xref: /csrg-svn/sys/kern/vfs_vnops.c (revision 38769)
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	7.9 (Berkeley) 08/26/89
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 = VOP_ACCESS(vp, VREAD, ndp->ni_cred))
94 				goto bad;
95 		}
96 		if (fmode & (FWRITE|FTRUNC)) {
97 			if (vp->v_type == VDIR) {
98 				error = EISDIR;
99 				goto bad;
100 			}
101 			if ((error = vn_writechk(vp)) ||
102 			    (error = VOP_ACCESS(vp, VWRITE, ndp->ni_cred)))
103 				goto bad;
104 		}
105 	}
106 	if (fmode & FTRUNC) {
107 		vattr_null(vap);
108 		vap->va_size = 0;
109 		if (error = VOP_SETATTR(vp, vap, ndp->ni_cred))
110 			goto bad;
111 	}
112 	VOP_UNLOCK(vp);
113 	if (setjmp(&u.u_qsave)) {
114 		if (error == 0)
115 			error = EINTR;
116 		return (error);
117 	}
118 	error = VOP_OPEN(vp, fmode, ndp->ni_cred);
119 	if (error)
120 		vrele(vp);
121 	return (error);
122 
123 bad:
124 	vput(vp);
125 	return(error);
126 }
127 
128 /*
129  * Check for write permissions on the specified vnode.
130  * The read-only status of the file system is checked.
131  * Also, prototype text segments cannot be written.
132  */
133 vn_writechk(vp)
134 	register struct vnode *vp;
135 {
136 
137 	/*
138 	 * Disallow write attempts on read-only file systems;
139 	 * unless the file is a socket or a block or character
140 	 * device resident on the file system.
141 	 */
142 	if ((vp->v_mount->m_flag & M_RDONLY) && vp->v_type != VCHR &&
143 	    vp->v_type != VBLK && vp->v_type != VSOCK)
144 		return (EROFS);
145 	/*
146 	 * If there's shared text associated with
147 	 * the vnode, try to free it up once.  If
148 	 * we fail, we can't allow writing.
149 	 */
150 	if (vp->v_flag & VTEXT)
151 		xrele(vp);
152 	if (vp->v_flag & VTEXT)
153 		return (ETXTBSY);
154 	return (0);
155 }
156 
157 /*
158  * Vnode version of rdwri() for calls on file systems.
159  */
160 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid)
161 	enum uio_rw rw;
162 	struct vnode *vp;
163 	caddr_t base;
164 	int len;
165 	off_t offset;
166 	enum uio_seg segflg;
167 	int ioflg;
168 	struct ucred *cred;
169 	int *aresid;
170 {
171 	struct uio auio;
172 	struct iovec aiov;
173 	int error;
174 
175 	auio.uio_iov = &aiov;
176 	auio.uio_iovcnt = 1;
177 	aiov.iov_base = base;
178 	aiov.iov_len = len;
179 	auio.uio_resid = len;
180 	auio.uio_offset = offset;
181 	auio.uio_segflg = segflg;
182 	auio.uio_rw = rw;
183 	if (rw == UIO_READ)
184 		error = VOP_READ(vp, &auio, &offset, ioflg, cred);
185 	else
186 		error = VOP_WRITE(vp, &auio, &offset, ioflg, cred);
187 	if (aresid)
188 		*aresid = auio.uio_resid;
189 	else
190 		if (auio.uio_resid && error == 0)
191 			error = EIO;
192 	return (error);
193 }
194 
195 vn_read(fp, uio, cred)
196 	struct file *fp;
197 	struct uio *uio;
198 	struct ucred *cred;
199 {
200 
201 	return (VOP_READ((struct vnode *)fp->f_data, uio, &(fp->f_offset),
202 		(fp->f_flag & FNDELAY) ? IO_NDELAY : 0, cred));
203 }
204 
205 vn_write(fp, uio, cred)
206 	struct file *fp;
207 	struct uio *uio;
208 	struct ucred *cred;
209 {
210 	register struct vnode *vp = (struct vnode *)fp->f_data;
211 	int ioflag = 0;
212 
213 	if (vp->v_type == VREG && (fp->f_flag & FAPPEND))
214 		ioflag |= IO_APPEND;
215 	if (fp->f_flag & FNDELAY)
216 		ioflag |= IO_NDELAY;
217 	return (VOP_WRITE(vp, uio, &(fp->f_offset), ioflag, cred));
218 }
219 
220 /*
221  * Get stat info for a vnode.
222  */
223 vn_stat(vp, sb)
224 	struct vnode *vp;
225 	register struct stat *sb;
226 {
227 	struct vattr vattr;
228 	register struct vattr *vap;
229 	int error;
230 	u_short mode;
231 
232 	vap = &vattr;
233 	error = VOP_GETATTR(vp, vap, u.u_cred);
234 	if (error)
235 		return (error);
236 	/*
237 	 * Copy from vattr table
238 	 */
239 	sb->st_dev = vap->va_fsid;
240 	sb->st_ino = vap->va_fileid;
241 	mode = vap->va_mode;
242 	switch (vp->v_type) {
243 	case VREG:
244 		mode |= IFREG;
245 		break;
246 	case VDIR:
247 		mode |= IFDIR;
248 		break;
249 	case VBLK:
250 		mode |= IFBLK;
251 		break;
252 	case VCHR:
253 		mode |= IFCHR;
254 		break;
255 	case VLNK:
256 		mode |= IFLNK;
257 		break;
258 	case VSOCK:
259 		mode |= IFSOCK;
260 		break;
261 	default:
262 		return (EBADF);
263 	};
264 	sb->st_mode = mode;
265 	sb->st_nlink = vap->va_nlink;
266 	sb->st_uid = vap->va_uid;
267 	sb->st_gid = vap->va_gid;
268 	sb->st_rdev = vap->va_rdev;
269 	sb->st_size = vap->va_size;
270 	sb->st_atime = vap->va_atime.tv_sec;
271 	sb->st_spare1 = 0;
272 	sb->st_mtime = vap->va_mtime.tv_sec;
273 	sb->st_spare2 = 0;
274 	sb->st_ctime = vap->va_ctime.tv_sec;
275 	sb->st_spare3 = 0;
276 	sb->st_blksize = vap->va_blocksize;
277 	sb->st_flags = vap->va_flags;
278 	sb->st_gen = vap->va_gen;
279 	sb->st_blocks = vap->va_bytes / S_BLKSIZE;
280 	return (0);
281 }
282 
283 /*
284  * Vnode ioctl call
285  */
286 vn_ioctl(fp, com, data)
287 	struct file *fp;
288 	int com;
289 	caddr_t data;
290 {
291 	register struct vnode *vp = ((struct vnode *)fp->f_data);
292 	struct vattr vattr;
293 	int error;
294 
295 	switch (vp->v_type) {
296 
297 	case VREG:
298 	case VDIR:
299 		if (com == FIONREAD) {
300 			if (error = VOP_GETATTR(vp, &vattr, u.u_cred))
301 				return (error);
302 			*(off_t *)data = vattr.va_size - fp->f_offset;
303 			return (0);
304 		}
305 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
306 			return (0);			/* XXX */
307 		/* fall into ... */
308 
309 	default:
310 		return (ENOTTY);
311 
312 	case VCHR:
313 	case VBLK:
314 		u.u_r.r_val1 = 0;
315 		if (setjmp(&u.u_qsave)) {
316 			if ((u.u_sigintr & sigmask(u.u_procp->p_cursig)) != 0)
317 				return(EINTR);
318 			u.u_eosys = RESTARTSYS;
319 			return (0);
320 		}
321 		return (VOP_IOCTL(vp, com, data, fp->f_flag, u.u_cred));
322 	}
323 }
324 
325 /*
326  * Vnode select call
327  */
328 vn_select(fp, which)
329 	struct file *fp;
330 	int which;
331 {
332 	return(VOP_SELECT(((struct vnode *)fp->f_data), which, u.u_cred));
333 }
334 
335 /*
336  * Vnode close call
337  */
338 vn_close(fp)
339 	register struct file *fp;
340 {
341 	struct vnode *vp = ((struct vnode *)fp->f_data);
342 	int error;
343 
344 	if (fp->f_flag & (FSHLOCK|FEXLOCK))
345 		vn_unlock(fp, FSHLOCK|FEXLOCK);
346 	/*
347 	 * Must delete vnode reference from this file entry
348 	 * before VOP_CLOSE, so that only other references
349 	 * will prevent close.
350 	 */
351 	fp->f_data = (caddr_t) 0;
352 	error = VOP_CLOSE(vp, fp->f_flag, u.u_cred);
353 	vrele(vp);
354 	return (error);
355 }
356 
357 /*
358  * Place an advisory lock on a vnode.
359  * !! THIS IMPLIES THAT ALL STATEFUL FILE SERVERS WILL USE file table entries
360  */
361 vn_lock(fp, cmd)
362 	register struct file *fp;
363 	int cmd;
364 {
365 	register int priority = PLOCK;
366 	register struct vnode *vp = (struct vnode *)fp->f_data;
367 
368 	if ((cmd & LOCK_EX) == 0)
369 		priority += 4;
370 	if (setjmp(&u.u_qsave)) {
371 		if ((u.u_sigintr & sigmask(u.u_procp->p_cursig)) != 0)
372 			return(EINTR);
373 		u.u_eosys = RESTARTSYS;
374 		return (0);
375 	}
376 	/*
377 	 * If there's a exclusive lock currently applied
378 	 * to the file, then we've gotta wait for the
379 	 * lock with everyone else.
380 	 */
381 again:
382 	while (vp->v_flag & VEXLOCK) {
383 		/*
384 		 * If we're holding an exclusive
385 		 * lock, then release it.
386 		 */
387 		if (fp->f_flag & FEXLOCK) {
388 			vn_unlock(fp, FEXLOCK);
389 			continue;
390 		}
391 		if (cmd & LOCK_NB)
392 			return (EWOULDBLOCK);
393 		vp->v_flag |= VLWAIT;
394 		sleep((caddr_t)&vp->v_exlockc, priority);
395 	}
396 	if ((cmd & LOCK_EX) && (vp->v_flag & VSHLOCK)) {
397 		/*
398 		 * Must wait for any shared locks to finish
399 		 * before we try to apply a exclusive lock.
400 		 *
401 		 * If we're holding a shared
402 		 * lock, then release it.
403 		 */
404 		if (fp->f_flag & FSHLOCK) {
405 			vn_unlock(fp, FSHLOCK);
406 			goto again;
407 		}
408 		if (cmd & LOCK_NB)
409 			return (EWOULDBLOCK);
410 		vp->v_flag |= VLWAIT;
411 		sleep((caddr_t)&vp->v_shlockc, PLOCK);
412 		goto again;
413 	}
414 	if (fp->f_flag & FEXLOCK)
415 		panic("vn_lock");
416 	if (cmd & LOCK_EX) {
417 		cmd &= ~LOCK_SH;
418 		vp->v_exlockc++;
419 		vp->v_flag |= VEXLOCK;
420 		fp->f_flag |= FEXLOCK;
421 	}
422 	if ((cmd & LOCK_SH) && (fp->f_flag & FSHLOCK) == 0) {
423 		vp->v_shlockc++;
424 		vp->v_flag |= VSHLOCK;
425 		fp->f_flag |= FSHLOCK;
426 	}
427 	return (0);
428 }
429 
430 /*
431  * Unlock a file.
432  */
433 vn_unlock(fp, kind)
434 	register struct file *fp;
435 	int kind;
436 {
437 	register struct vnode *vp = (struct vnode *)fp->f_data;
438 	int flags;
439 
440 	kind &= fp->f_flag;
441 	if (vp == NULL || kind == 0)
442 		return;
443 	flags = vp->v_flag;
444 	if (kind & FSHLOCK) {
445 		if ((flags & VSHLOCK) == 0)
446 			panic("vn_unlock: SHLOCK");
447 		if (--vp->v_shlockc == 0) {
448 			vp->v_flag &= ~VSHLOCK;
449 			if (flags & VLWAIT)
450 				wakeup((caddr_t)&vp->v_shlockc);
451 		}
452 		fp->f_flag &= ~FSHLOCK;
453 	}
454 	if (kind & FEXLOCK) {
455 		if ((flags & VEXLOCK) == 0)
456 			panic("vn_unlock: EXLOCK");
457 		if (--vp->v_exlockc == 0) {
458 			vp->v_flag &= ~(VEXLOCK|VLWAIT);
459 			if (flags & VLWAIT)
460 				wakeup((caddr_t)&vp->v_exlockc);
461 		}
462 		fp->f_flag &= ~FEXLOCK;
463 	}
464 }
465 
466 /*
467  * vn_fhtovp() - convert a fh to a vnode ptr (optionally locked)
468  * 	- look up fsid in mount list (if not found ret error)
469  *	- get vp by calling VFS_FHTOVP() macro
470  *	- if lockflag lock it with VOP_LOCK()
471  */
472 vn_fhtovp(fhp, lockflag, vpp)
473 	fhandle_t *fhp;
474 	int lockflag;
475 	struct vnode **vpp;
476 {
477 	register struct mount *mp;
478 	int error;
479 
480 	if ((mp = getvfs(&fhp->fh_fsid)) == NULL)
481 		return (ESTALE);
482 	if (VFS_FHTOVP(mp, &fhp->fh_fid, vpp))
483 		return (ESTALE);
484 	if (!lockflag)
485 		VOP_UNLOCK(*vpp);
486 	return (0);
487 }
488 
489 /*
490  * Revoke access the current tty by all processes.
491  * Used only by the super-user in init
492  * to give ``clean'' terminals at login.
493  */
494 vhangup()
495 {
496 
497 	if (u.u_error = suser(u.u_cred, &u.u_acflag))
498 		return;
499 	if (u.u_ttyp == NULL)
500 		return;
501 	forceclose(u.u_ttyd);
502 	if ((u.u_ttyp->t_state) & TS_ISOPEN)
503 		gsignal(u.u_ttyp->t_pgid, SIGHUP);
504 	u.u_ttyp->t_session = 0;
505 	u.u_ttyp->t_pgid = 0;
506 }
507 
508 forceclose(dev)
509 	dev_t dev;
510 {
511 	register struct file *fp;
512 	register struct vnode *vp;
513 
514 	for (fp = file; fp < fileNFILE; fp++) {
515 		if (fp->f_count == 0)
516 			continue;
517 		if (fp->f_type != DTYPE_VNODE)
518 			continue;
519 		vp = (struct vnode *)fp->f_data;
520 		if (vp == 0)
521 			continue;
522 		if (vp->v_type != VCHR)
523 			continue;
524 		if (vp->v_rdev != dev)
525 			continue;
526 		fp->f_flag &= ~(FREAD|FWRITE);
527 	}
528 }
529 
530 /*
531  * Vnode reference, just increment the count
532  */
533 void vref(vp)
534 	struct vnode *vp;
535 {
536 
537 	vp->v_count++;
538 }
539 
540 /*
541  * Vnode release, just decrement the count and call VOP_INACTIVE()
542  */
543 void vrele(vp)
544 	register struct vnode *vp;
545 {
546 
547 	if (vp == NULL)
548 		return;
549 	vp->v_count--;
550 	if (vp->v_count < 0)
551 		printf("inode %d, bad ref count %d\n",
552 			VTOI(vp)->i_number, vp->v_count);
553 	if (vp->v_count == 0)
554 		VOP_INACTIVE(vp);
555 }
556 
557 /*
558  * vput(), just unlock and vrele()
559  */
560 vput(vp)
561 	register struct vnode *vp;
562 {
563 	VOP_UNLOCK(vp);
564 	vrele(vp);
565 }
566 
567 /*
568  * Noop
569  */
570 vfs_noop()
571 {
572 
573 	return (ENXIO);
574 }
575 
576 /*
577  * Null op
578  */
579 vfs_nullop()
580 {
581 
582 	return (0);
583 }
584