xref: /original-bsd/sys/kern/vfs_vnops.c (revision a76afa45)
1 /*
2  * Copyright (c) 1982, 1986, 1989 Regents of the University of California.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms are permitted
6  * provided that the above copyright notice and this paragraph are
7  * duplicated in all such forms and that any documentation,
8  * advertising materials, and other materials related to such
9  * distribution and use acknowledge that the software was developed
10  * by the University of California, Berkeley.  The name of the
11  * University may not be used to endorse or promote products derived
12  * from this software without specific prior written permission.
13  * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
14  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
15  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
16  *
17  *	@(#)vfs_vnops.c	7.4 (Berkeley) 05/09/89
18  */
19 
20 #include "param.h"
21 #include "systm.h"
22 #include "user.h"
23 #include "kernel.h"
24 #include "file.h"
25 #include "stat.h"
26 #include "buf.h"
27 #include "proc.h"
28 #include "uio.h"
29 #include "socket.h"
30 #include "socketvar.h"
31 #include "mount.h"
32 #include "vnode.h"
33 #include "../ufs/inode.h"
34 #include "../ufs/fs.h"
35 #include "../ufs/quota.h"
36 #include "ioctl.h"
37 #include "tty.h"
38 
39 int	vn_read(), vn_write(), vn_ioctl(), vn_select(), vn_close();
40 struct 	fileops vnops =
41 	{ vn_read, vn_write, vn_ioctl, vn_select, vn_close };
42 
43 /*
44  * Common code for vnode open operations.
45  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
46  */
47 vn_open(ndp, fmode, cmode)
48 	register struct nameidata *ndp;
49 	int fmode, cmode;
50 {
51 	register struct vnode *vp;
52 	struct vattr vat;
53 	struct vattr *vap = &vat;
54 	int error;
55 
56 	if (fmode & FCREAT) {
57 		ndp->ni_nameiop = CREATE | LOCKPARENT | LOCKLEAF;
58 		if ((fmode & FEXCL) == 0)
59 			ndp->ni_nameiop |= FOLLOW;
60 		if (error = namei(ndp))
61 			return (error);
62 		if (ndp->ni_vp == NULL) {
63 			vattr_null(vap);
64 			vap->va_type = VREG;
65 			vap->va_mode = cmode;
66 			if (error = VOP_CREATE(ndp, vap))
67 				return (error);
68 			fmode &= ~FTRUNC;
69 			vp = ndp->ni_vp;
70 		} else {
71 			vp = ndp->ni_vp;
72 			ndp->ni_vp = 0;
73 			VOP_ABORTOP(ndp);
74 			ndp->ni_vp = vp;
75 			if (fmode & FEXCL) {
76 				error = EEXIST;
77 				goto bad;
78 			}
79 			fmode &= ~FCREAT;
80 		}
81 	} else {
82 		ndp->ni_nameiop = LOOKUP | FOLLOW | LOCKLEAF;
83 		if (error = namei(ndp))
84 			return (error);
85 		vp = ndp->ni_vp;
86 	}
87 	if (vp->v_type == VSOCK) {
88 		error = EOPNOTSUPP;
89 		goto bad;
90 	}
91 	if ((fmode & FCREAT) == 0) {
92 		if (fmode & FREAD) {
93 			if (error = vn_access(vp, VREAD, ndp->ni_cred))
94 				goto bad;
95 		}
96 		if (fmode & (FWRITE|FTRUNC)) {
97 			if (error = vn_access(vp, VWRITE, ndp->ni_cred))
98 				goto bad;
99 			if (vp->v_type == VDIR) {
100 				error = EISDIR;
101 				goto bad;
102 			}
103 		}
104 	}
105 	if (fmode & FTRUNC) {
106 		vattr_null(vap);
107 		vap->va_size = 0;
108 		if (error = VOP_SETATTR(vp, vap, ndp->ni_cred))
109 			goto bad;
110 	}
111 	VOP_UNLOCK(vp);
112 	if (setjmp(&u.u_qsave)) {
113 		if (error == 0)
114 			error = EINTR;
115 		return (error);
116 	}
117 	return (VOP_OPEN(vp, fmode, ndp->ni_cred));
118 
119 bad:
120 	vput(vp);
121 	return(error);
122 }
123 
124 /*
125  * Check mode permission on vnode pointer. Mode is READ, WRITE or EXEC.
126  * In the case of WRITE, the read-only status of the file system is
127  * checked. Also in WRITE, prototype text segments cannot be written.
128  */
129 vn_access(vp, mode, cred)
130 	register struct vnode *vp;
131 	int mode;
132 	struct ucred *cred;
133 {
134 
135 	if (mode & VWRITE) {
136 		/*
137 		 * Disallow write attempts on read-only file systems;
138 		 * unless the file is a socket or a block or character
139 		 * device resident on the file system.
140 		 */
141 		if ((vp->v_mount->m_flag & M_RDONLY) &&
142 			vp->v_type != VCHR &&
143 			vp->v_type != VBLK &&
144 			vp->v_type != VSOCK)
145 				return (EROFS);
146 		/*
147 		 * If there's shared text associated with
148 		 * the inode, try to free it up once.  If
149 		 * we fail, we can't allow writing.
150 		 */
151 		if (vp->v_flag & VTEXT)
152 			xrele(vp);
153 		if (vp->v_flag & VTEXT)
154 			return (ETXTBSY);
155 	}
156 	return (VOP_ACCESS(vp, mode, cred));
157 }
158 
159 /*
160  * Vnode version of rdwri() for calls on file systems.
161  */
162 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid)
163 	enum uio_rw rw;
164 	struct vnode *vp;
165 	caddr_t base;
166 	int len;
167 	off_t offset;
168 	enum uio_seg segflg;
169 	int ioflg;
170 	struct ucred *cred;
171 	int *aresid;
172 {
173 	struct uio auio;
174 	struct iovec aiov;
175 	int error;
176 
177 	auio.uio_iov = &aiov;
178 	auio.uio_iovcnt = 1;
179 	aiov.iov_base = base;
180 	aiov.iov_len = len;
181 	auio.uio_resid = len;
182 	auio.uio_offset = offset;
183 	auio.uio_segflg = segflg;
184 	auio.uio_rw = rw;
185 	if (rw == UIO_READ)
186 		error = VOP_READ(vp, &auio, &offset, ioflg, cred);
187 	else
188 		error = VOP_WRITE(vp, &auio, &offset, ioflg, cred);
189 	if (aresid)
190 		*aresid = auio.uio_resid;
191 	else
192 		if (auio.uio_resid && error == 0)
193 			error = EIO;
194 	return (error);
195 }
196 
197 vn_read(fp, uio, cred)
198 	struct file *fp;
199 	struct uio *uio;
200 	struct ucred *cred;
201 {
202 
203 	return (VOP_READ((struct vnode *)fp->f_data, uio, &(fp->f_offset),
204 		(fp->f_flag & FNDELAY) ? IO_NDELAY : 0, cred));
205 }
206 
207 vn_write(fp, uio, cred)
208 	struct file *fp;
209 	struct uio *uio;
210 	struct ucred *cred;
211 {
212 	register struct vnode *vp = (struct vnode *)fp->f_data;
213 	int ioflag = 0;
214 
215 	if (vp->v_type == VREG && (fp->f_flag & FAPPEND))
216 		ioflag |= IO_APPEND;
217 	if (fp->f_flag & FNDELAY)
218 		ioflag |= IO_NDELAY;
219 	return (VOP_WRITE(vp, uio, &(fp->f_offset), ioflag, cred));
220 }
221 
222 /*
223  * Get stat info for a vnode.
224  */
225 vn_stat(vp, sb)
226 	struct vnode *vp;
227 	register struct stat *sb;
228 {
229 	struct vattr vattr;
230 	register struct vattr *vap;
231 	int error;
232 	u_short mode;
233 
234 	vap = &vattr;
235 	error = VOP_GETATTR(vp, vap, u.u_cred);
236 	if (error)
237 		return (error);
238 	/*
239 	 * Copy from vattr table
240 	 */
241 	sb->st_dev = vap->va_fsid;
242 	sb->st_ino = vap->va_fileid;
243 	mode = vap->va_mode;
244 	switch (vp->v_type) {
245 	case VREG:
246 		mode |= IFREG;
247 		break;
248 	case VDIR:
249 		mode |= IFDIR;
250 		break;
251 	case VBLK:
252 		mode |= IFBLK;
253 		break;
254 	case VCHR:
255 		mode |= IFCHR;
256 		break;
257 	case VLNK:
258 		mode |= IFLNK;
259 		break;
260 	case VSOCK:
261 		mode |= IFSOCK;
262 		break;
263 	default:
264 		return (EBADF);
265 	};
266 	sb->st_mode = mode;
267 	sb->st_nlink = vap->va_nlink;
268 	sb->st_uid = vap->va_uid;
269 	sb->st_gid = vap->va_gid;
270 	sb->st_rdev = vap->va_rdev;
271 	sb->st_size = vap->va_size;
272 	sb->st_atime = vap->va_atime.tv_sec;
273 	sb->st_spare1 = 0;
274 	sb->st_mtime = vap->va_mtime.tv_sec;
275 	sb->st_spare2 = 0;
276 	sb->st_ctime = vap->va_ctime.tv_sec;
277 	sb->st_spare3 = 0;
278 	sb->st_blksize = vap->va_blocksize;
279 	sb->st_spare4[0] = sb->st_spare4[1] = 0;
280 	/*
281 	 * XXX THIS IS NOT CORRECT!!, but be sure to change ufs_getattr()
282 	 * if you change it.
283 	 */
284 	sb->st_blocks = vap->va_bytes;
285 	return (0);
286 }
287 
288 /*
289  * Vnode ioctl call
290  */
291 vn_ioctl(fp, com, data)
292 	struct file *fp;
293 	int com;
294 	caddr_t data;
295 {
296 	register struct vnode *vp = ((struct vnode *)fp->f_data);
297 	struct vattr vattr;
298 	int error;
299 
300 	switch (vp->v_type) {
301 
302 	case VREG:
303 	case VDIR:
304 		if (com == FIONREAD) {
305 			if (error = VOP_GETATTR(vp, &vattr, u.u_cred))
306 				return (error);
307 			*(off_t *)data = vattr.va_size - fp->f_offset;
308 			return (0);
309 		}
310 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
311 			return (0);			/* XXX */
312 		/* fall into ... */
313 
314 	default:
315 		return (ENOTTY);
316 
317 	case VCHR:
318 	case VBLK:
319 		u.u_r.r_val1 = 0;
320 		if (setjmp(&u.u_qsave)) {
321 			if ((u.u_sigintr & sigmask(u.u_procp->p_cursig)) != 0)
322 				return(EINTR);
323 			u.u_eosys = RESTARTSYS;
324 			return (0);
325 		}
326 		return (VOP_IOCTL(vp, com, data, fp->f_flag, u.u_cred));
327 	}
328 }
329 
330 /*
331  * Vnode select call
332  */
333 vn_select(fp, which)
334 	struct file *fp;
335 	int which;
336 {
337 	return(VOP_SELECT(((struct vnode *)fp->f_data), which, u.u_cred));
338 }
339 
340 /*
341  * Vnode close call
342  */
343 vn_close(fp)
344 	register struct file *fp;
345 {
346 	struct vnode *vp = ((struct vnode *)fp->f_data);
347 	int error;
348 
349 	if (fp->f_flag & (FSHLOCK|FEXLOCK))
350 		vn_unlock(fp, FSHLOCK|FEXLOCK);
351 	/*
352 	 * Must delete vnode reference from this file entry
353 	 * before VOP_CLOSE, so that only other references
354 	 * will prevent close.
355 	 */
356 	fp->f_data = (caddr_t) 0;
357 	error = VOP_CLOSE(vp, fp->f_flag, u.u_cred);
358 	vrele(vp);
359 	return (error);
360 }
361 
362 /*
363  * Place an advisory lock on a vnode.
364  * !! THIS IMPLIES THAT ALL STATEFUL FILE SERVERS WILL USE file table entries
365  */
366 vn_lock(fp, cmd)
367 	register struct file *fp;
368 	int cmd;
369 {
370 	register int priority = PLOCK;
371 	register struct vnode *vp = (struct vnode *)fp->f_data;
372 
373 	if ((cmd & LOCK_EX) == 0)
374 		priority += 4;
375 	if (setjmp(&u.u_qsave)) {
376 		if ((u.u_sigintr & sigmask(u.u_procp->p_cursig)) != 0)
377 			return(EINTR);
378 		u.u_eosys = RESTARTSYS;
379 		return (0);
380 	}
381 	/*
382 	 * If there's a exclusive lock currently applied
383 	 * to the file, then we've gotta wait for the
384 	 * lock with everyone else.
385 	 */
386 again:
387 	while (vp->v_flag & VEXLOCK) {
388 		/*
389 		 * If we're holding an exclusive
390 		 * lock, then release it.
391 		 */
392 		if (fp->f_flag & FEXLOCK) {
393 			vn_unlock(fp, FEXLOCK);
394 			continue;
395 		}
396 		if (cmd & LOCK_NB)
397 			return (EWOULDBLOCK);
398 		vp->v_flag |= VLWAIT;
399 		sleep((caddr_t)&vp->v_exlockc, priority);
400 	}
401 	if ((cmd & LOCK_EX) && (vp->v_flag & VSHLOCK)) {
402 		/*
403 		 * Must wait for any shared locks to finish
404 		 * before we try to apply a exclusive lock.
405 		 *
406 		 * If we're holding a shared
407 		 * lock, then release it.
408 		 */
409 		if (fp->f_flag & FSHLOCK) {
410 			vn_unlock(fp, FSHLOCK);
411 			goto again;
412 		}
413 		if (cmd & LOCK_NB)
414 			return (EWOULDBLOCK);
415 		vp->v_flag |= VLWAIT;
416 		sleep((caddr_t)&vp->v_shlockc, PLOCK);
417 		goto again;
418 	}
419 	if (fp->f_flag & FEXLOCK)
420 		panic("vn_lock");
421 	if (cmd & LOCK_EX) {
422 		cmd &= ~LOCK_SH;
423 		vp->v_exlockc++;
424 		vp->v_flag |= VEXLOCK;
425 		fp->f_flag |= FEXLOCK;
426 	}
427 	if ((cmd & LOCK_SH) && (fp->f_flag & FSHLOCK) == 0) {
428 		vp->v_shlockc++;
429 		vp->v_flag |= VSHLOCK;
430 		fp->f_flag |= FSHLOCK;
431 	}
432 	return (0);
433 }
434 
435 /*
436  * Unlock a file.
437  */
438 vn_unlock(fp, kind)
439 	register struct file *fp;
440 	int kind;
441 {
442 	register struct vnode *vp = (struct vnode *)fp->f_data;
443 	int flags;
444 
445 	kind &= fp->f_flag;
446 	if (vp == NULL || kind == 0)
447 		return;
448 	flags = vp->v_flag;
449 	if (kind & FSHLOCK) {
450 		if ((flags & VSHLOCK) == 0)
451 			panic("vn_unlock: SHLOCK");
452 		if (--vp->v_shlockc == 0) {
453 			vp->v_flag &= ~VSHLOCK;
454 			if (flags & VLWAIT)
455 				wakeup((caddr_t)&vp->v_shlockc);
456 		}
457 		fp->f_flag &= ~FSHLOCK;
458 	}
459 	if (kind & FEXLOCK) {
460 		if ((flags & VEXLOCK) == 0)
461 			panic("vn_unlock: EXLOCK");
462 		if (--vp->v_exlockc == 0) {
463 			vp->v_flag &= ~(VEXLOCK|VLWAIT);
464 			if (flags & VLWAIT)
465 				wakeup((caddr_t)&vp->v_exlockc);
466 		}
467 		fp->f_flag &= ~FEXLOCK;
468 	}
469 }
470 
471 /*
472  * vn_fhtovp() - convert a fh to a vnode ptr (optionally locked)
473  * 	- look up fsid in mount list (if not found ret error)
474  *	- get vp by calling VFS_FHTOVP() macro
475  *	- if lockflag lock it with VOP_LOCK()
476  */
477 vn_fhtovp(fhp, lockflag, vpp)
478 	fhandle_t *fhp;
479 	int lockflag;
480 	struct vnode **vpp;
481 {
482 	register struct mount *mp;
483 	int error;
484 
485 	if ((mp = getvfs(&fhp->fh_fsid)) == NULL)
486 		return (ESTALE);
487 	if (error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp))
488 		return (error);
489 	if (lockflag)
490 		VOP_LOCK(*vpp);
491 	return (0);
492 }
493 
494 /*
495  * Revoke access the current tty by all processes.
496  * Used only by the super-user in init
497  * to give ``clean'' terminals at login.
498  */
499 vhangup()
500 {
501 
502 	if (u.u_error = suser(u.u_cred, &u.u_acflag))
503 		return;
504 	if (u.u_ttyp == NULL)
505 		return;
506 	forceclose(u.u_ttyd);
507 	if ((u.u_ttyp->t_state) & TS_ISOPEN)
508 		gsignal(u.u_ttyp->t_pgid, SIGHUP);
509 	u.u_ttyp->t_session = 0;
510 	u.u_ttyp->t_pgid = 0;
511 }
512 
513 forceclose(dev)
514 	dev_t dev;
515 {
516 	register struct file *fp;
517 	register struct vnode *vp;
518 
519 	for (fp = file; fp < fileNFILE; fp++) {
520 		if (fp->f_count == 0)
521 			continue;
522 		if (fp->f_type != DTYPE_VNODE)
523 			continue;
524 		vp = (struct vnode *)fp->f_data;
525 		if (vp == 0)
526 			continue;
527 		if (vp->v_type != VCHR)
528 			continue;
529 		if (vp->v_rdev != dev)
530 			continue;
531 		fp->f_flag &= ~(FREAD|FWRITE);
532 	}
533 }
534 
535 /*
536  * Vnode release, just decrement the count and call VOP_INACTIVE()
537  */
538 void vrele(vp)
539 	register struct vnode *vp;
540 {
541 
542 	if (vp == NULL)
543 		return;
544 	vp->v_count--;
545 	if (vp->v_count < 0)
546 		printf("inode %d, bad ref count %d\n",
547 			VTOI(vp)->i_number, vp->v_count);
548 	if (vp->v_count == 0)
549 		VOP_INACTIVE(vp);
550 }
551 
552 /*
553  * vput(), just unlock and vrele()
554  */
555 vput(vp)
556 	register struct vnode *vp;
557 {
558 	VOP_UNLOCK(vp);
559 	vrele(vp);
560 }
561 
562 /*
563  * Noop
564  */
565 vfs_noop()
566 {
567 
568 	return (ENXIO);
569 }
570 
571 /*
572  * Null op
573  */
574 vfs_nullop()
575 {
576 
577 	return (0);
578 }
579