xref: /netbsd-src/sys/kern/vfs_init.c (revision cac8e449158efc7261bebc8657cbb0125a2cfdde)
1 /*	$NetBSD: vfs_init.c,v 1.40 2008/05/10 02:26:09 rumble 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 1.40 2008/05/10 02:26:09 rumble 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/malloc.h>
82 #include <sys/systm.h>
83 #include <sys/module.h>
84 
85 /*
86  * Sigh, such primitive tools are these...
87  */
88 #if 0
89 #define DODEBUG(A) A
90 #else
91 #define DODEBUG(A)
92 #endif
93 
94 /*
95  * The global list of vnode operations.
96  */
97 extern const struct vnodeop_desc * const vfs_op_descs[];
98 
99 /*
100  * These vnodeopv_descs are listed here because they are not
101  * associated with any particular file system, and thus cannot
102  * be initialized by vfs_attach().
103  */
104 extern const struct vnodeopv_desc dead_vnodeop_opv_desc;
105 extern const struct vnodeopv_desc fifo_vnodeop_opv_desc;
106 extern const struct vnodeopv_desc spec_vnodeop_opv_desc;
107 extern const struct vnodeopv_desc sync_vnodeop_opv_desc;
108 
109 const struct vnodeopv_desc * const vfs_special_vnodeopv_descs[] = {
110 	&dead_vnodeop_opv_desc,
111 	&fifo_vnodeop_opv_desc,
112 	&spec_vnodeop_opv_desc,
113 	&sync_vnodeop_opv_desc,
114 	NULL,
115 };
116 
117 struct vfs_list_head vfs_list =			/* vfs list */
118     LIST_HEAD_INITIALIZER(vfs_list);
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 /*
141  * vfs_init.c
142  *
143  * Allocate and fill in operations vectors.
144  *
145  * An undocumented feature of this approach to defining operations is that
146  * there can be multiple entries in vfs_opv_descs for the same operations
147  * vector. This allows third parties to extend the set of operations
148  * supported by another layer in a binary compatibile way. For example,
149  * assume that NFS needed to be modified to support Ficus. NFS has an entry
150  * (probably nfs_vnopdeop_decls) declaring all the operations NFS supports by
151  * default. Ficus could add another entry (ficus_nfs_vnodeop_decl_entensions)
152  * listing those new operations Ficus adds to NFS, all without modifying the
153  * NFS code. (Of couse, the OTW NFS protocol still needs to be munged, but
154  * that is a(whole)nother story.) This is a feature.
155  */
156 
157 /*
158  * Init the vector, if it needs it.
159  * Also handle backwards compatibility.
160  */
161 static void
162 vfs_opv_init_explicit(const struct vnodeopv_desc *vfs_opv_desc)
163 {
164 	int (**opv_desc_vector)(void *);
165 	const struct vnodeopv_entry_desc *opve_descp;
166 
167 	opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p);
168 
169 	for (opve_descp = vfs_opv_desc->opv_desc_ops;
170 	     opve_descp->opve_op;
171 	     opve_descp++) {
172 		/*
173 		 * Sanity check:  is this operation listed
174 		 * in the list of operations?  We check this
175 		 * by seeing if its offset is zero.  Since
176 		 * the default routine should always be listed
177 		 * first, it should be the only one with a zero
178 		 * offset.  Any other operation with a zero
179 		 * offset is probably not listed in
180 		 * vfs_op_descs, and so is probably an error.
181 		 *
182 		 * A panic here means the layer programmer
183 		 * has committed the all-too common bug
184 		 * of adding a new operation to the layer's
185 		 * list of vnode operations but
186 		 * not adding the operation to the system-wide
187 		 * list of supported operations.
188 		 */
189 		if (opve_descp->opve_op->vdesc_offset == 0 &&
190 		    opve_descp->opve_op->vdesc_offset != VOFFSET(vop_default)) {
191 			printf("operation %s not listed in %s.\n",
192 			    opve_descp->opve_op->vdesc_name, "vfs_op_descs");
193 			panic ("vfs_opv_init: bad operation");
194 		}
195 
196 		/*
197 		 * Fill in this entry.
198 		 */
199 		opv_desc_vector[opve_descp->opve_op->vdesc_offset] =
200 		    opve_descp->opve_impl;
201 	}
202 }
203 
204 static void
205 vfs_opv_init_default(const struct vnodeopv_desc *vfs_opv_desc)
206 {
207 	int j;
208 	int (**opv_desc_vector)(void *);
209 
210 	opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p);
211 
212 	/*
213 	 * Force every operations vector to have a default routine.
214 	 */
215 	if (opv_desc_vector[VOFFSET(vop_default)] == NULL)
216 		panic("vfs_opv_init: operation vector without default routine.");
217 
218 	for (j = 0; j < VNODE_OPS_COUNT; j++)
219 		if (opv_desc_vector[j] == NULL)
220 			opv_desc_vector[j] =
221 			    opv_desc_vector[VOFFSET(vop_default)];
222 }
223 
224 void
225 vfs_opv_init(const struct vnodeopv_desc * const *vopvdpp)
226 {
227 	int (**opv_desc_vector)(void *);
228 	int i;
229 
230 	/*
231 	 * Allocate the vectors.
232 	 */
233 	for (i = 0; vopvdpp[i] != NULL; i++) {
234 		/* XXX - shouldn't be M_VNODE */
235 		opv_desc_vector =
236 		    malloc(VNODE_OPS_COUNT * sizeof(PFI), M_VNODE, M_WAITOK);
237 		memset(opv_desc_vector, 0, VNODE_OPS_COUNT * sizeof(PFI));
238 		*(vopvdpp[i]->opv_desc_vector_p) = opv_desc_vector;
239 		DODEBUG(printf("vector at %p allocated\n",
240 		    opv_desc_vector_p));
241 	}
242 
243 	/*
244 	 * ...and fill them in.
245 	 */
246 	for (i = 0; vopvdpp[i] != NULL; i++)
247 		vfs_opv_init_explicit(vopvdpp[i]);
248 
249 	/*
250 	 * Finally, go back and replace unfilled routines
251 	 * with their default.
252 	 */
253 	for (i = 0; vopvdpp[i] != NULL; i++)
254 		vfs_opv_init_default(vopvdpp[i]);
255 }
256 
257 void
258 vfs_opv_free(const struct vnodeopv_desc * const *vopvdpp)
259 {
260 	int i;
261 
262 	/*
263 	 * Free the vectors allocated in vfs_opv_init().
264 	 */
265 	for (i = 0; vopvdpp[i] != NULL; i++) {
266 		/* XXX - shouldn't be M_VNODE */
267 		free(*(vopvdpp[i]->opv_desc_vector_p), M_VNODE);
268 		*(vopvdpp[i]->opv_desc_vector_p) = NULL;
269 	}
270 }
271 
272 #ifdef DEBUG
273 static void
274 vfs_op_check(void)
275 {
276 	int i;
277 
278 	DODEBUG(printf("Vnode_interface_init.\n"));
279 
280 	/*
281 	 * Check offset of each op.
282 	 */
283 	for (i = 0; vfs_op_descs[i]; i++) {
284 		if (vfs_op_descs[i]->vdesc_offset != i)
285 			panic("vfs_op_check: vfs_op_desc[] offset mismatch");
286 	}
287 
288 	if (i != VNODE_OPS_COUNT) {
289 		panic("vfs_op_check: vnode ops count mismatch (%d != %d)",
290 			i, VNODE_OPS_COUNT);
291 	}
292 
293 	DODEBUG(printf ("vfs_opv_numops=%d\n", VNODE_OPS_COUNT));
294 }
295 #endif /* DEBUG */
296 
297 /*
298  * Initialize the vnode structures and initialize each file system type.
299  */
300 void
301 vfsinit(void)
302 {
303 
304 	/*
305 	 * Initialize the namei pathname buffer pool and cache.
306 	 */
307 	pnbuf_cache = pool_cache_init(MAXPATHLEN, 0, 0, 0, "pnbufpl",
308 	    NULL, IPL_NONE, NULL, NULL, NULL);
309 	KASSERT(pnbuf_cache != NULL);
310 
311 	/*
312 	 * Initialize the vnode table
313 	 */
314 	vntblinit();
315 
316 	/*
317 	 * Initialize the vnode name cache
318 	 */
319 	nchinit();
320 
321 #ifdef DEBUG
322 	/*
323 	 * Check the list of vnode operations.
324 	 */
325 	vfs_op_check();
326 #endif
327 
328 	/*
329 	 * Initialize the special vnode operations.
330 	 */
331 	vfs_opv_init(vfs_special_vnodeopv_descs);
332 
333 	/*
334 	 * Initialise VFS hooks.
335 	 */
336 	vfs_hooks_init();
337 
338 	/*
339 	 * Establish each file system which was statically
340 	 * included in the kernel.
341 	 */
342 	module_init_class(MODULE_CLASS_VFS);
343 }
344 
345 /*
346  * Drop a reference to a file system type.
347  */
348 void
349 vfs_delref(struct vfsops *vfs)
350 {
351 
352 	mutex_enter(&vfs_list_lock);
353 	vfs->vfs_refcount--;
354 	mutex_exit(&vfs_list_lock);
355 }
356 
357 /*
358  * Establish a file system and initialize it.
359  */
360 int
361 vfs_attach(struct vfsops *vfs)
362 {
363 	struct vfsops *v;
364 	int error = 0;
365 
366 	mutex_enter(&vfs_list_lock);
367 
368 	/*
369 	 * Make sure this file system doesn't already exist.
370 	 */
371 	LIST_FOREACH(v, &vfs_list, vfs_list) {
372 		if (strcmp(vfs->vfs_name, v->vfs_name) == 0) {
373 			error = EEXIST;
374 			goto out;
375 		}
376 	}
377 
378 	/*
379 	 * Initialize the vnode operations for this file system.
380 	 */
381 	vfs_opv_init(vfs->vfs_opv_descs);
382 
383 	/*
384 	 * Now initialize the file system itself.
385 	 */
386 	(*vfs->vfs_init)();
387 
388 	/*
389 	 * ...and link it into the kernel's list.
390 	 */
391 	LIST_INSERT_HEAD(&vfs_list, vfs, vfs_list);
392 
393 	/*
394 	 * Sanity: make sure the reference count is 0.
395 	 */
396 	vfs->vfs_refcount = 0;
397  out:
398 	mutex_exit(&vfs_list_lock);
399 	return (error);
400 }
401 
402 /*
403  * Remove a file system from the kernel.
404  */
405 int
406 vfs_detach(struct vfsops *vfs)
407 {
408 	struct vfsops *v;
409 	int error = 0;
410 
411 	mutex_enter(&vfs_list_lock);
412 
413 	/*
414 	 * Make sure no one is using the filesystem.
415 	 */
416 	if (vfs->vfs_refcount != 0) {
417 		error = EBUSY;
418 		goto out;
419 	}
420 
421 	/*
422 	 * ...and remove it from the kernel's list.
423 	 */
424 	LIST_FOREACH(v, &vfs_list, vfs_list) {
425 		if (v == vfs) {
426 			LIST_REMOVE(v, vfs_list);
427 			break;
428 		}
429 	}
430 
431 	if (v == NULL) {
432 		error = ESRCH;
433 		goto out;
434 	}
435 
436 	/*
437 	 * Now run the file system-specific cleanups.
438 	 */
439 	(*vfs->vfs_done)();
440 
441 	/*
442 	 * Free the vnode operations vector.
443 	 */
444 	vfs_opv_free(vfs->vfs_opv_descs);
445  out:
446  	mutex_exit(&vfs_list_lock);
447 	return (error);
448 }
449 
450 void
451 vfs_reinit(void)
452 {
453 	struct vfsops *vfs;
454 
455 	mutex_enter(&vfs_list_lock);
456 	LIST_FOREACH(vfs, &vfs_list, vfs_list) {
457 		if (vfs->vfs_reinit) {
458 			vfs->vfs_refcount++;
459 			mutex_exit(&vfs_list_lock);
460 			(*vfs->vfs_reinit)();
461 			mutex_enter(&vfs_list_lock);
462 			vfs->vfs_refcount--;
463 		}
464 	}
465 	mutex_exit(&vfs_list_lock);
466 }
467