xref: /freebsd/sys/fs/fdescfs/fdesc_vnops.c (revision ae83180158c4c937f170e31eff311b18c0286a93)
1 /*
2  * Copyright (c) 1992, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software donated to Berkeley by
6  * Jan-Simon Pendry.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)fdesc_vnops.c	8.9 (Berkeley) 1/21/94
37  *
38  * $FreeBSD$
39  */
40 
41 /*
42  * /dev/fd Filesystem
43  */
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/conf.h>
48 #include <sys/dirent.h>
49 #include <sys/filedesc.h>
50 #include <sys/kernel.h>	/* boottime */
51 #include <sys/lock.h>
52 #include <sys/mutex.h>
53 #include <sys/malloc.h>
54 #include <sys/file.h>	/* Must come after sys/malloc.h */
55 #include <sys/mount.h>
56 #include <sys/namei.h>
57 #include <sys/proc.h>
58 #include <sys/stat.h>
59 #include <sys/vnode.h>
60 
61 #include <fs/fdescfs/fdesc.h>
62 
63 #define FDL_WANT	0x01
64 #define FDL_LOCKED	0x02
65 static int fdcache_lock;
66 
67 static vop_t **fdesc_vnodeop_p;
68 
69 #define	NFDCACHE 4
70 #define FD_NHASH(ix) \
71 	(&fdhashtbl[(ix) & fdhash])
72 static LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
73 static u_long fdhash;
74 
75 static int	fdesc_getattr(struct vop_getattr_args *ap);
76 static int	fdesc_inactive(struct vop_inactive_args *ap);
77 static int	fdesc_lookup(struct vop_lookup_args *ap);
78 static int	fdesc_open(struct vop_open_args *ap);
79 static int	fdesc_print(struct vop_print_args *ap);
80 static int	fdesc_readdir(struct vop_readdir_args *ap);
81 static int	fdesc_reclaim(struct vop_reclaim_args *ap);
82 static int	fdesc_poll(struct vop_poll_args *ap);
83 static int	fdesc_setattr(struct vop_setattr_args *ap);
84 
85 /*
86  * Initialise cache headers
87  */
88 int
89 fdesc_init(vfsp)
90 	struct vfsconf *vfsp;
91 {
92 
93 	fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash);
94 	return (0);
95 }
96 
97 int
98 fdesc_allocvp(ftype, ix, mp, vpp, td)
99 	fdntype ftype;
100 	int ix;
101 	struct mount *mp;
102 	struct vnode **vpp;
103 	struct thread *td;
104 {
105 	struct fdhashhead *fc;
106 	struct fdescnode *fd;
107 	int error = 0;
108 
109 	fc = FD_NHASH(ix);
110 loop:
111 	LIST_FOREACH(fd, fc, fd_hash) {
112 		if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
113 			if (vget(fd->fd_vnode, 0, td))
114 				goto loop;
115 			*vpp = fd->fd_vnode;
116 			return (error);
117 		}
118 	}
119 
120 	/*
121 	 * otherwise lock the array while we call getnewvnode
122 	 * since that can block.
123 	 */
124 	if (fdcache_lock & FDL_LOCKED) {
125 		fdcache_lock |= FDL_WANT;
126 		(void) tsleep((caddr_t) &fdcache_lock, PINOD, "fdalvp", 0);
127 		goto loop;
128 	}
129 	fdcache_lock |= FDL_LOCKED;
130 
131 	/*
132 	 * Do the MALLOC before the getnewvnode since doing so afterward
133 	 * might cause a bogus v_data pointer to get dereferenced
134 	 * elsewhere if MALLOC should block.
135 	 */
136 	MALLOC(fd, struct fdescnode *, sizeof(struct fdescnode), M_TEMP, M_WAITOK);
137 
138 	error = getnewvnode(VT_FDESC, mp, fdesc_vnodeop_p, vpp);
139 	if (error) {
140 		FREE(fd, M_TEMP);
141 		goto out;
142 	}
143 	(*vpp)->v_data = fd;
144 	fd->fd_vnode = *vpp;
145 	fd->fd_type = ftype;
146 	fd->fd_fd = -1;
147 	fd->fd_ix = ix;
148 	LIST_INSERT_HEAD(fc, fd, fd_hash);
149 
150 out:
151 	fdcache_lock &= ~FDL_LOCKED;
152 
153 	if (fdcache_lock & FDL_WANT) {
154 		fdcache_lock &= ~FDL_WANT;
155 		wakeup((caddr_t) &fdcache_lock);
156 	}
157 
158 	return (error);
159 }
160 
161 /*
162  * vp is the current namei directory
163  * ndp is the name to locate in that directory...
164  */
165 static int
166 fdesc_lookup(ap)
167 	struct vop_lookup_args /* {
168 		struct vnode * a_dvp;
169 		struct vnode ** a_vpp;
170 		struct componentname * a_cnp;
171 	} */ *ap;
172 {
173 	struct vnode **vpp = ap->a_vpp;
174 	struct vnode *dvp = ap->a_dvp;
175 	struct componentname *cnp = ap->a_cnp;
176 	char *pname = cnp->cn_nameptr;
177 	struct thread *td = cnp->cn_thread;
178 	struct file *fp;
179 	int nlen = cnp->cn_namelen;
180 	u_int fd;
181 	int error;
182 	struct vnode *fvp;
183 
184 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME) {
185 		error = EROFS;
186 		goto bad;
187 	}
188 
189 	VOP_UNLOCK(dvp, 0, td);
190 	if (cnp->cn_namelen == 1 && *pname == '.') {
191 		*vpp = dvp;
192 		VREF(dvp);
193 		vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
194 		return (0);
195 	}
196 
197 	if (VTOFDESC(dvp)->fd_type != Froot) {
198 		error = ENOTDIR;
199 		goto bad;
200 	}
201 
202 	fd = 0;
203 	/* the only time a leading 0 is acceptable is if it's "0" */
204 	if (*pname == '0' && nlen != 1) {
205 		error = ENOENT;
206 		goto bad;
207 	}
208 	while (nlen--) {
209 		if (*pname < '0' || *pname > '9') {
210 			error = ENOENT;
211 			goto bad;
212 		}
213 		fd = 10 * fd + *pname++ - '0';
214 	}
215 
216 	if ((error = fget(td, fd, &fp)) != 0)
217 		goto bad;
218 
219 	error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp, td);
220 	fdrop(fp, td);
221 	if (error)
222 		goto bad;
223 	VTOFDESC(fvp)->fd_fd = fd;
224 	vn_lock(fvp, LK_EXCLUSIVE | LK_RETRY, td);
225 	*vpp = fvp;
226 	return (0);
227 
228 bad:
229 	vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td);
230 	*vpp = NULL;
231 	return (error);
232 }
233 
234 static int
235 fdesc_open(ap)
236 	struct vop_open_args /* {
237 		struct vnode *a_vp;
238 		int  a_mode;
239 		struct ucred *a_cred;
240 		struct thread *a_td;
241 	} */ *ap;
242 {
243 	struct vnode *vp = ap->a_vp;
244 
245 	if (VTOFDESC(vp)->fd_type == Froot)
246 		return (0);
247 
248 	/*
249 	 * XXX Kludge: set td->td_proc->p_dupfd to contain the value of the the file
250 	 * descriptor being sought for duplication. The error return ensures
251 	 * that the vnode for this device will be released by vn_open. Open
252 	 * will detect this special error and take the actions in dupfdopen.
253 	 * Other callers of vn_open or VOP_OPEN will simply report the
254 	 * error.
255 	 */
256 	ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
257 	return (ENODEV);
258 }
259 
260 static int
261 fdesc_getattr(ap)
262 	struct vop_getattr_args /* {
263 		struct vnode *a_vp;
264 		struct vattr *a_vap;
265 		struct ucred *a_cred;
266 		struct thread *a_td;
267 	} */ *ap;
268 {
269 	struct vnode *vp = ap->a_vp;
270 	struct vattr *vap = ap->a_vap;
271 	struct file *fp;
272 	struct stat stb;
273 	u_int fd;
274 	int error = 0;
275 
276 	switch (VTOFDESC(vp)->fd_type) {
277 	case Froot:
278 		VATTR_NULL(vap);
279 
280 		vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
281 		vap->va_type = VDIR;
282 		vap->va_nlink = 2;
283 		vap->va_size = DEV_BSIZE;
284 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
285 		vap->va_uid = 0;
286 		vap->va_gid = 0;
287 		vap->va_blocksize = DEV_BSIZE;
288 		vap->va_atime.tv_sec = boottime.tv_sec;
289 		vap->va_atime.tv_nsec = 0;
290 		vap->va_mtime = vap->va_atime;
291 		vap->va_ctime = vap->va_mtime;
292 		vap->va_gen = 0;
293 		vap->va_flags = 0;
294 		vap->va_rdev = 0;
295 		vap->va_bytes = 0;
296 		break;
297 
298 	case Fdesc:
299 		fd = VTOFDESC(vp)->fd_fd;
300 
301 		if ((error = fget(ap->a_td, fd, &fp)) != 0)
302 			return (error);
303 
304 		bzero(&stb, sizeof(stb));
305 		error = fo_stat(fp, &stb, ap->a_td->td_ucred, ap->a_td);
306 		fdrop(fp, ap->a_td);
307 		if (error == 0) {
308 			VATTR_NULL(vap);
309 			vap->va_type = IFTOVT(stb.st_mode);
310 			vap->va_mode = stb.st_mode;
311 #define FDRX (VREAD|VEXEC)
312 			if (vap->va_type == VDIR)
313 				vap->va_mode &= ~((FDRX)|(FDRX>>3)|(FDRX>>6));
314 #undef FDRX
315 			vap->va_nlink = 1;
316 			vap->va_flags = 0;
317 			vap->va_bytes = stb.st_blocks * stb.st_blksize;
318 			vap->va_fileid = VTOFDESC(vp)->fd_ix;
319 			vap->va_size = stb.st_size;
320 			vap->va_blocksize = stb.st_blksize;
321 			vap->va_rdev = stb.st_rdev;
322 
323 			/*
324 			 * If no time data is provided, use the current time.
325 			 */
326 			if (stb.st_atimespec.tv_sec == 0 &&
327 			    stb.st_atimespec.tv_nsec == 0)
328 				nanotime(&stb.st_atimespec);
329 
330 			if (stb.st_ctimespec.tv_sec == 0 &&
331 			    stb.st_ctimespec.tv_nsec == 0)
332 				nanotime(&stb.st_ctimespec);
333 
334 			if (stb.st_mtimespec.tv_sec == 0 &&
335 			    stb.st_mtimespec.tv_nsec == 0)
336 				nanotime(&stb.st_mtimespec);
337 
338 			vap->va_atime = stb.st_atimespec;
339 			vap->va_mtime = stb.st_mtimespec;
340 			vap->va_ctime = stb.st_ctimespec;
341 			vap->va_uid = stb.st_uid;
342 			vap->va_gid = stb.st_gid;
343 		}
344 		break;
345 
346 	default:
347 		panic("fdesc_getattr");
348 		break;
349 	}
350 
351 	if (error == 0)
352 		vp->v_type = vap->va_type;
353 	return (error);
354 }
355 
356 static int
357 fdesc_setattr(ap)
358 	struct vop_setattr_args /* {
359 		struct vnode *a_vp;
360 		struct vattr *a_vap;
361 		struct ucred *a_cred;
362 		struct thread *a_td;
363 	} */ *ap;
364 {
365 	struct vattr *vap = ap->a_vap;
366 	struct vnode *vp;
367 	struct mount *mp;
368 	struct file *fp;
369 	unsigned fd;
370 	int error;
371 
372 	/*
373 	 * Can't mess with the root vnode
374 	 */
375 	if (VTOFDESC(ap->a_vp)->fd_type == Froot)
376 		return (EACCES);
377 
378 	fd = VTOFDESC(ap->a_vp)->fd_fd;
379 
380 	/*
381 	 * Allow setattr where there is an underlying vnode.
382 	 */
383 	error = getvnode(ap->a_td->td_proc->p_fd, fd, &fp);
384 	if (error) {
385 		/*
386 		 * getvnode() returns EINVAL if the file descriptor is not
387 		 * backed by a vnode.  Silently drop all changes except
388 		 * chflags(2) in this case.
389 		 */
390 		if (error == EINVAL) {
391 			if (vap->va_flags != VNOVAL)
392 				error = EOPNOTSUPP;
393 			else
394 				error = 0;
395 		}
396 		return (error);
397 	}
398 	vp = (struct vnode *)fp->f_data;
399 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
400 		fdrop(fp, ap->a_td);
401 		return (error);
402 	}
403 	error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred, ap->a_td);
404 	vn_finished_write(mp);
405 	fdrop(fp, ap->a_td);
406 	return (error);
407 }
408 
409 #define UIO_MX 16
410 
411 static int
412 fdesc_readdir(ap)
413 	struct vop_readdir_args /* {
414 		struct vnode *a_vp;
415 		struct uio *a_uio;
416 		struct ucred *a_cred;
417 		int *a_eofflag;
418 		u_long *a_cookies;
419 		int a_ncookies;
420 	} */ *ap;
421 {
422 	struct uio *uio = ap->a_uio;
423 	struct filedesc *fdp;
424 	struct dirent d;
425 	struct dirent *dp = &d;
426 	int error, i, off, fcnt;
427 
428 	/*
429 	 * We don't allow exporting fdesc mounts, and currently local
430 	 * requests do not need cookies.
431 	 */
432 	if (ap->a_ncookies)
433 		panic("fdesc_readdir: not hungry");
434 
435 	if (VTOFDESC(ap->a_vp)->fd_type != Froot)
436 		panic("fdesc_readdir: not dir");
437 
438 	off = (int)uio->uio_offset;
439 	if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 ||
440 	    uio->uio_resid < UIO_MX)
441 		return (EINVAL);
442 	i = (u_int)off / UIO_MX;
443 	fdp = uio->uio_td->td_proc->p_fd;
444 	error = 0;
445 
446 	fcnt = i - 2;		/* The first two nodes are `.' and `..' */
447 
448 	FILEDESC_LOCK(fdp);
449 	while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) {
450 		switch (i) {
451 		case 0:	/* `.' */
452 		case 1: /* `..' */
453 			bzero((caddr_t)dp, UIO_MX);
454 
455 			dp->d_fileno = i + FD_ROOT;
456 			dp->d_namlen = i + 1;
457 			dp->d_reclen = UIO_MX;
458 			bcopy("..", dp->d_name, dp->d_namlen);
459 			dp->d_name[i + 1] = '\0';
460 			dp->d_type = DT_DIR;
461 			break;
462 		default:
463 			if (fdp->fd_ofiles[fcnt] == NULL) {
464 				FILEDESC_UNLOCK(fdp);
465 				goto done;
466 			}
467 
468 			bzero((caddr_t) dp, UIO_MX);
469 			dp->d_namlen = sprintf(dp->d_name, "%d", fcnt);
470 			dp->d_reclen = UIO_MX;
471 			dp->d_type = DT_UNKNOWN;
472 			dp->d_fileno = i + FD_DESC;
473 			break;
474 		}
475 		/*
476 		 * And ship to userland
477 		 */
478 		FILEDESC_UNLOCK(fdp);
479 		error = uiomove((caddr_t) dp, UIO_MX, uio);
480 		if (error)
481 			goto done;
482 		FILEDESC_LOCK(fdp);
483 		i++;
484 		fcnt++;
485 	}
486 	FILEDESC_UNLOCK(fdp);
487 
488 done:
489 	uio->uio_offset = i * UIO_MX;
490 	return (error);
491 }
492 
493 static int
494 fdesc_poll(ap)
495 	struct vop_poll_args /* {
496 		struct vnode *a_vp;
497 		int  a_events;
498 		struct ucred *a_cred;
499 		struct thread *a_td;
500 	} */ *ap;
501 {
502 	return seltrue(0, ap->a_events, ap->a_td);
503 }
504 
505 static int
506 fdesc_inactive(ap)
507 	struct vop_inactive_args /* {
508 		struct vnode *a_vp;
509 		struct thread *a_td;
510 	} */ *ap;
511 {
512 	struct vnode *vp = ap->a_vp;
513 
514 	/*
515 	 * Clear out the v_type field to avoid
516 	 * nasty things happening in vgone().
517 	 */
518 	VOP_UNLOCK(vp, 0, ap->a_td);
519 	vp->v_type = VNON;
520 	return (0);
521 }
522 
523 static int
524 fdesc_reclaim(ap)
525 	struct vop_reclaim_args /* {
526 		struct vnode *a_vp;
527 	} */ *ap;
528 {
529 	struct vnode *vp = ap->a_vp;
530 	struct fdescnode *fd = VTOFDESC(vp);
531 
532 	LIST_REMOVE(fd, fd_hash);
533 	FREE(vp->v_data, M_TEMP);
534 	vp->v_data = 0;
535 
536 	return (0);
537 }
538 
539 /*
540  * Print out the contents of a /dev/fd vnode.
541  */
542 /* ARGSUSED */
543 static int
544 fdesc_print(ap)
545 	struct vop_print_args /* {
546 		struct vnode *a_vp;
547 	} */ *ap;
548 {
549 
550 	printf("tag VT_NON, fdesc vnode\n");
551 	return (0);
552 }
553 
554 static struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = {
555 	{ &vop_default_desc,		(vop_t *) vop_defaultop },
556 	{ &vop_access_desc,		(vop_t *) vop_null },
557 	{ &vop_getattr_desc,		(vop_t *) fdesc_getattr },
558 	{ &vop_inactive_desc,		(vop_t *) fdesc_inactive },
559 	{ &vop_lookup_desc,		(vop_t *) fdesc_lookup },
560 	{ &vop_open_desc,		(vop_t *) fdesc_open },
561 	{ &vop_pathconf_desc,		(vop_t *) vop_stdpathconf },
562 	{ &vop_poll_desc,		(vop_t *) fdesc_poll },
563 	{ &vop_print_desc,		(vop_t *) fdesc_print },
564 	{ &vop_readdir_desc,		(vop_t *) fdesc_readdir },
565 	{ &vop_reclaim_desc,		(vop_t *) fdesc_reclaim },
566 	{ &vop_setattr_desc,		(vop_t *) fdesc_setattr },
567 	{ &vop_lock_desc,		(vop_t *) vop_stdlock },
568 	{ &vop_unlock_desc,		(vop_t *) vop_stdunlock },
569 	{ &vop_islocked_desc,		(vop_t *) vop_stdislocked },
570 	{ NULL, NULL }
571 };
572 static struct vnodeopv_desc fdesc_vnodeop_opv_desc =
573 	{ &fdesc_vnodeop_p, fdesc_vnodeop_entries };
574 
575 VNODEOP_SET(fdesc_vnodeop_opv_desc);
576