vfs_default.c revision 185956 
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 185956 2008-12-12 00:57:38Z marcus $");
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/lockf.h>
48#include <sys/malloc.h>
49#include <sys/mount.h>
50#include <sys/mutex.h>
51#include <sys/unistd.h>
52#include <sys/vnode.h>
53#include <sys/poll.h>
54
55#include <vm/vm.h>
56#include <vm/vm_object.h>
57#include <vm/vm_extern.h>
58#include <vm/pmap.h>
59#include <vm/vm_map.h>
60#include <vm/vm_page.h>
61#include <vm/vm_pager.h>
62#include <vm/vnode_pager.h>
63
64static int	vop_nolookup(struct vop_lookup_args *);
65static int	vop_nostrategy(struct vop_strategy_args *);
66
67/*
68 * This vnode table stores what we want to do if the filesystem doesn't
69 * implement a particular VOP.
70 *
71 * If there is no specific entry here, we will return EOPNOTSUPP.
72 *
73 */
74
75struct vop_vector default_vnodeops = {
76	.vop_default =		NULL,
77	.vop_bypass =		VOP_EOPNOTSUPP,
78
79	.vop_advlock =		vop_stdadvlock,
80	.vop_advlockasync =	vop_stdadvlockasync,
81	.vop_bmap =		vop_stdbmap,
82	.vop_close =		VOP_NULL,
83	.vop_fsync =		VOP_NULL,
84	.vop_getpages =		vop_stdgetpages,
85	.vop_getwritemount = 	vop_stdgetwritemount,
86	.vop_inactive =		VOP_NULL,
87	.vop_ioctl =		VOP_ENOTTY,
88	.vop_kqfilter =		vop_stdkqfilter,
89	.vop_islocked =		vop_stdislocked,
90	.vop_lease =		VOP_NULL,
91	.vop_lock1 =		vop_stdlock,
92	.vop_lookup =		vop_nolookup,
93	.vop_open =		VOP_NULL,
94	.vop_pathconf =		VOP_EINVAL,
95	.vop_poll =		vop_nopoll,
96	.vop_putpages =		vop_stdputpages,
97	.vop_readlink =		VOP_EINVAL,
98	.vop_revoke =		VOP_PANIC,
99	.vop_strategy =		vop_nostrategy,
100	.vop_unlock =		vop_stdunlock,
101	.vop_vptocnp =		VOP_ENOENT,
102	.vop_vptofh =		vop_stdvptofh,
103};
104
105/*
106 * Series of placeholder functions for various error returns for
107 * VOPs.
108 */
109
110int
111vop_eopnotsupp(struct vop_generic_args *ap)
112{
113	/*
114	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
115	*/
116
117	return (EOPNOTSUPP);
118}
119
120int
121vop_ebadf(struct vop_generic_args *ap)
122{
123
124	return (EBADF);
125}
126
127int
128vop_enotty(struct vop_generic_args *ap)
129{
130
131	return (ENOTTY);
132}
133
134int
135vop_einval(struct vop_generic_args *ap)
136{
137
138	return (EINVAL);
139}
140
141int
142vop_enoent(struct vop_generic_args *ap)
143{
144
145	return (ENOENT);
146}
147
148int
149vop_null(struct vop_generic_args *ap)
150{
151
152	return (0);
153}
154
155/*
156 * Helper function to panic on some bad VOPs in some filesystems.
157 */
158int
159vop_panic(struct vop_generic_args *ap)
160{
161
162	panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
163}
164
165/*
166 * vop_std<something> and vop_no<something> are default functions for use by
167 * filesystems that need the "default reasonable" implementation for a
168 * particular operation.
169 *
170 * The documentation for the operations they implement exists (if it exists)
171 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
172 */
173
174/*
175 * Default vop for filesystems that do not support name lookup
176 */
177static int
178vop_nolookup(ap)
179	struct vop_lookup_args /* {
180		struct vnode *a_dvp;
181		struct vnode **a_vpp;
182		struct componentname *a_cnp;
183	} */ *ap;
184{
185
186	*ap->a_vpp = NULL;
187	return (ENOTDIR);
188}
189
190/*
191 *	vop_nostrategy:
192 *
193 *	Strategy routine for VFS devices that have none.
194 *
195 *	BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
196 *	routine.  Typically this is done for a BIO_READ strategy call.
197 *	Typically B_INVAL is assumed to already be clear prior to a write
198 *	and should not be cleared manually unless you just made the buffer
199 *	invalid.  BIO_ERROR should be cleared either way.
200 */
201
202static int
203vop_nostrategy (struct vop_strategy_args *ap)
204{
205	printf("No strategy for buffer at %p\n", ap->a_bp);
206	vprint("vnode", ap->a_vp);
207	ap->a_bp->b_ioflags |= BIO_ERROR;
208	ap->a_bp->b_error = EOPNOTSUPP;
209	bufdone(ap->a_bp);
210	return (EOPNOTSUPP);
211}
212
213/*
214 * Advisory record locking support
215 */
216int
217vop_stdadvlock(struct vop_advlock_args *ap)
218{
219	struct vnode *vp;
220	struct ucred *cred;
221	struct vattr vattr;
222	int error;
223
224	vp = ap->a_vp;
225	cred = curthread->td_ucred;
226	vn_lock(vp, LK_SHARED | LK_RETRY);
227	error = VOP_GETATTR(vp, &vattr, cred);
228	VOP_UNLOCK(vp, 0);
229	if (error)
230		return (error);
231
232	return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
233}
234
235int
236vop_stdadvlockasync(struct vop_advlockasync_args *ap)
237{
238	struct vnode *vp;
239	struct ucred *cred;
240	struct vattr vattr;
241	int error;
242
243	vp = ap->a_vp;
244	cred = curthread->td_ucred;
245	vn_lock(vp, LK_SHARED | LK_RETRY);
246	error = VOP_GETATTR(vp, &vattr, cred);
247	VOP_UNLOCK(vp, 0);
248	if (error)
249		return (error);
250
251	return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
252}
253
254/*
255 * vop_stdpathconf:
256 *
257 * Standard implementation of POSIX pathconf, to get information about limits
258 * for a filesystem.
259 * Override per filesystem for the case where the filesystem has smaller
260 * limits.
261 */
262int
263vop_stdpathconf(ap)
264	struct vop_pathconf_args /* {
265	struct vnode *a_vp;
266	int a_name;
267	int *a_retval;
268	} */ *ap;
269{
270
271	switch (ap->a_name) {
272		case _PC_NAME_MAX:
273			*ap->a_retval = NAME_MAX;
274			return (0);
275		case _PC_PATH_MAX:
276			*ap->a_retval = PATH_MAX;
277			return (0);
278		case _PC_LINK_MAX:
279			*ap->a_retval = LINK_MAX;
280			return (0);
281		case _PC_MAX_CANON:
282			*ap->a_retval = MAX_CANON;
283			return (0);
284		case _PC_MAX_INPUT:
285			*ap->a_retval = MAX_INPUT;
286			return (0);
287		case _PC_PIPE_BUF:
288			*ap->a_retval = PIPE_BUF;
289			return (0);
290		case _PC_CHOWN_RESTRICTED:
291			*ap->a_retval = 1;
292			return (0);
293		case _PC_VDISABLE:
294			*ap->a_retval = _POSIX_VDISABLE;
295			return (0);
296		default:
297			return (EINVAL);
298	}
299	/* NOTREACHED */
300}
301
302/*
303 * Standard lock, unlock and islocked functions.
304 */
305int
306vop_stdlock(ap)
307	struct vop_lock1_args /* {
308		struct vnode *a_vp;
309		int a_flags;
310		char *file;
311		int line;
312	} */ *ap;
313{
314	struct vnode *vp = ap->a_vp;
315
316	return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
317	    LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
318	    ap->a_line));
319}
320
321/* See above. */
322int
323vop_stdunlock(ap)
324	struct vop_unlock_args /* {
325		struct vnode *a_vp;
326		int a_flags;
327	} */ *ap;
328{
329	struct vnode *vp = ap->a_vp;
330
331	return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
332}
333
334/* See above. */
335int
336vop_stdislocked(ap)
337	struct vop_islocked_args /* {
338		struct vnode *a_vp;
339	} */ *ap;
340{
341
342	return (lockstatus(ap->a_vp->v_vnlock));
343}
344
345/*
346 * Return true for select/poll.
347 */
348int
349vop_nopoll(ap)
350	struct vop_poll_args /* {
351		struct vnode *a_vp;
352		int  a_events;
353		struct ucred *a_cred;
354		struct thread *a_td;
355	} */ *ap;
356{
357	/*
358	 * Return true for read/write.  If the user asked for something
359	 * special, return POLLNVAL, so that clients have a way of
360	 * determining reliably whether or not the extended
361	 * functionality is present without hard-coding knowledge
362	 * of specific filesystem implementations.
363	 * Stay in sync with kern_conf.c::no_poll().
364	 */
365	if (ap->a_events & ~POLLSTANDARD)
366		return (POLLNVAL);
367
368	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
369}
370
371/*
372 * Implement poll for local filesystems that support it.
373 */
374int
375vop_stdpoll(ap)
376	struct vop_poll_args /* {
377		struct vnode *a_vp;
378		int  a_events;
379		struct ucred *a_cred;
380		struct thread *a_td;
381	} */ *ap;
382{
383	if (ap->a_events & ~POLLSTANDARD)
384		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
385	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
386}
387
388/*
389 * Return our mount point, as we will take charge of the writes.
390 */
391int
392vop_stdgetwritemount(ap)
393	struct vop_getwritemount_args /* {
394		struct vnode *a_vp;
395		struct mount **a_mpp;
396	} */ *ap;
397{
398	struct mount *mp;
399
400	/*
401	 * XXX Since this is called unlocked we may be recycled while
402	 * attempting to ref the mount.  If this is the case or mountpoint
403	 * will be set to NULL.  We only have to prevent this call from
404	 * returning with a ref to an incorrect mountpoint.  It is not
405	 * harmful to return with a ref to our previous mountpoint.
406	 */
407	mp = ap->a_vp->v_mount;
408	if (mp != NULL) {
409		vfs_ref(mp);
410		if (mp != ap->a_vp->v_mount) {
411			vfs_rel(mp);
412			mp = NULL;
413		}
414	}
415	*(ap->a_mpp) = mp;
416	return (0);
417}
418
419/* XXX Needs good comment and VOP_BMAP(9) manpage */
420int
421vop_stdbmap(ap)
422	struct vop_bmap_args /* {
423		struct vnode *a_vp;
424		daddr_t  a_bn;
425		struct bufobj **a_bop;
426		daddr_t *a_bnp;
427		int *a_runp;
428		int *a_runb;
429	} */ *ap;
430{
431
432	if (ap->a_bop != NULL)
433		*ap->a_bop = &ap->a_vp->v_bufobj;
434	if (ap->a_bnp != NULL)
435		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
436	if (ap->a_runp != NULL)
437		*ap->a_runp = 0;
438	if (ap->a_runb != NULL)
439		*ap->a_runb = 0;
440	return (0);
441}
442
443int
444vop_stdfsync(ap)
445	struct vop_fsync_args /* {
446		struct vnode *a_vp;
447		struct ucred *a_cred;
448		int a_waitfor;
449		struct thread *a_td;
450	} */ *ap;
451{
452	struct vnode *vp = ap->a_vp;
453	struct buf *bp;
454	struct bufobj *bo;
455	struct buf *nbp;
456	int error = 0;
457	int maxretry = 1000;     /* large, arbitrarily chosen */
458
459	bo = &vp->v_bufobj;
460	BO_LOCK(bo);
461loop1:
462	/*
463	 * MARK/SCAN initialization to avoid infinite loops.
464	 */
465        TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
466                bp->b_vflags &= ~BV_SCANNED;
467		bp->b_error = 0;
468	}
469
470	/*
471	 * Flush all dirty buffers associated with a vnode.
472	 */
473loop2:
474	TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
475		if ((bp->b_vflags & BV_SCANNED) != 0)
476			continue;
477		bp->b_vflags |= BV_SCANNED;
478		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
479			continue;
480		BO_UNLOCK(bo);
481		KASSERT(bp->b_bufobj == bo,
482		    ("bp %p wrong b_bufobj %p should be %p",
483		    bp, bp->b_bufobj, bo));
484		if ((bp->b_flags & B_DELWRI) == 0)
485			panic("fsync: not dirty");
486		if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
487			vfs_bio_awrite(bp);
488		} else {
489			bremfree(bp);
490			bawrite(bp);
491		}
492		BO_LOCK(bo);
493		goto loop2;
494	}
495
496	/*
497	 * If synchronous the caller expects us to completely resolve all
498	 * dirty buffers in the system.  Wait for in-progress I/O to
499	 * complete (which could include background bitmap writes), then
500	 * retry if dirty blocks still exist.
501	 */
502	if (ap->a_waitfor == MNT_WAIT) {
503		bufobj_wwait(bo, 0, 0);
504		if (bo->bo_dirty.bv_cnt > 0) {
505			/*
506			 * If we are unable to write any of these buffers
507			 * then we fail now rather than trying endlessly
508			 * to write them out.
509			 */
510			TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
511				if ((error = bp->b_error) == 0)
512					continue;
513			if (error == 0 && --maxretry >= 0)
514				goto loop1;
515			error = EAGAIN;
516		}
517	}
518	BO_UNLOCK(bo);
519	if (error == EAGAIN)
520		vprint("fsync: giving up on dirty", vp);
521
522	return (error);
523}
524
525/* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
526int
527vop_stdgetpages(ap)
528	struct vop_getpages_args /* {
529		struct vnode *a_vp;
530		vm_page_t *a_m;
531		int a_count;
532		int a_reqpage;
533		vm_ooffset_t a_offset;
534	} */ *ap;
535{
536
537	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
538	    ap->a_count, ap->a_reqpage);
539}
540
541int
542vop_stdkqfilter(struct vop_kqfilter_args *ap)
543{
544	return vfs_kqfilter(ap);
545}
546
547/* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
548int
549vop_stdputpages(ap)
550	struct vop_putpages_args /* {
551		struct vnode *a_vp;
552		vm_page_t *a_m;
553		int a_count;
554		int a_sync;
555		int *a_rtvals;
556		vm_ooffset_t a_offset;
557	} */ *ap;
558{
559
560	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
561	     ap->a_sync, ap->a_rtvals);
562}
563
564int
565vop_stdvptofh(struct vop_vptofh_args *ap)
566{
567	return (EOPNOTSUPP);
568}
569
570/*
571 * vfs default ops
572 * used to fill the vfs function table to get reasonable default return values.
573 */
574int
575vfs_stdroot (mp, flags, vpp, td)
576	struct mount *mp;
577	int flags;
578	struct vnode **vpp;
579	struct thread *td;
580{
581
582	return (EOPNOTSUPP);
583}
584
585int
586vfs_stdstatfs (mp, sbp, td)
587	struct mount *mp;
588	struct statfs *sbp;
589	struct thread *td;
590{
591
592	return (EOPNOTSUPP);
593}
594
595int
596vfs_stdquotactl (mp, cmds, uid, arg, td)
597	struct mount *mp;
598	int cmds;
599	uid_t uid;
600	void *arg;
601	struct thread *td;
602{
603
604	return (EOPNOTSUPP);
605}
606
607int
608vfs_stdsync(mp, waitfor, td)
609	struct mount *mp;
610	int waitfor;
611	struct thread *td;
612{
613	struct vnode *vp, *mvp;
614	int error, lockreq, allerror = 0;
615
616	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
617	if (waitfor != MNT_WAIT)
618		lockreq |= LK_NOWAIT;
619	/*
620	 * Force stale buffer cache information to be flushed.
621	 */
622	MNT_ILOCK(mp);
623loop:
624	MNT_VNODE_FOREACH(vp, mp, mvp) {
625		/* bv_cnt is an acceptable race here. */
626		if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
627			continue;
628		VI_LOCK(vp);
629		MNT_IUNLOCK(mp);
630		if ((error = vget(vp, lockreq, td)) != 0) {
631			MNT_ILOCK(mp);
632			if (error == ENOENT) {
633				MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
634				goto loop;
635			}
636			continue;
637		}
638		error = VOP_FSYNC(vp, waitfor, td);
639		if (error)
640			allerror = error;
641
642		/* Do not turn this into vput.  td is not always curthread. */
643		VOP_UNLOCK(vp, 0);
644		vrele(vp);
645		MNT_ILOCK(mp);
646	}
647	MNT_IUNLOCK(mp);
648	return (allerror);
649}
650
651int
652vfs_stdnosync (mp, waitfor, td)
653	struct mount *mp;
654	int waitfor;
655	struct thread *td;
656{
657
658	return (0);
659}
660
661int
662vfs_stdvget (mp, ino, flags, vpp)
663	struct mount *mp;
664	ino_t ino;
665	int flags;
666	struct vnode **vpp;
667{
668
669	return (EOPNOTSUPP);
670}
671
672int
673vfs_stdfhtovp (mp, fhp, vpp)
674	struct mount *mp;
675	struct fid *fhp;
676	struct vnode **vpp;
677{
678
679	return (EOPNOTSUPP);
680}
681
682int
683vfs_stdinit (vfsp)
684	struct vfsconf *vfsp;
685{
686
687	return (0);
688}
689
690int
691vfs_stduninit (vfsp)
692	struct vfsconf *vfsp;
693{
694
695	return(0);
696}
697
698int
699vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td)
700	struct mount *mp;
701	int cmd;
702	struct vnode *filename_vp;
703	int attrnamespace;
704	const char *attrname;
705	struct thread *td;
706{
707
708	if (filename_vp != NULL)
709		VOP_UNLOCK(filename_vp, 0);
710	return (EOPNOTSUPP);
711}
712
713int
714vfs_stdsysctl(mp, op, req)
715	struct mount *mp;
716	fsctlop_t op;
717	struct sysctl_req *req;
718{
719
720	return (EOPNOTSUPP);
721}
722
723/* end of vfs default ops */
724