xref: /freebsd/sys/kern/vfs_default.c (revision f856af0466c076beef4ea9b15d088e1119a945b8)
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 		char *file;
260 		int line;
261 	} */ *ap;
262 {
263 	struct vnode *vp = ap->a_vp;
264 
265 	return (_lockmgr(vp->v_vnlock, ap->a_flags, VI_MTX(vp), ap->a_td, ap->a_file, ap->a_line));
266 }
267 
268 /* See above. */
269 int
270 vop_stdunlock(ap)
271 	struct vop_unlock_args /* {
272 		struct vnode *a_vp;
273 		int a_flags;
274 		struct thread *a_td;
275 	} */ *ap;
276 {
277 	struct vnode *vp = ap->a_vp;
278 
279 	return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp),
280 	    ap->a_td));
281 }
282 
283 /* See above. */
284 int
285 vop_stdislocked(ap)
286 	struct vop_islocked_args /* {
287 		struct vnode *a_vp;
288 		struct thread *a_td;
289 	} */ *ap;
290 {
291 
292 	return (lockstatus(ap->a_vp->v_vnlock, ap->a_td));
293 }
294 
295 /*
296  * Return true for select/poll.
297  */
298 int
299 vop_nopoll(ap)
300 	struct vop_poll_args /* {
301 		struct vnode *a_vp;
302 		int  a_events;
303 		struct ucred *a_cred;
304 		struct thread *a_td;
305 	} */ *ap;
306 {
307 	/*
308 	 * Return true for read/write.  If the user asked for something
309 	 * special, return POLLNVAL, so that clients have a way of
310 	 * determining reliably whether or not the extended
311 	 * functionality is present without hard-coding knowledge
312 	 * of specific filesystem implementations.
313 	 * Stay in sync with kern_conf.c::no_poll().
314 	 */
315 	if (ap->a_events & ~POLLSTANDARD)
316 		return (POLLNVAL);
317 
318 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
319 }
320 
321 /*
322  * Implement poll for local filesystems that support it.
323  */
324 int
325 vop_stdpoll(ap)
326 	struct vop_poll_args /* {
327 		struct vnode *a_vp;
328 		int  a_events;
329 		struct ucred *a_cred;
330 		struct thread *a_td;
331 	} */ *ap;
332 {
333 	if (ap->a_events & ~POLLSTANDARD)
334 		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
335 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
336 }
337 
338 /*
339  * Return our mount point, as we will take charge of the writes.
340  */
341 int
342 vop_stdgetwritemount(ap)
343 	struct vop_getwritemount_args /* {
344 		struct vnode *a_vp;
345 		struct mount **a_mpp;
346 	} */ *ap;
347 {
348 	struct mount *mp;
349 
350 	/*
351 	 * XXX Since this is called unlocked we may be recycled while
352 	 * attempting to ref the mount.  If this is the case or mountpoint
353 	 * will be set to NULL.  We only have to prevent this call from
354 	 * returning with a ref to an incorrect mountpoint.  It is not
355 	 * harmful to return with a ref to our previous mountpoint.
356 	 */
357 	mp = ap->a_vp->v_mount;
358 	if (mp != NULL) {
359 		vfs_ref(mp);
360 		if (mp != ap->a_vp->v_mount) {
361 			vfs_rel(mp);
362 			mp = NULL;
363 		}
364 	}
365 	*(ap->a_mpp) = mp;
366 	return (0);
367 }
368 
369 /* XXX Needs good comment and VOP_BMAP(9) manpage */
370 int
371 vop_stdbmap(ap)
372 	struct vop_bmap_args /* {
373 		struct vnode *a_vp;
374 		daddr_t  a_bn;
375 		struct bufobj **a_bop;
376 		daddr_t *a_bnp;
377 		int *a_runp;
378 		int *a_runb;
379 	} */ *ap;
380 {
381 
382 	if (ap->a_bop != NULL)
383 		*ap->a_bop = &ap->a_vp->v_bufobj;
384 	if (ap->a_bnp != NULL)
385 		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
386 	if (ap->a_runp != NULL)
387 		*ap->a_runp = 0;
388 	if (ap->a_runb != NULL)
389 		*ap->a_runb = 0;
390 	return (0);
391 }
392 
393 int
394 vop_stdfsync(ap)
395 	struct vop_fsync_args /* {
396 		struct vnode *a_vp;
397 		struct ucred *a_cred;
398 		int a_waitfor;
399 		struct thread *a_td;
400 	} */ *ap;
401 {
402 	struct vnode *vp = ap->a_vp;
403 	struct buf *bp;
404 	struct bufobj *bo;
405 	struct buf *nbp;
406 	int error = 0;
407 	int maxretry = 1000;     /* large, arbitrarily chosen */
408 
409 	VI_LOCK(vp);
410 loop1:
411 	/*
412 	 * MARK/SCAN initialization to avoid infinite loops.
413 	 */
414         TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs) {
415                 bp->b_vflags &= ~BV_SCANNED;
416 		bp->b_error = 0;
417 	}
418 
419 	/*
420 	 * Flush all dirty buffers associated with a vnode.
421 	 */
422 loop2:
423 	TAILQ_FOREACH_SAFE(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs, nbp) {
424 		if ((bp->b_vflags & BV_SCANNED) != 0)
425 			continue;
426 		bp->b_vflags |= BV_SCANNED;
427 		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
428 			continue;
429 		VI_UNLOCK(vp);
430 		KASSERT(bp->b_bufobj == &vp->v_bufobj,
431 		    ("bp %p wrong b_bufobj %p should be %p",
432 		    bp, bp->b_bufobj, &vp->v_bufobj));
433 		if ((bp->b_flags & B_DELWRI) == 0)
434 			panic("fsync: not dirty");
435 		if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
436 			vfs_bio_awrite(bp);
437 		} else {
438 			bremfree(bp);
439 			bawrite(bp);
440 		}
441 		VI_LOCK(vp);
442 		goto loop2;
443 	}
444 
445 	/*
446 	 * If synchronous the caller expects us to completely resolve all
447 	 * dirty buffers in the system.  Wait for in-progress I/O to
448 	 * complete (which could include background bitmap writes), then
449 	 * retry if dirty blocks still exist.
450 	 */
451 	if (ap->a_waitfor == MNT_WAIT) {
452 		bo = &vp->v_bufobj;
453 		bufobj_wwait(bo, 0, 0);
454 		if (bo->bo_dirty.bv_cnt > 0) {
455 			/*
456 			 * If we are unable to write any of these buffers
457 			 * then we fail now rather than trying endlessly
458 			 * to write them out.
459 			 */
460 			TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
461 				if ((error = bp->b_error) == 0)
462 					continue;
463 			if (error == 0 && --maxretry >= 0)
464 				goto loop1;
465 			error = EAGAIN;
466 		}
467 	}
468 	VI_UNLOCK(vp);
469 	if (error == EAGAIN)
470 		vprint("fsync: giving up on dirty", vp);
471 
472 	return (error);
473 }
474 
475 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
476 int
477 vop_stdgetpages(ap)
478 	struct vop_getpages_args /* {
479 		struct vnode *a_vp;
480 		vm_page_t *a_m;
481 		int a_count;
482 		int a_reqpage;
483 		vm_ooffset_t a_offset;
484 	} */ *ap;
485 {
486 
487 	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
488 	    ap->a_count, ap->a_reqpage);
489 }
490 
491 int
492 vop_stdkqfilter(struct vop_kqfilter_args *ap)
493 {
494 	return vfs_kqfilter(ap);
495 }
496 
497 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
498 int
499 vop_stdputpages(ap)
500 	struct vop_putpages_args /* {
501 		struct vnode *a_vp;
502 		vm_page_t *a_m;
503 		int a_count;
504 		int a_sync;
505 		int *a_rtvals;
506 		vm_ooffset_t a_offset;
507 	} */ *ap;
508 {
509 
510 	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
511 	     ap->a_sync, ap->a_rtvals);
512 }
513 
514 /*
515  * vfs default ops
516  * used to fill the vfs function table to get reasonable default return values.
517  */
518 int
519 vfs_stdroot (mp, flags, vpp, td)
520 	struct mount *mp;
521 	int flags;
522 	struct vnode **vpp;
523 	struct thread *td;
524 {
525 
526 	return (EOPNOTSUPP);
527 }
528 
529 int
530 vfs_stdstatfs (mp, sbp, td)
531 	struct mount *mp;
532 	struct statfs *sbp;
533 	struct thread *td;
534 {
535 
536 	return (EOPNOTSUPP);
537 }
538 
539 int
540 vfs_stdvptofh (vp, fhp)
541 	struct vnode *vp;
542 	struct fid *fhp;
543 {
544 
545 	return (EOPNOTSUPP);
546 }
547 
548 int
549 vfs_stdquotactl (mp, cmds, uid, arg, td)
550 	struct mount *mp;
551 	int cmds;
552 	uid_t uid;
553 	void *arg;
554 	struct thread *td;
555 {
556 
557 	return (EOPNOTSUPP);
558 }
559 
560 int
561 vfs_stdsync(mp, waitfor, td)
562 	struct mount *mp;
563 	int waitfor;
564 	struct thread *td;
565 {
566 	struct vnode *vp, *mvp;
567 	int error, lockreq, allerror = 0;
568 
569 	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
570 	if (waitfor != MNT_WAIT)
571 		lockreq |= LK_NOWAIT;
572 	/*
573 	 * Force stale buffer cache information to be flushed.
574 	 */
575 	MNT_ILOCK(mp);
576 loop:
577 	MNT_VNODE_FOREACH(vp, mp, mvp) {
578 
579 		VI_LOCK(vp);
580 		if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
581 			VI_UNLOCK(vp);
582 			continue;
583 		}
584 		MNT_IUNLOCK(mp);
585 
586 		if ((error = vget(vp, lockreq, td)) != 0) {
587 			MNT_ILOCK(mp);
588 			if (error == ENOENT) {
589 				MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
590 				goto loop;
591 			}
592 			continue;
593 		}
594 		error = VOP_FSYNC(vp, waitfor, td);
595 		if (error)
596 			allerror = error;
597 
598 		/* Do not turn this into vput.  td is not always curthread. */
599 		VOP_UNLOCK(vp, 0, td);
600 		vrele(vp);
601 		MNT_ILOCK(mp);
602 	}
603 	MNT_IUNLOCK(mp);
604 	return (allerror);
605 }
606 
607 int
608 vfs_stdnosync (mp, waitfor, td)
609 	struct mount *mp;
610 	int waitfor;
611 	struct thread *td;
612 {
613 
614 	return (0);
615 }
616 
617 int
618 vfs_stdvget (mp, ino, flags, vpp)
619 	struct mount *mp;
620 	ino_t ino;
621 	int flags;
622 	struct vnode **vpp;
623 {
624 
625 	return (EOPNOTSUPP);
626 }
627 
628 int
629 vfs_stdfhtovp (mp, fhp, vpp)
630 	struct mount *mp;
631 	struct fid *fhp;
632 	struct vnode **vpp;
633 {
634 
635 	return (EOPNOTSUPP);
636 }
637 
638 int
639 vfs_stdinit (vfsp)
640 	struct vfsconf *vfsp;
641 {
642 
643 	return (0);
644 }
645 
646 int
647 vfs_stduninit (vfsp)
648 	struct vfsconf *vfsp;
649 {
650 
651 	return(0);
652 }
653 
654 int
655 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td)
656 	struct mount *mp;
657 	int cmd;
658 	struct vnode *filename_vp;
659 	int attrnamespace;
660 	const char *attrname;
661 	struct thread *td;
662 {
663 
664 	if (filename_vp != NULL)
665 		VOP_UNLOCK(filename_vp, 0, td);
666 	return (EOPNOTSUPP);
667 }
668 
669 int
670 vfs_stdsysctl(mp, op, req)
671 	struct mount *mp;
672 	fsctlop_t op;
673 	struct sysctl_req *req;
674 {
675 
676 	return (EOPNOTSUPP);
677 }
678 
679 /* end of vfs default ops */
680