xref: /netbsd-src/sys/kern/vfs_vnops.c (revision 4472dbe5e3bd91ef2540bada7a7ca7384627ff9b)
1 /*	$NetBSD: vfs_vnops.c,v 1.42 2000/04/11 04:37:51 chs 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 	struct nameidata *ndp;
77 	int fmode, cmode;
78 {
79 	struct vnode *vp;
80 	struct proc *p = ndp->ni_cnd.cn_proc;
81 	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 	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  * Mark a vnode as being the text image of a running process.
190  */
191 void
192 vn_marktext(vp)
193 	struct vnode *vp;
194 {
195 	vp->v_flag |= VTEXT;
196 }
197 
198 /*
199  * Vnode close call
200  *
201  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
202  */
203 int
204 vn_close(vp, flags, cred, p)
205 	struct vnode *vp;
206 	int flags;
207 	struct ucred *cred;
208 	struct proc *p;
209 {
210 	int error;
211 
212 	if (flags & FWRITE)
213 		vp->v_writecount--;
214 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
215 	error = VOP_CLOSE(vp, flags, cred, p);
216 	vput(vp);
217 	return (error);
218 }
219 
220 /*
221  * Package up an I/O request on a vnode into a uio and do it.
222  */
223 int
224 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p)
225 	enum uio_rw rw;
226 	struct vnode *vp;
227 	caddr_t base;
228 	int len;
229 	off_t offset;
230 	enum uio_seg segflg;
231 	int ioflg;
232 	struct ucred *cred;
233 	size_t *aresid;
234 	struct proc *p;
235 {
236 	struct uio auio;
237 	struct iovec aiov;
238 	int error;
239 
240 	if ((ioflg & IO_NODELOCKED) == 0)
241 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
242 	auio.uio_iov = &aiov;
243 	auio.uio_iovcnt = 1;
244 	aiov.iov_base = base;
245 	aiov.iov_len = len;
246 	auio.uio_resid = len;
247 	auio.uio_offset = offset;
248 	auio.uio_segflg = segflg;
249 	auio.uio_rw = rw;
250 	auio.uio_procp = p;
251 	if (rw == UIO_READ) {
252 		error = VOP_READ(vp, &auio, ioflg, cred);
253 	} else {
254 		error = VOP_WRITE(vp, &auio, ioflg, cred);
255 	}
256 	if (aresid)
257 		*aresid = auio.uio_resid;
258 	else
259 		if (auio.uio_resid && error == 0)
260 			error = EIO;
261 	if ((ioflg & IO_NODELOCKED) == 0)
262 		VOP_UNLOCK(vp, 0);
263 	return (error);
264 }
265 
266 int
267 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies)
268 	struct file *fp;
269 	char *buf;
270 	int segflg, *done, *ncookies;
271 	u_int count;
272 	struct proc *p;
273 	off_t **cookies;
274 {
275 	struct vnode *vp = (struct vnode *)fp->f_data;
276 	struct iovec aiov;
277 	struct uio auio;
278 	int error, eofflag;
279 
280 unionread:
281 	if (vp->v_type != VDIR)
282 		return (EINVAL);
283 	aiov.iov_base = buf;
284 	aiov.iov_len = count;
285 	auio.uio_iov = &aiov;
286 	auio.uio_iovcnt = 1;
287 	auio.uio_rw = UIO_READ;
288 	auio.uio_segflg = segflg;
289 	auio.uio_procp = p;
290 	auio.uio_resid = count;
291 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
292 	auio.uio_offset = fp->f_offset;
293 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
294 		    ncookies);
295 	fp->f_offset = auio.uio_offset;
296 	VOP_UNLOCK(vp, 0);
297 	if (error)
298 		return (error);
299 
300 #ifdef UNION
301 {
302 	extern struct vnode *union_dircache __P((struct vnode *));
303 
304 	if (count == auio.uio_resid && (vp->v_op == union_vnodeop_p)) {
305 		struct vnode *lvp;
306 
307 		lvp = union_dircache(vp);
308 		if (lvp != NULLVP) {
309 			struct vattr va;
310 
311 			/*
312 			 * If the directory is opaque,
313 			 * then don't show lower entries
314 			 */
315 			error = VOP_GETATTR(vp, &va, fp->f_cred, p);
316 			if (va.va_flags & OPAQUE) {
317 				vput(lvp);
318 				lvp = NULL;
319 			}
320 		}
321 
322 		if (lvp != NULLVP) {
323 			error = VOP_OPEN(lvp, FREAD, fp->f_cred, p);
324 			if (error) {
325 				vput(lvp);
326 				return (error);
327 			}
328 			VOP_UNLOCK(lvp, 0);
329 			fp->f_data = (caddr_t) lvp;
330 			fp->f_offset = 0;
331 			error = vn_close(vp, FREAD, fp->f_cred, p);
332 			if (error)
333 				return (error);
334 			vp = lvp;
335 			goto unionread;
336 		}
337 	}
338 }
339 #endif /* UNION */
340 
341 	if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
342 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
343 		struct vnode *tvp = vp;
344 		vp = vp->v_mount->mnt_vnodecovered;
345 		VREF(vp);
346 		fp->f_data = (caddr_t) vp;
347 		fp->f_offset = 0;
348 		vrele(tvp);
349 		goto unionread;
350 	}
351 	*done = count - auio.uio_resid;
352 	return error;
353 }
354 
355 /*
356  * File table vnode read routine.
357  */
358 int
359 vn_read(fp, offset, uio, cred, flags)
360 	struct file *fp;
361 	off_t *offset;
362 	struct uio *uio;
363 	struct ucred *cred;
364 	int flags;
365 {
366 	struct vnode *vp = (struct vnode *)fp->f_data;
367 	int count, error, ioflag = 0;
368 
369 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
370 	if (fp->f_flag & FNONBLOCK)
371 		ioflag |= IO_NDELAY;
372 	if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
373 		ioflag |= IO_SYNC;
374 	if (fp->f_flag & FALTIO)
375 		ioflag |= IO_ALTSEMANTICS;
376 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
377 	uio->uio_offset = *offset;
378 	count = uio->uio_resid;
379 	error = VOP_READ(vp, uio, ioflag, cred);
380 	if (flags & FOF_UPDATE_OFFSET)
381 		*offset += count - uio->uio_resid;
382 	VOP_UNLOCK(vp, 0);
383 	return (error);
384 }
385 
386 /*
387  * File table vnode write routine.
388  */
389 int
390 vn_write(fp, offset, uio, cred, flags)
391 	struct file *fp;
392 	off_t *offset;
393 	struct uio *uio;
394 	struct ucred *cred;
395 	int flags;
396 {
397 	struct vnode *vp = (struct vnode *)fp->f_data;
398 	int count, error, ioflag = IO_UNIT;
399 
400 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
401 		ioflag |= IO_APPEND;
402 	if (fp->f_flag & FNONBLOCK)
403 		ioflag |= IO_NDELAY;
404 	if (fp->f_flag & FFSYNC ||
405 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
406 		ioflag |= IO_SYNC;
407 	else if (fp->f_flag & FDSYNC)
408 		ioflag |= IO_DSYNC;
409 	if (fp->f_flag & FALTIO)
410 		ioflag |= IO_ALTSEMANTICS;
411 	VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
412 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
413 	uio->uio_offset = *offset;
414 	count = uio->uio_resid;
415 	error = VOP_WRITE(vp, uio, ioflag, cred);
416 	if (flags & FOF_UPDATE_OFFSET) {
417 		if (ioflag & IO_APPEND)
418 			*offset = uio->uio_offset;
419 		else
420 			*offset += count - uio->uio_resid;
421 	}
422 	VOP_UNLOCK(vp, 0);
423 	return (error);
424 }
425 
426 /*
427  * File table vnode stat routine.
428  */
429 int
430 vn_stat(vp, sb, p)
431 	struct vnode *vp;
432 	struct stat *sb;
433 	struct proc *p;
434 {
435 	struct vattr va;
436 	int error;
437 	mode_t mode;
438 
439 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
440 	if (error)
441 		return (error);
442 	/*
443 	 * Copy from vattr table
444 	 */
445 	sb->st_dev = va.va_fsid;
446 	sb->st_ino = va.va_fileid;
447 	mode = va.va_mode;
448 	switch (vp->v_type) {
449 	case VREG:
450 		mode |= S_IFREG;
451 		break;
452 	case VDIR:
453 		mode |= S_IFDIR;
454 		break;
455 	case VBLK:
456 		mode |= S_IFBLK;
457 		break;
458 	case VCHR:
459 		mode |= S_IFCHR;
460 		break;
461 	case VLNK:
462 		mode |= S_IFLNK;
463 		break;
464 	case VSOCK:
465 		mode |= S_IFSOCK;
466 		break;
467 	case VFIFO:
468 		mode |= S_IFIFO;
469 		break;
470 	default:
471 		return (EBADF);
472 	};
473 	sb->st_mode = mode;
474 	sb->st_nlink = va.va_nlink;
475 	sb->st_uid = va.va_uid;
476 	sb->st_gid = va.va_gid;
477 	sb->st_rdev = va.va_rdev;
478 	sb->st_size = va.va_size;
479 	sb->st_atimespec = va.va_atime;
480 	sb->st_mtimespec = va.va_mtime;
481 	sb->st_ctimespec = va.va_ctime;
482 	sb->st_blksize = va.va_blocksize;
483 	sb->st_flags = va.va_flags;
484 	sb->st_gen = 0;
485 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
486 	return (0);
487 }
488 
489 /*
490  * File table vnode fcntl routine.
491  */
492 int
493 vn_fcntl(fp, com, data, p)
494 	struct file *fp;
495 	u_int com;
496 	caddr_t data;
497 	struct proc *p;
498 {
499 	struct vnode *vp = ((struct vnode *)fp->f_data);
500 	int error;
501 
502 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
503 	error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p);
504 	VOP_UNLOCK(vp, 0);
505 	return (error);
506 }
507 
508 /*
509  * File table vnode ioctl routine.
510  */
511 int
512 vn_ioctl(fp, com, data, p)
513 	struct file *fp;
514 	u_long com;
515 	caddr_t data;
516 	struct proc *p;
517 {
518 	struct vnode *vp = ((struct vnode *)fp->f_data);
519 	struct vattr vattr;
520 	int error;
521 
522 	switch (vp->v_type) {
523 
524 	case VREG:
525 	case VDIR:
526 		if (com == FIONREAD) {
527 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
528 			if (error)
529 				return (error);
530 			*(int *)data = vattr.va_size - fp->f_offset;
531 			return (0);
532 		}
533 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
534 			return (0);			/* XXX */
535 		/* fall into ... */
536 
537 	default:
538 		return (ENOTTY);
539 
540 	case VFIFO:
541 	case VCHR:
542 	case VBLK:
543 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
544 		if (error == 0 && com == TIOCSCTTY) {
545 			if (p->p_session->s_ttyvp)
546 				vrele(p->p_session->s_ttyvp);
547 			p->p_session->s_ttyvp = vp;
548 			VREF(vp);
549 		}
550 		return (error);
551 	}
552 }
553 
554 /*
555  * File table vnode poll routine.
556  */
557 int
558 vn_poll(fp, events, p)
559 	struct file *fp;
560 	int events;
561 	struct proc *p;
562 {
563 
564 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
565 }
566 
567 /*
568  * Check that the vnode is still valid, and if so
569  * acquire requested lock.
570  */
571 int
572 vn_lock(vp, flags)
573 	struct vnode *vp;
574 	int flags;
575 {
576 	int error;
577 
578 	do {
579 		if ((flags & LK_INTERLOCK) == 0)
580 			simple_lock(&vp->v_interlock);
581 		if (vp->v_flag & VXLOCK) {
582 			vp->v_flag |= VXWANT;
583 			simple_unlock(&vp->v_interlock);
584 			tsleep((caddr_t)vp, PINOD, "vn_lock", 0);
585 			error = ENOENT;
586 		} else {
587 			error = VOP_LOCK(vp, flags | LK_INTERLOCK);
588 			if (error == 0 || error == EDEADLK)
589 				return (error);
590 		}
591 		flags &= ~LK_INTERLOCK;
592 	} while (flags & LK_RETRY);
593 	return (error);
594 }
595 
596 /*
597  * File table vnode close routine.
598  */
599 int
600 vn_closefile(fp, p)
601 	struct file *fp;
602 	struct proc *p;
603 {
604 
605 	return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
606 		fp->f_cred, p));
607 }
608 
609 /*
610  * Enable LK_CANRECURSE on lock. Return prior status.
611  */
612 u_int
613 vn_setrecurse(vp)
614 	struct vnode *vp;
615 {
616 	struct lock *lkp = &vp->v_lock;
617 	u_int retval = lkp->lk_flags & LK_CANRECURSE;
618 
619 	lkp->lk_flags |= LK_CANRECURSE;
620 	return retval;
621 }
622 
623 /*
624  * Called when done with locksetrecurse.
625  */
626 void
627 vn_restorerecurse(vp, flags)
628 	struct vnode *vp;
629 	u_int flags;
630 {
631 	struct lock *lkp = &vp->v_lock;
632 
633 	lkp->lk_flags &= ~LK_CANRECURSE;
634 	lkp->lk_flags |= flags;
635 }
636