vfs_default.c revision 50477
1/*
2 * Copyright (c) 1989, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed
6 * to Berkeley by John Heidemann of the UCLA Ficus project.
7 *
8 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 *    must display the following acknowledgement:
20 *	This product includes software developed by the University of
21 *	California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 *    may be used to endorse or promote products derived from this software
24 *    without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 *
39 * $FreeBSD: head/sys/kern/vfs_default.c 50477 1999-08-28 01:08:13Z peter $
40 */
41
42#include <sys/param.h>
43#include <sys/systm.h>
44#include <sys/buf.h>
45#include <sys/kernel.h>
46#include <sys/lock.h>
47#include <sys/malloc.h>
48#include <sys/unistd.h>
49#include <sys/vnode.h>
50#include <sys/poll.h>
51
52static int vop_nostrategy __P((struct vop_strategy_args *));
53
54/*
55 * This vnode table stores what we want to do if the filesystem doesn't
56 * implement a particular VOP.
57 *
58 * If there is no specific entry here, we will return EOPNOTSUPP.
59 *
60 */
61
62vop_t **default_vnodeop_p;
63static struct vnodeopv_entry_desc default_vnodeop_entries[] = {
64	{ &vop_default_desc,		(vop_t *) vop_eopnotsupp },
65	{ &vop_abortop_desc,		(vop_t *) vop_null },
66	{ &vop_advlock_desc,		(vop_t *) vop_einval },
67	{ &vop_bwrite_desc,		(vop_t *) vop_stdbwrite },
68	{ &vop_close_desc,		(vop_t *) vop_null },
69	{ &vop_fsync_desc,		(vop_t *) vop_null },
70	{ &vop_ioctl_desc,		(vop_t *) vop_enotty },
71	{ &vop_islocked_desc,		(vop_t *) vop_noislocked },
72	{ &vop_lease_desc,		(vop_t *) vop_null },
73	{ &vop_lock_desc,		(vop_t *) vop_nolock },
74	{ &vop_mmap_desc,		(vop_t *) vop_einval },
75	{ &vop_open_desc,		(vop_t *) vop_null },
76	{ &vop_pathconf_desc,		(vop_t *) vop_einval },
77	{ &vop_poll_desc,		(vop_t *) vop_nopoll },
78	{ &vop_readlink_desc,		(vop_t *) vop_einval },
79	{ &vop_reallocblks_desc,	(vop_t *) vop_eopnotsupp },
80	{ &vop_revoke_desc,		(vop_t *) vop_revoke },
81	{ &vop_strategy_desc,		(vop_t *) vop_nostrategy },
82	{ &vop_unlock_desc,		(vop_t *) vop_nounlock },
83	{ NULL, NULL }
84};
85
86static struct vnodeopv_desc default_vnodeop_opv_desc =
87        { &default_vnodeop_p, default_vnodeop_entries };
88
89VNODEOP_SET(default_vnodeop_opv_desc);
90
91int
92vop_eopnotsupp(struct vop_generic_args *ap)
93{
94	/*
95	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
96	*/
97
98	return (EOPNOTSUPP);
99}
100
101int
102vop_ebadf(struct vop_generic_args *ap)
103{
104
105	return (EBADF);
106}
107
108int
109vop_enotty(struct vop_generic_args *ap)
110{
111
112	return (ENOTTY);
113}
114
115int
116vop_einval(struct vop_generic_args *ap)
117{
118
119	return (EINVAL);
120}
121
122int
123vop_null(struct vop_generic_args *ap)
124{
125
126	return (0);
127}
128
129int
130vop_defaultop(struct vop_generic_args *ap)
131{
132
133	return (VOCALL(default_vnodeop_p, ap->a_desc->vdesc_offset, ap));
134}
135
136int
137vop_panic(struct vop_generic_args *ap)
138{
139
140	panic("illegal vnode op called");
141}
142
143/*
144 *	vop_nostrategy:
145 *
146 *	Strategy routine for VFS devices that have none.
147 *
148 *	B_ERROR and B_INVAL must be cleared prior to calling any strategy
149 *	routine.  Typically this is done for a B_READ strategy call.  Typically
150 *	B_INVAL is assumed to already be clear prior to a write and should not
151 *	be cleared manually unless you just made the buffer invalid.  B_ERROR
152 *	should be cleared either way.
153 */
154
155static int
156vop_nostrategy (struct vop_strategy_args *ap)
157{
158	printf("No strategy for buffer at %p\n", ap->a_bp);
159	vprint("", ap->a_vp);
160	vprint("", ap->a_bp->b_vp);
161	ap->a_bp->b_flags |= B_ERROR;
162	ap->a_bp->b_error = EOPNOTSUPP;
163	biodone(ap->a_bp);
164	return (EOPNOTSUPP);
165}
166
167int
168vop_stdpathconf(ap)
169	struct vop_pathconf_args /* {
170	struct vnode *a_vp;
171	int a_name;
172	int *a_retval;
173	} */ *ap;
174{
175
176	switch (ap->a_name) {
177		case _PC_LINK_MAX:
178			*ap->a_retval = LINK_MAX;
179			return (0);
180		case _PC_MAX_CANON:
181			*ap->a_retval = MAX_CANON;
182			return (0);
183		case _PC_MAX_INPUT:
184			*ap->a_retval = MAX_INPUT;
185			return (0);
186		case _PC_PIPE_BUF:
187			*ap->a_retval = PIPE_BUF;
188			return (0);
189		case _PC_CHOWN_RESTRICTED:
190			*ap->a_retval = 1;
191			return (0);
192		case _PC_VDISABLE:
193			*ap->a_retval = _POSIX_VDISABLE;
194			return (0);
195		default:
196			return (EINVAL);
197	}
198	/* NOTREACHED */
199}
200
201/*
202 * Standard lock, unlock and islocked functions.
203 *
204 * These depend on the lock structure being the first element in the
205 * inode, ie: vp->v_data points to the the lock!
206 */
207int
208vop_stdlock(ap)
209	struct vop_lock_args /* {
210		struct vnode *a_vp;
211		int a_flags;
212		struct proc *a_p;
213	} */ *ap;
214{
215	struct lock *l;
216
217	if ((l = (struct lock *)ap->a_vp->v_data) == NULL) {
218		if (ap->a_flags & LK_INTERLOCK)
219			simple_unlock(&ap->a_vp->v_interlock);
220		return 0;
221	}
222
223#ifndef	DEBUG_LOCKS
224	return (lockmgr(l, ap->a_flags, &ap->a_vp->v_interlock, ap->a_p));
225#else
226	return (debuglockmgr(l, ap->a_flags, &ap->a_vp->v_interlock, ap->a_p,
227	    "vop_stdlock", ap->a_vp->filename, ap->a_vp->line));
228#endif
229}
230
231int
232vop_stdunlock(ap)
233	struct vop_unlock_args /* {
234		struct vnode *a_vp;
235		int a_flags;
236		struct proc *a_p;
237	} */ *ap;
238{
239	struct lock *l;
240
241	if ((l = (struct lock *)ap->a_vp->v_data) == NULL) {
242		if (ap->a_flags & LK_INTERLOCK)
243			simple_unlock(&ap->a_vp->v_interlock);
244		return 0;
245	}
246
247	return (lockmgr(l, ap->a_flags | LK_RELEASE, &ap->a_vp->v_interlock,
248	    ap->a_p));
249}
250
251int
252vop_stdislocked(ap)
253	struct vop_islocked_args /* {
254		struct vnode *a_vp;
255	} */ *ap;
256{
257	struct lock *l;
258
259	if ((l = (struct lock *)ap->a_vp->v_data) == NULL)
260		return 0;
261
262	return (lockstatus(l));
263}
264
265/*
266 * Return true for select/poll.
267 */
268int
269vop_nopoll(ap)
270	struct vop_poll_args /* {
271		struct vnode *a_vp;
272		int  a_events;
273		struct ucred *a_cred;
274		struct proc *a_p;
275	} */ *ap;
276{
277	/*
278	 * Return true for read/write.  If the user asked for something
279	 * special, return POLLNVAL, so that clients have a way of
280	 * determining reliably whether or not the extended
281	 * functionality is present without hard-coding knowledge
282	 * of specific filesystem implementations.
283	 */
284	if (ap->a_events & ~POLLSTANDARD)
285		return (POLLNVAL);
286
287	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
288}
289
290/*
291 * Implement poll for local filesystems that support it.
292 */
293int
294vop_stdpoll(ap)
295	struct vop_poll_args /* {
296		struct vnode *a_vp;
297		int  a_events;
298		struct ucred *a_cred;
299		struct proc *a_p;
300	} */ *ap;
301{
302	if ((ap->a_events & ~POLLSTANDARD) == 0)
303		return (ap->a_events & (POLLRDNORM|POLLWRNORM));
304	return (vn_pollrecord(ap->a_vp, ap->a_p, ap->a_events));
305}
306
307int
308vop_stdbwrite(ap)
309	struct vop_bwrite_args *ap;
310{
311	return (bwrite(ap->a_bp));
312}
313
314/*
315 * Stubs to use when there is no locking to be done on the underlying object.
316 * A minimal shared lock is necessary to ensure that the underlying object
317 * is not revoked while an operation is in progress. So, an active shared
318 * count is maintained in an auxillary vnode lock structure.
319 */
320int
321vop_sharedlock(ap)
322	struct vop_lock_args /* {
323		struct vnode *a_vp;
324		int a_flags;
325		struct proc *a_p;
326	} */ *ap;
327{
328	/*
329	 * This code cannot be used until all the non-locking filesystems
330	 * (notably NFS) are converted to properly lock and release nodes.
331	 * Also, certain vnode operations change the locking state within
332	 * the operation (create, mknod, remove, link, rename, mkdir, rmdir,
333	 * and symlink). Ideally these operations should not change the
334	 * lock state, but should be changed to let the caller of the
335	 * function unlock them. Otherwise all intermediate vnode layers
336	 * (such as union, umapfs, etc) must catch these functions to do
337	 * the necessary locking at their layer. Note that the inactive
338	 * and lookup operations also change their lock state, but this
339	 * cannot be avoided, so these two operations will always need
340	 * to be handled in intermediate layers.
341	 */
342	struct vnode *vp = ap->a_vp;
343	int vnflags, flags = ap->a_flags;
344
345	if (vp->v_vnlock == NULL) {
346		if ((flags & LK_TYPE_MASK) == LK_DRAIN)
347			return (0);
348		MALLOC(vp->v_vnlock, struct lock *, sizeof(struct lock),
349		    M_VNODE, M_WAITOK);
350		lockinit(vp->v_vnlock, PVFS, "vnlock", 0, LK_NOPAUSE);
351	}
352	switch (flags & LK_TYPE_MASK) {
353	case LK_DRAIN:
354		vnflags = LK_DRAIN;
355		break;
356	case LK_EXCLUSIVE:
357#ifdef DEBUG_VFS_LOCKS
358		/*
359		 * Normally, we use shared locks here, but that confuses
360		 * the locking assertions.
361		 */
362		vnflags = LK_EXCLUSIVE;
363		break;
364#endif
365	case LK_SHARED:
366		vnflags = LK_SHARED;
367		break;
368	case LK_UPGRADE:
369	case LK_EXCLUPGRADE:
370	case LK_DOWNGRADE:
371		return (0);
372	case LK_RELEASE:
373	default:
374		panic("vop_sharedlock: bad operation %d", flags & LK_TYPE_MASK);
375	}
376	if (flags & LK_INTERLOCK)
377		vnflags |= LK_INTERLOCK;
378#ifndef	DEBUG_LOCKS
379	return (lockmgr(vp->v_vnlock, vnflags, &vp->v_interlock, ap->a_p));
380#else
381	return (debuglockmgr(vp->v_vnlock, vnflags, &vp->v_interlock, ap->a_p,
382	    "vop_sharedlock", vp->filename, vp->line));
383#endif
384}
385
386/*
387 * Stubs to use when there is no locking to be done on the underlying object.
388 * A minimal shared lock is necessary to ensure that the underlying object
389 * is not revoked while an operation is in progress. So, an active shared
390 * count is maintained in an auxillary vnode lock structure.
391 */
392int
393vop_nolock(ap)
394	struct vop_lock_args /* {
395		struct vnode *a_vp;
396		int a_flags;
397		struct proc *a_p;
398	} */ *ap;
399{
400#ifdef notyet
401	/*
402	 * This code cannot be used until all the non-locking filesystems
403	 * (notably NFS) are converted to properly lock and release nodes.
404	 * Also, certain vnode operations change the locking state within
405	 * the operation (create, mknod, remove, link, rename, mkdir, rmdir,
406	 * and symlink). Ideally these operations should not change the
407	 * lock state, but should be changed to let the caller of the
408	 * function unlock them. Otherwise all intermediate vnode layers
409	 * (such as union, umapfs, etc) must catch these functions to do
410	 * the necessary locking at their layer. Note that the inactive
411	 * and lookup operations also change their lock state, but this
412	 * cannot be avoided, so these two operations will always need
413	 * to be handled in intermediate layers.
414	 */
415	struct vnode *vp = ap->a_vp;
416	int vnflags, flags = ap->a_flags;
417
418	if (vp->v_vnlock == NULL) {
419		if ((flags & LK_TYPE_MASK) == LK_DRAIN)
420			return (0);
421		MALLOC(vp->v_vnlock, struct lock *, sizeof(struct lock),
422		    M_VNODE, M_WAITOK);
423		lockinit(vp->v_vnlock, PVFS, "vnlock", 0, LK_NOPAUSE);
424	}
425	switch (flags & LK_TYPE_MASK) {
426	case LK_DRAIN:
427		vnflags = LK_DRAIN;
428		break;
429	case LK_EXCLUSIVE:
430	case LK_SHARED:
431		vnflags = LK_SHARED;
432		break;
433	case LK_UPGRADE:
434	case LK_EXCLUPGRADE:
435	case LK_DOWNGRADE:
436		return (0);
437	case LK_RELEASE:
438	default:
439		panic("vop_nolock: bad operation %d", flags & LK_TYPE_MASK);
440	}
441	if (flags & LK_INTERLOCK)
442		vnflags |= LK_INTERLOCK;
443	return(lockmgr(vp->v_vnlock, vnflags, &vp->v_interlock, ap->a_p));
444#else /* for now */
445	/*
446	 * Since we are not using the lock manager, we must clear
447	 * the interlock here.
448	 */
449	if (ap->a_flags & LK_INTERLOCK)
450		simple_unlock(&ap->a_vp->v_interlock);
451	return (0);
452#endif
453}
454
455/*
456 * Do the inverse of vop_nolock, handling the interlock in a compatible way.
457 */
458int
459vop_nounlock(ap)
460	struct vop_unlock_args /* {
461		struct vnode *a_vp;
462		int a_flags;
463		struct proc *a_p;
464	} */ *ap;
465{
466	struct vnode *vp = ap->a_vp;
467
468	if (vp->v_vnlock == NULL) {
469		if (ap->a_flags & LK_INTERLOCK)
470			simple_unlock(&ap->a_vp->v_interlock);
471		return (0);
472	}
473	return (lockmgr(vp->v_vnlock, LK_RELEASE | ap->a_flags,
474		&ap->a_vp->v_interlock, ap->a_p));
475}
476
477/*
478 * Return whether or not the node is in use.
479 */
480int
481vop_noislocked(ap)
482	struct vop_islocked_args /* {
483		struct vnode *a_vp;
484	} */ *ap;
485{
486	struct vnode *vp = ap->a_vp;
487
488	if (vp->v_vnlock == NULL)
489		return (0);
490	return (lockstatus(vp->v_vnlock));
491}
492
493