xref: /netbsd-src/sys/miscfs/fdesc/fdesc_vnops.c (revision 5b84b3983f71fd20a534cfa5d1556623a8aaa717)
1 /*	$NetBSD: fdesc_vnops.c,v 1.86 2005/08/30 20:08:01 xtraeme Exp $	*/
2 
3 /*
4  * Copyright (c) 1992, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software donated to Berkeley by
8  * Jan-Simon Pendry.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)fdesc_vnops.c	8.17 (Berkeley) 5/22/95
35  *
36  * #Id: fdesc_vnops.c,v 1.12 1993/04/06 16:17:17 jsp Exp #
37  */
38 
39 /*
40  * /dev/fd Filesystem
41  */
42 
43 #include <sys/cdefs.h>
44 __KERNEL_RCSID(0, "$NetBSD: fdesc_vnops.c,v 1.86 2005/08/30 20:08:01 xtraeme Exp $");
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/time.h>
49 #include <sys/proc.h>
50 #include <sys/kernel.h>	/* boottime */
51 #include <sys/resourcevar.h>
52 #include <sys/socketvar.h>
53 #include <sys/filedesc.h>
54 #include <sys/vnode.h>
55 #include <sys/malloc.h>
56 #include <sys/conf.h>
57 #include <sys/file.h>
58 #include <sys/stat.h>
59 #include <sys/mount.h>
60 #include <sys/namei.h>
61 #include <sys/buf.h>
62 #include <sys/dirent.h>
63 #include <sys/tty.h>
64 
65 #include <miscfs/fdesc/fdesc.h>
66 #include <miscfs/genfs/genfs.h>
67 
68 #define cttyvp(p) ((p)->p_flag & P_CONTROLT ? (p)->p_session->s_ttyvp : NULL)
69 
70 #define FDL_WANT	0x01
71 #define FDL_LOCKED	0x02
72 static int fdcache_lock;
73 
74 dev_t devctty;
75 
76 #if (FD_STDIN != FD_STDOUT-1) || (FD_STDOUT != FD_STDERR-1)
77 FD_STDIN, FD_STDOUT, FD_STDERR must be a sequence n, n+1, n+2
78 #endif
79 
80 #define	NFDCACHE 4
81 
82 #define FD_NHASH(ix) \
83 	(&fdhashtbl[(ix) & fdhash])
84 LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
85 u_long fdhash;
86 
87 int	fdesc_lookup(void *);
88 #define	fdesc_create	genfs_eopnotsupp
89 #define	fdesc_mknod	genfs_eopnotsupp
90 int	fdesc_open(void *);
91 #define	fdesc_close	genfs_nullop
92 #define	fdesc_access	genfs_nullop
93 int	fdesc_getattr(void *);
94 int	fdesc_setattr(void *);
95 int	fdesc_read(void *);
96 int	fdesc_write(void *);
97 int	fdesc_ioctl(void *);
98 int	fdesc_poll(void *);
99 int	fdesc_kqfilter(void *);
100 #define	fdesc_mmap	genfs_eopnotsupp
101 #define	fdesc_fcntl	genfs_fcntl
102 #define	fdesc_fsync	genfs_nullop
103 #define	fdesc_seek	genfs_seek
104 #define	fdesc_remove	genfs_eopnotsupp
105 int	fdesc_link(void *);
106 #define	fdesc_rename	genfs_eopnotsupp
107 #define	fdesc_mkdir	genfs_eopnotsupp
108 #define	fdesc_rmdir	genfs_eopnotsupp
109 int	fdesc_symlink(void *);
110 int	fdesc_readdir(void *);
111 int	fdesc_readlink(void *);
112 #define	fdesc_abortop	genfs_abortop
113 int	fdesc_inactive(void *);
114 int	fdesc_reclaim(void *);
115 #define	fdesc_lock	genfs_lock
116 #define	fdesc_unlock	genfs_unlock
117 #define	fdesc_bmap	genfs_badop
118 #define	fdesc_strategy	genfs_badop
119 int	fdesc_print(void *);
120 int	fdesc_pathconf(void *);
121 #define	fdesc_islocked	genfs_islocked
122 #define	fdesc_advlock	genfs_einval
123 #define	fdesc_blkatoff	genfs_eopnotsupp
124 #define	fdesc_valloc	genfs_eopnotsupp
125 #define	fdesc_vfree	genfs_nullop
126 #define	fdesc_truncate	genfs_eopnotsupp
127 #define	fdesc_update	genfs_nullop
128 #define	fdesc_bwrite	genfs_eopnotsupp
129 #define fdesc_revoke	genfs_revoke
130 #define fdesc_putpages	genfs_null_putpages
131 
132 static int fdesc_attr(int, struct vattr *, struct ucred *, struct proc *);
133 
134 int (**fdesc_vnodeop_p)(void *);
135 const struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = {
136 	{ &vop_default_desc, vn_default_error },
137 	{ &vop_lookup_desc, fdesc_lookup },		/* lookup */
138 	{ &vop_create_desc, fdesc_create },		/* create */
139 	{ &vop_mknod_desc, fdesc_mknod },		/* mknod */
140 	{ &vop_open_desc, fdesc_open },			/* open */
141 	{ &vop_close_desc, fdesc_close },		/* close */
142 	{ &vop_access_desc, fdesc_access },		/* access */
143 	{ &vop_getattr_desc, fdesc_getattr },		/* getattr */
144 	{ &vop_setattr_desc, fdesc_setattr },		/* setattr */
145 	{ &vop_read_desc, fdesc_read },			/* read */
146 	{ &vop_write_desc, fdesc_write },		/* write */
147 	{ &vop_ioctl_desc, fdesc_ioctl },		/* ioctl */
148 	{ &vop_fcntl_desc, fdesc_fcntl },		/* fcntl */
149 	{ &vop_poll_desc, fdesc_poll },			/* poll */
150 	{ &vop_kqfilter_desc, fdesc_kqfilter },		/* kqfilter */
151 	{ &vop_revoke_desc, fdesc_revoke },		/* revoke */
152 	{ &vop_mmap_desc, fdesc_mmap },			/* mmap */
153 	{ &vop_fsync_desc, fdesc_fsync },		/* fsync */
154 	{ &vop_seek_desc, fdesc_seek },			/* seek */
155 	{ &vop_remove_desc, fdesc_remove },		/* remove */
156 	{ &vop_link_desc, fdesc_link },			/* link */
157 	{ &vop_rename_desc, fdesc_rename },		/* rename */
158 	{ &vop_mkdir_desc, fdesc_mkdir },		/* mkdir */
159 	{ &vop_rmdir_desc, fdesc_rmdir },		/* rmdir */
160 	{ &vop_symlink_desc, fdesc_symlink },		/* symlink */
161 	{ &vop_readdir_desc, fdesc_readdir },		/* readdir */
162 	{ &vop_readlink_desc, fdesc_readlink },		/* readlink */
163 	{ &vop_abortop_desc, fdesc_abortop },		/* abortop */
164 	{ &vop_inactive_desc, fdesc_inactive },		/* inactive */
165 	{ &vop_reclaim_desc, fdesc_reclaim },		/* reclaim */
166 	{ &vop_lock_desc, fdesc_lock },			/* lock */
167 	{ &vop_unlock_desc, fdesc_unlock },		/* unlock */
168 	{ &vop_bmap_desc, fdesc_bmap },			/* bmap */
169 	{ &vop_strategy_desc, fdesc_strategy },		/* strategy */
170 	{ &vop_print_desc, fdesc_print },		/* print */
171 	{ &vop_islocked_desc, fdesc_islocked },		/* islocked */
172 	{ &vop_pathconf_desc, fdesc_pathconf },		/* pathconf */
173 	{ &vop_advlock_desc, fdesc_advlock },		/* advlock */
174 	{ &vop_blkatoff_desc, fdesc_blkatoff },		/* blkatoff */
175 	{ &vop_valloc_desc, fdesc_valloc },		/* valloc */
176 	{ &vop_vfree_desc, fdesc_vfree },		/* vfree */
177 	{ &vop_truncate_desc, fdesc_truncate },		/* truncate */
178 	{ &vop_update_desc, fdesc_update },		/* update */
179 	{ &vop_bwrite_desc, fdesc_bwrite },		/* bwrite */
180 	{ &vop_putpages_desc, fdesc_putpages },		/* putpages */
181 	{ NULL, NULL }
182 };
183 
184 const struct vnodeopv_desc fdesc_vnodeop_opv_desc =
185 	{ &fdesc_vnodeop_p, fdesc_vnodeop_entries };
186 
187 extern const struct cdevsw ctty_cdevsw;
188 
189 /*
190  * Initialise cache headers
191  */
192 void
193 fdesc_init()
194 {
195 	int cttymajor;
196 
197 	/* locate the major number */
198 	cttymajor = cdevsw_lookup_major(&ctty_cdevsw);
199 	devctty = makedev(cttymajor, 0);
200 	fdhashtbl = hashinit(NFDCACHE, HASH_LIST, M_CACHE, M_NOWAIT, &fdhash);
201 }
202 
203 /*
204  * Free hash table.
205  */
206 void
207 fdesc_done()
208 {
209 	hashdone(fdhashtbl, M_CACHE);
210 }
211 
212 /*
213  * Return a locked vnode of the correct type.
214  */
215 int
216 fdesc_allocvp(ftype, ix, mp, vpp)
217 	fdntype ftype;
218 	int ix;
219 	struct mount *mp;
220 	struct vnode **vpp;
221 {
222 	struct fdhashhead *fc;
223 	struct fdescnode *fd;
224 	int error = 0;
225 
226 	fc = FD_NHASH(ix);
227 loop:
228 	for (fd = fc->lh_first; fd != 0; fd = fd->fd_hash.le_next) {
229 		if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
230 			if (vget(fd->fd_vnode, LK_EXCLUSIVE))
231 				goto loop;
232 			*vpp = fd->fd_vnode;
233 			return (error);
234 		}
235 	}
236 
237 	/*
238 	 * otherwise lock the array while we call getnewvnode
239 	 * since that can block.
240 	 */
241 	if (fdcache_lock & FDL_LOCKED) {
242 		fdcache_lock |= FDL_WANT;
243 		(void) tsleep(&fdcache_lock, PINOD, "fdcache", 0);
244 		goto loop;
245 	}
246 	fdcache_lock |= FDL_LOCKED;
247 
248 	error = getnewvnode(VT_FDESC, mp, fdesc_vnodeop_p, vpp);
249 	if (error)
250 		goto out;
251 	MALLOC(fd, void *, sizeof(struct fdescnode), M_TEMP, M_WAITOK);
252 	(*vpp)->v_data = fd;
253 	fd->fd_vnode = *vpp;
254 	fd->fd_type = ftype;
255 	fd->fd_fd = -1;
256 	fd->fd_link = 0;
257 	fd->fd_ix = ix;
258 	VOP_LOCK(*vpp, LK_EXCLUSIVE);
259 	LIST_INSERT_HEAD(fc, fd, fd_hash);
260 
261 out:;
262 	fdcache_lock &= ~FDL_LOCKED;
263 
264 	if (fdcache_lock & FDL_WANT) {
265 		fdcache_lock &= ~FDL_WANT;
266 		wakeup(&fdcache_lock);
267 	}
268 
269 	return (error);
270 }
271 
272 /*
273  * vp is the current namei directory
274  * ndp is the name to locate in that directory...
275  */
276 int
277 fdesc_lookup(v)
278 	void *v;
279 {
280 	struct vop_lookup_args /* {
281 		struct vnode * a_dvp;
282 		struct vnode ** a_vpp;
283 		struct componentname * a_cnp;
284 	} */ *ap = v;
285 	struct vnode **vpp = ap->a_vpp;
286 	struct vnode *dvp = ap->a_dvp;
287 	struct componentname *cnp = ap->a_cnp;
288 	struct proc *p = cnp->cn_proc;
289 	const char *pname = cnp->cn_nameptr;
290 	int numfiles = p->p_fd->fd_nfiles;
291 	unsigned fd = 0;
292 	int error;
293 	struct vnode *fvp;
294 	const char *ln;
295 
296 	if (cnp->cn_namelen == 1 && *pname == '.') {
297 		*vpp = dvp;
298 		VREF(dvp);
299 		return (0);
300 	}
301 
302 	switch (VTOFDESC(dvp)->fd_type) {
303 	default:
304 	case Flink:
305 	case Fdesc:
306 	case Fctty:
307 		error = ENOTDIR;
308 		goto bad;
309 
310 	case Froot:
311 		if (cnp->cn_namelen == 2 && memcmp(pname, "fd", 2) == 0) {
312 			error = fdesc_allocvp(Fdevfd, FD_DEVFD, dvp->v_mount, &fvp);
313 			if (error)
314 				goto bad;
315 			*vpp = fvp;
316 			fvp->v_type = VDIR;
317 			goto good;
318 		}
319 
320 		if (cnp->cn_namelen == 3 && memcmp(pname, "tty", 3) == 0) {
321 			struct vnode *ttyvp = cttyvp(p);
322 			if (ttyvp == NULL) {
323 				error = ENXIO;
324 				goto bad;
325 			}
326 			error = fdesc_allocvp(Fctty, FD_CTTY, dvp->v_mount, &fvp);
327 			if (error)
328 				goto bad;
329 			*vpp = fvp;
330 			fvp->v_type = VCHR;
331 			goto good;
332 		}
333 
334 		ln = 0;
335 		switch (cnp->cn_namelen) {
336 		case 5:
337 			if (memcmp(pname, "stdin", 5) == 0) {
338 				ln = "fd/0";
339 				fd = FD_STDIN;
340 			}
341 			break;
342 		case 6:
343 			if (memcmp(pname, "stdout", 6) == 0) {
344 				ln = "fd/1";
345 				fd = FD_STDOUT;
346 			} else
347 			if (memcmp(pname, "stderr", 6) == 0) {
348 				ln = "fd/2";
349 				fd = FD_STDERR;
350 			}
351 			break;
352 		}
353 
354 		if (ln) {
355 			error = fdesc_allocvp(Flink, fd, dvp->v_mount, &fvp);
356 			if (error)
357 				goto bad;
358 			/* XXXUNCONST */
359 			VTOFDESC(fvp)->fd_link = __UNCONST(ln);
360 			*vpp = fvp;
361 			fvp->v_type = VLNK;
362 			goto good;
363 		} else {
364 			error = ENOENT;
365 			goto bad;
366 		}
367 
368 		/* FALL THROUGH */
369 
370 	case Fdevfd:
371 		if (cnp->cn_namelen == 2 && memcmp(pname, "..", 2) == 0) {
372 			VOP_UNLOCK(dvp, 0);
373 			cnp->cn_flags |= PDIRUNLOCK;
374 			error = fdesc_root(dvp->v_mount, vpp);
375 			if (error)
376 				goto bad;
377 			/*
378 			 * If we're at the last component and need the
379 			 * parent locked, undo the unlock above.
380 			 */
381 			if (((~cnp->cn_flags & (ISLASTCN | LOCKPARENT)) == 0) &&
382 				   ((error = vn_lock(dvp, LK_EXCLUSIVE)) == 0))
383 				cnp->cn_flags &= ~PDIRUNLOCK;
384 			return (error);
385 		}
386 
387 		fd = 0;
388 		while (*pname >= '0' && *pname <= '9') {
389 			fd = 10 * fd + *pname++ - '0';
390 			if (fd >= numfiles)
391 				break;
392 		}
393 
394 		if (*pname != '\0') {
395 			error = ENOENT;
396 			goto bad;
397 		}
398 
399 		if (fd >= numfiles || p->p_fd->fd_ofiles[fd] == NULL ||
400 		    FILE_IS_USABLE(p->p_fd->fd_ofiles[fd]) == 0) {
401 			error = EBADF;
402 			goto bad;
403 		}
404 
405 		error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp);
406 		if (error)
407 			goto bad;
408 		VTOFDESC(fvp)->fd_fd = fd;
409 		*vpp = fvp;
410 		goto good;
411 	}
412 
413 bad:;
414 	*vpp = NULL;
415 	return (error);
416 
417 good:;
418 	/*
419 	 * As "." was special cased above, we now unlock the parent if we're
420 	 * suppoed to. We're only supposed to not unlock if this is the
421 	 * last component, and the caller requested LOCKPARENT. So if either
422 	 * condition is false, unlock.
423 	 */
424 	if (((~cnp->cn_flags) & (ISLASTCN | LOCKPARENT)) != 0) {
425 		VOP_UNLOCK(dvp, 0);
426 		cnp->cn_flags |= PDIRUNLOCK;
427 	}
428 	return (0);
429 }
430 
431 int
432 fdesc_open(v)
433 	void *v;
434 {
435 	struct vop_open_args /* {
436 		struct vnode *a_vp;
437 		int  a_mode;
438 		struct ucred *a_cred;
439 		struct proc *a_p;
440 	} */ *ap = v;
441 	struct vnode *vp = ap->a_vp;
442 
443 	switch (VTOFDESC(vp)->fd_type) {
444 	case Fdesc:
445 		/*
446 		 * XXX Kludge: set dupfd to contain the value of the
447 		 * the file descriptor being sought for duplication. The error
448 		 * return ensures that the vnode for this device will be
449 		 * released by vn_open. Open will detect this special error and
450 		 * take the actions in dupfdopen.  Other callers of vn_open or
451 		 * VOP_OPEN will simply report the error.
452 		 */
453 		curlwp->l_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
454 		return EDUPFD;
455 
456 	case Fctty:
457 		return ((*ctty_cdevsw.d_open)(devctty, ap->a_mode, 0, ap->a_p));
458 	case Froot:
459 	case Fdevfd:
460 	case Flink:
461 		break;
462 	}
463 
464 	return (0);
465 }
466 
467 static int
468 fdesc_attr(fd, vap, cred, p)
469 	int fd;
470 	struct vattr *vap;
471 	struct ucred *cred;
472 	struct proc *p;
473 {
474 	struct filedesc *fdp = p->p_fd;
475 	struct file *fp;
476 	struct stat stb;
477 	int error;
478 
479 	if ((fp = fd_getfile(fdp, fd)) == NULL)
480 		return (EBADF);
481 
482 	switch (fp->f_type) {
483 	case DTYPE_VNODE:
484 		simple_unlock(&fp->f_slock);
485 		error = VOP_GETATTR((struct vnode *) fp->f_data, vap, cred, p);
486 		if (error == 0 && vap->va_type == VDIR) {
487 			/*
488 			 * directories can cause loops in the namespace,
489 			 * so turn off the 'x' bits to avoid trouble.
490 			 */
491 			vap->va_mode &= ~(S_IXUSR|S_IXGRP|S_IXOTH);
492 		}
493 		break;
494 
495 	default:
496 		FILE_USE(fp);
497 		memset(&stb, 0, sizeof(stb));
498 		error = (*fp->f_ops->fo_stat)(fp, &stb, p);
499 		FILE_UNUSE(fp, p);
500 		if (error)
501 			break;
502 
503 		vattr_null(vap);
504 		switch(fp->f_type) {
505 		case DTYPE_SOCKET:
506 			vap->va_type = VSOCK;
507 			break;
508 		case DTYPE_PIPE:
509 			vap->va_type = VFIFO;
510 			break;
511 		default:
512 			/* use VNON perhaps? */
513 			vap->va_type = VBAD;
514 			break;
515 		}
516 		vap->va_mode = stb.st_mode;
517 		vap->va_nlink = stb.st_nlink;
518 		vap->va_uid = stb.st_uid;
519 		vap->va_gid = stb.st_gid;
520 		vap->va_fsid = stb.st_dev;
521 		vap->va_fileid = stb.st_ino;
522 		vap->va_size = stb.st_size;
523 		vap->va_blocksize = stb.st_blksize;
524 		vap->va_atime = stb.st_atimespec;
525 		vap->va_mtime = stb.st_mtimespec;
526 		vap->va_ctime = stb.st_ctimespec;
527 		vap->va_gen = stb.st_gen;
528 		vap->va_flags = stb.st_flags;
529 		vap->va_rdev = stb.st_rdev;
530 		vap->va_bytes = stb.st_blocks * stb.st_blksize;
531 		break;
532 	}
533 
534 	return (error);
535 }
536 
537 int
538 fdesc_getattr(v)
539 	void *v;
540 {
541 	struct vop_getattr_args /* {
542 		struct vnode *a_vp;
543 		struct vattr *a_vap;
544 		struct ucred *a_cred;
545 		struct proc *a_p;
546 	} */ *ap = v;
547 	struct vnode *vp = ap->a_vp;
548 	struct vattr *vap = ap->a_vap;
549 	unsigned fd;
550 	int error = 0;
551 
552 	switch (VTOFDESC(vp)->fd_type) {
553 	case Froot:
554 	case Fdevfd:
555 	case Flink:
556 	case Fctty:
557 		VATTR_NULL(vap);
558 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
559 
560 #define R_ALL (S_IRUSR|S_IRGRP|S_IROTH)
561 #define W_ALL (S_IWUSR|S_IWGRP|S_IWOTH)
562 #define X_ALL (S_IXUSR|S_IXGRP|S_IXOTH)
563 
564 		switch (VTOFDESC(vp)->fd_type) {
565 		case Flink:
566 			vap->va_mode = R_ALL|X_ALL;
567 			vap->va_type = VLNK;
568 			vap->va_rdev = 0;
569 			vap->va_nlink = 1;
570 			vap->va_size = strlen(VTOFDESC(vp)->fd_link);
571 			break;
572 
573 		case Fctty:
574 			vap->va_mode = R_ALL|W_ALL;
575 			vap->va_type = VCHR;
576 			vap->va_rdev = devctty;
577 			vap->va_nlink = 1;
578 			vap->va_size = 0;
579 			break;
580 
581 		default:
582 			vap->va_mode = R_ALL|X_ALL;
583 			vap->va_type = VDIR;
584 			vap->va_rdev = 0;
585 			vap->va_nlink = 2;
586 			vap->va_size = DEV_BSIZE;
587 			break;
588 		}
589 		vap->va_uid = 0;
590 		vap->va_gid = 0;
591 		vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0];
592 		vap->va_blocksize = DEV_BSIZE;
593 		vap->va_atime.tv_sec = boottime.tv_sec;
594 		vap->va_atime.tv_nsec = 0;
595 		vap->va_mtime = vap->va_atime;
596 		vap->va_ctime = vap->va_mtime;
597 		vap->va_gen = 0;
598 		vap->va_flags = 0;
599 		vap->va_bytes = 0;
600 		break;
601 
602 	case Fdesc:
603 		fd = VTOFDESC(vp)->fd_fd;
604 		error = fdesc_attr(fd, vap, ap->a_cred, ap->a_p);
605 		break;
606 
607 	default:
608 		panic("fdesc_getattr");
609 		break;
610 	}
611 
612 	if (error == 0)
613 		vp->v_type = vap->va_type;
614 
615 	return (error);
616 }
617 
618 int
619 fdesc_setattr(v)
620 	void *v;
621 {
622 	struct vop_setattr_args /* {
623 		struct vnode *a_vp;
624 		struct vattr *a_vap;
625 		struct ucred *a_cred;
626 		struct proc *a_p;
627 	} */ *ap = v;
628 	struct filedesc *fdp = ap->a_p->p_fd;
629 	struct file *fp;
630 	unsigned fd;
631 
632 	/*
633 	 * Can't mess with the root vnode
634 	 */
635 	switch (VTOFDESC(ap->a_vp)->fd_type) {
636 	case Fdesc:
637 		break;
638 
639 	case Fctty:
640 		return (0);
641 
642 	default:
643 		return (EACCES);
644 	}
645 
646 	fd = VTOFDESC(ap->a_vp)->fd_fd;
647 	if ((fp = fd_getfile(fdp, fd)) == NULL)
648 		return (EBADF);
649 
650 	/*
651 	 * XXX: Can't reasonably set the attr's on any types currently.
652 	 *      On vnode's this will cause truncation and socket/pipes make
653 	 *      no sense.
654 	 */
655 	simple_unlock(&fp->f_slock);
656 	return (0);
657 }
658 
659 
660 struct fdesc_target {
661 	ino_t ft_fileno;
662 	u_char ft_type;
663 	u_char ft_namlen;
664 	const char *ft_name;
665 } fdesc_targets[] = {
666 #define N(s) sizeof(s)-1, s
667 	{ FD_DEVFD,  DT_DIR,     N("fd")     },
668 	{ FD_STDIN,  DT_LNK,     N("stdin")  },
669 	{ FD_STDOUT, DT_LNK,     N("stdout") },
670 	{ FD_STDERR, DT_LNK,     N("stderr") },
671 	{ FD_CTTY,   DT_UNKNOWN, N("tty")    },
672 #undef N
673 #define UIO_MX _DIRENT_RECLEN((struct dirent *)NULL, sizeof("stderr") - 1)
674 };
675 static int nfdesc_targets = sizeof(fdesc_targets) / sizeof(fdesc_targets[0]);
676 
677 int
678 fdesc_readdir(v)
679 	void *v;
680 {
681 	struct vop_readdir_args /* {
682 		struct vnode *a_vp;
683 		struct uio *a_uio;
684 		struct ucred *a_cred;
685 		int *a_eofflag;
686 		off_t **a_cookies;
687 		int *a_ncookies;
688 	} */ *ap = v;
689 	struct uio *uio = ap->a_uio;
690 	struct dirent d;
691 	struct filedesc *fdp;
692 	off_t i;
693 	int error;
694 	off_t *cookies = NULL;
695 	int ncookies = 0;
696 
697 	switch (VTOFDESC(ap->a_vp)->fd_type) {
698 	case Fctty:
699 		return (0);
700 
701 	case Fdesc:
702 		return (ENOTDIR);
703 
704 	default:
705 		break;
706 	}
707 
708 	fdp = uio->uio_procp->p_fd;
709 
710 	if (uio->uio_resid < UIO_MX)
711 		return (EINVAL);
712 	if (uio->uio_offset < 0)
713 		return (EINVAL);
714 
715 	error = 0;
716 	i = uio->uio_offset;
717 	memset(&d, 0, UIO_MX);
718 	d.d_reclen = UIO_MX;
719 	if (ap->a_ncookies)
720 		ncookies = (uio->uio_resid / UIO_MX);
721 
722 	if (VTOFDESC(ap->a_vp)->fd_type == Froot) {
723 		struct fdesc_target *ft;
724 
725 		if (i >= nfdesc_targets)
726 			return 0;
727 
728 		if (ap->a_ncookies) {
729 			ncookies = min(ncookies, (nfdesc_targets - i));
730 			cookies = malloc(ncookies * sizeof(off_t),
731 			    M_TEMP, M_WAITOK);
732 			*ap->a_cookies = cookies;
733 			*ap->a_ncookies = ncookies;
734 		}
735 
736 		for (ft = &fdesc_targets[i];
737 		     uio->uio_resid >= UIO_MX && i < nfdesc_targets; ft++, i++) {
738 			switch (ft->ft_fileno) {
739 			case FD_CTTY:
740 				if (cttyvp(uio->uio_procp) == NULL)
741 					continue;
742 				break;
743 
744 			case FD_STDIN:
745 			case FD_STDOUT:
746 			case FD_STDERR:
747 				if ((ft->ft_fileno - FD_STDIN) >= fdp->fd_nfiles)
748 					continue;
749 				if (fdp->fd_ofiles[ft->ft_fileno - FD_STDIN] == NULL
750 				    || FILE_IS_USABLE(fdp->fd_ofiles[ft->ft_fileno - FD_STDIN]) == 0)
751 					continue;
752 				break;
753 			}
754 
755 			d.d_fileno = ft->ft_fileno;
756 			d.d_namlen = ft->ft_namlen;
757 			memcpy(d.d_name, ft->ft_name, ft->ft_namlen + 1);
758 			d.d_type = ft->ft_type;
759 
760 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
761 				break;
762 			if (cookies)
763 				*cookies++ = i + 1;
764 		}
765 	} else {
766 		if (ap->a_ncookies) {
767 			ncookies = min(ncookies, (fdp->fd_nfiles + 2));
768 			cookies = malloc(ncookies * sizeof(off_t),
769 			    M_TEMP, M_WAITOK);
770 			*ap->a_cookies = cookies;
771 			*ap->a_ncookies = ncookies;
772 		}
773 		for (; i - 2 < fdp->fd_nfiles && uio->uio_resid >= UIO_MX;
774 		     i++) {
775 			switch (i) {
776 			case 0:
777 			case 1:
778 				d.d_fileno = FD_ROOT;		/* XXX */
779 				d.d_namlen = i + 1;
780 				memcpy(d.d_name, "..", d.d_namlen);
781 				d.d_name[i + 1] = '\0';
782 				d.d_type = DT_DIR;
783 				break;
784 
785 			default:
786 				if (fdp->fd_ofiles[i - 2] == NULL ||
787 				    FILE_IS_USABLE(fdp->fd_ofiles[i - 2]) == 0)
788 					continue;
789 				d.d_fileno = i - 2 + FD_STDIN;
790 				d.d_namlen = sprintf(d.d_name, "%d", (int) i - 2);
791 				d.d_type = DT_UNKNOWN;
792 				break;
793 			}
794 
795 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
796 				break;
797 			if (cookies)
798 				*cookies++ = i + 1;
799 		}
800 	}
801 
802 	if (ap->a_ncookies && error) {
803 		free(*ap->a_cookies, M_TEMP);
804 		*ap->a_ncookies = 0;
805 		*ap->a_cookies = NULL;
806 	}
807 
808 	uio->uio_offset = i;
809 	return (error);
810 }
811 
812 int
813 fdesc_readlink(v)
814 	void *v;
815 {
816 	struct vop_readlink_args /* {
817 		struct vnode *a_vp;
818 		struct uio *a_uio;
819 		struct ucred *a_cred;
820 	} */ *ap = v;
821 	struct vnode *vp = ap->a_vp;
822 	int error;
823 
824 	if (vp->v_type != VLNK)
825 		return (EPERM);
826 
827 	if (VTOFDESC(vp)->fd_type == Flink) {
828 		char *ln = VTOFDESC(vp)->fd_link;
829 		error = uiomove(ln, strlen(ln), ap->a_uio);
830 	} else {
831 		error = EOPNOTSUPP;
832 	}
833 
834 	return (error);
835 }
836 
837 int
838 fdesc_read(v)
839 	void *v;
840 {
841 	struct vop_read_args /* {
842 		struct vnode *a_vp;
843 		struct uio *a_uio;
844 		int  a_ioflag;
845 		struct ucred *a_cred;
846 	} */ *ap = v;
847 	int error = EOPNOTSUPP;
848 	struct vnode *vp = ap->a_vp;
849 
850 	switch (VTOFDESC(vp)->fd_type) {
851 	case Fctty:
852 		VOP_UNLOCK(vp, 0);
853 		error = (*ctty_cdevsw.d_read)(devctty, ap->a_uio, ap->a_ioflag);
854 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
855 		break;
856 
857 	default:
858 		error = EOPNOTSUPP;
859 		break;
860 	}
861 
862 	return (error);
863 }
864 
865 int
866 fdesc_write(v)
867 	void *v;
868 {
869 	struct vop_write_args /* {
870 		struct vnode *a_vp;
871 		struct uio *a_uio;
872 		int  a_ioflag;
873 		struct ucred *a_cred;
874 	} */ *ap = v;
875 	int error = EOPNOTSUPP;
876 	struct vnode *vp = ap->a_vp;
877 
878 	switch (VTOFDESC(vp)->fd_type) {
879 	case Fctty:
880 		VOP_UNLOCK(vp, 0);
881 		error = (*ctty_cdevsw.d_write)(devctty, ap->a_uio,
882 					       ap->a_ioflag);
883 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
884 		break;
885 
886 	default:
887 		error = EOPNOTSUPP;
888 		break;
889 	}
890 
891 	return (error);
892 }
893 
894 int
895 fdesc_ioctl(v)
896 	void *v;
897 {
898 	struct vop_ioctl_args /* {
899 		struct vnode *a_vp;
900 		u_long a_command;
901 		void *a_data;
902 		int  a_fflag;
903 		struct ucred *a_cred;
904 		struct proc *a_p;
905 	} */ *ap = v;
906 	int error = EOPNOTSUPP;
907 
908 	switch (VTOFDESC(ap->a_vp)->fd_type) {
909 	case Fctty:
910 		error = (*ctty_cdevsw.d_ioctl)(devctty, ap->a_command,
911 					       ap->a_data, ap->a_fflag,
912 					       ap->a_p);
913 		break;
914 
915 	default:
916 		error = EOPNOTSUPP;
917 		break;
918 	}
919 
920 	return (error);
921 }
922 
923 int
924 fdesc_poll(v)
925 	void *v;
926 {
927 	struct vop_poll_args /* {
928 		struct vnode *a_vp;
929 		int a_events;
930 		struct proc *a_p;
931 	} */ *ap = v;
932 	int revents;
933 
934 	switch (VTOFDESC(ap->a_vp)->fd_type) {
935 	case Fctty:
936 		revents = (*ctty_cdevsw.d_poll)(devctty, ap->a_events, ap->a_p);
937 		break;
938 
939 	default:
940 		revents = genfs_poll(v);
941 		break;
942 	}
943 
944 	return (revents);
945 }
946 
947 int
948 fdesc_kqfilter(v)
949 	void *v;
950 {
951 	struct vop_kqfilter_args /* {
952 		struct vnode *a_vp;
953 		struct knote *a_kn;
954 	} */ *ap = v;
955 	int error;
956 	struct proc *p;
957 	struct file *fp;
958 
959 	switch (VTOFDESC(ap->a_vp)->fd_type) {
960 	case Fctty:
961 		error = (*ctty_cdevsw.d_kqfilter)(devctty, ap->a_kn);
962 		break;
963 
964 	case Fdesc:
965 		/* just invoke kqfilter for the underlying descriptor */
966 		p = curproc;	/* XXX hopefully ok to use curproc here */
967 		if ((fp = fd_getfile(p->p_fd, VTOFDESC(ap->a_vp)->fd_fd)) == NULL)
968 			return (1);
969 
970 		FILE_USE(fp);
971 		error = (*fp->f_ops->fo_kqfilter)(fp, ap->a_kn);
972 		FILE_UNUSE(fp, p);
973 		break;
974 
975 	default:
976 		return (genfs_kqfilter(v));
977 	}
978 
979 	return (error);
980 }
981 
982 int
983 fdesc_inactive(v)
984 	void *v;
985 {
986 	struct vop_inactive_args /* {
987 		struct vnode *a_vp;
988 		struct proc *a_p;
989 	} */ *ap = v;
990 	struct vnode *vp = ap->a_vp;
991 
992 	/*
993 	 * Clear out the v_type field to avoid
994 	 * nasty things happening in vgone().
995 	 */
996 	VOP_UNLOCK(vp, 0);
997 	vp->v_type = VNON;
998 	return (0);
999 }
1000 
1001 int
1002 fdesc_reclaim(v)
1003 	void *v;
1004 {
1005 	struct vop_reclaim_args /* {
1006 		struct vnode *a_vp;
1007 	} */ *ap = v;
1008 	struct vnode *vp = ap->a_vp;
1009 	struct fdescnode *fd = VTOFDESC(vp);
1010 
1011 	LIST_REMOVE(fd, fd_hash);
1012 	FREE(vp->v_data, M_TEMP);
1013 	vp->v_data = 0;
1014 
1015 	return (0);
1016 }
1017 
1018 /*
1019  * Return POSIX pathconf information applicable to special devices.
1020  */
1021 int
1022 fdesc_pathconf(v)
1023 	void *v;
1024 {
1025 	struct vop_pathconf_args /* {
1026 		struct vnode *a_vp;
1027 		int a_name;
1028 		register_t *a_retval;
1029 	} */ *ap = v;
1030 
1031 	switch (ap->a_name) {
1032 	case _PC_LINK_MAX:
1033 		*ap->a_retval = LINK_MAX;
1034 		return (0);
1035 	case _PC_MAX_CANON:
1036 		*ap->a_retval = MAX_CANON;
1037 		return (0);
1038 	case _PC_MAX_INPUT:
1039 		*ap->a_retval = MAX_INPUT;
1040 		return (0);
1041 	case _PC_PIPE_BUF:
1042 		*ap->a_retval = PIPE_BUF;
1043 		return (0);
1044 	case _PC_CHOWN_RESTRICTED:
1045 		*ap->a_retval = 1;
1046 		return (0);
1047 	case _PC_VDISABLE:
1048 		*ap->a_retval = _POSIX_VDISABLE;
1049 		return (0);
1050 	case _PC_SYNC_IO:
1051 		*ap->a_retval = 1;
1052 		return (0);
1053 	default:
1054 		return (EINVAL);
1055 	}
1056 	/* NOTREACHED */
1057 }
1058 
1059 /*
1060  * Print out the contents of a /dev/fd vnode.
1061  */
1062 /* ARGSUSED */
1063 int
1064 fdesc_print(v)
1065 	void *v;
1066 {
1067 	printf("tag VT_NON, fdesc vnode\n");
1068 	return (0);
1069 }
1070 
1071 int
1072 fdesc_link(v)
1073 	void *v;
1074 {
1075 	struct vop_link_args /* {
1076 		struct vnode *a_dvp;
1077 		struct vnode *a_vp;
1078 		struct componentname *a_cnp;
1079 	} */ *ap = v;
1080 
1081 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1082 	vput(ap->a_dvp);
1083 	return (EROFS);
1084 }
1085 
1086 int
1087 fdesc_symlink(v)
1088 	void *v;
1089 {
1090 	struct vop_symlink_args /* {
1091 		struct vnode *a_dvp;
1092 		struct vnode **a_vpp;
1093 		struct componentname *a_cnp;
1094 		struct vattr *a_vap;
1095 		char *a_target;
1096 	} */ *ap = v;
1097 
1098 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1099 	vput(ap->a_dvp);
1100 	return (EROFS);
1101 }
1102