xref: /netbsd-src/sys/kern/vfs_vnops.c (revision 89c5a767f8fc7a4633b2d409966e2becbb98ff92)
1 /*	$NetBSD: vfs_vnops.c,v 1.39 2000/02/14 22:00:21 fvdl Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *	This product includes software developed by the University of
23  *	California, Berkeley and its contributors.
24  * 4. Neither the name of the University nor the names of its contributors
25  *    may be used to endorse or promote products derived from this software
26  *    without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38  * SUCH DAMAGE.
39  *
40  *	@(#)vfs_vnops.c	8.14 (Berkeley) 6/15/95
41  */
42 
43 #include "fs_union.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/kernel.h>
48 #include <sys/file.h>
49 #include <sys/stat.h>
50 #include <sys/buf.h>
51 #include <sys/proc.h>
52 #include <sys/mount.h>
53 #include <sys/namei.h>
54 #include <sys/vnode.h>
55 #include <sys/ioctl.h>
56 #include <sys/tty.h>
57 #include <sys/poll.h>
58 
59 #include <vm/vm.h>
60 
61 #include <uvm/uvm_extern.h>
62 
63 #ifdef UNION
64 #include <miscfs/union/union.h>
65 #endif
66 
67 struct 	fileops vnops =
68 	{ vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, vn_closefile };
69 
70 /*
71  * Common code for vnode open operations.
72  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
73  */
74 int
75 vn_open(ndp, fmode, cmode)
76 	register struct nameidata *ndp;
77 	int fmode, cmode;
78 {
79 	register struct vnode *vp;
80 	register struct proc *p = ndp->ni_cnd.cn_proc;
81 	register struct ucred *cred = p->p_ucred;
82 	struct vattr va;
83 	int error;
84 
85 	if (fmode & O_CREAT) {
86 		ndp->ni_cnd.cn_nameiop = CREATE;
87 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
88 		if ((fmode & O_EXCL) == 0 &&
89 		    ((fmode & FNOSYMLINK) == 0))
90 			ndp->ni_cnd.cn_flags |= FOLLOW;
91 		if ((error = namei(ndp)) != 0)
92 			return (error);
93 		if (ndp->ni_vp == NULL) {
94 			VATTR_NULL(&va);
95 			va.va_type = VREG;
96 			va.va_mode = cmode;
97 			if (fmode & O_EXCL)
98 				 va.va_vaflags |= VA_EXCLUSIVE;
99 			VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
100 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
101 					   &ndp->ni_cnd, &va);
102 			if (error)
103 				return (error);
104 			fmode &= ~O_TRUNC;
105 			vp = ndp->ni_vp;
106 		} else {
107 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
108 			if (ndp->ni_dvp == ndp->ni_vp)
109 				vrele(ndp->ni_dvp);
110 			else
111 				vput(ndp->ni_dvp);
112 			ndp->ni_dvp = NULL;
113 			vp = ndp->ni_vp;
114 			if (fmode & O_EXCL) {
115 				error = EEXIST;
116 				goto bad;
117 			}
118 			if (ndp->ni_vp->v_type == VLNK) {
119 				error = EFTYPE;
120 				goto bad;
121 			}
122 			fmode &= ~O_CREAT;
123 		}
124 	} else {
125 		ndp->ni_cnd.cn_nameiop = LOOKUP;
126 		ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF;
127 		if ((error = namei(ndp)) != 0)
128 			return (error);
129 		vp = ndp->ni_vp;
130 	}
131 	if (vp->v_type == VSOCK) {
132 		error = EOPNOTSUPP;
133 		goto bad;
134 	}
135 	if ((fmode & O_CREAT) == 0) {
136 		if (fmode & FREAD) {
137 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
138 				goto bad;
139 		}
140 		if (fmode & (FWRITE | O_TRUNC)) {
141 			if (vp->v_type == VDIR) {
142 				error = EISDIR;
143 				goto bad;
144 			}
145 			if ((error = vn_writechk(vp)) != 0 ||
146 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
147 				goto bad;
148 		}
149 	}
150 	if (fmode & O_TRUNC) {
151 		VOP_UNLOCK(vp, 0);			/* XXX */
152 		VOP_LEASE(vp, p, cred, LEASE_WRITE);
153 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
154 		VATTR_NULL(&va);
155 		va.va_size = 0;
156 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
157 			goto bad;
158 	}
159 	if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
160 		goto bad;
161 	if (fmode & FWRITE)
162 		vp->v_writecount++;
163 	return (0);
164 bad:
165 	vput(vp);
166 	return (error);
167 }
168 
169 /*
170  * Check for write permissions on the specified vnode.
171  * Prototype text segments cannot be written.
172  */
173 int
174 vn_writechk(vp)
175 	register struct vnode *vp;
176 {
177 
178 	/*
179 	 * If there's shared text associated with
180 	 * the vnode, try to free it up once.  If
181 	 * we fail, we can't allow writing.
182 	 */
183 	if ((vp->v_flag & VTEXT) && !uvm_vnp_uncache(vp))
184 		return (ETXTBSY);
185 	return (0);
186 }
187 
188 /*
189  * Vnode close call
190  *
191  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
192  */
193 int
194 vn_close(vp, flags, cred, p)
195 	register struct vnode *vp;
196 	int flags;
197 	struct ucred *cred;
198 	struct proc *p;
199 {
200 	int error;
201 
202 	if (flags & FWRITE)
203 		vp->v_writecount--;
204 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
205 	error = VOP_CLOSE(vp, flags, cred, p);
206 	vput(vp);
207 	return (error);
208 }
209 
210 /*
211  * Package up an I/O request on a vnode into a uio and do it.
212  */
213 int
214 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p)
215 	enum uio_rw rw;
216 	struct vnode *vp;
217 	caddr_t base;
218 	int len;
219 	off_t offset;
220 	enum uio_seg segflg;
221 	int ioflg;
222 	struct ucred *cred;
223 	size_t *aresid;
224 	struct proc *p;
225 {
226 	struct uio auio;
227 	struct iovec aiov;
228 	int error;
229 
230 	if ((ioflg & IO_NODELOCKED) == 0)
231 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
232 	auio.uio_iov = &aiov;
233 	auio.uio_iovcnt = 1;
234 	aiov.iov_base = base;
235 	aiov.iov_len = len;
236 	auio.uio_resid = len;
237 	auio.uio_offset = offset;
238 	auio.uio_segflg = segflg;
239 	auio.uio_rw = rw;
240 	auio.uio_procp = p;
241 	if (rw == UIO_READ) {
242 		error = VOP_READ(vp, &auio, ioflg, cred);
243 	} else {
244 		error = VOP_WRITE(vp, &auio, ioflg, cred);
245 	}
246 	if (aresid)
247 		*aresid = auio.uio_resid;
248 	else
249 		if (auio.uio_resid && error == 0)
250 			error = EIO;
251 	if ((ioflg & IO_NODELOCKED) == 0)
252 		VOP_UNLOCK(vp, 0);
253 	return (error);
254 }
255 
256 int
257 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies)
258 	struct file *fp;
259 	char *buf;
260 	int segflg, *done, *ncookies;
261 	u_int count;
262 	struct proc *p;
263 	off_t **cookies;
264 {
265 	struct vnode *vp = (struct vnode *)fp->f_data;
266 	struct iovec aiov;
267 	struct uio auio;
268 	int error, eofflag;
269 
270 unionread:
271 	if (vp->v_type != VDIR)
272 		return (EINVAL);
273 	aiov.iov_base = buf;
274 	aiov.iov_len = count;
275 	auio.uio_iov = &aiov;
276 	auio.uio_iovcnt = 1;
277 	auio.uio_rw = UIO_READ;
278 	auio.uio_segflg = segflg;
279 	auio.uio_procp = p;
280 	auio.uio_resid = count;
281 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
282 	auio.uio_offset = fp->f_offset;
283 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
284 		    ncookies);
285 	fp->f_offset = auio.uio_offset;
286 	VOP_UNLOCK(vp, 0);
287 	if (error)
288 		return (error);
289 
290 #ifdef UNION
291 {
292 	extern int (**union_vnodeop_p) __P((void *));
293 	extern struct vnode *union_dircache __P((struct vnode *));
294 
295 	if (count == auio.uio_resid && (vp->v_op == union_vnodeop_p)) {
296 		struct vnode *lvp;
297 
298 		lvp = union_dircache(vp);
299 		if (lvp != NULLVP) {
300 			struct vattr va;
301 
302 			/*
303 			 * If the directory is opaque,
304 			 * then don't show lower entries
305 			 */
306 			error = VOP_GETATTR(vp, &va, fp->f_cred, p);
307 			if (va.va_flags & OPAQUE) {
308 				vput(lvp);
309 				lvp = NULL;
310 			}
311 		}
312 
313 		if (lvp != NULLVP) {
314 			error = VOP_OPEN(lvp, FREAD, fp->f_cred, p);
315 			if (error) {
316 				vput(lvp);
317 				return (error);
318 			}
319 			VOP_UNLOCK(lvp, 0);
320 			fp->f_data = (caddr_t) lvp;
321 			fp->f_offset = 0;
322 			error = vn_close(vp, FREAD, fp->f_cred, p);
323 			if (error)
324 				return (error);
325 			vp = lvp;
326 			goto unionread;
327 		}
328 	}
329 }
330 #endif /* UNION */
331 
332 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
333 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
334 		struct vnode *tvp = vp;
335 		vp = vp->v_mount->mnt_vnodecovered;
336 		VREF(vp);
337 		fp->f_data = (caddr_t) vp;
338 		fp->f_offset = 0;
339 		vrele(tvp);
340 		goto unionread;
341 	}
342 	*done = count - auio.uio_resid;
343 	return error;
344 }
345 
346 /*
347  * File table vnode read routine.
348  */
349 int
350 vn_read(fp, offset, uio, cred, flags)
351 	struct file *fp;
352 	off_t *offset;
353 	struct uio *uio;
354 	struct ucred *cred;
355 	int flags;
356 {
357 	struct vnode *vp = (struct vnode *)fp->f_data;
358 	int count, error, ioflag = 0;
359 
360 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
361 	if (fp->f_flag & FNONBLOCK)
362 		ioflag |= IO_NDELAY;
363 	if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
364 		ioflag |= IO_SYNC;
365 	if (fp->f_flag & FALTIO)
366 		ioflag |= IO_ALTSEMANTICS;
367 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
368 	uio->uio_offset = *offset;
369 	count = uio->uio_resid;
370 	error = VOP_READ(vp, uio, ioflag, cred);
371 	if (flags & FOF_UPDATE_OFFSET)
372 		*offset += count - uio->uio_resid;
373 	VOP_UNLOCK(vp, 0);
374 	return (error);
375 }
376 
377 /*
378  * File table vnode write routine.
379  */
380 int
381 vn_write(fp, offset, uio, cred, flags)
382 	struct file *fp;
383 	off_t *offset;
384 	struct uio *uio;
385 	struct ucred *cred;
386 	int flags;
387 {
388 	struct vnode *vp = (struct vnode *)fp->f_data;
389 	int count, error, ioflag = IO_UNIT;
390 
391 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
392 		ioflag |= IO_APPEND;
393 	if (fp->f_flag & FNONBLOCK)
394 		ioflag |= IO_NDELAY;
395 	if (fp->f_flag & FFSYNC ||
396 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
397 		ioflag |= IO_SYNC;
398 	else if (fp->f_flag & FDSYNC)
399 		ioflag |= IO_DSYNC;
400 	if (fp->f_flag & FALTIO)
401 		ioflag |= IO_ALTSEMANTICS;
402 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
403 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
404 	uio->uio_offset = *offset;
405 	count = uio->uio_resid;
406 	error = VOP_WRITE(vp, uio, ioflag, cred);
407 	if (flags & FOF_UPDATE_OFFSET) {
408 		if (ioflag & IO_APPEND)
409 			*offset = uio->uio_offset;
410 		else
411 			*offset += count - uio->uio_resid;
412 	}
413 	VOP_UNLOCK(vp, 0);
414 	return (error);
415 }
416 
417 /*
418  * File table vnode stat routine.
419  */
420 int
421 vn_stat(vp, sb, p)
422 	struct vnode *vp;
423 	register struct stat *sb;
424 	struct proc *p;
425 {
426 	struct vattr va;
427 	int error;
428 	mode_t mode;
429 
430 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
431 	if (error)
432 		return (error);
433 	/*
434 	 * Copy from vattr table
435 	 */
436 	sb->st_dev = va.va_fsid;
437 	sb->st_ino = va.va_fileid;
438 	mode = va.va_mode;
439 	switch (vp->v_type) {
440 	case VREG:
441 		mode |= S_IFREG;
442 		break;
443 	case VDIR:
444 		mode |= S_IFDIR;
445 		break;
446 	case VBLK:
447 		mode |= S_IFBLK;
448 		break;
449 	case VCHR:
450 		mode |= S_IFCHR;
451 		break;
452 	case VLNK:
453 		mode |= S_IFLNK;
454 		break;
455 	case VSOCK:
456 		mode |= S_IFSOCK;
457 		break;
458 	case VFIFO:
459 		mode |= S_IFIFO;
460 		break;
461 	default:
462 		return (EBADF);
463 	};
464 	sb->st_mode = mode;
465 	sb->st_nlink = va.va_nlink;
466 	sb->st_uid = va.va_uid;
467 	sb->st_gid = va.va_gid;
468 	sb->st_rdev = va.va_rdev;
469 	sb->st_size = va.va_size;
470 	sb->st_atimespec = va.va_atime;
471 	sb->st_mtimespec = va.va_mtime;
472 	sb->st_ctimespec = va.va_ctime;
473 	sb->st_blksize = va.va_blocksize;
474 	sb->st_flags = va.va_flags;
475 	sb->st_gen = 0;
476 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
477 	return (0);
478 }
479 
480 /*
481  * File table vnode fcntl routine.
482  */
483 int
484 vn_fcntl(fp, com, data, p)
485 	struct file *fp;
486 	u_int com;
487 	caddr_t data;
488 	struct proc *p;
489 {
490 	register struct vnode *vp = ((struct vnode *)fp->f_data);
491 	int error;
492 
493 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
494 	error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p);
495 	VOP_UNLOCK(vp, 0);
496 	return (error);
497 }
498 
499 /*
500  * File table vnode ioctl routine.
501  */
502 int
503 vn_ioctl(fp, com, data, p)
504 	struct file *fp;
505 	u_long com;
506 	caddr_t data;
507 	struct proc *p;
508 {
509 	register struct vnode *vp = ((struct vnode *)fp->f_data);
510 	struct vattr vattr;
511 	int error;
512 
513 	switch (vp->v_type) {
514 
515 	case VREG:
516 	case VDIR:
517 		if (com == FIONREAD) {
518 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
519 			if (error)
520 				return (error);
521 			*(int *)data = vattr.va_size - fp->f_offset;
522 			return (0);
523 		}
524 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
525 			return (0);			/* XXX */
526 		/* fall into ... */
527 
528 	default:
529 		return (ENOTTY);
530 
531 	case VFIFO:
532 	case VCHR:
533 	case VBLK:
534 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
535 		if (error == 0 && com == TIOCSCTTY) {
536 			if (p->p_session->s_ttyvp)
537 				vrele(p->p_session->s_ttyvp);
538 			p->p_session->s_ttyvp = vp;
539 			VREF(vp);
540 		}
541 		return (error);
542 	}
543 }
544 
545 /*
546  * File table vnode poll routine.
547  */
548 int
549 vn_poll(fp, events, p)
550 	struct file *fp;
551 	int events;
552 	struct proc *p;
553 {
554 
555 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
556 }
557 
558 /*
559  * Check that the vnode is still valid, and if so
560  * acquire requested lock.
561  */
562 int
563 vn_lock(vp, flags)
564 	struct vnode *vp;
565 	int flags;
566 {
567 	int error;
568 
569 	do {
570 		if ((flags & LK_INTERLOCK) == 0)
571 			simple_lock(&vp->v_interlock);
572 		if (vp->v_flag & VXLOCK) {
573 			vp->v_flag |= VXWANT;
574 			simple_unlock(&vp->v_interlock);
575 			tsleep((caddr_t)vp, PINOD, "vn_lock", 0);
576 			error = ENOENT;
577 		} else {
578 			error = VOP_LOCK(vp, flags | LK_INTERLOCK);
579 			if (error == 0 || error == EDEADLK)
580 				return (error);
581 		}
582 		flags &= ~LK_INTERLOCK;
583 	} while (flags & LK_RETRY);
584 	return (error);
585 }
586 
587 /*
588  * File table vnode close routine.
589  */
590 int
591 vn_closefile(fp, p)
592 	struct file *fp;
593 	struct proc *p;
594 {
595 
596 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
597 		fp->f_cred, p));
598 }
599 
600 /*
601  * Enable LK_CANRECURSE on lock. Return prior status.
602  */
603 u_int
604 vn_setrecurse(vp)
605 	struct vnode *vp;
606 {
607 	struct lock *lkp = &vp->v_lock;
608 	u_int retval = lkp->lk_flags & LK_CANRECURSE;
609 
610 	lkp->lk_flags |= LK_CANRECURSE;
611 	return retval;
612 }
613 
614 /*
615  * Called when done with locksetrecurse.
616  */
617 void
618 vn_restorerecurse(vp, flags)
619 	struct vnode *vp;
620 	u_int flags;
621 {
622 	struct lock *lkp = &vp->v_lock;
623 
624 	lkp->lk_flags &= ~LK_CANRECURSE;
625 	lkp->lk_flags |= flags;
626 }
627