated with any particular file system, and thus cannot
104  * be initialized by vfs_attach().
105  */
106 extern const struct vnodeopv_desc dead_vnodeop_opv_desc;
107 extern const struct vnodeopv_desc fifo_vnodeop_opv_desc;
108 extern const struct vnodeopv_desc spec_vnodeop_opv_desc;
109 extern const struct vnodeopv_desc sync_vnodeop_opv_desc;
110 
111 const struct vnodeopv_desc * const vfs_special_vnodeopv_descs[] = {
112 	&dead_vnodeop_opv_desc,
113 	&fifo_vnodeop_opv_desc,
114 	&spec_vnodeop_opv_desc,
115 	&sync_vnodeop_opv_desc,
116 	NULL,
117 };
118 
119 struct vfs_list_head vfs_list =			/* vfs list */
120     LIST_HEAD_INITIALIZER(vfs_list);
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 	extern int vfs_magiclinks;
151 
152 	sysctl_createv(&vfs_sysctllog, 0, NULL, NULL,
153 		       CTLFLAG_PERMANENT,
154 		       CTLTYPE_NODE, "vfs", NULL,
155 		       NULL, 0, NULL, 0,
156 		       CTL_VFS, CTL_EOL);
157 	sysctl_createv(&vfs_sysctllog, 0, NULL, NULL,
158 		       CTLFLAG_PERMANENT,
159 		       CTLTYPE_NODE, "generic",
160 		       SYSCTL_DESCR("Non-specific vfs related information"),
161 		       NULL, 0, NULL, 0,
162 		       CTL_VFS, VFS_GENERIC, CTL_EOL);
163 	sysctl_createv(&vfs_sysctllog, 0, NULL, NULL,
164 		       CTLFLAG_PERMANENT,
165 		       CTLTYPE_STRING, "fstypes",
166 		       SYSCTL_DESCR("List of file systems present"),
167 		       sysctl_vfs_generic_fstypes, 0, NULL, 0,
168 		       CTL_VFS, VFS_GENERIC, CTL_CREATE, CTL_EOL);
169 	sysctl_createv(&vfs_sysctllog, 0, NULL, NULL,
170 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
171 		       CTLTYPE_INT, "magiclinks",
172 		       SYSCTL_DESCR("Whether \"magic\" symlinks are expanded"),
173 		       NULL, 0, &vfs_magiclinks, 0,
174 		       CTL_VFS, VFS_GENERIC, VFS_MAGICLINKS, CTL_EOL);
175 }
176 
177 
178 /*
179  * vfs_init.c
180  *
181  * Allocate and fill in operations vectors.
182  *
183  * An undocumented feature of this approach to defining operations is that
184  * there can be multiple entries in vfs_opv_descs for the same operations
185  * vector. This allows third parties to extend the set of operations
186  * supported by another layer in a binary compatibile way. For example,
187  * assume that NFS needed to be modified to support Ficus. NFS has an entry
188  * (probably nfs_vnopdeop_decls) declaring all the operations NFS supports by
189  * default. Ficus could add another entry (ficus_nfs_vnodeop_decl_entensions)
190  * listing those new operations Ficus adds to NFS, all without modifying the
191  * NFS code. (Of couse, the OTW NFS protocol still needs to be munged, but
192  * that is a(whole)nother story.) This is a feature.
193  */
194 
195 /*
196  * Init the vector, if it needs it.
197  * Also handle backwards compatibility.
198  */
199 static void
200 vfs_opv_init_explicit(const struct vnodeopv_desc *vfs_opv_desc)
201 {
202 	int (**opv_desc_vector)(void *);
203 	const struct vnodeopv_entry_desc *opve_descp;
204 
205 	opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p);
206 
207 	for (opve_descp = vfs_opv_desc->opv_desc_ops;
208 	     opve_descp->opve_op;
209 	     opve_descp++) {
210 		/*
211 		 * Sanity check:  is this operation listed
212 		 * in the list of operations?  We check this
213 		 * by seeing if its offset is zero.  Since
214 		 * the default routine should always be listed
215 		 * first, it should be the only one with a zero
216 		 * offset.  Any other operation with a zero
217 		 * offset is probably not listed in
218 		 * vfs_op_descs, and so is probably an error.
219 		 *
220 		 * A panic here means the layer programmer
221 		 * has committed the all-too common bug
222 		 * of adding a new operation to the layer's
223 		 * list of vnode operations but
224 		 * not adding the operation to the system-wide
225 		 * list of supported operations.
226 		 */
227 		if (opve_descp->opve_op->vdesc_offset == 0 &&
228 		    opve_descp->opve_op->vdesc_offset != VOFFSET(vop_default)) {
229 			printf("operation %s not listed in %s.\n",
230 			    opve_descp->opve_op->vdesc_name, "vfs_op_descs");
231 			panic ("vfs_opv_init: bad operation");
232 		}
233 
234 		/*
235 		 * Fill in this entry.
236 		 */
237 		opv_desc_vector[opve_descp->opve_op->vdesc_offset] =
238 		    opve_descp->opve_impl;
239 	}
240 }
241 
242 static void
243 vfs_opv_init_default(const struct vnodeopv_desc *vfs_opv_desc)
244 {
245 	int j;
246 	int (**opv_desc_vector)(void *);
247 
248 	opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p);
249 
250 	/*
251 	 * Force every operations vector to have a default routine.
252 	 */
253 	if (opv_desc_vector[VOFFSET(vop_default)] == NULL)
254 		panic("vfs_opv_init: operation vector without default routine.");
255 
256 	for (j = 0; j < VNODE_OPS_COUNT; j++)
257 		if (opv_desc_vector[j] == NULL)
258 			opv_desc_vector[j] =
259 			    opv_desc_vector[VOFFSET(vop_default)];
260 }
261 
262 void
263 vfs_opv_init(const struct vnodeopv_desc * const *vopvdpp)
264 {
265 	int (**opv_desc_vector)(void *);
266 	int i;
267 
268 	/*
269 	 * Allocate the vectors.
270 	 */
271 	for (i = 0; vopvdpp[i] != NULL; i++) {
272 		opv_desc_vector =
273 		    kmem_alloc(VNODE_OPS_COUNT * sizeof(PFI), KM_SLEEP);
274 		memset(opv_desc_vector, 0, VNODE_OPS_COUNT * sizeof(PFI));
275 		*(vopvdpp[i]->opv_desc_vector_p) = opv_desc_vector;
276 		DODEBUG(printf("vector at %p allocated\n",
277 		    opv_desc_vector_p));
278 	}
279 
280 	/*
281 	 * ...and fill them in.
282 	 */
283 	for (i = 0; vopvdpp[i] != NULL; i++)
284 		vfs_opv_init_explicit(vopvdpp[i]);
285 
286 	/*
287 	 * Finally, go back and replace unfilled routines
288 	 * with their default.
289 	 */
290 	for (i = 0; vopvdpp[i] != NULL; i++)
291 		vfs_opv_init_default(vopvdpp[i]);
292 }
293 
294 void
295 vfs_opv_free(const struct vnodeopv_desc * const *vopvdpp)
296 {
297 	int i;
298 
299 	/*
300 	 * Free the vectors allocated in vfs_opv_init().
301 	 */
302 	for (i = 0; vopvdpp[i] != NULL; i++) {
303 		kmem_free(*(vopvdpp[i]->opv_desc_vector_p),
304 		    VNODE_OPS_COUNT * sizeof(PFI));
305 		*(vopvdpp[i]->opv_desc_vector_p) = NULL;
306 	}
307 }
308 
309 #ifdef DEBUG
310 static void
311 vfs_op_check(void)
312 {
313 	int i;
314 
315 	DODEBUG(printf("Vnode_interface_init.\n"));
316 
317 	/*
318 	 * Check offset of each op.
319 	 */
320 	for (i = 0; vfs_op_descs[i]; i++) {
321 		if (vfs_op_descs[i]->vdesc_offset != i)
322 			panic("vfs_op_check: vfs_op_desc[] offset mismatch");
323 	}
324 
325 	if (i != VNODE_OPS_COUNT) {
326 		panic("vfs_op_check: vnode ops count mismatch (%d != %d)",
327 			i, VNODE_OPS_COUNT);
328 	}
329 
330 	DODEBUG(printf ("vfs_opv_numops=%d\n", VNODE_OPS_COUNT));
331 }
332 #endif /* DEBUG */
333 
334 /*
335  * Initialize the vnode structures and initialize each file system type.
336  */
337 void
338 vfsinit(void)
339 {
340 
341 	/*
342 	 * Attach sysctl nodes
343 	 */
344 	sysctl_vfs_setup();
345 
346 	/*
347 	 * Initialize the namei pathname buffer pool and cache.
348 	 */
349 	pnbuf_cache = pool_cache_init(MAXPATHLEN, 0, 0, 0, "pnbufpl",
350 	    NULL, IPL_NONE, NULL, NULL, NULL);
351 	KASSERT(pnbuf_cache != NULL);
352 
353 	/*
354 	 * Initialize the vnode table
355 	 */
356 	vntblinit();
357 
358 	/*
359 	 * Initialize the vnode name cache
360 	 */
361 	nchinit();
362 
363 #ifdef DEBUG
364 	/*
365 	 * Check the list of vnode operations.
366 	 */
367 	vfs_op_check();
368 #endif
369 
370 	/*
371 	 * Initialize the special vnode operations.
372 	 */
373 	vfs_opv_init(vfs_special_vnodeopv_descs);
374 
375 	/*
376 	 * Initialise generic dirhash.
377 	 */
378 	dirhash_init();
379 
380 	/*
381 	 * Initialise VFS hooks.
382 	 */
383 	vfs_hooks_init();
384 
385 	/*
386 	 * Establish each file system which was statically
387 	 * included in the kernel.
388 	 */
389 	module_init_class(MODULE_CLASS_VFS);
390 }
391 
392 /*
393  * Drop a reference to a file system type.
394  */
395 void
396 vfs_delref(struct vfsops *vfs)
397 {
398 
399 	mutex_enter(&vfs_list_lock);
400 	vfs->vfs_refcount--;
401 	mutex_exit(&vfs_list_lock);
402 }
403 
404 /*
405  * Establish a file system and initialize it.
406  */
407 int
408 vfs_attach(struct vfsops *vfs)
409 {
410 	struct vfsops *v;
411 	int error = 0;
412 
413 	mutex_enter(&vfs_list_lock);
414 
415 	/*
416 	 * Make sure this file system doesn't already exist.
417 	 */
418 	LIST_FOREACH(v, &vfs_list, vfs_list) {
419 		if (strcmp(vfs->vfs_name, v->vfs_name) == 0) {
420 			error = EEXIST;
421 			goto out;
422 		}
423 	}
424 
425 	/*
426 	 * Initialize the vnode operations for this file system.
427 	 */
428 	vfs_opv_init(vfs->vfs_opv_descs);
429 
430 	/*
431 	 * Now initialize the file system itself.
432 	 */
433 	(*vfs->vfs_init)();
434 
435 	/*
436 	 * ...and link it into the kernel's list.
437 	 */
438 	LIST_INSERT_HEAD(&vfs_list, vfs, vfs_list);
439 
440 	/*
441 	 * Sanity: make sure the reference count is 0.
442 	 */
443 	vfs->vfs_refcount = 0;
444  out:
445 	mutex_exit(&vfs_list_lock);
446 	return (error);
447 }
448 
449 /*
450  * Remove a file system from the kernel.
451  */
452 int
453 vfs_detach(struct vfsops *vfs)
454 {
455 	struct vfsops *v;
456 	int error = 0;
457 
458 	mutex_enter(&vfs_list_lock);
459 
460 	/*
461 	 * Make sure no one is using the filesystem.
462 	 */
463 	if (vfs->vfs_refcount != 0) {
464 		error = EBUSY;
465 		goto out;
466 	}
467 
468 	/*
469 	 * ...and remove it from the kernel's list.
470 	 */
471 	LIST_FOREACH(v, &vfs_list, vfs_list) {
472 		if (v == vfs) {
473 			LIST_REMOVE(v, vfs_list);
474 			break;
475 		}
476 	}
477 
478 	if (v == NULL) {
479 		error = ESRCH;
480 		goto out;
481 	}
482 
483 	/*
484 	 * Now run the file system-specific cleanups.
485 	 */
486 	(*vfs->vfs_done)();
487 
488 	/*
489 	 * Free the vnode operations vector.
490 	 */
491 	vfs_opv_free(vfs->vfs_opv_descs);
492  out:
493  	mutex_exit(&vfs_list_lock);
494 	return (error);
495 }
496 
497 void
498 vfs_reinit(void)
499 {
500 	struct vfsops *vfs;
501 
502 	mutex_enter(&vfs_list_lock);
503 	LIST_FOREACH(vfs, &vfs_list, vfs_list) {
504 		if (vfs->vfs_reinit) {
505 			vfs->vfs_refcount++;
506 			mutex_exit(&vfs_list_lock);
507 			(*vfs->vfs_reinit)();
508 			mutex_enter(&vfs_list_lock);
509 			vfs->vfs_refcount--;
510 		}
511 	}
512 	mutex_exit(&vfs_list_lock);
513 }
514