xref: /netbsd-src/sys/miscfs/fdesc/fdesc_vnops.c (revision 404fbe5fb94ca1e054339640cabb2801ce52dd30)
1 /*	$NetBSD: fdesc_vnops.c,v 1.104 2008/12/17 20:51:36 cegger 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 1.104 2008/12/17 20:51:36 cegger 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 #include <sys/kauth.h>
65 #include <sys/atomic.h>
66 
67 #include <miscfs/fdesc/fdesc.h>
68 #include <miscfs/genfs/genfs.h>
69 
70 #define cttyvp(p) ((p)->p_lflag & PL_CONTROLT ? (p)->p_session->s_ttyvp : NULL)
71 
72 #define FDL_WANT	0x01
73 #define FDL_LOCKED	0x02
74 static int fdcache_lock;
75 
76 dev_t devctty;
77 
78 #if (FD_STDIN != FD_STDOUT-1) || (FD_STDOUT != FD_STDERR-1)
79 FD_STDIN, FD_STDOUT, FD_STDERR must be a sequence n, n+1, n+2
80 #endif
81 
82 #define	NFDCACHE 4
83 
84 #define FD_NHASH(ix) \
85 	(&fdhashtbl[(ix) & fdhash])
86 LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
87 u_long fdhash;
88 
89 int	fdesc_lookup(void *);
90 #define	fdesc_create	genfs_eopnotsupp
91 #define	fdesc_mknod	genfs_eopnotsupp
92 int	fdesc_open(void *);
93 #define	fdesc_close	genfs_nullop
94 #define	fdesc_access	genfs_nullop
95 int	fdesc_getattr(void *);
96 int	fdesc_setattr(void *);
97 int	fdesc_read(void *);
98 int	fdesc_write(void *);
99 int	fdesc_ioctl(void *);
100 int	fdesc_poll(void *);
101 int	fdesc_kqfilter(void *);
102 #define	fdesc_mmap	genfs_eopnotsupp
103 #define	fdesc_fcntl	genfs_fcntl
104 #define	fdesc_fsync	genfs_nullop
105 #define	fdesc_seek	genfs_seek
106 #define	fdesc_remove	genfs_eopnotsupp
107 int	fdesc_link(void *);
108 #define	fdesc_rename	genfs_eopnotsupp
109 #define	fdesc_mkdir	genfs_eopnotsupp
110 #define	fdesc_rmdir	genfs_eopnotsupp
111 int	fdesc_symlink(void *);
112 int	fdesc_readdir(void *);
113 int	fdesc_readlink(void *);
114 #define	fdesc_abortop	genfs_abortop
115 int	fdesc_inactive(void *);
116 int	fdesc_reclaim(void *);
117 #define	fdesc_lock	genfs_lock
118 #define	fdesc_unlock	genfs_unlock
119 #define	fdesc_bmap	genfs_badop
120 #define	fdesc_strategy	genfs_badop
121 int	fdesc_print(void *);
122 int	fdesc_pathconf(void *);
123 #define	fdesc_islocked	genfs_islocked
124 #define	fdesc_advlock	genfs_einval
125 #define	fdesc_bwrite	genfs_eopnotsupp
126 #define fdesc_revoke	genfs_revoke
127 #define fdesc_putpages	genfs_null_putpages
128 
129 static int fdesc_attr(int, struct vattr *, kauth_cred_t);
130 
131 int (**fdesc_vnodeop_p)(void *);
132 const struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = {
133 	{ &vop_default_desc, vn_default_error },
134 	{ &vop_lookup_desc, fdesc_lookup },		/* lookup */
135 	{ &vop_create_desc, fdesc_create },		/* create */
136 	{ &vop_mknod_desc, fdesc_mknod },		/* mknod */
137 	{ &vop_open_desc, fdesc_open },			/* open */
138 	{ &vop_close_desc, fdesc_close },		/* close */
139 	{ &vop_access_desc, fdesc_access },		/* access */
140 	{ &vop_getattr_desc, fdesc_getattr },		/* getattr */
141 	{ &vop_setattr_desc, fdesc_setattr },		/* setattr */
142 	{ &vop_read_desc, fdesc_read },			/* read */
143 	{ &vop_write_desc, fdesc_write },		/* write */
144 	{ &vop_ioctl_desc, fdesc_ioctl },		/* ioctl */
145 	{ &vop_fcntl_desc, fdesc_fcntl },		/* fcntl */
146 	{ &vop_poll_desc, fdesc_poll },			/* poll */
147 	{ &vop_kqfilter_desc, fdesc_kqfilter },		/* kqfilter */
148 	{ &vop_revoke_desc, fdesc_revoke },		/* revoke */
149 	{ &vop_mmap_desc, fdesc_mmap },			/* mmap */
150 	{ &vop_fsync_desc, fdesc_fsync },		/* fsync */
151 	{ &vop_seek_desc, fdesc_seek },			/* seek */
152 	{ &vop_remove_desc, fdesc_remove },		/* remove */
153 	{ &vop_link_desc, fdesc_link },			/* link */
154 	{ &vop_rename_desc, fdesc_rename },		/* rename */
155 	{ &vop_mkdir_desc, fdesc_mkdir },		/* mkdir */
156 	{ &vop_rmdir_desc, fdesc_rmdir },		/* rmdir */
157 	{ &vop_symlink_desc, fdesc_symlink },		/* symlink */
158 	{ &vop_readdir_desc, fdesc_readdir },		/* readdir */
159 	{ &vop_readlink_desc, fdesc_readlink },		/* readlink */
160 	{ &vop_abortop_desc, fdesc_abortop },		/* abortop */
161 	{ &vop_inactive_desc, fdesc_inactive },		/* inactive */
162 	{ &vop_reclaim_desc, fdesc_reclaim },		/* reclaim */
163 	{ &vop_lock_desc, fdesc_lock },			/* lock */
164 	{ &vop_unlock_desc, fdesc_unlock },		/* unlock */
165 	{ &vop_bmap_desc, fdesc_bmap },			/* bmap */
166 	{ &vop_strategy_desc, fdesc_strategy },		/* strategy */
167 	{ &vop_print_desc, fdesc_print },		/* print */
168 	{ &vop_islocked_desc, fdesc_islocked },		/* islocked */
169 	{ &vop_pathconf_desc, fdesc_pathconf },		/* pathconf */
170 	{ &vop_advlock_desc, fdesc_advlock },		/* advlock */
171 	{ &vop_bwrite_desc, fdesc_bwrite },		/* bwrite */
172 	{ &vop_putpages_desc, fdesc_putpages },		/* putpages */
173 	{ NULL, NULL }
174 };
175 
176 const struct vnodeopv_desc fdesc_vnodeop_opv_desc =
177 	{ &fdesc_vnodeop_p, fdesc_vnodeop_entries };
178 
179 extern const struct cdevsw ctty_cdevsw;
180 
181 /*
182  * Initialise cache headers
183  */
184 void
185 fdesc_init()
186 {
187 	int cttymajor;
188 
189 	/* locate the major number */
190 	cttymajor = cdevsw_lookup_major(&ctty_cdevsw);
191 	devctty = makedev(cttymajor, 0);
192 	fdhashtbl = hashinit(NFDCACHE, HASH_LIST, true, &fdhash);
193 }
194 
195 /*
196  * Free hash table.
197  */
198 void
199 fdesc_done()
200 {
201 	hashdone(fdhashtbl, HASH_LIST, fdhash);
202 }
203 
204 /*
205  * Return a locked vnode of the correct type.
206  */
207 int
208 fdesc_allocvp(ftype, ix, mp, vpp)
209 	fdntype ftype;
210 	int ix;
211 	struct mount *mp;
212 	struct vnode **vpp;
213 {
214 	struct fdhashhead *fc;
215 	struct fdescnode *fd;
216 	int error = 0;
217 
218 	fc = FD_NHASH(ix);
219 loop:
220 	for (fd = fc->lh_first; fd != 0; fd = fd->fd_hash.le_next) {
221 		if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
222 			if (vget(fd->fd_vnode, LK_EXCLUSIVE))
223 				goto loop;
224 			*vpp = fd->fd_vnode;
225 			return (error);
226 		}
227 	}
228 
229 	/*
230 	 * otherwise lock the array while we call getnewvnode
231 	 * since that can block.
232 	 */
233 	if (fdcache_lock & FDL_LOCKED) {
234 		fdcache_lock |= FDL_WANT;
235 		(void) tsleep(&fdcache_lock, PINOD, "fdcache", 0);
236 		goto loop;
237 	}
238 	fdcache_lock |= FDL_LOCKED;
239 
240 	error = getnewvnode(VT_FDESC, mp, fdesc_vnodeop_p, vpp);
241 	if (error)
242 		goto out;
243 	fd = malloc(sizeof(struct fdescnode), M_TEMP, M_WAITOK);
244 	(*vpp)->v_data = fd;
245 	fd->fd_vnode = *vpp;
246 	fd->fd_type = ftype;
247 	fd->fd_fd = -1;
248 	fd->fd_link = 0;
249 	fd->fd_ix = ix;
250 	uvm_vnp_setsize(*vpp, 0);
251 	VOP_LOCK(*vpp, LK_EXCLUSIVE);
252 	LIST_INSERT_HEAD(fc, fd, fd_hash);
253 
254 out:;
255 	fdcache_lock &= ~FDL_LOCKED;
256 
257 	if (fdcache_lock & FDL_WANT) {
258 		fdcache_lock &= ~FDL_WANT;
259 		wakeup(&fdcache_lock);
260 	}
261 
262 	return (error);
263 }
264 
265 /*
266  * vp is the current namei directory
267  * ndp is the name to locate in that directory...
268  */
269 int
270 fdesc_lookup(v)
271 	void *v;
272 {
273 	struct vop_lookup_args /* {
274 		struct vnode * a_dvp;
275 		struct vnode ** a_vpp;
276 		struct componentname * a_cnp;
277 	} */ *ap = v;
278 	struct vnode **vpp = ap->a_vpp;
279 	struct vnode *dvp = ap->a_dvp;
280 	struct componentname *cnp = ap->a_cnp;
281 	struct lwp *l = curlwp;
282 	const char *pname = cnp->cn_nameptr;
283 	struct proc *p = l->l_proc;
284 	filedesc_t *fdp = p->p_fd;
285 	int numfiles = fdp->fd_nfiles;
286 	unsigned fd = 0;
287 	int error;
288 	struct vnode *fvp;
289 	const char *ln;
290 
291 	if (cnp->cn_namelen == 1 && *pname == '.') {
292 		*vpp = dvp;
293 		VREF(dvp);
294 		return (0);
295 	}
296 
297 	switch (VTOFDESC(dvp)->fd_type) {
298 	default:
299 	case Flink:
300 	case Fdesc:
301 	case Fctty:
302 		error = ENOTDIR;
303 		goto bad;
304 
305 	case Froot:
306 		if (cnp->cn_namelen == 2 && memcmp(pname, "fd", 2) == 0) {
307 			error = fdesc_allocvp(Fdevfd, FD_DEVFD, dvp->v_mount, &fvp);
308 			if (error)
309 				goto bad;
310 			*vpp = fvp;
311 			fvp->v_type = VDIR;
312 			goto good;
313 		}
314 
315 		if (cnp->cn_namelen == 3 && memcmp(pname, "tty", 3) == 0) {
316 			struct vnode *ttyvp = cttyvp(p);
317 			if (ttyvp == NULL) {
318 				error = ENXIO;
319 				goto bad;
320 			}
321 			error = fdesc_allocvp(Fctty, FD_CTTY, dvp->v_mount, &fvp);
322 			if (error)
323 				goto bad;
324 			*vpp = fvp;
325 			fvp->v_type = VCHR;
326 			goto good;
327 		}
328 
329 		ln = 0;
330 		switch (cnp->cn_namelen) {
331 		case 5:
332 			if (memcmp(pname, "stdin", 5) == 0) {
333 				ln = "fd/0";
334 				fd = FD_STDIN;
335 			}
336 			break;
337 		case 6:
338 			if (memcmp(pname, "stdout", 6) == 0) {
339 				ln = "fd/1";
340 				fd = FD_STDOUT;
341 			} else
342 			if (memcmp(pname, "stderr", 6) == 0) {
343 				ln = "fd/2";
344 				fd = FD_STDERR;
345 			}
346 			break;
347 		}
348 
349 		if (ln) {
350 			error = fdesc_allocvp(Flink, fd, dvp->v_mount, &fvp);
351 			if (error)
352 				goto bad;
353 			/* XXXUNCONST */
354 			VTOFDESC(fvp)->fd_link = __UNCONST(ln);
355 			*vpp = fvp;
356 			fvp->v_type = VLNK;
357 			goto good;
358 		} else {
359 			error = ENOENT;
360 			goto bad;
361 		}
362 
363 		/* FALL THROUGH */
364 
365 	case Fdevfd:
366 		if (cnp->cn_namelen == 2 && memcmp(pname, "..", 2) == 0) {
367 			VOP_UNLOCK(dvp, 0);
368 			error = fdesc_root(dvp->v_mount, vpp);
369 			vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
370 			if (error)
371 				goto bad;
372 			return (error);
373 		}
374 
375 		fd = 0;
376 		while (*pname >= '0' && *pname <= '9') {
377 			fd = 10 * fd + *pname++ - '0';
378 			if (fd >= numfiles)
379 				break;
380 		}
381 
382 		if (*pname != '\0') {
383 			error = ENOENT;
384 			goto bad;
385 		}
386 
387 		mutex_enter(&fdp->fd_lock);
388 		if (fd >= numfiles ||fdp->fd_ofiles[fd] == NULL ||
389 		    fdp->fd_ofiles[fd]->ff_file == NULL) {
390 			mutex_exit(&fdp->fd_lock);
391 			error = EBADF;
392 			goto bad;
393 		}
394 		mutex_exit(&fdp->fd_lock);
395 
396 		error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp);
397 		if (error)
398 			goto bad;
399 		VTOFDESC(fvp)->fd_fd = fd;
400 		*vpp = fvp;
401 		goto good;
402 	}
403 
404 bad:
405 	*vpp = NULL;
406 	return (error);
407 
408 good:
409 	return (0);
410 }
411 
412 int
413 fdesc_open(v)
414 	void *v;
415 {
416 	struct vop_open_args /* {
417 		struct vnode *a_vp;
418 		int  a_mode;
419 		kauth_cred_t a_cred;
420 	} */ *ap = v;
421 	struct vnode *vp = ap->a_vp;
422 
423 	switch (VTOFDESC(vp)->fd_type) {
424 	case Fdesc:
425 		/*
426 		 * XXX Kludge: set dupfd to contain the value of the
427 		 * the file descriptor being sought for duplication. The error
428 		 * return ensures that the vnode for this device will be
429 		 * released by vn_open. Open will detect this special error and
430 		 * take the actions in dupfdopen.  Other callers of vn_open or
431 		 * VOP_OPEN will simply report the error.
432 		 */
433 		curlwp->l_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
434 		return EDUPFD;
435 
436 	case Fctty:
437 		return cdev_open(devctty, ap->a_mode, 0, curlwp);
438 	case Froot:
439 	case Fdevfd:
440 	case Flink:
441 		break;
442 	}
443 
444 	return (0);
445 }
446 
447 static int
448 fdesc_attr(fd, vap, cred)
449 	int fd;
450 	struct vattr *vap;
451 	kauth_cred_t cred;
452 {
453 	file_t *fp;
454 	struct stat stb;
455 	int error;
456 
457 	if ((fp = fd_getfile(fd)) == NULL)
458 		return (EBADF);
459 
460 	switch (fp->f_type) {
461 	case DTYPE_VNODE:
462 		error = VOP_GETATTR((struct vnode *) fp->f_data, vap, cred);
463 		if (error == 0 && vap->va_type == VDIR) {
464 			/*
465 			 * directories can cause loops in the namespace,
466 			 * so turn off the 'x' bits to avoid trouble.
467 			 */
468 			vap->va_mode &= ~(S_IXUSR|S_IXGRP|S_IXOTH);
469 		}
470 		break;
471 
472 	default:
473 		memset(&stb, 0, sizeof(stb));
474 		error = (*fp->f_ops->fo_stat)(fp, &stb);
475 		if (error)
476 			break;
477 
478 		vattr_null(vap);
479 		switch(fp->f_type) {
480 		case DTYPE_SOCKET:
481 			vap->va_type = VSOCK;
482 			break;
483 		case DTYPE_PIPE:
484 			vap->va_type = VFIFO;
485 			break;
486 		default:
487 			/* use VNON perhaps? */
488 			vap->va_type = VBAD;
489 			break;
490 		}
491 		vap->va_mode = stb.st_mode;
492 		vap->va_nlink = stb.st_nlink;
493 		vap->va_uid = stb.st_uid;
494 		vap->va_gid = stb.st_gid;
495 		vap->va_fsid = stb.st_dev;
496 		vap->va_fileid = stb.st_ino;
497 		vap->va_size = stb.st_size;
498 		vap->va_blocksize = stb.st_blksize;
499 		vap->va_atime = stb.st_atimespec;
500 		vap->va_mtime = stb.st_mtimespec;
501 		vap->va_ctime = stb.st_ctimespec;
502 		vap->va_gen = stb.st_gen;
503 		vap->va_flags = stb.st_flags;
504 		vap->va_rdev = stb.st_rdev;
505 		vap->va_bytes = stb.st_blocks * stb.st_blksize;
506 		break;
507 	}
508 
509 	fd_putfile(fd);
510 	return (error);
511 }
512 
513 int
514 fdesc_getattr(v)
515 	void *v;
516 {
517 	struct vop_getattr_args /* {
518 		struct vnode *a_vp;
519 		struct vattr *a_vap;
520 		kauth_cred_t a_cred;
521 		struct lwp *a_l;
522 	} */ *ap = v;
523 	struct vnode *vp = ap->a_vp;
524 	struct vattr *vap = ap->a_vap;
525 	unsigned fd;
526 	int error = 0;
527 
528 	switch (VTOFDESC(vp)->fd_type) {
529 	case Froot:
530 	case Fdevfd:
531 	case Flink:
532 	case Fctty:
533 		VATTR_NULL(vap);
534 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
535 
536 #define R_ALL (S_IRUSR|S_IRGRP|S_IROTH)
537 #define W_ALL (S_IWUSR|S_IWGRP|S_IWOTH)
538 #define X_ALL (S_IXUSR|S_IXGRP|S_IXOTH)
539 
540 		switch (VTOFDESC(vp)->fd_type) {
541 		case Flink:
542 			vap->va_mode = R_ALL|X_ALL;
543 			vap->va_type = VLNK;
544 			vap->va_rdev = 0;
545 			vap->va_nlink = 1;
546 			vap->va_size = strlen(VTOFDESC(vp)->fd_link);
547 			break;
548 
549 		case Fctty:
550 			vap->va_mode = R_ALL|W_ALL;
551 			vap->va_type = VCHR;
552 			vap->va_rdev = devctty;
553 			vap->va_nlink = 1;
554 			vap->va_size = 0;
555 			break;
556 
557 		default:
558 			vap->va_mode = R_ALL|X_ALL;
559 			vap->va_type = VDIR;
560 			vap->va_rdev = 0;
561 			vap->va_nlink = 2;
562 			vap->va_size = DEV_BSIZE;
563 			break;
564 		}
565 		vap->va_uid = 0;
566 		vap->va_gid = 0;
567 		vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0];
568 		vap->va_blocksize = DEV_BSIZE;
569 		vap->va_atime.tv_sec = boottime.tv_sec;
570 		vap->va_atime.tv_nsec = 0;
571 		vap->va_mtime = vap->va_atime;
572 		vap->va_ctime = vap->va_mtime;
573 		vap->va_gen = 0;
574 		vap->va_flags = 0;
575 		vap->va_bytes = 0;
576 		break;
577 
578 	case Fdesc:
579 		fd = VTOFDESC(vp)->fd_fd;
580 		error = fdesc_attr(fd, vap, ap->a_cred);
581 		break;
582 
583 	default:
584 		panic("fdesc_getattr");
585 		break;
586 	}
587 
588 	if (error == 0)
589 		vp->v_type = vap->va_type;
590 
591 	return (error);
592 }
593 
594 int
595 fdesc_setattr(v)
596 	void *v;
597 {
598 	struct vop_setattr_args /* {
599 		struct vnode *a_vp;
600 		struct vattr *a_vap;
601 		kauth_cred_t a_cred;
602 	} */ *ap = v;
603 	file_t *fp;
604 	unsigned fd;
605 
606 	/*
607 	 * Can't mess with the root vnode
608 	 */
609 	switch (VTOFDESC(ap->a_vp)->fd_type) {
610 	case Fdesc:
611 		break;
612 
613 	case Fctty:
614 		return (0);
615 
616 	default:
617 		return (EACCES);
618 	}
619 
620 	fd = VTOFDESC(ap->a_vp)->fd_fd;
621 	if ((fp = fd_getfile(fd)) == NULL)
622 		return (EBADF);
623 
624 	/*
625 	 * XXX: Can't reasonably set the attr's on any types currently.
626 	 *      On vnode's this will cause truncation and socket/pipes make
627 	 *      no sense.
628 	 */
629 	fd_putfile(fd);
630 	return (0);
631 }
632 
633 
634 struct fdesc_target {
635 	ino_t ft_fileno;
636 	u_char ft_type;
637 	u_char ft_namlen;
638 	const char *ft_name;
639 } fdesc_targets[] = {
640 #define N(s) sizeof(s)-1, s
641 	{ FD_DEVFD,  DT_DIR,     N("fd")     },
642 	{ FD_STDIN,  DT_LNK,     N("stdin")  },
643 	{ FD_STDOUT, DT_LNK,     N("stdout") },
644 	{ FD_STDERR, DT_LNK,     N("stderr") },
645 	{ FD_CTTY,   DT_UNKNOWN, N("tty")    },
646 #undef N
647 #define UIO_MX _DIRENT_RECLEN((struct dirent *)NULL, sizeof("stderr") - 1)
648 };
649 static int nfdesc_targets = sizeof(fdesc_targets) / sizeof(fdesc_targets[0]);
650 
651 int
652 fdesc_readdir(v)
653 	void *v;
654 {
655 	struct vop_readdir_args /* {
656 		struct vnode *a_vp;
657 		struct uio *a_uio;
658 		kauth_cred_t a_cred;
659 		int *a_eofflag;
660 		off_t **a_cookies;
661 		int *a_ncookies;
662 	} */ *ap = v;
663 	struct uio *uio = ap->a_uio;
664 	struct dirent d;
665 	filedesc_t *fdp;
666 	off_t i;
667 	int j;
668 	int error;
669 	off_t *cookies = NULL;
670 	int ncookies;
671 
672 	switch (VTOFDESC(ap->a_vp)->fd_type) {
673 	case Fctty:
674 		return 0;
675 
676 	case Fdesc:
677 		return ENOTDIR;
678 
679 	default:
680 		break;
681 	}
682 
683 	fdp = curproc->p_fd;
684 
685 	if (uio->uio_resid < UIO_MX)
686 		return EINVAL;
687 	if (uio->uio_offset < 0)
688 		return EINVAL;
689 
690 	error = 0;
691 	i = uio->uio_offset;
692 	(void)memset(&d, 0, UIO_MX);
693 	d.d_reclen = UIO_MX;
694 	if (ap->a_ncookies)
695 		ncookies = uio->uio_resid / UIO_MX;
696 	else
697 		ncookies = 0;
698 
699 	if (VTOFDESC(ap->a_vp)->fd_type == Froot) {
700 		struct fdesc_target *ft;
701 
702 		if (i >= nfdesc_targets)
703 			return 0;
704 
705 		if (ap->a_ncookies) {
706 			ncookies = min(ncookies, (nfdesc_targets - i));
707 			cookies = malloc(ncookies * sizeof(off_t),
708 			    M_TEMP, M_WAITOK);
709 			*ap->a_cookies = cookies;
710 			*ap->a_ncookies = ncookies;
711 		}
712 
713 		for (ft = &fdesc_targets[i]; uio->uio_resid >= UIO_MX &&
714 		    i < nfdesc_targets; ft++, i++) {
715 			switch (ft->ft_fileno) {
716 			case FD_CTTY:
717 				if (cttyvp(curproc) == NULL)
718 					continue;
719 				break;
720 
721 			case FD_STDIN:
722 			case FD_STDOUT:
723 			case FD_STDERR:
724 				if (fdp == NULL)
725 					continue;
726 				if ((ft->ft_fileno - FD_STDIN) >=
727 				    fdp->fd_nfiles)
728 					continue;
729 				membar_consumer();
730 				if (fdp->fd_ofiles[ft->ft_fileno - FD_STDIN]
731 				    == NULL || fdp->fd_ofiles[ft->ft_fileno -
732 				    FD_STDIN]->ff_file == NULL)
733 					continue;
734 				break;
735 			}
736 
737 			d.d_fileno = ft->ft_fileno;
738 			d.d_namlen = ft->ft_namlen;
739 			(void)memcpy(d.d_name, ft->ft_name, ft->ft_namlen + 1);
740 			d.d_type = ft->ft_type;
741 
742 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
743 				break;
744 			if (cookies)
745 				*cookies++ = i + 1;
746 		}
747 	} else {
748 		int nfdp = fdp ? fdp->fd_nfiles : 0;
749 		membar_consumer();
750 		if (ap->a_ncookies) {
751 			ncookies = min(ncookies, nfdp + 2);
752 			cookies = malloc(ncookies * sizeof(off_t),
753 			    M_TEMP, M_WAITOK);
754 			*ap->a_cookies = cookies;
755 			*ap->a_ncookies = ncookies;
756 		}
757 		for (; i - 2 < nfdp && uio->uio_resid >= UIO_MX; i++) {
758 			switch (i) {
759 			case 0:
760 			case 1:
761 				d.d_fileno = FD_ROOT;		/* XXX */
762 				d.d_namlen = i + 1;
763 				(void)memcpy(d.d_name, "..", d.d_namlen);
764 				d.d_name[i + 1] = '\0';
765 				d.d_type = DT_DIR;
766 				break;
767 
768 			default:
769 				KASSERT(fdp != NULL);
770 				j = (int)i - 2;
771 				if (fdp == NULL || fdp->fd_ofiles[j] == NULL ||
772 				    fdp->fd_ofiles[j]->ff_file == NULL)
773 					continue;
774 				d.d_fileno = j + FD_STDIN;
775 				d.d_namlen = sprintf(d.d_name, "%d", j);
776 				d.d_type = DT_UNKNOWN;
777 				break;
778 			}
779 
780 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
781 				break;
782 			if (cookies)
783 				*cookies++ = i + 1;
784 		}
785 	}
786 
787 	if (ap->a_ncookies && error) {
788 		free(*ap->a_cookies, M_TEMP);
789 		*ap->a_ncookies = 0;
790 		*ap->a_cookies = NULL;
791 	}
792 
793 	uio->uio_offset = i;
794 	return error;
795 }
796 
797 int
798 fdesc_readlink(v)
799 	void *v;
800 {
801 	struct vop_readlink_args /* {
802 		struct vnode *a_vp;
803 		struct uio *a_uio;
804 		kauth_cred_t a_cred;
805 	} */ *ap = v;
806 	struct vnode *vp = ap->a_vp;
807 	int error;
808 
809 	if (vp->v_type != VLNK)
810 		return (EPERM);
811 
812 	if (VTOFDESC(vp)->fd_type == Flink) {
813 		char *ln = VTOFDESC(vp)->fd_link;
814 		error = uiomove(ln, strlen(ln), ap->a_uio);
815 	} else {
816 		error = EOPNOTSUPP;
817 	}
818 
819 	return (error);
820 }
821 
822 int
823 fdesc_read(v)
824 	void *v;
825 {
826 	struct vop_read_args /* {
827 		struct vnode *a_vp;
828 		struct uio *a_uio;
829 		int  a_ioflag;
830 		kauth_cred_t a_cred;
831 	} */ *ap = v;
832 	int error = EOPNOTSUPP;
833 	struct vnode *vp = ap->a_vp;
834 
835 	switch (VTOFDESC(vp)->fd_type) {
836 	case Fctty:
837 		VOP_UNLOCK(vp, 0);
838 		error = cdev_read(devctty, ap->a_uio, ap->a_ioflag);
839 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
840 		break;
841 
842 	default:
843 		error = EOPNOTSUPP;
844 		break;
845 	}
846 
847 	return (error);
848 }
849 
850 int
851 fdesc_write(v)
852 	void *v;
853 {
854 	struct vop_write_args /* {
855 		struct vnode *a_vp;
856 		struct uio *a_uio;
857 		int  a_ioflag;
858 		kauth_cred_t a_cred;
859 	} */ *ap = v;
860 	int error = EOPNOTSUPP;
861 	struct vnode *vp = ap->a_vp;
862 
863 	switch (VTOFDESC(vp)->fd_type) {
864 	case Fctty:
865 		VOP_UNLOCK(vp, 0);
866 		error = cdev_write(devctty, ap->a_uio, ap->a_ioflag);
867 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
868 		break;
869 
870 	default:
871 		error = EOPNOTSUPP;
872 		break;
873 	}
874 
875 	return (error);
876 }
877 
878 int
879 fdesc_ioctl(v)
880 	void *v;
881 {
882 	struct vop_ioctl_args /* {
883 		struct vnode *a_vp;
884 		u_long a_command;
885 		void *a_data;
886 		int  a_fflag;
887 		kauth_cred_t a_cred;
888 	} */ *ap = v;
889 	int error = EOPNOTSUPP;
890 
891 	switch (VTOFDESC(ap->a_vp)->fd_type) {
892 	case Fctty:
893 		error = cdev_ioctl(devctty, ap->a_command, ap->a_data,
894 		    ap->a_fflag, curlwp);
895 		break;
896 
897 	default:
898 		error = EOPNOTSUPP;
899 		break;
900 	}
901 
902 	return (error);
903 }
904 
905 int
906 fdesc_poll(v)
907 	void *v;
908 {
909 	struct vop_poll_args /* {
910 		struct vnode *a_vp;
911 		int a_events;
912 	} */ *ap = v;
913 	int revents;
914 
915 	switch (VTOFDESC(ap->a_vp)->fd_type) {
916 	case Fctty:
917 		revents = cdev_poll(devctty, ap->a_events, curlwp);
918 		break;
919 
920 	default:
921 		revents = genfs_poll(v);
922 		break;
923 	}
924 
925 	return (revents);
926 }
927 
928 int
929 fdesc_kqfilter(v)
930 	void *v;
931 {
932 	struct vop_kqfilter_args /* {
933 		struct vnode *a_vp;
934 		struct knote *a_kn;
935 	} */ *ap = v;
936 	int error, fd;
937 	file_t *fp;
938 
939 	switch (VTOFDESC(ap->a_vp)->fd_type) {
940 	case Fctty:
941 		error = cdev_kqfilter(devctty, ap->a_kn);
942 		break;
943 
944 	case Fdesc:
945 		/* just invoke kqfilter for the underlying descriptor */
946 		fd = VTOFDESC(ap->a_vp)->fd_fd;
947 		if ((fp = fd_getfile(fd)) == NULL)
948 			return (1);
949 		error = (*fp->f_ops->fo_kqfilter)(fp, ap->a_kn);
950 		fd_putfile(fd);
951 		break;
952 
953 	default:
954 		return (genfs_kqfilter(v));
955 	}
956 
957 	return (error);
958 }
959 
960 int
961 fdesc_inactive(v)
962 	void *v;
963 {
964 	struct vop_inactive_args /* {
965 		struct vnode *a_vp;
966 	} */ *ap = v;
967 	struct vnode *vp = ap->a_vp;
968 
969 	/*
970 	 * Clear out the v_type field to avoid
971 	 * nasty things happening in vgone().
972 	 */
973 	VOP_UNLOCK(vp, 0);
974 	vp->v_type = VNON;
975 	return (0);
976 }
977 
978 int
979 fdesc_reclaim(v)
980 	void *v;
981 {
982 	struct vop_reclaim_args /* {
983 		struct vnode *a_vp;
984 	} */ *ap = v;
985 	struct vnode *vp = ap->a_vp;
986 	struct fdescnode *fd = VTOFDESC(vp);
987 
988 	LIST_REMOVE(fd, fd_hash);
989 	free(vp->v_data, M_TEMP);
990 	vp->v_data = 0;
991 
992 	return (0);
993 }
994 
995 /*
996  * Return POSIX pathconf information applicable to special devices.
997  */
998 int
999 fdesc_pathconf(v)
1000 	void *v;
1001 {
1002 	struct vop_pathconf_args /* {
1003 		struct vnode *a_vp;
1004 		int a_name;
1005 		register_t *a_retval;
1006 	} */ *ap = v;
1007 
1008 	switch (ap->a_name) {
1009 	case _PC_LINK_MAX:
1010 		*ap->a_retval = LINK_MAX;
1011 		return (0);
1012 	case _PC_MAX_CANON:
1013 		*ap->a_retval = MAX_CANON;
1014 		return (0);
1015 	case _PC_MAX_INPUT:
1016 		*ap->a_retval = MAX_INPUT;
1017 		return (0);
1018 	case _PC_PIPE_BUF:
1019 		*ap->a_retval = PIPE_BUF;
1020 		return (0);
1021 	case _PC_CHOWN_RESTRICTED:
1022 		*ap->a_retval = 1;
1023 		return (0);
1024 	case _PC_VDISABLE:
1025 		*ap->a_retval = _POSIX_VDISABLE;
1026 		return (0);
1027 	case _PC_SYNC_IO:
1028 		*ap->a_retval = 1;
1029 		return (0);
1030 	default:
1031 		return (EINVAL);
1032 	}
1033 	/* NOTREACHED */
1034 }
1035 
1036 /*
1037  * Print out the contents of a /dev/fd vnode.
1038  */
1039 /* ARGSUSED */
1040 int
1041 fdesc_print(void *v)
1042 {
1043 	printf("tag VT_NON, fdesc vnode\n");
1044 	return (0);
1045 }
1046 
1047 int
1048 fdesc_link(v)
1049 	void *v;
1050 {
1051 	struct vop_link_args /* {
1052 		struct vnode *a_dvp;
1053 		struct vnode *a_vp;
1054 		struct componentname *a_cnp;
1055 	} */ *ap = v;
1056 
1057 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1058 	vput(ap->a_dvp);
1059 	return (EROFS);
1060 }
1061 
1062 int
1063 fdesc_symlink(v)
1064 	void *v;
1065 {
1066 	struct vop_symlink_args /* {
1067 		struct vnode *a_dvp;
1068 		struct vnode **a_vpp;
1069 		struct componentname *a_cnp;
1070 		struct vattr *a_vap;
1071 		char *a_target;
1072 	} */ *ap = v;
1073 
1074 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1075 	vput(ap->a_dvp);
1076 	return (EROFS);
1077 }
1078