xref: /dragonfly/sys/kern/vfs_init.c (revision b40e316c)
1 /*
2  * Copyright (c) 2003,2004 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Matthew Dillon <dillon@backplane.com>
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in
15  *    the documentation and/or other materials provided with the
16  *    distribution.
17  * 3. Neither the name of The DragonFly Project nor the names of its
18  *    contributors may be used to endorse or promote products derived
19  *    from this software without specific, prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *
35  * Copyright (c) 1989, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * This code is derived from software contributed
39  * to Berkeley by John Heidemann of the UCLA Ficus project.
40  *
41  * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project
42  *
43  * Redistribution and use in source and binary forms, with or without
44  * modification, are permitted provided that the following conditions
45  * are met:
46  * 1. Redistributions of source code must retain the above copyright
47  *    notice, this list of conditions and the following disclaimer.
48  * 2. Redistributions in binary form must reproduce the above copyright
49  *    notice, this list of conditions and the following disclaimer in the
50  *    documentation and/or other materials provided with the distribution.
51  * 3. All advertising materials mentioning features or use of this software
52  *    must display the following acknowledgement:
53  *	This product includes software developed by the University of
54  *	California, Berkeley and its contributors.
55  * 4. Neither the name of the University nor the names of its contributors
56  *    may be used to endorse or promote products derived from this software
57  *    without specific prior written permission.
58  *
59  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69  * SUCH DAMAGE.
70  *
71  *	@(#)vfs_init.c	8.3 (Berkeley) 1/4/94
72  * $FreeBSD: src/sys/kern/vfs_init.c,v 1.59 2002/04/30 18:44:32 dillon Exp $
73  * $DragonFly: src/sys/kern/vfs_init.c,v 1.8 2004/12/17 00:18:07 dillon Exp $
74  */
75 /*
76  * Manage vnode VOP operations vectors
77  */
78 #include <sys/param.h>
79 #include <sys/systm.h>
80 #include <sys/kernel.h>
81 #include <sys/mount.h>
82 #include <sys/sysctl.h>
83 #include <sys/vnode.h>
84 #include <sys/malloc.h>
85 #include <vm/vm_zone.h>
86 
87 static MALLOC_DEFINE(M_VNODEOP, "vnodeops", "vnode operations vectors");
88 
89 /*
90  * Zone for namei
91  */
92 struct vm_zone *namei_zone;
93 
94 /*
95  * vfs_init() will set maxvfsconf
96  * to the highest defined type number.
97  */
98 int maxvfsconf;
99 struct vfsconf *vfsconf;
100 
101 static TAILQ_HEAD(, vnodeopv_node) vnodeopv_list;
102 static void vfs_recalc_vnodeops(void);
103 
104 /*
105  * Add a vnode operations (vnops) vector to the global list.
106  */
107 void
108 vfs_add_vnodeops_sysinit(const void *data)
109 {
110 	const struct vnodeopv_desc *vdesc = data;
111 
112 	vfs_add_vnodeops(NULL, vdesc->opv_desc_vector, vdesc->opv_desc_ops);
113 }
114 
115 /*
116  * Unlink previously added vnode operations vector.
117  */
118 void
119 vfs_rm_vnodeops_sysinit(const void *data)
120 {
121 	const struct vnodeopv_desc *vdesc = data;
122 
123 	vfs_rm_vnodeops(vdesc->opv_desc_vector);
124 }
125 
126 void
127 vfs_add_vnodeops(struct mount *mp, struct vop_ops **vops_pp,
128 		struct vnodeopv_entry_desc *descs)
129 {
130 	struct vnodeopv_node *node;
131 	struct vop_ops *ops;
132 
133 	node = malloc(sizeof(*node), M_VNODEOP, M_ZERO|M_WAITOK);
134 	if ((ops = *vops_pp) == NULL) {
135 		ops = malloc(sizeof(struct vop_ops),
136 				M_VNODEOP, M_ZERO|M_WAITOK);
137 		*vops_pp = ops;
138 	}
139 	node->ops = ops;
140 	node->descs = descs;
141 	ops->vv_mount = mp;
142 	++ops->vv_refs;
143 	TAILQ_INSERT_TAIL(&vnodeopv_list, node, entry);
144 	vfs_recalc_vnodeops();
145 	if (mp) {
146 		if (mp->mnt_vn_coherency_ops)
147 			mp->mnt_vn_use_ops = mp->mnt_vn_coherency_ops;
148 		else if (mp->mnt_vn_journal_ops)
149 			mp->mnt_vn_use_ops = mp->mnt_vn_journal_ops;
150 		else
151 			mp->mnt_vn_use_ops = mp->mnt_vn_norm_ops;
152 	}
153 }
154 
155 void
156 vfs_rm_vnodeops(struct vop_ops **vops_pp)
157 {
158 	struct vop_ops *ops = *vops_pp;
159 	struct vnodeopv_node *node;
160 	struct mount *mp;
161 
162 	if (ops == NULL)
163 		return;
164 
165 	TAILQ_FOREACH(node, &vnodeopv_list, entry) {
166 		if (node->ops == ops)
167 			break;
168 	}
169 	if (node == NULL) {
170 		printf("vfs_rm_vnodeops: unable to find ops: %p\n", ops);
171 		return;
172 	}
173 	TAILQ_REMOVE(&vnodeopv_list, node, entry);
174 	free(node, M_VNODEOP);
175 	KKASSERT(ops != NULL && ops->vv_refs > 0);
176 	if (--ops->vv_refs == 0) {
177 		*vops_pp = NULL;
178 		if ((mp = ops->vv_mount) != NULL) {
179 			if (mp->mnt_vn_coherency_ops)
180 				mp->mnt_vn_use_ops = mp->mnt_vn_coherency_ops;
181 			else if (mp->mnt_vn_journal_ops)
182 				mp->mnt_vn_use_ops = mp->mnt_vn_journal_ops;
183 			else
184 				mp->mnt_vn_use_ops = mp->mnt_vn_norm_ops;
185 		}
186 		free(ops, M_VNODEOP);
187 	}
188 	vfs_recalc_vnodeops();
189 }
190 
191 /*
192  * Recalculate VFS operations vectors
193  */
194 static void
195 vfs_recalc_vnodeops(void)
196 {
197 	struct vnodeopv_node *node;
198 	struct vnodeopv_entry_desc *desc;
199 	struct vop_ops *ops;
200 	struct vop_ops *vnew;
201 	int off;
202 
203 	/*
204 	 * Because vop_ops may be active we can't just blow them away, we
205 	 * have to generate new vop_ops and then copy them into the running
206 	 * vop_ops.  Any missing entries will be assigned to the default
207 	 * entry.  If the default entry itself is missing it will be assigned
208 	 * to vop_eopnotsupp.
209 	 */
210 	TAILQ_FOREACH(node, &vnodeopv_list, entry) {
211 		ops = node->ops;
212 		if ((vnew = ops->vv_new) == NULL) {
213 			vnew = malloc(sizeof(struct vop_ops),
214 					M_VNODEOP, M_ZERO|M_WAITOK);
215 			ops->vv_new = vnew;
216 			vnew->vop_default = vop_eopnotsupp;
217 		}
218 		for (desc = node->descs; desc->opve_op; ++desc) {
219 			off = desc->opve_op->vdesc_offset;
220 			*(void **)((char *)vnew + off) = desc->opve_func;
221 		}
222 		for (off = __offsetof(struct vop_ops, vop_ops_first_field);
223 		     off <= __offsetof(struct vop_ops, vop_ops_last_field);
224 		     off += sizeof(void **)
225 		) {
226 			if (*(void **)((char *)vnew + off) == NULL)
227 			    *(void **)((char *)vnew + off) = vnew->vop_default;
228 		}
229 	}
230 
231 	/*
232 	 * Copy the temporary ops into the running configuration and then
233 	 * delete them.
234 	 */
235 	TAILQ_FOREACH(node, &vnodeopv_list, entry) {
236 		ops = node->ops;
237 		if ((vnew = ops->vv_new) == NULL)
238 			continue;
239 		for (off = __offsetof(struct vop_ops, vop_ops_first_field);
240 		     off <= __offsetof(struct vop_ops, vop_ops_last_field);
241 		     off += sizeof(void **)
242 		) {
243 			*(void **)((char *)ops + off) =
244 				*(void **)((char *)vnew + off);
245 		}
246 		ops->vv_new = NULL;
247 		free(vnew, M_VNODEOP);
248 	}
249 }
250 
251 /*
252  * Routines having to do with the management of the vnode table.
253  */
254 struct vattr va_null;
255 
256 /*
257  * Initialize the vnode structures and initialize each file system type.
258  */
259 /* ARGSUSED*/
260 static void
261 vfsinit(void *dummy)
262 {
263 	TAILQ_INIT(&vnodeopv_list);
264 	namei_zone = zinit("NAMEI", MAXPATHLEN, 0, 0, 2);
265 
266 	/*
267 	 * Initialize the vnode table
268 	 */
269 	vfs_subr_init();
270 	vfs_mount_init();
271 	vfs_lock_init();
272 	vfs_sync_init();
273 	/*
274 	 * Initialize the vnode name cache
275 	 */
276 	nchinit();
277 	/*
278 	 * Initialize each file system type.
279 	 * Vfs type numbers must be distinct from VFS_GENERIC (and VFS_VFSCONF).
280 	 */
281 	vattr_null(&va_null);
282 	maxvfsconf = VFS_GENERIC + 1;
283 }
284 SYSINIT(vfs, SI_SUB_VFS, SI_ORDER_FIRST, vfsinit, NULL)
285 
286 /*
287  * Register a VFS.
288  *
289  * After doing general initialisation, this function will
290  * call the filesystem specific initialisation vector op,
291  * i.e. vfsops->vfs_init().
292  */
293 int
294 vfs_register(struct vfsconf *vfc)
295 {
296 	struct sysctl_oid *oidp;
297 	struct vfsconf *vfsp;
298 
299 	vfsp = NULL;
300 	if (vfsconf)
301 		for (vfsp = vfsconf; vfsp->vfc_next; vfsp = vfsp->vfc_next)
302 			if (strcmp(vfc->vfc_name, vfsp->vfc_name) == 0)
303 				return EEXIST;
304 
305 	vfc->vfc_typenum = maxvfsconf++;
306 	if (vfsp)
307 		vfsp->vfc_next = vfc;
308 	else
309 		vfsconf = vfc;
310 	vfc->vfc_next = NULL;
311 
312 	/*
313 	 * If this filesystem has a sysctl node under vfs
314 	 * (i.e. vfs.xxfs), then change the oid number of that node to
315 	 * match the filesystem's type number.  This allows user code
316 	 * which uses the type number to read sysctl variables defined
317 	 * by the filesystem to continue working. Since the oids are
318 	 * in a sorted list, we need to make sure the order is
319 	 * preserved by re-registering the oid after modifying its
320 	 * number.
321 	 */
322 	SLIST_FOREACH(oidp, &sysctl__vfs_children, oid_link)
323 		if (strcmp(oidp->oid_name, vfc->vfc_name) == 0) {
324 			sysctl_unregister_oid(oidp);
325 			oidp->oid_number = vfc->vfc_typenum;
326 			sysctl_register_oid(oidp);
327 		}
328 
329 	/*
330 	 * Call init function for this VFS...
331 	 */
332 	(*(vfc->vfc_vfsops->vfs_init))(vfc);
333 
334 	return 0;
335 }
336 
337 
338 /*
339  * Remove previously registered VFS.
340  *
341  * After doing general de-registration like removing sysctl
342  * nodes etc, it will call the filesystem specific vector
343  * op, i.e. vfsops->vfs_uninit().
344  *
345  */
346 int
347 vfs_unregister(struct vfsconf *vfc)
348 {
349 	struct vfsconf *vfsp, *prev_vfsp;
350 	int error, i, maxtypenum;
351 
352 	i = vfc->vfc_typenum;
353 
354 	prev_vfsp = NULL;
355 	for (vfsp = vfsconf; vfsp;
356 			prev_vfsp = vfsp, vfsp = vfsp->vfc_next) {
357 		if (!strcmp(vfc->vfc_name, vfsp->vfc_name))
358 			break;
359 	}
360 	if (vfsp == NULL)
361 		return EINVAL;
362 	if (vfsp->vfc_refcount)
363 		return EBUSY;
364 	if (vfc->vfc_vfsops->vfs_uninit != NULL) {
365 		error = (*vfc->vfc_vfsops->vfs_uninit)(vfsp);
366 		if (error)
367 			return (error);
368 	}
369 	if (prev_vfsp)
370 		prev_vfsp->vfc_next = vfsp->vfc_next;
371 	else
372 		vfsconf = vfsp->vfc_next;
373 	maxtypenum = VFS_GENERIC;
374 	for (vfsp = vfsconf; vfsp != NULL; vfsp = vfsp->vfc_next)
375 		if (maxtypenum < vfsp->vfc_typenum)
376 			maxtypenum = vfsp->vfc_typenum;
377 	maxvfsconf = maxtypenum + 1;
378 	return 0;
379 }
380 
381 int
382 vfs_modevent(module_t mod, int type, void *data)
383 {
384 	struct vfsconf *vfc;
385 	int error = 0;
386 
387 	vfc = (struct vfsconf *)data;
388 
389 	switch (type) {
390 	case MOD_LOAD:
391 		if (vfc)
392 			error = vfs_register(vfc);
393 		break;
394 
395 	case MOD_UNLOAD:
396 		if (vfc)
397 			error = vfs_unregister(vfc);
398 		break;
399 	default:	/* including MOD_SHUTDOWN */
400 		break;
401 	}
402 	return (error);
403 }
404