xref: /freebsd/sys/kern/vfs_default.c (revision acd3428b7d3e94cef0e1881c868cb4b131d4ff41)
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("$FreeBSD$");
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 
63 static int	vop_nolookup(struct vop_lookup_args *);
64 static 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 
74 struct 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 
106 int
107 vop_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 
116 int
117 vop_ebadf(struct vop_generic_args *ap)
118 {
119 
120 	return (EBADF);
121 }
122 
123 int
124 vop_enotty(struct vop_generic_args *ap)
125 {
126 
127 	return (ENOTTY);
128 }
129 
130 int
131 vop_einval(struct vop_generic_args *ap)
132 {
133 
134 	return (EINVAL);
135 }
136 
137 int
138 vop_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  */
147 int
148 vop_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  */
166 static int
167 vop_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 
191 static int
192 vop_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  */
210 int
211 vop_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  */
253 int
254 vop_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. */
267 int
268 vop_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. */
282 int
283 vop_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  */
296 int
297 vop_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  */
322 int
323 vop_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  */
339 int
340 vop_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 	if (mp != NULL) {
357 		vfs_ref(mp);
358 		if (mp != ap->a_vp->v_mount) {
359 			vfs_rel(mp);
360 			mp = NULL;
361 		}
362 	}
363 	*(ap->a_mpp) = mp;
364 	return (0);
365 }
366 
367 /* XXX Needs good comment and VOP_BMAP(9) manpage */
368 int
369 vop_stdbmap(ap)
370 	struct vop_bmap_args /* {
371 		struct vnode *a_vp;
372 		daddr_t  a_bn;
373 		struct bufobj **a_bop;
374 		daddr_t *a_bnp;
375 		int *a_runp;
376 		int *a_runb;
377 	} */ *ap;
378 {
379 
380 	if (ap->a_bop != NULL)
381 		*ap->a_bop = &ap->a_vp->v_bufobj;
382 	if (ap->a_bnp != NULL)
383 		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
384 	if (ap->a_runp != NULL)
385 		*ap->a_runp = 0;
386 	if (ap->a_runb != NULL)
387 		*ap->a_runb = 0;
388 	return (0);
389 }
390 
391 int
392 vop_stdfsync(ap)
393 	struct vop_fsync_args /* {
394 		struct vnode *a_vp;
395 		struct ucred *a_cred;
396 		int a_waitfor;
397 		struct thread *a_td;
398 	} */ *ap;
399 {
400 	struct vnode *vp = ap->a_vp;
401 	struct buf *bp;
402 	struct bufobj *bo;
403 	struct buf *nbp;
404 	int error = 0;
405 	int maxretry = 1000;     /* large, arbitrarily chosen */
406 
407 	VI_LOCK(vp);
408 loop1:
409 	/*
410 	 * MARK/SCAN initialization to avoid infinite loops.
411 	 */
412         TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs) {
413                 bp->b_vflags &= ~BV_SCANNED;
414 		bp->b_error = 0;
415 	}
416 
417 	/*
418 	 * Flush all dirty buffers associated with a vnode.
419 	 */
420 loop2:
421 	TAILQ_FOREACH_SAFE(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs, nbp) {
422 		if ((bp->b_vflags & BV_SCANNED) != 0)
423 			continue;
424 		bp->b_vflags |= BV_SCANNED;
425 		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
426 			continue;
427 		VI_UNLOCK(vp);
428 		KASSERT(bp->b_bufobj == &vp->v_bufobj,
429 		    ("bp %p wrong b_bufobj %p should be %p",
430 		    bp, bp->b_bufobj, &vp->v_bufobj));
431 		if ((bp->b_flags & B_DELWRI) == 0)
432 			panic("fsync: not dirty");
433 		if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
434 			vfs_bio_awrite(bp);
435 		} else {
436 			bremfree(bp);
437 			bawrite(bp);
438 		}
439 		VI_LOCK(vp);
440 		goto loop2;
441 	}
442 
443 	/*
444 	 * If synchronous the caller expects us to completely resolve all
445 	 * dirty buffers in the system.  Wait for in-progress I/O to
446 	 * complete (which could include background bitmap writes), then
447 	 * retry if dirty blocks still exist.
448 	 */
449 	if (ap->a_waitfor == MNT_WAIT) {
450 		bo = &vp->v_bufobj;
451 		bufobj_wwait(bo, 0, 0);
452 		if (bo->bo_dirty.bv_cnt > 0) {
453 			/*
454 			 * If we are unable to write any of these buffers
455 			 * then we fail now rather than trying endlessly
456 			 * to write them out.
457 			 */
458 			TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
459 				if ((error = bp->b_error) == 0)
460 					continue;
461 			if (error == 0 && --maxretry >= 0)
462 				goto loop1;
463 			error = EAGAIN;
464 		}
465 	}
466 	VI_UNLOCK(vp);
467 	if (error == EAGAIN)
468 		vprint("fsync: giving up on dirty", vp);
469 
470 	return (error);
471 }
472 
473 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
474 int
475 vop_stdgetpages(ap)
476 	struct vop_getpages_args /* {
477 		struct vnode *a_vp;
478 		vm_page_t *a_m;
479 		int a_count;
480 		int a_reqpage;
481 		vm_ooffset_t a_offset;
482 	} */ *ap;
483 {
484 
485 	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
486 	    ap->a_count, ap->a_reqpage);
487 }
488 
489 int
490 vop_stdkqfilter(struct vop_kqfilter_args *ap)
491 {
492 	return vfs_kqfilter(ap);
493 }
494 
495 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
496 int
497 vop_stdputpages(ap)
498 	struct vop_putpages_args /* {
499 		struct vnode *a_vp;
500 		vm_page_t *a_m;
501 		int a_count;
502 		int a_sync;
503 		int *a_rtvals;
504 		vm_ooffset_t a_offset;
505 	} */ *ap;
506 {
507 
508 	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
509 	     ap->a_sync, ap->a_rtvals);
510 }
511 
512 /*
513  * vfs default ops
514  * used to fill the vfs function table to get reasonable default return values.
515  */
516 int
517 vfs_stdroot (mp, flags, vpp, td)
518 	struct mount *mp;
519 	int flags;
520 	struct vnode **vpp;
521 	struct thread *td;
522 {
523 
524 	return (EOPNOTSUPP);
525 }
526 
527 int
528 vfs_stdstatfs (mp, sbp, td)
529 	struct mount *mp;
530 	struct statfs *sbp;
531 	struct thread *td;
532 {
533 
534 	return (EOPNOTSUPP);
535 }
536 
537 int
538 vfs_stdvptofh (vp, fhp)
539 	struct vnode *vp;
540 	struct fid *fhp;
541 {
542 
543 	return (EOPNOTSUPP);
544 }
545 
546 int
547 vfs_stdquotactl (mp, cmds, uid, arg, td)
548 	struct mount *mp;
549 	int cmds;
550 	uid_t uid;
551 	void *arg;
552 	struct thread *td;
553 {
554 
555 	return (EOPNOTSUPP);
556 }
557 
558 int
559 vfs_stdsync(mp, waitfor, td)
560 	struct mount *mp;
561 	int waitfor;
562 	struct thread *td;
563 {
564 	struct vnode *vp, *mvp;
565 	int error, lockreq, allerror = 0;
566 
567 	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
568 	if (waitfor != MNT_WAIT)
569 		lockreq |= LK_NOWAIT;
570 	/*
571 	 * Force stale buffer cache information to be flushed.
572 	 */
573 	MNT_ILOCK(mp);
574 loop:
575 	MNT_VNODE_FOREACH(vp, mp, mvp) {
576 
577 		VI_LOCK(vp);
578 		if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
579 			VI_UNLOCK(vp);
580 			continue;
581 		}
582 		MNT_IUNLOCK(mp);
583 
584 		if ((error = vget(vp, lockreq, td)) != 0) {
585 			MNT_ILOCK(mp);
586 			if (error == ENOENT) {
587 				MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
588 				goto loop;
589 			}
590 			continue;
591 		}
592 		error = VOP_FSYNC(vp, waitfor, td);
593 		if (error)
594 			allerror = error;
595 
596 		/* Do not turn this into vput.  td is not always curthread. */
597 		VOP_UNLOCK(vp, 0, td);
598 		vrele(vp);
599 		MNT_ILOCK(mp);
600 	}
601 	MNT_IUNLOCK(mp);
602 	return (allerror);
603 }
604 
605 int
606 vfs_stdnosync (mp, waitfor, td)
607 	struct mount *mp;
608 	int waitfor;
609 	struct thread *td;
610 {
611 
612 	return (0);
613 }
614 
615 int
616 vfs_stdvget (mp, ino, flags, vpp)
617 	struct mount *mp;
618 	ino_t ino;
619 	int flags;
620 	struct vnode **vpp;
621 {
622 
623 	return (EOPNOTSUPP);
624 }
625 
626 int
627 vfs_stdfhtovp (mp, fhp, vpp)
628 	struct mount *mp;
629 	struct fid *fhp;
630 	struct vnode **vpp;
631 {
632 
633 	return (EOPNOTSUPP);
634 }
635 
636 int
637 vfs_stdinit (vfsp)
638 	struct vfsconf *vfsp;
639 {
640 
641 	return (0);
642 }
643 
644 int
645 vfs_stduninit (vfsp)
646 	struct vfsconf *vfsp;
647 {
648 
649 	return(0);
650 }
651 
652 int
653 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td)
654 	struct mount *mp;
655 	int cmd;
656 	struct vnode *filename_vp;
657 	int attrnamespace;
658 	const char *attrname;
659 	struct thread *td;
660 {
661 
662 	if (filename_vp != NULL)
663 		VOP_UNLOCK(filename_vp, 0, td);
664 	return (EOPNOTSUPP);
665 }
666 
667 int
668 vfs_stdsysctl(mp, op, req)
669 	struct mount *mp;
670 	fsctlop_t op;
671 	struct sysctl_req *req;
672 {
673 
674 	return (EOPNOTSUPP);
675 }
676 
677 /* end of vfs default ops */
678