xref: /netbsd-src/sys/kern/vfs_init.c (revision 96230fab84e26a6435963032070e916a951a8b2e)
1 /*	$NetBSD: vfs_init.c,v 1.41 2008/09/27 13:01:07 reinoud 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 1.41 2008/09/27 13:01:07 reinoud 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 #include <sys/dirhash.h>
85 
86 /*
87  * Sigh, such primitive tools are these...
88  */
89 #if 0
90 #define DODEBUG(A) A
91 #else
92 #define DODEBUG(A)
93 #endif
94 
95 /*
96  * The global list of vnode operations.
97  */
98 extern const struct vnodeop_desc * const vfs_op_descs[];
99 
100 /*
101  * These vnodeopv_descs are listed here because they are not
102  * associated with any particular file system, and thus cannot
103  * be initialized by vfs_attach().
104  */
105 extern const struct vnodeopv_desc dead_vnodeop_opv_desc;
106 extern const struct vnodeopv_desc fifo_vnodeop_opv_desc;
107 extern const struct vnodeopv_desc spec_vnodeop_opv_desc;
108 extern const struct vnodeopv_desc sync_vnodeop_opv_desc;
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 	&sync_vnodeop_opv_desc,
115 	NULL,
116 };
117 
118 struct vfs_list_head vfs_list =			/* vfs list */
119     LIST_HEAD_INITIALIZER(vfs_list);
120 
121 /*
122  * This code doesn't work if the defn is **vnodop_defns with cc.
123  * The problem is because of the compiler sometimes putting in an
124  * extra level of indirection for arrays.  It's an interesting
125  * "feature" of C.
126  */
127 typedef int (*PFI)(void *);
128 
129 /*
130  * A miscellaneous routine.
131  * A generic "default" routine that just returns an error.
132  */
133 /*ARGSUSED*/
134 int
135 vn_default_error(void *v)
136 {
137 
138 	return (EOPNOTSUPP);
139 }
140 
141 /*
142  * vfs_init.c
143  *
144  * Allocate and fill in operations vectors.
145  *
146  * An undocumented feature of this approach to defining operations is that
147  * there can be multiple entries in vfs_opv_descs for the same operations
148  * vector. This allows third parties to extend the set of operations
149  * supported by another layer in a binary compatibile way. For example,
150  * assume that NFS needed to be modified to support Ficus. NFS has an entry
151  * (probably nfs_vnopdeop_decls) declaring all the operations NFS supports by
152  * default. Ficus could add another entry (ficus_nfs_vnodeop_decl_entensions)
153  * listing those new operations Ficus adds to NFS, all without modifying the
154  * NFS code. (Of couse, the OTW NFS protocol still needs to be munged, but
155  * that is a(whole)nother story.) This is a feature.
156  */
157 
158 /*
159  * Init the vector, if it needs it.
160  * Also handle backwards compatibility.
161  */
162 static void
163 vfs_opv_init_explicit(const struct vnodeopv_desc *vfs_opv_desc)
164 {
165 	int (**opv_desc_vector)(void *);
166 	const struct vnodeopv_entry_desc *opve_descp;
167 
168 	opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p);
169 
170 	for (opve_descp = vfs_opv_desc->opv_desc_ops;
171 	     opve_descp->opve_op;
172 	     opve_descp++) {
173 		/*
174 		 * Sanity check:  is this operation listed
175 		 * in the list of operations?  We check this
176 		 * by seeing if its offset is zero.  Since
177 		 * the default routine should always be listed
178 		 * first, it should be the only one with a zero
179 		 * offset.  Any other operation with a zero
180 		 * offset is probably not listed in
181 		 * vfs_op_descs, and so is probably an error.
182 		 *
183 		 * A panic here means the layer programmer
184 		 * has committed the all-too common bug
185 		 * of adding a new operation to the layer's
186 		 * list of vnode operations but
187 		 * not adding the operation to the system-wide
188 		 * list of supported operations.
189 		 */
190 		if (opve_descp->opve_op->vdesc_offset == 0 &&
191 		    opve_descp->opve_op->vdesc_offset != VOFFSET(vop_default)) {
192 			printf("operation %s not listed in %s.\n",
193 			    opve_descp->opve_op->vdesc_name, "vfs_op_descs");
194 			panic ("vfs_opv_init: bad operation");
195 		}
196 
197 		/*
198 		 * Fill in this entry.
199 		 */
200 		opv_desc_vector[opve_descp->opve_op->vdesc_offset] =
201 		    opve_descp->opve_impl;
202 	}
203 }
204 
205 static void
206 vfs_opv_init_default(const struct vnodeopv_desc *vfs_opv_desc)
207 {
208 	int j;
209 	int (**opv_desc_vector)(void *);
210 
211 	opv_desc_vector = *(vfs_opv_desc->opv_desc_vector_p);
212 
213 	/*
214 	 * Force every operations vector to have a default routine.
215 	 */
216 	if (opv_desc_vector[VOFFSET(vop_default)] == NULL)
217 		panic("vfs_opv_init: operation vector without default routine.");
218 
219 	for (j = 0; j < VNODE_OPS_COUNT; j++)
220 		if (opv_desc_vector[j] == NULL)
221 			opv_desc_vector[j] =
222 			    opv_desc_vector[VOFFSET(vop_default)];
223 }
224 
225 void
226 vfs_opv_init(const struct vnodeopv_desc * const *vopvdpp)
227 {
228 	int (**opv_desc_vector)(void *);
229 	int i;
230 
231 	/*
232 	 * Allocate the vectors.
233 	 */
234 	for (i = 0; vopvdpp[i] != NULL; i++) {
235 		/* XXX - shouldn't be M_VNODE */
236 		opv_desc_vector =
237 		    malloc(VNODE_OPS_COUNT * sizeof(PFI), M_VNODE, M_WAITOK);
238 		memset(opv_desc_vector, 0, VNODE_OPS_COUNT * sizeof(PFI));
239 		*(vopvdpp[i]->opv_desc_vector_p) = opv_desc_vector;
240 		DODEBUG(printf("vector at %p allocated\n",
241 		    opv_desc_vector_p));
242 	}
243 
244 	/*
245 	 * ...and fill them in.
246 	 */
247 	for (i = 0; vopvdpp[i] != NULL; i++)
248 		vfs_opv_init_explicit(vopvdpp[i]);
249 
250 	/*
251 	 * Finally, go back and replace unfilled routines
252 	 * with their default.
253 	 */
254 	for (i = 0; vopvdpp[i] != NULL; i++)
255 		vfs_opv_init_default(vopvdpp[i]);
256 }
257 
258 void
259 vfs_opv_free(const struct vnodeopv_desc * const *vopvdpp)
260 {
261 	int i;
262 
263 	/*
264 	 * Free the vectors allocated in vfs_opv_init().
265 	 */
266 	for (i = 0; vopvdpp[i] != NULL; i++) {
267 		/* XXX - shouldn't be M_VNODE */
268 		free(*(vopvdpp[i]->opv_desc_vector_p), M_VNODE);
269 		*(vopvdpp[i]->opv_desc_vector_p) = NULL;
270 	}
271 }
272 
273 #ifdef DEBUG
274 static void
275 vfs_op_check(void)
276 {
277 	int i;
278 
279 	DODEBUG(printf("Vnode_interface_init.\n"));
280 
281 	/*
282 	 * Check offset of each op.
283 	 */
284 	for (i = 0; vfs_op_descs[i]; i++) {
285 		if (vfs_op_descs[i]->vdesc_offset != i)
286 			panic("vfs_op_check: vfs_op_desc[] offset mismatch");
287 	}
288 
289 	if (i != VNODE_OPS_COUNT) {
290 		panic("vfs_op_check: vnode ops count mismatch (%d != %d)",
291 			i, VNODE_OPS_COUNT);
292 	}
293 
294 	DODEBUG(printf ("vfs_opv_numops=%d\n", VNODE_OPS_COUNT));
295 }
296 #endif /* DEBUG */
297 
298 /*
299  * Initialize the vnode structures and initialize each file system type.
300  */
301 void
302 vfsinit(void)
303 {
304 
305 	/*
306 	 * Initialize the namei pathname buffer pool and cache.
307 	 */
308 	pnbuf_cache = pool_cache_init(MAXPATHLEN, 0, 0, 0, "pnbufpl",
309 	    NULL, IPL_NONE, NULL, NULL, NULL);
310 	KASSERT(pnbuf_cache != NULL);
311 
312 	/*
313 	 * Initialize the vnode table
314 	 */
315 	vntblinit();
316 
317 	/*
318 	 * Initialize the vnode name cache
319 	 */
320 	nchinit();
321 
322 #ifdef DEBUG
323 	/*
324 	 * Check the list of vnode operations.
325 	 */
326 	vfs_op_check();
327 #endif
328 
329 	/*
330 	 * Initialize the special vnode operations.
331 	 */
332 	vfs_opv_init(vfs_special_vnodeopv_descs);
333 
334 	/*
335 	 * Initialise generic dirhash.
336 	 */
337 	dirhash_init();
338 
339 	/*
340 	 * Initialise VFS hooks.
341 	 */
342 	vfs_hooks_init();
343 
344 	/*
345 	 * Establish each file system which was statically
346 	 * included in the kernel.
347 	 */
348 	module_init_class(MODULE_CLASS_VFS);
349 }
350 
351 /*
352  * Drop a reference to a file system type.
353  */
354 void
355 vfs_delref(struct vfsops *vfs)
356 {
357 
358 	mutex_enter(&vfs_list_lock);
359 	vfs->vfs_refcount--;
360 	mutex_exit(&vfs_list_lock);
361 }
362 
363 /*
364  * Establish a file system and initialize it.
365  */
366 int
367 vfs_attach(struct vfsops *vfs)
368 {
369 	struct vfsops *v;
370 	int error = 0;
371 
372 	mutex_enter(&vfs_list_lock);
373 
374 	/*
375 	 * Make sure this file system doesn't already exist.
376 	 */
377 	LIST_FOREACH(v, &vfs_list, vfs_list) {
378 		if (strcmp(vfs->vfs_name, v->vfs_name) == 0) {
379 			error = EEXIST;
380 			goto out;
381 		}
382 	}
383 
384 	/*
385 	 * Initialize the vnode operations for this file system.
386 	 */
387 	vfs_opv_init(vfs->vfs_opv_descs);
388 
389 	/*
390 	 * Now initialize the file system itself.
391 	 */
392 	(*vfs->vfs_init)();
393 
394 	/*
395 	 * ...and link it into the kernel's list.
396 	 */
397 	LIST_INSERT_HEAD(&vfs_list, vfs, vfs_list);
398 
399 	/*
400 	 * Sanity: make sure the reference count is 0.
401 	 */
402 	vfs->vfs_refcount = 0;
403  out:
404 	mutex_exit(&vfs_list_lock);
405 	return (error);
406 }
407 
408 /*
409  * Remove a file system from the kernel.
410  */
411 int
412 vfs_detach(struct vfsops *vfs)
413 {
414 	struct vfsops *v;
415 	int error = 0;
416 
417 	mutex_enter(&vfs_list_lock);
418 
419 	/*
420 	 * Make sure no one is using the filesystem.
421 	 */
422 	if (vfs->vfs_refcount != 0) {
423 		error = EBUSY;
424 		goto out;
425 	}
426 
427 	/*
428 	 * ...and remove it from the kernel's list.
429 	 */
430 	LIST_FOREACH(v, &vfs_list, vfs_list) {
431 		if (v == vfs) {
432 			LIST_REMOVE(v, vfs_list);
433 			break;
434 		}
435 	}
436 
437 	if (v == NULL) {
438 		error = ESRCH;
439 		goto out;
440 	}
441 
442 	/*
443 	 * Now run the file system-specific cleanups.
444 	 */
445 	(*vfs->vfs_done)();
446 
447 	/*
448 	 * Free the vnode operations vector.
449 	 */
450 	vfs_opv_free(vfs->vfs_opv_descs);
451  out:
452  	mutex_exit(&vfs_list_lock);
453 	return (error);
454 }
455 
456 void
457 vfs_reinit(void)
458 {
459 	struct vfsops *vfs;
460 
461 	mutex_enter(&vfs_list_lock);
462 	LIST_FOREACH(vfs, &vfs_list, vfs_list) {
463 		if (vfs->vfs_reinit) {
464 			vfs->vfs_refcount++;
465 			mutex_exit(&vfs_list_lock);
466 			(*vfs->vfs_reinit)();
467 			mutex_enter(&vfs_list_lock);
468 			vfs->vfs_refcount--;
469 		}
470 	}
471 	mutex_exit(&vfs_list_lock);
472 }
473