vfs_default.c revision 157323 
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 * 4. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: head/sys/kern/vfs_default.c 157323 2006-03-31 03:52:24Z jeff $");
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/bio.h>
41#include <sys/buf.h>
42#include <sys/conf.h>
43#include <sys/event.h>
44#include <sys/kernel.h>
45#include <sys/limits.h>
46#include <sys/lock.h>
47#include <sys/malloc.h>
48#include <sys/mount.h>
49#include <sys/mutex.h>
50#include <sys/unistd.h>
51#include <sys/vnode.h>
52#include <sys/poll.h>
53
54#include <vm/vm.h>
55#include <vm/vm_object.h>
56#include <vm/vm_extern.h>
57#include <vm/pmap.h>
58#include <vm/vm_map.h>
59#include <vm/vm_page.h>
60#include <vm/vm_pager.h>
61#include <vm/vnode_pager.h>
62
63static int	vop_nolookup(struct vop_lookup_args *);
64static int	vop_nostrategy(struct vop_strategy_args *);
65
66/*
67 * This vnode table stores what we want to do if the filesystem doesn't
68 * implement a particular VOP.
69 *
70 * If there is no specific entry here, we will return EOPNOTSUPP.
71 *
72 */
73
74struct vop_vector default_vnodeops = {
75	.vop_default =		NULL,
76	.vop_bypass =		VOP_EOPNOTSUPP,
77
78	.vop_advlock =		VOP_EINVAL,
79	.vop_bmap =		vop_stdbmap,
80	.vop_close =		VOP_NULL,
81	.vop_fsync =		VOP_NULL,
82	.vop_getpages =		vop_stdgetpages,
83	.vop_getwritemount = 	vop_stdgetwritemount,
84	.vop_inactive =		VOP_NULL,
85	.vop_ioctl =		VOP_ENOTTY,
86	.vop_kqfilter =		vop_stdkqfilter,
87	.vop_islocked =		vop_stdislocked,
88	.vop_lease =		VOP_NULL,
89	.vop_lock =		vop_stdlock,
90	.vop_lookup =		vop_nolookup,
91	.vop_open =		VOP_NULL,
92	.vop_pathconf =		VOP_EINVAL,
93	.vop_poll =		vop_nopoll,
94	.vop_putpages =		vop_stdputpages,
95	.vop_readlink =		VOP_EINVAL,
96	.vop_revoke =		VOP_PANIC,
97	.vop_strategy =		vop_nostrategy,
98	.vop_unlock =		vop_stdunlock,
99};
100
101/*
102 * Series of placeholder functions for various error returns for
103 * VOPs.
104 */
105
106int
107vop_eopnotsupp(struct vop_generic_args *ap)
108{
109	/*
110	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
111	*/
112
113	return (EOPNOTSUPP);
114}
115
116int
117vop_ebadf(struct vop_generic_args *ap)
118{
119
120	return (EBADF);
121}
122
123int
124vop_enotty(struct vop_generic_args *ap)
125{
126
127	return (ENOTTY);
128}
129
130int
131vop_einval(struct vop_generic_args *ap)
132{
133
134	return (EINVAL);
135}
136
137int
138vop_null(struct vop_generic_args *ap)
139{
140
141	return (0);
142}
143
144/*
145 * Helper function to panic on some bad VOPs in some filesystems.
146 */
147int
148vop_panic(struct vop_generic_args *ap)
149{
150
151	panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
152}
153
154/*
155 * vop_std<something> and vop_no<something> are default functions for use by
156 * filesystems that need the "default reasonable" implementation for a
157 * particular operation.
158 *
159 * The documentation for the operations they implement exists (if it exists)
160 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
161 */
162
163/*
164 * Default vop for filesystems that do not support name lookup
165 */
166static int
167vop_nolookup(ap)
168	struct vop_lookup_args /* {
169		struct vnode *a_dvp;
170		struct vnode **a_vpp;
171		struct componentname *a_cnp;
172	} */ *ap;
173{
174
175	*ap->a_vpp = NULL;
176	return (ENOTDIR);
177}
178
179/*
180 *	vop_nostrategy:
181 *
182 *	Strategy routine for VFS devices that have none.
183 *
184 *	BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
185 *	routine.  Typically this is done for a BIO_READ strategy call.
186 *	Typically B_INVAL is assumed to already be clear prior to a write
187 *	and should not be cleared manually unless you just made the buffer
188 *	invalid.  BIO_ERROR should be cleared either way.
189 */
190
191static int
192vop_nostrategy (struct vop_strategy_args *ap)
193{
194	printf("No strategy for buffer at %p\n", ap->a_bp);
195	vprint("vnode", ap->a_vp);
196	ap->a_bp->b_ioflags |= BIO_ERROR;
197	ap->a_bp->b_error = EOPNOTSUPP;
198	bufdone(ap->a_bp);
199	return (EOPNOTSUPP);
200}
201
202/*
203 * vop_stdpathconf:
204 *
205 * Standard implementation of POSIX pathconf, to get information about limits
206 * for a filesystem.
207 * Override per filesystem for the case where the filesystem has smaller
208 * limits.
209 */
210int
211vop_stdpathconf(ap)
212	struct vop_pathconf_args /* {
213	struct vnode *a_vp;
214	int a_name;
215	int *a_retval;
216	} */ *ap;
217{
218
219	switch (ap->a_name) {
220		case _PC_NAME_MAX:
221			*ap->a_retval = NAME_MAX;
222			return (0);
223		case _PC_PATH_MAX:
224			*ap->a_retval = PATH_MAX;
225			return (0);
226		case _PC_LINK_MAX:
227			*ap->a_retval = LINK_MAX;
228			return (0);
229		case _PC_MAX_CANON:
230			*ap->a_retval = MAX_CANON;
231			return (0);
232		case _PC_MAX_INPUT:
233			*ap->a_retval = MAX_INPUT;
234			return (0);
235		case _PC_PIPE_BUF:
236			*ap->a_retval = PIPE_BUF;
237			return (0);
238		case _PC_CHOWN_RESTRICTED:
239			*ap->a_retval = 1;
240			return (0);
241		case _PC_VDISABLE:
242			*ap->a_retval = _POSIX_VDISABLE;
243			return (0);
244		default:
245			return (EINVAL);
246	}
247	/* NOTREACHED */
248}
249
250/*
251 * Standard lock, unlock and islocked functions.
252 */
253int
254vop_stdlock(ap)
255	struct vop_lock_args /* {
256		struct vnode *a_vp;
257		int a_flags;
258		struct thread *a_td;
259	} */ *ap;
260{
261	struct vnode *vp = ap->a_vp;
262
263	return (lockmgr(vp->v_vnlock, ap->a_flags, VI_MTX(vp), ap->a_td));
264}
265
266/* See above. */
267int
268vop_stdunlock(ap)
269	struct vop_unlock_args /* {
270		struct vnode *a_vp;
271		int a_flags;
272		struct thread *a_td;
273	} */ *ap;
274{
275	struct vnode *vp = ap->a_vp;
276
277	return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp),
278	    ap->a_td));
279}
280
281/* See above. */
282int
283vop_stdislocked(ap)
284	struct vop_islocked_args /* {
285		struct vnode *a_vp;
286		struct thread *a_td;
287	} */ *ap;
288{
289
290	return (lockstatus(ap->a_vp->v_vnlock, ap->a_td));
291}
292
293/*
294 * Return true for select/poll.
295 */
296int
297vop_nopoll(ap)
298	struct vop_poll_args /* {
299		struct vnode *a_vp;
300		int  a_events;
301		struct ucred *a_cred;
302		struct thread *a_td;
303	} */ *ap;
304{
305	/*
306	 * Return true for read/write.  If the user asked for something
307	 * special, return POLLNVAL, so that clients have a way of
308	 * determining reliably whether or not the extended
309	 * functionality is present without hard-coding knowledge
310	 * of specific filesystem implementations.
311	 * Stay in sync with kern_conf.c::no_poll().
312	 */
313	if (ap->a_events & ~POLLSTANDARD)
314		return (POLLNVAL);
315
316	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
317}
318
319/*
320 * Implement poll for local filesystems that support it.
321 */
322int
323vop_stdpoll(ap)
324	struct vop_poll_args /* {
325		struct vnode *a_vp;
326		int  a_events;
327		struct ucred *a_cred;
328		struct thread *a_td;
329	} */ *ap;
330{
331	if (ap->a_events & ~POLLSTANDARD)
332		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
333	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
334}
335
336/*
337 * Return our mount point, as we will take charge of the writes.
338 */
339int
340vop_stdgetwritemount(ap)
341	struct vop_getwritemount_args /* {
342		struct vnode *a_vp;
343		struct mount **a_mpp;
344	} */ *ap;
345{
346	struct mount *mp;
347
348	/*
349	 * XXX Since this is called unlocked we may be recycled while
350	 * attempting to ref the mount.  If this is the case or mountpoint
351	 * will be set to NULL.  We only have to prevent this call from
352	 * returning with a ref to an incorrect mountpoint.  It is not
353	 * harmful to return with a ref to our previous mountpoint.
354	 */
355	mp = ap->a_vp->v_mount;
356	vfs_ref(mp);
357	if (mp != ap->a_vp->v_mount) {
358		vfs_rel(mp);
359		mp = NULL;
360	}
361	*(ap->a_mpp) = mp;
362	return (0);
363}
364
365/* XXX Needs good comment and VOP_BMAP(9) manpage */
366int
367vop_stdbmap(ap)
368	struct vop_bmap_args /* {
369		struct vnode *a_vp;
370		daddr_t  a_bn;
371		struct bufobj **a_bop;
372		daddr_t *a_bnp;
373		int *a_runp;
374		int *a_runb;
375	} */ *ap;
376{
377
378	if (ap->a_bop != NULL)
379		*ap->a_bop = &ap->a_vp->v_bufobj;
380	if (ap->a_bnp != NULL)
381		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
382	if (ap->a_runp != NULL)
383		*ap->a_runp = 0;
384	if (ap->a_runb != NULL)
385		*ap->a_runb = 0;
386	return (0);
387}
388
389int
390vop_stdfsync(ap)
391	struct vop_fsync_args /* {
392		struct vnode *a_vp;
393		struct ucred *a_cred;
394		int a_waitfor;
395		struct thread *a_td;
396	} */ *ap;
397{
398	struct vnode *vp = ap->a_vp;
399	struct buf *bp;
400	struct bufobj *bo;
401	struct buf *nbp;
402	int error = 0;
403	int maxretry = 1000;     /* large, arbitrarily chosen */
404
405	VI_LOCK(vp);
406loop1:
407	/*
408	 * MARK/SCAN initialization to avoid infinite loops.
409	 */
410        TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs) {
411                bp->b_vflags &= ~BV_SCANNED;
412		bp->b_error = 0;
413	}
414
415	/*
416	 * Flush all dirty buffers associated with a vnode.
417	 */
418loop2:
419	TAILQ_FOREACH_SAFE(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs, nbp) {
420		if ((bp->b_vflags & BV_SCANNED) != 0)
421			continue;
422		bp->b_vflags |= BV_SCANNED;
423		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
424			continue;
425		VI_UNLOCK(vp);
426		KASSERT(bp->b_bufobj == &vp->v_bufobj,
427		    ("bp %p wrong b_bufobj %p should be %p",
428		    bp, bp->b_bufobj, &vp->v_bufobj));
429		if ((bp->b_flags & B_DELWRI) == 0)
430			panic("fsync: not dirty");
431		if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
432			vfs_bio_awrite(bp);
433		} else {
434			bremfree(bp);
435			bawrite(bp);
436		}
437		VI_LOCK(vp);
438		goto loop2;
439	}
440
441	/*
442	 * If synchronous the caller expects us to completely resolve all
443	 * dirty buffers in the system.  Wait for in-progress I/O to
444	 * complete (which could include background bitmap writes), then
445	 * retry if dirty blocks still exist.
446	 */
447	if (ap->a_waitfor == MNT_WAIT) {
448		bo = &vp->v_bufobj;
449		bufobj_wwait(bo, 0, 0);
450		if (bo->bo_dirty.bv_cnt > 0) {
451			/*
452			 * If we are unable to write any of these buffers
453			 * then we fail now rather than trying endlessly
454			 * to write them out.
455			 */
456			TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
457				if ((error = bp->b_error) == 0)
458					continue;
459			if (error == 0 && --maxretry >= 0)
460				goto loop1;
461			error = EAGAIN;
462		}
463	}
464	VI_UNLOCK(vp);
465	if (error == EAGAIN)
466		vprint("fsync: giving up on dirty", vp);
467
468	return (error);
469}
470
471/* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
472int
473vop_stdgetpages(ap)
474	struct vop_getpages_args /* {
475		struct vnode *a_vp;
476		vm_page_t *a_m;
477		int a_count;
478		int a_reqpage;
479		vm_ooffset_t a_offset;
480	} */ *ap;
481{
482
483	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
484	    ap->a_count, ap->a_reqpage);
485}
486
487int
488vop_stdkqfilter(struct vop_kqfilter_args *ap)
489{
490	return vfs_kqfilter(ap);
491}
492
493/* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
494int
495vop_stdputpages(ap)
496	struct vop_putpages_args /* {
497		struct vnode *a_vp;
498		vm_page_t *a_m;
499		int a_count;
500		int a_sync;
501		int *a_rtvals;
502		vm_ooffset_t a_offset;
503	} */ *ap;
504{
505
506	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
507	     ap->a_sync, ap->a_rtvals);
508}
509
510/*
511 * vfs default ops
512 * used to fill the vfs function table to get reasonable default return values.
513 */
514int
515vfs_stdroot (mp, flags, vpp, td)
516	struct mount *mp;
517	int flags;
518	struct vnode **vpp;
519	struct thread *td;
520{
521
522	return (EOPNOTSUPP);
523}
524
525int
526vfs_stdstatfs (mp, sbp, td)
527	struct mount *mp;
528	struct statfs *sbp;
529	struct thread *td;
530{
531
532	return (EOPNOTSUPP);
533}
534
535int
536vfs_stdvptofh (vp, fhp)
537	struct vnode *vp;
538	struct fid *fhp;
539{
540
541	return (EOPNOTSUPP);
542}
543
544int
545vfs_stdquotactl (mp, cmds, uid, arg, td)
546	struct mount *mp;
547	int cmds;
548	uid_t uid;
549	void *arg;
550	struct thread *td;
551{
552
553	return (EOPNOTSUPP);
554}
555
556int
557vfs_stdsync(mp, waitfor, td)
558	struct mount *mp;
559	int waitfor;
560	struct thread *td;
561{
562	struct vnode *vp, *mvp;
563	int error, lockreq, allerror = 0;
564
565	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
566	if (waitfor != MNT_WAIT)
567		lockreq |= LK_NOWAIT;
568	/*
569	 * Force stale buffer cache information to be flushed.
570	 */
571	MNT_ILOCK(mp);
572loop:
573	MNT_VNODE_FOREACH(vp, mp, mvp) {
574
575		VI_LOCK(vp);
576		if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
577			VI_UNLOCK(vp);
578			continue;
579		}
580		MNT_IUNLOCK(mp);
581
582		if ((error = vget(vp, lockreq, td)) != 0) {
583			MNT_ILOCK(mp);
584			if (error == ENOENT) {
585				MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
586				goto loop;
587			}
588			continue;
589		}
590		error = VOP_FSYNC(vp, waitfor, td);
591		if (error)
592			allerror = error;
593
594		/* Do not turn this into vput.  td is not always curthread. */
595		VOP_UNLOCK(vp, 0, td);
596		vrele(vp);
597		MNT_ILOCK(mp);
598	}
599	MNT_IUNLOCK(mp);
600	return (allerror);
601}
602
603int
604vfs_stdnosync (mp, waitfor, td)
605	struct mount *mp;
606	int waitfor;
607	struct thread *td;
608{
609
610	return (0);
611}
612
613int
614vfs_stdvget (mp, ino, flags, vpp)
615	struct mount *mp;
616	ino_t ino;
617	int flags;
618	struct vnode **vpp;
619{
620
621	return (EOPNOTSUPP);
622}
623
624int
625vfs_stdfhtovp (mp, fhp, vpp)
626	struct mount *mp;
627	struct fid *fhp;
628	struct vnode **vpp;
629{
630
631	return (EOPNOTSUPP);
632}
633
634int
635vfs_stdinit (vfsp)
636	struct vfsconf *vfsp;
637{
638
639	return (0);
640}
641
642int
643vfs_stduninit (vfsp)
644	struct vfsconf *vfsp;
645{
646
647	return(0);
648}
649
650int
651vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td)
652	struct mount *mp;
653	int cmd;
654	struct vnode *filename_vp;
655	int attrnamespace;
656	const char *attrname;
657	struct thread *td;
658{
659
660	if (filename_vp != NULL)
661		VOP_UNLOCK(filename_vp, 0, td);
662	return (EOPNOTSUPP);
663}
664
665int
666vfs_stdsysctl(mp, op, req)
667	struct mount *mp;
668	fsctlop_t op;
669	struct sysctl_req *req;
670{
671
672	return (EOPNOTSUPP);
673}
674
675/* end of vfs default ops */
676