xref: /openbsd-src/sys/kern/vfs_vnops.c (revision 2b0358df1d88d06ef4139321dd05bd5e05d91eaf)
1 /*	$OpenBSD: vfs_vnops.c,v 1.62 2009/01/29 22:08:45 guenther Exp $	*/
2 /*	$NetBSD: vfs_vnops.c,v 1.20 1996/02/04 02:18:41 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1989, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  * (c) UNIX System Laboratories, Inc.
8  * All or some portions of this file are derived from material licensed
9  * to the University of California by American Telephone and Telegraph
10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11  * the permission of UNIX System Laboratories, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)vfs_vnops.c	8.5 (Berkeley) 12/8/94
38  */
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/malloc.h>
44 #include <sys/file.h>
45 #include <sys/stat.h>
46 #include <sys/buf.h>
47 #include <sys/proc.h>
48 #include <sys/mount.h>
49 #include <sys/namei.h>
50 #include <sys/vnode.h>
51 #include <sys/ioctl.h>
52 #include <sys/tty.h>
53 #include <sys/cdio.h>
54 #include <sys/poll.h>
55 #include <sys/filedesc.h>
56 
57 #include <uvm/uvm_extern.h>
58 #include <miscfs/specfs/specdev.h>
59 
60 int vn_read(struct file *, off_t *, struct uio *, struct ucred *);
61 int vn_write(struct file *, off_t *, struct uio *, struct ucred *);
62 int vn_poll(struct file *, int, struct proc *);
63 int vn_kqfilter(struct file *, struct knote *);
64 int vn_closefile(struct file *, struct proc *);
65 
66 struct 	fileops vnops =
67 	{ vn_read, vn_write, vn_ioctl, vn_poll, vn_kqfilter, vn_statfile,
68 	  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(struct nameidata *ndp, int fmode, int cmode)
76 {
77 	struct vnode *vp;
78 	struct proc *p = ndp->ni_cnd.cn_proc;
79 	struct ucred *cred = p->p_ucred;
80 	struct vattr va;
81 	struct cloneinfo *cip;
82 	int error;
83 
84 	if ((fmode & (FREAD|FWRITE)) == 0)
85 		return (EINVAL);
86 	if ((fmode & (O_TRUNC | FWRITE)) == O_TRUNC)
87 		return (EINVAL);
88 	if (fmode & O_CREAT) {
89 		ndp->ni_cnd.cn_nameiop = CREATE;
90 		ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
91 		if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
92 			ndp->ni_cnd.cn_flags |= FOLLOW;
93 		if ((error = namei(ndp)) != 0)
94 			return (error);
95 
96 		if (ndp->ni_vp == NULL) {
97 			VATTR_NULL(&va);
98 			va.va_type = VREG;
99 			va.va_mode = cmode;
100 			if (fmode & O_EXCL)
101 				va.va_vaflags |= VA_EXCLUSIVE;
102 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
103 					   &ndp->ni_cnd, &va);
104 			if (error)
105 				return (error);
106 			fmode &= ~O_TRUNC;
107 			vp = ndp->ni_vp;
108 		} else {
109 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
110 			if (ndp->ni_dvp == ndp->ni_vp)
111 				vrele(ndp->ni_dvp);
112 			else
113 				vput(ndp->ni_dvp);
114 			ndp->ni_dvp = NULL;
115 			vp = ndp->ni_vp;
116 			if (fmode & O_EXCL) {
117 				error = EEXIST;
118 				goto bad;
119 			}
120 			fmode &= ~O_CREAT;
121 		}
122 	} else {
123 		ndp->ni_cnd.cn_nameiop = LOOKUP;
124 		ndp->ni_cnd.cn_flags =
125 		    ((fmode & O_NOFOLLOW) ? NOFOLLOW : FOLLOW) | LOCKLEAF;
126 		if ((error = namei(ndp)) != 0)
127 			return (error);
128 		vp = ndp->ni_vp;
129 	}
130 	if (vp->v_type == VSOCK) {
131 		error = EOPNOTSUPP;
132 		goto bad;
133 	}
134 	if (vp->v_type == VLNK) {
135 		error = EMLINK;
136 		goto bad;
137 	}
138 	if ((fmode & O_CREAT) == 0) {
139 		if (fmode & FREAD) {
140 			if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
141 				goto bad;
142 		}
143 		if (fmode & FWRITE) {
144 			if (vp->v_type == VDIR) {
145 				error = EISDIR;
146 				goto bad;
147 			}
148 			if ((error = vn_writechk(vp)) != 0 ||
149 			    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
150 				goto bad;
151 		}
152 	}
153 	if ((fmode & O_TRUNC) && vp->v_type == VREG) {
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 
162 	if (vp->v_flag & VCLONED) {
163 		cip = (struct cloneinfo *)vp->v_data;
164 
165 		vp->v_flag &= ~VCLONED;
166 
167 		ndp->ni_vp = cip->ci_vp;	/* return cloned vnode */
168 		vp->v_data = cip->ci_data;	/* restore v_data */
169 		VOP_UNLOCK(vp, 0, p);		/* keep a reference */
170 		vp = ndp->ni_vp;		/* for the increment below */
171 
172 		free(cip, M_TEMP);
173 	}
174 
175 	if (fmode & FWRITE)
176 		vp->v_writecount++;
177 	return (0);
178 bad:
179 	vput(vp);
180 	return (error);
181 }
182 
183 /*
184  * Check for write permissions on the specified vnode.
185  * Prototype text segments cannot be written.
186  */
187 int
188 vn_writechk(struct vnode *vp)
189 {
190 	/*
191 	 * Disallow write attempts on read-only file systems;
192 	 * unless the file is a socket or a block or character
193 	 * device resident on the file system.
194 	 */
195 	if (vp->v_mount->mnt_flag & MNT_RDONLY) {
196 		switch (vp->v_type) {
197 		case VREG:
198 		case VDIR:
199 		case VLNK:
200 			return (EROFS);
201 		case VNON:
202 		case VCHR:
203 		case VSOCK:
204 		case VFIFO:
205 		case VBAD:
206 		case VBLK:
207 			break;
208 		}
209 	}
210 	/*
211 	 * If there's shared text associated with
212 	 * the vnode, try to free it up once.  If
213 	 * we fail, we can't allow writing.
214 	 */
215 	if ((vp->v_flag & VTEXT) && !uvm_vnp_uncache(vp))
216 		return (ETXTBSY);
217 
218 	return (0);
219 }
220 
221 /*
222  * Mark a vnode as being the text image of a running process.
223  */
224 void
225 vn_marktext(struct vnode *vp)
226 {
227 	vp->v_flag |= VTEXT;
228 }
229 
230 /*
231  * Vnode close call
232  */
233 int
234 vn_close(struct vnode *vp, int flags, struct ucred *cred, struct proc *p)
235 {
236 	int error;
237 
238 	if (flags & FWRITE)
239 		vp->v_writecount--;
240 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
241 	error = VOP_CLOSE(vp, flags, cred, p);
242 	vput(vp);
243 	return (error);
244 }
245 
246 /*
247  * Package up an I/O request on a vnode into a uio and do it.
248  */
249 int
250 vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len, off_t offset,
251     enum uio_seg segflg, int ioflg, struct ucred *cred, size_t *aresid,
252     struct proc *p)
253 {
254 	struct uio auio;
255 	struct iovec aiov;
256 	int error;
257 
258 	if ((ioflg & IO_NODELOCKED) == 0)
259 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
260 	auio.uio_iov = &aiov;
261 	auio.uio_iovcnt = 1;
262 	aiov.iov_base = base;
263 	aiov.iov_len = len;
264 	auio.uio_resid = len;
265 	auio.uio_offset = offset;
266 	auio.uio_segflg = segflg;
267 	auio.uio_rw = rw;
268 	auio.uio_procp = p;
269 	if (rw == UIO_READ) {
270 		error = VOP_READ(vp, &auio, ioflg, cred);
271 	} else {
272 		error = VOP_WRITE(vp, &auio, ioflg, cred);
273 	}
274 	if (aresid)
275 		*aresid = auio.uio_resid;
276 	else
277 		if (auio.uio_resid && error == 0)
278 			error = EIO;
279 	if ((ioflg & IO_NODELOCKED) == 0)
280 		VOP_UNLOCK(vp, 0, p);
281 	return (error);
282 }
283 
284 /*
285  * File table vnode read routine.
286  */
287 int
288 vn_read(struct file *fp, off_t *poff, struct uio *uio, struct ucred *cred)
289 {
290 	struct vnode *vp = (struct vnode *)fp->f_data;
291 	int error = 0;
292 	size_t count;
293 	struct proc *p = uio->uio_procp;
294 
295 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
296 	uio->uio_offset = *poff;
297 	count = uio->uio_resid;
298 	if (vp->v_type != VDIR)
299 		error = VOP_READ(vp, uio,
300 		    (fp->f_flag & FNONBLOCK) ? IO_NDELAY : 0, cred);
301 	*poff += count - uio->uio_resid;
302 	VOP_UNLOCK(vp, 0, p);
303 	return (error);
304 }
305 
306 /*
307  * File table vnode write routine.
308  */
309 int
310 vn_write(struct file *fp, off_t *poff, struct uio *uio, struct ucred *cred)
311 {
312 	struct vnode *vp = (struct vnode *)fp->f_data;
313 	struct proc *p = uio->uio_procp;
314 	int error, ioflag = IO_UNIT;
315 	size_t count;
316 
317 	if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
318 		ioflag |= IO_APPEND;
319 	if (fp->f_flag & FNONBLOCK)
320 		ioflag |= IO_NDELAY;
321 	if ((fp->f_flag & FFSYNC) ||
322 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
323 		ioflag |= IO_SYNC;
324 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
325 	uio->uio_offset = *poff;
326 	count = uio->uio_resid;
327 	error = VOP_WRITE(vp, uio, ioflag, cred);
328 	if (ioflag & IO_APPEND)
329 		*poff = uio->uio_offset;
330 	else
331 		*poff += count - uio->uio_resid;
332 	VOP_UNLOCK(vp, 0, p);
333 	return (error);
334 }
335 
336 /*
337  * File table wrapper for vn_stat
338  */
339 int
340 vn_statfile(struct file *fp, struct stat *sb, struct proc *p)
341 {
342 	struct vnode *vp = (struct vnode *)fp->f_data;
343 	return vn_stat(vp, sb, p);
344 }
345 
346 /*
347  * vnode stat routine.
348  */
349 int
350 vn_stat(struct vnode *vp, struct stat *sb, struct proc *p)
351 {
352 	struct vattr va;
353 	int error;
354 	mode_t mode;
355 
356 	error = VOP_GETATTR(vp, &va, p->p_ucred, p);
357 	if (error)
358 		return (error);
359 	/*
360 	 * Copy from vattr table
361 	 */
362 	sb->st_dev = va.va_fsid;
363 	sb->st_ino = va.va_fileid;
364 	mode = va.va_mode;
365 	switch (vp->v_type) {
366 	case VREG:
367 		mode |= S_IFREG;
368 		break;
369 	case VDIR:
370 		mode |= S_IFDIR;
371 		break;
372 	case VBLK:
373 		mode |= S_IFBLK;
374 		break;
375 	case VCHR:
376 		mode |= S_IFCHR;
377 		break;
378 	case VLNK:
379 		mode |= S_IFLNK;
380 		break;
381 	case VSOCK:
382 		mode |= S_IFSOCK;
383 		break;
384 	case VFIFO:
385 		mode |= S_IFIFO;
386 		break;
387 	default:
388 		return (EBADF);
389 	}
390 	sb->st_mode = mode;
391 	sb->st_nlink = va.va_nlink;
392 	sb->st_uid = va.va_uid;
393 	sb->st_gid = va.va_gid;
394 	sb->st_rdev = va.va_rdev;
395 	sb->st_size = va.va_size;
396 	sb->st_atim = va.va_atime;
397 	sb->st_mtim = va.va_mtime;
398 	sb->st_ctim = va.va_ctime;
399 	sb->st_blksize = va.va_blocksize;
400 	sb->st_flags = va.va_flags;
401 	sb->st_gen = va.va_gen;
402 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
403 	return (0);
404 }
405 
406 /*
407  * File table vnode ioctl routine.
408  */
409 int
410 vn_ioctl(struct file *fp, u_long com, caddr_t data, struct proc *p)
411 {
412 	struct vnode *vp = ((struct vnode *)fp->f_data);
413 	struct vattr vattr;
414 	int error;
415 
416 	switch (vp->v_type) {
417 
418 	case VREG:
419 	case VDIR:
420 		if (com == FIONREAD) {
421 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
422 			if (error)
423 				return (error);
424 			*(int *)data = vattr.va_size - fp->f_offset;
425 			return (0);
426 		}
427 		if (com == FIONBIO || com == FIOASYNC)  /* XXX */
428 			return (0);			/* XXX */
429 		/* FALLTHROUGH */
430 	default:
431 		return (ENOTTY);
432 
433 	case VFIFO:
434 	case VCHR:
435 	case VBLK:
436 		error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
437 		if (error == 0 && com == TIOCSCTTY) {
438 			if (p->p_session->s_ttyvp)
439 				vrele(p->p_session->s_ttyvp);
440 			p->p_session->s_ttyvp = vp;
441 			VREF(vp);
442 		}
443 		return (error);
444 	}
445 }
446 
447 /*
448  * File table vnode poll routine.
449  */
450 int
451 vn_poll(struct file *fp, int events, struct proc *p)
452 {
453 	return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
454 }
455 
456 /*
457  * Check that the vnode is still valid, and if so
458  * acquire requested lock.
459  */
460 int
461 vn_lock(struct vnode *vp, int flags, struct proc *p)
462 {
463 	int error;
464 
465 	if ((flags & LK_RECURSEFAIL) == 0)
466 		flags |= LK_CANRECURSE;
467 
468 	do {
469 		if (vp->v_flag & VXLOCK) {
470 			vp->v_flag |= VXWANT;
471 			tsleep(vp, PINOD, "vn_lock", 0);
472 			error = ENOENT;
473 		} else {
474 			error = VOP_LOCK(vp, flags, p);
475 			if (error == 0)
476 				return (error);
477 		}
478 	} while (flags & LK_RETRY);
479 	return (error);
480 }
481 
482 /*
483  * File table vnode close routine.
484  */
485 int
486 vn_closefile(struct file *fp, struct proc *p)
487 {
488 	struct vnode *vp = fp->f_data;
489 	struct flock lf;
490 
491 	if ((fp->f_flag & FHASLOCK)) {
492 		lf.l_whence = SEEK_SET;
493 		lf.l_start = 0;
494 		lf.l_len = 0;
495 		lf.l_type = F_UNLCK;
496 		(void) VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK);
497 	}
498 
499 	return (vn_close(vp, fp->f_flag, fp->f_cred, p));
500 }
501 
502 int
503 vn_kqfilter(struct file *fp, struct knote *kn)
504 {
505 	return (VOP_KQFILTER(((struct vnode *)fp->f_data), kn));
506 }
507 
508 /*
509  * Common code for vnode access operations.
510  */
511 
512 /* Check if a directory can be found inside another in the hierarchy */
513 int
514 vn_isunder(struct vnode *lvp, struct vnode *rvp, struct proc *p)
515 {
516 	int error;
517 
518 	error = vfs_getcwd_common(lvp, rvp, NULL, NULL, MAXPATHLEN/2, 0, p);
519 
520 	if (!error)
521 		return (1);
522 
523 	return (0);
524 }
525