xref: /freebsd/sys/kern/vfs_vnops.c (revision c1462236787ec09d00d5e2d222edc3e34bce1e69)
1 /*
2  * Copyright (c) 1982, 1986, 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
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  *	@(#)vfs_vnops.c	8.2 (Berkeley) 1/21/94
39  * $FreeBSD$
40  */
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/fcntl.h>
45 #include <sys/file.h>
46 #include <sys/stat.h>
47 #include <sys/proc.h>
48 #include <sys/mount.h>
49 #include <sys/namei.h>
50 #include <sys/vnode.h>
51 #include <sys/bio.h>
52 #include <sys/buf.h>
53 #include <sys/filio.h>
54 #include <sys/ttycom.h>
55 #include <sys/conf.h>
56 
57 #include <ufs/ufs/quota.h>
58 #include <ufs/ufs/inode.h>
59 
60 static int vn_closefile __P((struct file *fp, struct proc *p));
61 static int vn_ioctl __P((struct file *fp, u_long com, caddr_t data,
62 		struct proc *p));
63 static int vn_read __P((struct file *fp, struct uio *uio,
64 		struct ucred *cred, int flags, struct proc *p));
65 static int vn_poll __P((struct file *fp, int events, struct ucred *cred,
66 		struct proc *p));
67 static int vn_statfile __P((struct file *fp, struct stat *sb, struct proc *p));
68 static int vn_write __P((struct file *fp, struct uio *uio,
69 		struct ucred *cred, int flags, struct proc *p));
70 
71 struct 	fileops vnops =
72 	{ vn_read, vn_write, vn_ioctl, vn_poll, vn_statfile, vn_closefile };
73 
74 static int	filt_nullattach(struct knote *kn);
75 static int	filt_vnattach(struct knote *kn);
76 static void	filt_vndetach(struct knote *kn);
77 static int	filt_vnode(struct knote *kn, long hint);
78 static int	filt_vnread(struct knote *kn, long hint);
79 
80 struct filterops vn_filtops =
81 	{ 1, filt_vnattach, filt_vndetach, filt_vnode };
82 
83 /*
84  * XXX
85  * filt_vnread is ufs-specific, so the attach routine should really
86  * switch out to different filterops based on the vn filetype
87  */
88 struct filterops vn_rwfiltops[] = {
89 	{ 1, filt_vnattach, filt_vndetach, filt_vnread },
90 	{ 1, filt_nullattach, NULL, NULL },
91 };
92 
93 /*
94  * Common code for vnode open operations.
95  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
96  *
97  * Note that this does NOT free nameidata for the successful case,
98  * due to the NDINIT being done elsewhere.
99  */
100 int
101 vn_open(ndp, flagp, cmode)
102 	register struct nameidata *ndp;
103 	int *flagp, cmode;
104 {
105 	struct vnode *vp;
106 	struct mount *mp;
107 	struct proc *p = ndp->ni_cnd.cn_proc;
108 	struct ucred *cred = p->p_ucred;
109 	struct vattr vat;
110 	struct vattr *vap = &vat;
111 	int mode, fmode, error;
112 
113 restart:
114 	fmode = *flagp;
115 	if (fmode & O_CREAT) {
116 		ndp->ni_cnd.cn_nameiop = CREATE;
117 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
118 		if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
119 			ndp->ni_cnd.cn_flags |= FOLLOW;
120 		bwillwrite();
121 		if ((error = namei(ndp)) != 0)
122 			return (error);
123 		if (ndp->ni_vp == NULL) {
124 			VATTR_NULL(vap);
125 			vap->va_type = VREG;
126 			vap->va_mode = cmode;
127 			if (fmode & O_EXCL)
128 				vap->va_vaflags |= VA_EXCLUSIVE;
129 			if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) {
130 				NDFREE(ndp, NDF_ONLY_PNBUF);
131 				vput(ndp->ni_dvp);
132 				if ((error = vn_start_write(NULL, &mp,
133 				    V_XSLEEP | PCATCH)) != 0)
134 					return (error);
135 				goto restart;
136 			}
137 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
138 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
139 					   &ndp->ni_cnd, vap);
140 			vput(ndp->ni_dvp);
141 			vn_finished_write(mp);
142 			if (error) {
143 				NDFREE(ndp, NDF_ONLY_PNBUF);
144 				return (error);
145 			}
146 			ASSERT_VOP_UNLOCKED(ndp->ni_dvp, "create");
147 			ASSERT_VOP_LOCKED(ndp->ni_vp, "create");
148 			fmode &= ~O_TRUNC;
149 			vp = ndp->ni_vp;
150 		} else {
151 			if (ndp->ni_dvp == ndp->ni_vp)
152 				vrele(ndp->ni_dvp);
153 			else
154 				vput(ndp->ni_dvp);
155 			ndp->ni_dvp = NULL;
156 			vp = ndp->ni_vp;
157 			if (fmode & O_EXCL) {
158 				error = EEXIST;
159 				goto bad;
160 			}
161 			fmode &= ~O_CREAT;
162 		}
163 	} else {
164 		ndp->ni_cnd.cn_nameiop = LOOKUP;
165 		ndp->ni_cnd.cn_flags =
166 		    ((fmode & O_NOFOLLOW) ? NOFOLLOW : FOLLOW) | LOCKLEAF;
167 		if ((error = namei(ndp)) != 0)
168 			return (error);
169 		vp = ndp->ni_vp;
170 	}
171 	if (vp->v_type == VLNK) {
172 		error = EMLINK;
173 		goto bad;
174 	}
175 	if (vp->v_type == VSOCK) {
176 		error = EOPNOTSUPP;
177 		goto bad;
178 	}
179 	if ((fmode & O_CREAT) == 0) {
180 		mode = 0;
181 		if (fmode & (FWRITE | O_TRUNC)) {
182 			if (vp->v_type == VDIR) {
183 				error = EISDIR;
184 				goto bad;
185 			}
186 			error = vn_writechk(vp);
187 			if (error)
188 				goto bad;
189 			mode |= VWRITE;
190 		}
191 		if (fmode & FREAD)
192 			mode |= VREAD;
193 		if (mode) {
194 		        error = VOP_ACCESS(vp, mode, cred, p);
195 			if (error)
196 				goto bad;
197 		}
198 	}
199 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
200 		goto bad;
201 	/*
202 	 * Make sure that a VM object is created for VMIO support.
203 	 */
204 	if (vn_canvmio(vp) == TRUE) {
205 		if ((error = vfs_object_create(vp, p, cred)) != 0)
206 			goto bad;
207 	}
208 
209 	if (fmode & FWRITE)
210 		vp->v_writecount++;
211 	*flagp = fmode;
212 	return (0);
213 bad:
214 	NDFREE(ndp, NDF_ONLY_PNBUF);
215 	vput(vp);
216 	*flagp = fmode;
217 	return (error);
218 }
219 
220 /*
221  * Check for write permissions on the specified vnode.
222  * Prototype text segments cannot be written.
223  */
224 int
225 vn_writechk(vp)
226 	register struct vnode *vp;
227 {
228 
229 	/*
230 	 * If there's shared text associated with
231 	 * the vnode, try to free it up once.  If
232 	 * we fail, we can't allow writing.
233 	 */
234 	if (vp->v_flag & VTEXT)
235 		return (ETXTBSY);
236 	return (0);
237 }
238 
239 /*
240  * Vnode close call
241  */
242 int
243 vn_close(vp, flags, cred, p)
244 	register struct vnode *vp;
245 	int flags;
246 	struct ucred *cred;
247 	struct proc *p;
248 {
249 	int error;
250 
251 	if (flags & FWRITE)
252 		vp->v_writecount--;
253 	error = VOP_CLOSE(vp, flags, cred, p);
254 	vrele(vp);
255 	return (error);
256 }
257 
258 static __inline
259 int
260 sequential_heuristic(struct uio *uio, struct file *fp)
261 {
262 	/*
263 	 * Sequential heuristic - detect sequential operation
264 	 */
265 	if ((uio->uio_offset == 0 && fp->f_seqcount > 0) ||
266 	    uio->uio_offset == fp->f_nextoff) {
267 		/*
268 		 * XXX we assume that the filesystem block size is
269 		 * the default.  Not true, but still gives us a pretty
270 		 * good indicator of how sequential the read operations
271 		 * are.
272 		 */
273 		fp->f_seqcount += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE;
274 		if (fp->f_seqcount >= 127)
275 			fp->f_seqcount = 127;
276 		return(fp->f_seqcount << 16);
277 	}
278 
279 	/*
280 	 * Not sequential, quick draw-down of seqcount
281 	 */
282 	if (fp->f_seqcount > 1)
283 		fp->f_seqcount = 1;
284 	else
285 		fp->f_seqcount = 0;
286 	return(0);
287 }
288 
289 /*
290  * Package up an I/O request on a vnode into a uio and do it.
291  */
292 int
293 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p)
294 	enum uio_rw rw;
295 	struct vnode *vp;
296 	caddr_t base;
297 	int len;
298 	off_t offset;
299 	enum uio_seg segflg;
300 	int ioflg;
301 	struct ucred *cred;
302 	int *aresid;
303 	struct proc *p;
304 {
305 	struct uio auio;
306 	struct iovec aiov;
307 	struct mount *mp;
308 	int error;
309 
310 	if ((ioflg & IO_NODELOCKED) == 0) {
311 		mp = NULL;
312 		if (rw == UIO_WRITE &&
313 		    vp->v_type != VCHR && vp->v_type != VBLK &&
314 		    (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
315 			return (error);
316 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
317 	}
318 	auio.uio_iov = &aiov;
319 	auio.uio_iovcnt = 1;
320 	aiov.iov_base = base;
321 	aiov.iov_len = len;
322 	auio.uio_resid = len;
323 	auio.uio_offset = offset;
324 	auio.uio_segflg = segflg;
325 	auio.uio_rw = rw;
326 	auio.uio_procp = p;
327 	if (rw == UIO_READ) {
328 		error = VOP_READ(vp, &auio, ioflg, cred);
329 	} else {
330 		error = VOP_WRITE(vp, &auio, ioflg, cred);
331 	}
332 	if (aresid)
333 		*aresid = auio.uio_resid;
334 	else
335 		if (auio.uio_resid && error == 0)
336 			error = EIO;
337 	if ((ioflg & IO_NODELOCKED) == 0) {
338 		vn_finished_write(mp);
339 		VOP_UNLOCK(vp, 0, p);
340 	}
341 	return (error);
342 }
343 
344 /*
345  * File table vnode read routine.
346  */
347 static int
348 vn_read(fp, uio, cred, flags, p)
349 	struct file *fp;
350 	struct uio *uio;
351 	struct ucred *cred;
352 	struct proc *p;
353 	int flags;
354 {
355 	struct vnode *vp;
356 	int error, ioflag;
357 
358 	KASSERT(uio->uio_procp == p, ("uio_procp %p is not p %p",
359 	    uio->uio_procp, p));
360 	vp = (struct vnode *)fp->f_data;
361 	ioflag = 0;
362 	if (fp->f_flag & FNONBLOCK)
363 		ioflag |= IO_NDELAY;
364 	VOP_LEASE(vp, p, cred, LEASE_READ);
365 	vn_lock(vp, LK_SHARED | LK_NOPAUSE | LK_RETRY, p);
366 	if ((flags & FOF_OFFSET) == 0)
367 		uio->uio_offset = fp->f_offset;
368 
369 	ioflag |= sequential_heuristic(uio, fp);
370 
371 	error = VOP_READ(vp, uio, ioflag, cred);
372 	if ((flags & FOF_OFFSET) == 0)
373 		fp->f_offset = uio->uio_offset;
374 	fp->f_nextoff = uio->uio_offset;
375 	VOP_UNLOCK(vp, 0, p);
376 	return (error);
377 }
378 
379 /*
380  * File table vnode write routine.
381  */
382 static int
383 vn_write(fp, uio, cred, flags, p)
384 	struct file *fp;
385 	struct uio *uio;
386 	struct ucred *cred;
387 	struct proc *p;
388 	int flags;
389 {
390 	struct vnode *vp;
391 	struct mount *mp;
392 	int error, ioflag;
393 
394 	KASSERT(uio->uio_procp == p, ("uio_procp %p is not p %p",
395 	    uio->uio_procp, p));
396 	vp = (struct vnode *)fp->f_data;
397 	if (vp->v_type == VREG)
398 		bwillwrite();
399 	vp = (struct vnode *)fp->f_data;	/* XXX needed? */
400 	ioflag = IO_UNIT;
401 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
402 		ioflag |= IO_APPEND;
403 	if (fp->f_flag & FNONBLOCK)
404 		ioflag |= IO_NDELAY;
405 	if ((fp->f_flag & O_FSYNC) ||
406 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
407 		ioflag |= IO_SYNC;
408 	mp = NULL;
409 	if (vp->v_type != VCHR && vp->v_type != VBLK &&
410 	    (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
411 		return (error);
412 	VOP_LEASE(vp, p, cred, LEASE_WRITE);
413 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
414 	if ((flags & FOF_OFFSET) == 0)
415 		uio->uio_offset = fp->f_offset;
416 	ioflag |= sequential_heuristic(uio, fp);
417 	error = VOP_WRITE(vp, uio, ioflag, cred);
418 	if ((flags & FOF_OFFSET) == 0)
419 		fp->f_offset = uio->uio_offset;
420 	fp->f_nextoff = uio->uio_offset;
421 	VOP_UNLOCK(vp, 0, p);
422 	vn_finished_write(mp);
423 	return (error);
424 }
425 
426 /*
427  * File table vnode stat routine.
428  */
429 static int
430 vn_statfile(fp, sb, p)
431 	struct file *fp;
432 	struct stat *sb;
433 	struct proc *p;
434 {
435 	struct vnode *vp = (struct vnode *)fp->f_data;
436 
437 	return vn_stat(vp, sb, p);
438 }
439 
440 int
441 vn_stat(vp, sb, p)
442 	struct vnode *vp;
443 	register struct stat *sb;
444 	struct proc *p;
445 {
446 	struct vattr vattr;
447 	register struct vattr *vap;
448 	int error;
449 	u_short mode;
450 
451 	vap = &vattr;
452 	error = VOP_GETATTR(vp, vap, p->p_ucred, p);
453 	if (error)
454 		return (error);
455 
456 	/*
457 	 * Zero the spare stat fields
458 	 */
459 	sb->st_lspare = 0;
460 	sb->st_qspare[0] = 0;
461 	sb->st_qspare[1] = 0;
462 
463 	/*
464 	 * Copy from vattr table
465 	 */
466 	if (vap->va_fsid != VNOVAL)
467 		sb->st_dev = vap->va_fsid;
468 	else
469 		sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
470 	sb->st_ino = vap->va_fileid;
471 	mode = vap->va_mode;
472 	switch (vap->va_type) {
473 	case VREG:
474 		mode |= S_IFREG;
475 		break;
476 	case VDIR:
477 		mode |= S_IFDIR;
478 		break;
479 	case VBLK:
480 		mode |= S_IFBLK;
481 		break;
482 	case VCHR:
483 		mode |= S_IFCHR;
484 		break;
485 	case VLNK:
486 		mode |= S_IFLNK;
487 		/* This is a cosmetic change, symlinks do not have a mode. */
488 		if (vp->v_mount->mnt_flag & MNT_NOSYMFOLLOW)
489 			sb->st_mode &= ~ACCESSPERMS;	/* 0000 */
490 		else
491 			sb->st_mode |= ACCESSPERMS;	/* 0777 */
492 		break;
493 	case VSOCK:
494 		mode |= S_IFSOCK;
495 		break;
496 	case VFIFO:
497 		mode |= S_IFIFO;
498 		break;
499 	default:
500 		return (EBADF);
501 	};
502 	sb->st_mode = mode;
503 	sb->st_nlink = vap->va_nlink;
504 	sb->st_uid = vap->va_uid;
505 	sb->st_gid = vap->va_gid;
506 	sb->st_rdev = vap->va_rdev;
507 	sb->st_size = vap->va_size;
508 	sb->st_atimespec = vap->va_atime;
509 	sb->st_mtimespec = vap->va_mtime;
510 	sb->st_ctimespec = vap->va_ctime;
511 
512         /*
513 	 * According to www.opengroup.org, the meaning of st_blksize is
514 	 *   "a filesystem-specific preferred I/O block size for this
515 	 *    object.  In some filesystem types, this may vary from file
516 	 *    to file"
517 	 * Default to zero to catch bogus uses of this field.
518 	 */
519 
520 	if (vap->va_type == VREG) {
521 		sb->st_blksize = vap->va_blocksize;
522 	} else if (vn_isdisk(vp, NULL)) {
523 		sb->st_blksize = vp->v_rdev->si_bsize_best;
524 		if (sb->st_blksize < vp->v_rdev->si_bsize_phys)
525 			sb->st_blksize = vp->v_rdev->si_bsize_phys;
526 		if (sb->st_blksize < BLKDEV_IOSIZE)
527 			sb->st_blksize = BLKDEV_IOSIZE;
528 	} else {
529 		sb->st_blksize = 0;
530 	}
531 
532 	sb->st_flags = vap->va_flags;
533 	if (suser_xxx(p->p_ucred, 0, 0))
534 		sb->st_gen = 0;
535 	else
536 		sb->st_gen = vap->va_gen;
537 
538 #if (S_BLKSIZE == 512)
539 	/* Optimize this case */
540 	sb->st_blocks = vap->va_bytes >> 9;
541 #else
542 	sb->st_blocks = vap->va_bytes / S_BLKSIZE;
543 #endif
544 	return (0);
545 }
546 
547 /*
548  * File table vnode ioctl routine.
549  */
550 static int
551 vn_ioctl(fp, com, data, p)
552 	struct file *fp;
553 	u_long com;
554 	caddr_t data;
555 	struct proc *p;
556 {
557 	register struct vnode *vp = ((struct vnode *)fp->f_data);
558 	struct vattr vattr;
559 	int error;
560 
561 	switch (vp->v_type) {
562 
563 	case VREG:
564 	case VDIR:
565 		if (com == FIONREAD) {
566 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
567 			if (error)
568 				return (error);
569 			*(int *)data = vattr.va_size - fp->f_offset;
570 			return (0);
571 		}
572 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
573 			return (0);			/* XXX */
574 		/* fall into ... */
575 
576 	default:
577 #if 0
578 		return (ENOTTY);
579 #endif
580 	case VFIFO:
581 	case VCHR:
582 	case VBLK:
583 		if (com == FIODTYPE) {
584 			if (vp->v_type != VCHR && vp->v_type != VBLK)
585 				return (ENOTTY);
586 			*(int *)data = devsw(vp->v_rdev)->d_flags & D_TYPEMASK;
587 			return (0);
588 		}
589 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
590 		if (error == 0 && com == TIOCSCTTY) {
591 
592 			/* Do nothing if reassigning same control tty */
593 			if (p->p_session->s_ttyvp == vp)
594 				return (0);
595 
596 			/* Get rid of reference to old control tty */
597 			if (p->p_session->s_ttyvp)
598 				vrele(p->p_session->s_ttyvp);
599 
600 			p->p_session->s_ttyvp = vp;
601 			VREF(vp);
602 		}
603 		return (error);
604 	}
605 }
606 
607 /*
608  * File table vnode poll routine.
609  */
610 static int
611 vn_poll(fp, events, cred, p)
612 	struct file *fp;
613 	int events;
614 	struct ucred *cred;
615 	struct proc *p;
616 {
617 
618 	return (VOP_POLL(((struct vnode *)fp->f_data), events, cred, p));
619 }
620 
621 /*
622  * Check that the vnode is still valid, and if so
623  * acquire requested lock.
624  */
625 int
626 #ifndef	DEBUG_LOCKS
627 vn_lock(vp, flags, p)
628 #else
629 debug_vn_lock(vp, flags, p, filename, line)
630 #endif
631 	struct vnode *vp;
632 	int flags;
633 	struct proc *p;
634 #ifdef	DEBUG_LOCKS
635 	const char *filename;
636 	int line;
637 #endif
638 {
639 	int error;
640 
641 	do {
642 		if ((flags & LK_INTERLOCK) == 0)
643 			simple_lock(&vp->v_interlock);
644 		if (vp->v_flag & VXLOCK) {
645 			vp->v_flag |= VXWANT;
646 			simple_unlock(&vp->v_interlock);
647 			tsleep((caddr_t)vp, PINOD, "vn_lock", 0);
648 			error = ENOENT;
649 		} else {
650 #ifdef	DEBUG_LOCKS
651 			vp->filename = filename;
652 			vp->line = line;
653 #endif
654 			error = VOP_LOCK(vp,
655 				    flags | LK_NOPAUSE | LK_INTERLOCK, p);
656 			if (error == 0)
657 				return (error);
658 		}
659 		flags &= ~LK_INTERLOCK;
660 	} while (flags & LK_RETRY);
661 	return (error);
662 }
663 
664 /*
665  * File table vnode close routine.
666  */
667 static int
668 vn_closefile(fp, p)
669 	struct file *fp;
670 	struct proc *p;
671 {
672 
673 	fp->f_ops = &badfileops;
674 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
675 		fp->f_cred, p));
676 }
677 
678 /*
679  * Preparing to start a filesystem write operation. If the operation is
680  * permitted, then we bump the count of operations in progress and
681  * proceed. If a suspend request is in progress, we wait until the
682  * suspension is over, and then proceed.
683  */
684 int
685 vn_start_write(vp, mpp, flags)
686 	struct vnode *vp;
687 	struct mount **mpp;
688 	int flags;
689 {
690 	struct mount *mp;
691 	int error;
692 
693 	/*
694 	 * If a vnode is provided, get and return the mount point that
695 	 * to which it will write.
696 	 */
697 	if (vp != NULL) {
698 		if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) {
699 			*mpp = NULL;
700 			if (error != EOPNOTSUPP)
701 				return (error);
702 			return (0);
703 		}
704 	}
705 	if ((mp = *mpp) == NULL)
706 		return (0);
707 	/*
708 	 * Check on status of suspension.
709 	 */
710 	while ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
711 		if (flags & V_NOWAIT)
712 			return (EWOULDBLOCK);
713 		error = tsleep(&mp->mnt_flag, (PUSER - 1) | (flags & PCATCH),
714 		    "suspfs", 0);
715 		if (error)
716 			return (error);
717 	}
718 	if (flags & V_XSLEEP)
719 		return (0);
720 	mp->mnt_writeopcount++;
721 	return (0);
722 }
723 
724 /*
725  * Secondary suspension. Used by operations such as vop_inactive
726  * routines that are needed by the higher level functions. These
727  * are allowed to proceed until all the higher level functions have
728  * completed (indicated by mnt_writeopcount dropping to zero). At that
729  * time, these operations are halted until the suspension is over.
730  */
731 int
732 vn_write_suspend_wait(vp, mp, flags)
733 	struct vnode *vp;
734 	struct mount *mp;
735 	int flags;
736 {
737 	int error;
738 
739 	if (vp != NULL) {
740 		if ((error = VOP_GETWRITEMOUNT(vp, &mp)) != 0) {
741 			if (error != EOPNOTSUPP)
742 				return (error);
743 			return (0);
744 		}
745 	}
746 	/*
747 	 * If we are not suspended or have not yet reached suspended
748 	 * mode, then let the operation proceed.
749 	 */
750 	if (mp == NULL || (mp->mnt_kern_flag & MNTK_SUSPENDED) == 0)
751 		return (0);
752 	if (flags & V_NOWAIT)
753 		return (EWOULDBLOCK);
754 	/*
755 	 * Wait for the suspension to finish.
756 	 */
757 	return (tsleep(&mp->mnt_flag, (PUSER - 1) | (flags & PCATCH),
758 	    "suspfs", 0));
759 }
760 
761 /*
762  * Filesystem write operation has completed. If we are suspending and this
763  * operation is the last one, notify the suspender that the suspension is
764  * now in effect.
765  */
766 void
767 vn_finished_write(mp)
768 	struct mount *mp;
769 {
770 
771 	if (mp == NULL)
772 		return;
773 	mp->mnt_writeopcount--;
774 	if (mp->mnt_writeopcount < 0)
775 		panic("vn_finished_write: neg cnt");
776 	if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 &&
777 	    mp->mnt_writeopcount <= 0)
778 		wakeup(&mp->mnt_writeopcount);
779 }
780 
781 /*
782  * Request a filesystem to suspend write operations.
783  */
784 void
785 vfs_write_suspend(mp)
786 	struct mount *mp;
787 {
788 	struct proc *p = curproc;
789 
790 	if (mp->mnt_kern_flag & MNTK_SUSPEND)
791 		return;
792 	mp->mnt_kern_flag |= MNTK_SUSPEND;
793 	if (mp->mnt_writeopcount > 0)
794 		(void) tsleep(&mp->mnt_writeopcount, PUSER - 1, "suspwt", 0);
795 	VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p);
796 	mp->mnt_kern_flag |= MNTK_SUSPENDED;
797 }
798 
799 /*
800  * Request a filesystem to resume write operations.
801  */
802 void
803 vfs_write_resume(mp)
804 	struct mount *mp;
805 {
806 
807 	if ((mp->mnt_kern_flag & MNTK_SUSPEND) == 0)
808 		return;
809 	mp->mnt_kern_flag &= ~(MNTK_SUSPEND | MNTK_SUSPENDED);
810 	wakeup(&mp->mnt_writeopcount);
811 	wakeup(&mp->mnt_flag);
812 }
813 
814 static int
815 filt_vnattach(struct knote *kn)
816 {
817 	struct vnode *vp;
818 
819 	if (kn->kn_fp->f_type != DTYPE_VNODE &&
820 	    kn->kn_fp->f_type != DTYPE_FIFO)
821 		return (EBADF);
822 
823 	vp = (struct vnode *)kn->kn_fp->f_data;
824 
825 	/*
826 	 * XXX
827 	 * this is a hack simply to cause the filter attach to fail
828 	 * for non-ufs filesystems, until the support for them is done.
829 	 */
830 	if ((vp)->v_tag != VT_UFS)
831 		return (EOPNOTSUPP);
832 
833         simple_lock(&vp->v_pollinfo.vpi_lock);
834 	SLIST_INSERT_HEAD(&vp->v_pollinfo.vpi_selinfo.si_note, kn, kn_selnext);
835         simple_unlock(&vp->v_pollinfo.vpi_lock);
836 
837 	return (0);
838 }
839 
840 static void
841 filt_vndetach(struct knote *kn)
842 {
843 	struct vnode *vp = (struct vnode *)kn->kn_fp->f_data;
844 
845         simple_lock(&vp->v_pollinfo.vpi_lock);
846 	SLIST_REMOVE(&vp->v_pollinfo.vpi_selinfo.si_note,
847 	    kn, knote, kn_selnext);
848         simple_unlock(&vp->v_pollinfo.vpi_lock);
849 }
850 
851 static int
852 filt_vnode(struct knote *kn, long hint)
853 {
854 
855 	if (kn->kn_sfflags & hint)
856 		kn->kn_fflags |= hint;
857 	return (kn->kn_fflags != 0);
858 }
859 
860 static int
861 filt_nullattach(struct knote *kn)
862 {
863 	return (ENXIO);
864 }
865 
866 /*ARGSUSED*/
867 static int
868 filt_vnread(struct knote *kn, long hint)
869 {
870 	struct vnode *vp = (struct vnode *)kn->kn_fp->f_data;
871 	struct inode *ip = VTOI(vp);
872 
873 	kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
874 	return (kn->kn_data != 0);
875 }
876