xref: /csrg-svn/sys/kern/vfs_vnops.c (revision 40284)
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	7.16 (Berkeley) 03/05/90
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/fs.h"
34 #include "../ufs/quota.h"
35 #include "ioctl.h"
36 #include "tty.h"
37 #include "tsleep.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 			if (ndp->ni_dvp == ndp->ni_vp)
72 				vrele(ndp->ni_dvp);
73 			else if (ndp->ni_dvp != NULL)
74 				vput(ndp->ni_dvp);
75 			ndp->ni_dvp = NULL;
76 			vp = ndp->ni_vp;
77 			if (fmode & FEXCL) {
78 				error = EEXIST;
79 				goto bad;
80 			}
81 			fmode &= ~FCREAT;
82 		}
83 	} else {
84 		ndp->ni_nameiop = LOOKUP | FOLLOW | LOCKLEAF;
85 		if (error = namei(ndp))
86 			return (error);
87 		vp = ndp->ni_vp;
88 	}
89 	if (vp->v_type == VSOCK) {
90 		error = EOPNOTSUPP;
91 		goto bad;
92 	}
93 	if ((fmode & FCREAT) == 0) {
94 		if (fmode & FREAD) {
95 			if (error = VOP_ACCESS(vp, VREAD, ndp->ni_cred))
96 				goto bad;
97 		}
98 		if (fmode & (FWRITE|FTRUNC)) {
99 			if (vp->v_type == VDIR) {
100 				error = EISDIR;
101 				goto bad;
102 			}
103 			if ((error = vn_writechk(vp)) ||
104 			    (error = VOP_ACCESS(vp, VWRITE, ndp->ni_cred)))
105 				goto bad;
106 		}
107 	}
108 	if (fmode & FTRUNC) {
109 		vattr_null(vap);
110 		vap->va_size = 0;
111 		if (error = VOP_SETATTR(vp, vap, ndp->ni_cred))
112 			goto bad;
113 	}
114 	VOP_UNLOCK(vp);
115 	if (setjmp(&u.u_qsave)) {
116 		if (error == 0)
117 			error = EINTR;
118 		return (error);
119 	}
120 	error = VOP_OPEN(vp, fmode, ndp->ni_cred);
121 	if (error)
122 		vrele(vp);
123 	return (error);
124 
125 bad:
126 	vput(vp);
127 	return(error);
128 }
129 
130 /*
131  * Check for write permissions on the specified vnode.
132  * The read-only status of the file system is checked.
133  * Also, prototype text segments cannot be written.
134  */
135 vn_writechk(vp)
136 	register struct vnode *vp;
137 {
138 
139 	/*
140 	 * Disallow write attempts on read-only file systems;
141 	 * unless the file is a socket or a block or character
142 	 * device resident on the file system.
143 	 */
144 	if ((vp->v_mount->m_flag & M_RDONLY) && vp->v_type != VCHR &&
145 	    vp->v_type != VBLK && vp->v_type != VSOCK)
146 		return (EROFS);
147 	/*
148 	 * If there's shared text associated with
149 	 * the vnode, try to free it up once.  If
150 	 * we fail, we can't allow writing.
151 	 */
152 	if (vp->v_flag & VTEXT)
153 		xrele(vp);
154 	if (vp->v_flag & VTEXT)
155 		return (ETXTBSY);
156 	return (0);
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 	if ((ioflg & IO_NODELOCKED) == 0)
178 		VOP_LOCK(vp);
179 	auio.uio_iov = &aiov;
180 	auio.uio_iovcnt = 1;
181 	aiov.iov_base = base;
182 	aiov.iov_len = len;
183 	auio.uio_resid = len;
184 	auio.uio_offset = offset;
185 	auio.uio_segflg = segflg;
186 	auio.uio_rw = rw;
187 	if (rw == UIO_READ)
188 		error = VOP_READ(vp, &auio, ioflg, cred);
189 	else
190 		error = VOP_WRITE(vp, &auio, ioflg, cred);
191 	if (aresid)
192 		*aresid = auio.uio_resid;
193 	else
194 		if (auio.uio_resid && error == 0)
195 			error = EIO;
196 	if ((ioflg & IO_NODELOCKED) == 0)
197 		VOP_UNLOCK(vp);
198 	return (error);
199 }
200 
201 vn_read(fp, uio, cred)
202 	struct file *fp;
203 	struct uio *uio;
204 	struct ucred *cred;
205 {
206 	register struct vnode *vp = (struct vnode *)fp->f_data;
207 	int count, error;
208 
209 	VOP_LOCK(vp);
210 	uio->uio_offset = fp->f_offset;
211 	count = uio->uio_resid;
212 	error = VOP_READ(vp, uio, (fp->f_flag & FNDELAY) ? IO_NDELAY : 0, cred);
213 	fp->f_offset += count - uio->uio_resid;
214 	VOP_UNLOCK(vp);
215 	return (error);
216 }
217 
218 vn_write(fp, uio, cred)
219 	struct file *fp;
220 	struct uio *uio;
221 	struct ucred *cred;
222 {
223 	register struct vnode *vp = (struct vnode *)fp->f_data;
224 	int count, error, ioflag = 0;
225 
226 	if (vp->v_type == VREG && (fp->f_flag & FAPPEND))
227 		ioflag |= IO_APPEND;
228 	if (fp->f_flag & FNDELAY)
229 		ioflag |= IO_NDELAY;
230 	VOP_LOCK(vp);
231 	uio->uio_offset = fp->f_offset;
232 	count = uio->uio_resid;
233 	error = VOP_WRITE(vp, uio, ioflag, cred);
234 	if (ioflag & IO_APPEND)
235 		fp->f_offset = uio->uio_offset;
236 	else
237 		fp->f_offset += count - uio->uio_resid;
238 	VOP_UNLOCK(vp);
239 	return (error);
240 }
241 
242 /*
243  * Get stat info for a vnode.
244  */
245 vn_stat(vp, sb)
246 	struct vnode *vp;
247 	register struct stat *sb;
248 {
249 	struct vattr vattr;
250 	register struct vattr *vap;
251 	int error;
252 	u_short mode;
253 
254 	vap = &vattr;
255 	error = VOP_GETATTR(vp, vap, u.u_cred);
256 	if (error)
257 		return (error);
258 	/*
259 	 * Copy from vattr table
260 	 */
261 	sb->st_dev = vap->va_fsid;
262 	sb->st_ino = vap->va_fileid;
263 	mode = vap->va_mode;
264 	switch (vp->v_type) {
265 	case VREG:
266 		mode |= S_IFREG;
267 		break;
268 	case VDIR:
269 		mode |= S_IFDIR;
270 		break;
271 	case VBLK:
272 		mode |= S_IFBLK;
273 		break;
274 	case VCHR:
275 		mode |= S_IFCHR;
276 		break;
277 	case VLNK:
278 		mode |= S_IFLNK;
279 		break;
280 	case VSOCK:
281 		mode |= S_IFSOCK;
282 		break;
283 	case VFIFO:
284 		mode |= S_IFIFO;
285 		break;
286 	default:
287 		return (EBADF);
288 	};
289 	sb->st_mode = mode;
290 	sb->st_nlink = vap->va_nlink;
291 	sb->st_uid = vap->va_uid;
292 	sb->st_gid = vap->va_gid;
293 	sb->st_rdev = vap->va_rdev;
294 	sb->st_size = vap->va_size;
295 	sb->st_atime = vap->va_atime.tv_sec;
296 	sb->st_spare1 = 0;
297 	sb->st_mtime = vap->va_mtime.tv_sec;
298 	sb->st_spare2 = 0;
299 	sb->st_ctime = vap->va_ctime.tv_sec;
300 	sb->st_spare3 = 0;
301 	sb->st_blksize = vap->va_blocksize;
302 	sb->st_flags = vap->va_flags;
303 	sb->st_gen = vap->va_gen;
304 	sb->st_blocks = vap->va_bytes / S_BLKSIZE;
305 	return (0);
306 }
307 
308 /*
309  * Vnode ioctl call
310  */
311 vn_ioctl(fp, com, data)
312 	struct file *fp;
313 	int com;
314 	caddr_t data;
315 {
316 	register struct vnode *vp = ((struct vnode *)fp->f_data);
317 	struct vattr vattr;
318 	int error;
319 
320 	switch (vp->v_type) {
321 
322 	case VREG:
323 	case VDIR:
324 		if (com == FIONREAD) {
325 			if (error = VOP_GETATTR(vp, &vattr, u.u_cred))
326 				return (error);
327 			*(off_t *)data = vattr.va_size - fp->f_offset;
328 			return (0);
329 		}
330 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
331 			return (0);			/* XXX */
332 		/* fall into ... */
333 
334 	default:
335 		return (ENOTTY);
336 
337 	case VFIFO:
338 	case VCHR:
339 	case VBLK:
340 		u.u_r.r_val1 = 0;
341 		if (setjmp(&u.u_qsave)) {
342 			if ((u.u_sigintr & sigmask(u.u_procp->p_cursig)) != 0)
343 				return(EINTR);
344 			u.u_eosys = RESTARTSYS;
345 			return (0);
346 		}
347 		error = VOP_IOCTL(vp, com, data, fp->f_flag, u.u_cred);
348 		if (error == 0 && com == TIOCSCTTY) {
349 			u.u_procp->p_session->s_ttyvp = vp;
350 			VREF(vp);
351 		}
352 		return (error);
353 	}
354 }
355 
356 /*
357  * Vnode select call
358  */
359 vn_select(fp, which)
360 	struct file *fp;
361 	int which;
362 {
363 	return(VOP_SELECT(((struct vnode *)fp->f_data), which, fp->f_flag,
364 		u.u_cred));
365 }
366 
367 /*
368  * Vnode close call
369  */
370 vn_close(fp)
371 	register struct file *fp;
372 {
373 	struct vnode *vp = ((struct vnode *)fp->f_data);
374 	int error;
375 
376 	if (fp->f_flag & (FSHLOCK|FEXLOCK))
377 		vn_unlock(fp, FSHLOCK|FEXLOCK);
378 	/*
379 	 * Must delete vnode reference from this file entry
380 	 * before VOP_CLOSE, so that only other references
381 	 * will prevent close.
382 	 */
383 	fp->f_data = (caddr_t) 0;
384 	error = VOP_CLOSE(vp, fp->f_flag, u.u_cred);
385 	vrele(vp);
386 	return (error);
387 }
388 
389 /*
390  * Place an advisory lock on a vnode.
391  * !! THIS IMPLIES THAT ALL STATEFUL FILE SERVERS WILL USE file table entries
392  */
393 vn_lock(fp, cmd)
394 	register struct file *fp;
395 	int cmd;
396 {
397 	register int priority = PLOCK;
398 	register struct vnode *vp = (struct vnode *)fp->f_data;
399 
400 	if ((cmd & LOCK_EX) == 0)
401 		priority += 4;
402 	if (setjmp(&u.u_qsave)) {
403 		if ((u.u_sigintr & sigmask(u.u_procp->p_cursig)) != 0)
404 			return(EINTR);
405 		u.u_eosys = RESTARTSYS;
406 		return (0);
407 	}
408 	/*
409 	 * If there's a exclusive lock currently applied
410 	 * to the file, then we've gotta wait for the
411 	 * lock with everyone else.
412 	 */
413 again:
414 	while (vp->v_flag & VEXLOCK) {
415 		/*
416 		 * If we're holding an exclusive
417 		 * lock, then release it.
418 		 */
419 		if (fp->f_flag & FEXLOCK) {
420 			vn_unlock(fp, FEXLOCK);
421 			continue;
422 		}
423 		if (cmd & LOCK_NB)
424 			return (EWOULDBLOCK);
425 		vp->v_flag |= VLWAIT;
426 		tsleep((caddr_t)&vp->v_exlockc, priority, SLP_EXLCK, 0);
427 	}
428 	if ((cmd & LOCK_EX) && (vp->v_flag & VSHLOCK)) {
429 		/*
430 		 * Must wait for any shared locks to finish
431 		 * before we try to apply a exclusive lock.
432 		 *
433 		 * If we're holding a shared
434 		 * lock, then release it.
435 		 */
436 		if (fp->f_flag & FSHLOCK) {
437 			vn_unlock(fp, FSHLOCK);
438 			goto again;
439 		}
440 		if (cmd & LOCK_NB)
441 			return (EWOULDBLOCK);
442 		vp->v_flag |= VLWAIT;
443 		tsleep((caddr_t)&vp->v_shlockc, PLOCK, SLP_SHLCK, 0);
444 		goto again;
445 	}
446 	if (fp->f_flag & FEXLOCK)
447 		panic("vn_lock");
448 	if (cmd & LOCK_EX) {
449 		cmd &= ~LOCK_SH;
450 		vp->v_exlockc++;
451 		vp->v_flag |= VEXLOCK;
452 		fp->f_flag |= FEXLOCK;
453 	}
454 	if ((cmd & LOCK_SH) && (fp->f_flag & FSHLOCK) == 0) {
455 		vp->v_shlockc++;
456 		vp->v_flag |= VSHLOCK;
457 		fp->f_flag |= FSHLOCK;
458 	}
459 	return (0);
460 }
461 
462 /*
463  * Unlock a file.
464  */
465 vn_unlock(fp, kind)
466 	register struct file *fp;
467 	int kind;
468 {
469 	register struct vnode *vp = (struct vnode *)fp->f_data;
470 	int flags;
471 
472 	kind &= fp->f_flag;
473 	if (vp == NULL || kind == 0)
474 		return;
475 	flags = vp->v_flag;
476 	if (kind & FSHLOCK) {
477 		if ((flags & VSHLOCK) == 0)
478 			panic("vn_unlock: SHLOCK");
479 		if (--vp->v_shlockc == 0) {
480 			vp->v_flag &= ~VSHLOCK;
481 			if (flags & VLWAIT)
482 				wakeup((caddr_t)&vp->v_shlockc);
483 		}
484 		fp->f_flag &= ~FSHLOCK;
485 	}
486 	if (kind & FEXLOCK) {
487 		if ((flags & VEXLOCK) == 0)
488 			panic("vn_unlock: EXLOCK");
489 		if (--vp->v_exlockc == 0) {
490 			vp->v_flag &= ~(VEXLOCK|VLWAIT);
491 			if (flags & VLWAIT)
492 				wakeup((caddr_t)&vp->v_exlockc);
493 		}
494 		fp->f_flag &= ~FEXLOCK;
495 	}
496 }
497 
498 /*
499  * vn_fhtovp() - convert a fh to a vnode ptr (optionally locked)
500  * 	- look up fsid in mount list (if not found ret error)
501  *	- get vp by calling VFS_FHTOVP() macro
502  *	- if lockflag lock it with VOP_LOCK()
503  */
504 vn_fhtovp(fhp, lockflag, vpp)
505 	fhandle_t *fhp;
506 	int lockflag;
507 	struct vnode **vpp;
508 {
509 	register struct mount *mp;
510 
511 	if ((mp = getvfs(&fhp->fh_fsid)) == NULL)
512 		return (ESTALE);
513 	if (VFS_FHTOVP(mp, &fhp->fh_fid, vpp))
514 		return (ESTALE);
515 	if (!lockflag)
516 		VOP_UNLOCK(*vpp);
517 	return (0);
518 }
519 
520 /*
521  * Noop
522  */
523 vfs_noop()
524 {
525 
526 	return (ENXIO);
527 }
528 
529 /*
530  * Null op
531  */
532 vfs_nullop()
533 {
534 
535 	return (0);
536 }
537