xref: /netbsd-src/sys/miscfs/fdesc/fdesc_vnops.c (revision 6a493d6bc668897c91594964a732d38505b70cbb)
1 /*	$NetBSD: fdesc_vnops.c,v 1.114 2011/10/16 12:26:16 hannken 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 1.114 2011/10/16 12:26:16 hannken 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 static kmutex_t 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_bwrite	genfs_eopnotsupp
124 #define fdesc_revoke	genfs_revoke
125 #define fdesc_putpages	genfs_null_putpages
126 
127 static int fdesc_attr(int, struct vattr *, kauth_cred_t);
128 
129 int (**fdesc_vnodeop_p)(void *);
130 const struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = {
131 	{ &vop_default_desc, vn_default_error },
132 	{ &vop_lookup_desc, fdesc_lookup },		/* lookup */
133 	{ &vop_create_desc, fdesc_create },		/* create */
134 	{ &vop_mknod_desc, fdesc_mknod },		/* mknod */
135 	{ &vop_open_desc, fdesc_open },			/* open */
136 	{ &vop_close_desc, fdesc_close },		/* close */
137 	{ &vop_access_desc, fdesc_access },		/* access */
138 	{ &vop_getattr_desc, fdesc_getattr },		/* getattr */
139 	{ &vop_setattr_desc, fdesc_setattr },		/* setattr */
140 	{ &vop_read_desc, fdesc_read },			/* read */
141 	{ &vop_write_desc, fdesc_write },		/* write */
142 	{ &vop_ioctl_desc, fdesc_ioctl },		/* ioctl */
143 	{ &vop_fcntl_desc, fdesc_fcntl },		/* fcntl */
144 	{ &vop_poll_desc, fdesc_poll },			/* poll */
145 	{ &vop_kqfilter_desc, fdesc_kqfilter },		/* kqfilter */
146 	{ &vop_revoke_desc, fdesc_revoke },		/* revoke */
147 	{ &vop_mmap_desc, fdesc_mmap },			/* mmap */
148 	{ &vop_fsync_desc, fdesc_fsync },		/* fsync */
149 	{ &vop_seek_desc, fdesc_seek },			/* seek */
150 	{ &vop_remove_desc, fdesc_remove },		/* remove */
151 	{ &vop_link_desc, fdesc_link },			/* link */
152 	{ &vop_rename_desc, fdesc_rename },		/* rename */
153 	{ &vop_mkdir_desc, fdesc_mkdir },		/* mkdir */
154 	{ &vop_rmdir_desc, fdesc_rmdir },		/* rmdir */
155 	{ &vop_symlink_desc, fdesc_symlink },		/* symlink */
156 	{ &vop_readdir_desc, fdesc_readdir },		/* readdir */
157 	{ &vop_readlink_desc, fdesc_readlink },		/* readlink */
158 	{ &vop_abortop_desc, fdesc_abortop },		/* abortop */
159 	{ &vop_inactive_desc, fdesc_inactive },		/* inactive */
160 	{ &vop_reclaim_desc, fdesc_reclaim },		/* reclaim */
161 	{ &vop_lock_desc, fdesc_lock },			/* lock */
162 	{ &vop_unlock_desc, fdesc_unlock },		/* unlock */
163 	{ &vop_bmap_desc, fdesc_bmap },			/* bmap */
164 	{ &vop_strategy_desc, fdesc_strategy },		/* strategy */
165 	{ &vop_print_desc, fdesc_print },		/* print */
166 	{ &vop_islocked_desc, fdesc_islocked },		/* islocked */
167 	{ &vop_pathconf_desc, fdesc_pathconf },		/* pathconf */
168 	{ &vop_advlock_desc, fdesc_advlock },		/* advlock */
169 	{ &vop_bwrite_desc, fdesc_bwrite },		/* bwrite */
170 	{ &vop_putpages_desc, fdesc_putpages },		/* putpages */
171 	{ NULL, NULL }
172 };
173 
174 const struct vnodeopv_desc fdesc_vnodeop_opv_desc =
175 	{ &fdesc_vnodeop_p, fdesc_vnodeop_entries };
176 
177 /*
178  * Initialise cache headers
179  */
180 void
181 fdesc_init(void)
182 {
183 	int cttymajor;
184 
185 	/* locate the major number */
186 	cttymajor = devsw_name2chr("ctty", NULL, 0);
187 	devctty = makedev(cttymajor, 0);
188 	mutex_init(&fdcache_lock, MUTEX_DEFAULT, IPL_NONE);
189 	fdhashtbl = hashinit(NFDCACHE, HASH_LIST, true, &fdhash);
190 }
191 
192 /*
193  * Free hash table.
194  */
195 void
196 fdesc_done(void)
197 {
198 	hashdone(fdhashtbl, HASH_LIST, fdhash);
199 	mutex_destroy(&fdcache_lock);
200 }
201 
202 /*
203  * Return a locked vnode of the correct type.
204  */
205 int
206 fdesc_allocvp(fdntype ftype, int ix, struct mount *mp, struct vnode **vpp)
207 {
208 	struct fdhashhead *fc;
209 	struct fdescnode *fd;
210 	int error = 0;
211 
212 	fc = FD_NHASH(ix);
213 loop:
214 	mutex_enter(&fdcache_lock);
215 	LIST_FOREACH(fd, fc, fd_hash) {
216 		if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
217 			mutex_enter(fd->fd_vnode->v_interlock);
218 			mutex_exit(&fdcache_lock);
219 			if (vget(fd->fd_vnode, LK_EXCLUSIVE))
220 				goto loop;
221 			*vpp = fd->fd_vnode;
222 			return 0;
223 		}
224 	}
225 	mutex_exit(&fdcache_lock);
226 
227 	error = getnewvnode(VT_FDESC, mp, fdesc_vnodeop_p, NULL, vpp);
228 	if (error)
229 		return error;
230 
231 	mutex_enter(&fdcache_lock);
232 	LIST_FOREACH(fd, fc, fd_hash) {
233 		if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
234 			/*
235 			 * Another thread beat us, push back freshly
236 			 * allocated vnode and retry.
237 			 */
238 			mutex_exit(&fdcache_lock);
239 			ungetnewvnode(*vpp);
240 			goto loop;
241 		}
242 	}
243 
244 	fd = malloc(sizeof(struct fdescnode), M_TEMP, M_WAITOK);
245 	(*vpp)->v_data = fd;
246 	fd->fd_vnode = *vpp;
247 	fd->fd_type = ftype;
248 	fd->fd_fd = -1;
249 	fd->fd_link = 0;
250 	fd->fd_ix = ix;
251 	uvm_vnp_setsize(*vpp, 0);
252 	VOP_LOCK(*vpp, LK_EXCLUSIVE);
253 	LIST_INSERT_HEAD(fc, fd, fd_hash);
254 	mutex_exit(&fdcache_lock);
255 
256 	return 0;
257 }
258 
259 /*
260  * vp is the current namei directory
261  * ndp is the name to locate in that directory...
262  */
263 int
264 fdesc_lookup(void *v)
265 {
266 	struct vop_lookup_args /* {
267 		struct vnode * a_dvp;
268 		struct vnode ** a_vpp;
269 		struct componentname * a_cnp;
270 	} */ *ap = v;
271 	struct vnode **vpp = ap->a_vpp;
272 	struct vnode *dvp = ap->a_dvp;
273 	struct componentname *cnp = ap->a_cnp;
274 	struct lwp *l = curlwp;
275 	const char *pname = cnp->cn_nameptr;
276 	struct proc *p = l->l_proc;
277 	unsigned fd = 0;
278 	int error;
279 	struct vnode *fvp;
280 	const char *ln;
281 	fdtab_t *dt;
282 
283 	dt = curlwp->l_fd->fd_dt;
284 
285 	if (cnp->cn_namelen == 1 && *pname == '.') {
286 		*vpp = dvp;
287 		vref(dvp);
288 		return (0);
289 	}
290 
291 	switch (VTOFDESC(dvp)->fd_type) {
292 	default:
293 	case Flink:
294 	case Fdesc:
295 	case Fctty:
296 		error = ENOTDIR;
297 		goto bad;
298 
299 	case Froot:
300 		if (cnp->cn_namelen == 2 && memcmp(pname, "fd", 2) == 0) {
301 			error = fdesc_allocvp(Fdevfd, FD_DEVFD, dvp->v_mount, &fvp);
302 			if (error)
303 				goto bad;
304 			*vpp = fvp;
305 			fvp->v_type = VDIR;
306 			goto good;
307 		}
308 
309 		if (cnp->cn_namelen == 3 && memcmp(pname, "tty", 3) == 0) {
310 			struct vnode *ttyvp = cttyvp(p);
311 			if (ttyvp == NULL) {
312 				error = ENXIO;
313 				goto bad;
314 			}
315 			error = fdesc_allocvp(Fctty, FD_CTTY, dvp->v_mount, &fvp);
316 			if (error)
317 				goto bad;
318 			*vpp = fvp;
319 			fvp->v_type = VCHR;
320 			goto good;
321 		}
322 
323 		ln = 0;
324 		switch (cnp->cn_namelen) {
325 		case 5:
326 			if (memcmp(pname, "stdin", 5) == 0) {
327 				ln = "fd/0";
328 				fd = FD_STDIN;
329 			}
330 			break;
331 		case 6:
332 			if (memcmp(pname, "stdout", 6) == 0) {
333 				ln = "fd/1";
334 				fd = FD_STDOUT;
335 			} else
336 			if (memcmp(pname, "stderr", 6) == 0) {
337 				ln = "fd/2";
338 				fd = FD_STDERR;
339 			}
340 			break;
341 		}
342 
343 		if (ln) {
344 			error = fdesc_allocvp(Flink, fd, dvp->v_mount, &fvp);
345 			if (error)
346 				goto bad;
347 			/* XXXUNCONST */
348 			VTOFDESC(fvp)->fd_link = __UNCONST(ln);
349 			*vpp = fvp;
350 			fvp->v_type = VLNK;
351 			goto good;
352 		} else {
353 			error = ENOENT;
354 			goto bad;
355 		}
356 
357 		/* FALL THROUGH */
358 
359 	case Fdevfd:
360 		if (cnp->cn_namelen == 2 && memcmp(pname, "..", 2) == 0) {
361 			VOP_UNLOCK(dvp);
362 			error = fdesc_root(dvp->v_mount, vpp);
363 			vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
364 			if (error)
365 				goto bad;
366 			return (error);
367 		}
368 
369 		fd = 0;
370 		while (*pname >= '0' && *pname <= '9') {
371 			fd = 10 * fd + *pname++ - '0';
372 			if (fd >= dt->dt_nfiles)
373 				break;
374 		}
375 
376 		if (*pname != '\0') {
377 			error = ENOENT;
378 			goto bad;
379 		}
380 
381 		if (fd >= dt->dt_nfiles || dt->dt_ff[fd] == NULL ||
382 		    dt->dt_ff[fd]->ff_file == NULL) {
383 			error = EBADF;
384 			goto bad;
385 		}
386 
387 		error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp);
388 		if (error)
389 			goto bad;
390 		VTOFDESC(fvp)->fd_fd = fd;
391 		*vpp = fvp;
392 		goto good;
393 	}
394 
395 bad:
396 	*vpp = NULL;
397 	return (error);
398 
399 good:
400 	return (0);
401 }
402 
403 int
404 fdesc_open(void *v)
405 {
406 	struct vop_open_args /* {
407 		struct vnode *a_vp;
408 		int  a_mode;
409 		kauth_cred_t a_cred;
410 	} */ *ap = v;
411 	struct vnode *vp = ap->a_vp;
412 
413 	switch (VTOFDESC(vp)->fd_type) {
414 	case Fdesc:
415 		/*
416 		 * XXX Kludge: set dupfd to contain the value of the
417 		 * the file descriptor being sought for duplication. The error
418 		 * return ensures that the vnode for this device will be
419 		 * released by vn_open. Open will detect this special error and
420 		 * take the actions in dupfdopen.  Other callers of vn_open or
421 		 * VOP_OPEN will simply report the error.
422 		 */
423 		curlwp->l_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
424 		return EDUPFD;
425 
426 	case Fctty:
427 		return cdev_open(devctty, ap->a_mode, 0, curlwp);
428 	case Froot:
429 	case Fdevfd:
430 	case Flink:
431 		break;
432 	}
433 
434 	return (0);
435 }
436 
437 static int
438 fdesc_attr(int fd, struct vattr *vap, kauth_cred_t cred)
439 {
440 	file_t *fp;
441 	struct stat stb;
442 	int error;
443 
444 	if ((fp = fd_getfile(fd)) == NULL)
445 		return (EBADF);
446 
447 	switch (fp->f_type) {
448 	case DTYPE_VNODE:
449 		vn_lock((struct vnode *) fp->f_data, LK_SHARED | LK_RETRY);
450 		error = VOP_GETATTR((struct vnode *) fp->f_data, vap, cred);
451 		VOP_UNLOCK((struct vnode *) fp->f_data);
452 		if (error == 0 && vap->va_type == VDIR) {
453 			/*
454 			 * directories can cause loops in the namespace,
455 			 * so turn off the 'x' bits to avoid trouble.
456 			 */
457 			vap->va_mode &= ~(S_IXUSR|S_IXGRP|S_IXOTH);
458 		}
459 		break;
460 
461 	default:
462 		memset(&stb, 0, sizeof(stb));
463 		error = (*fp->f_ops->fo_stat)(fp, &stb);
464 		if (error)
465 			break;
466 
467 		vattr_null(vap);
468 		switch(fp->f_type) {
469 		case DTYPE_SOCKET:
470 			vap->va_type = VSOCK;
471 			break;
472 		case DTYPE_PIPE:
473 			vap->va_type = VFIFO;
474 			break;
475 		default:
476 			/* use VNON perhaps? */
477 			vap->va_type = VBAD;
478 			break;
479 		}
480 		vap->va_mode = stb.st_mode;
481 		vap->va_nlink = stb.st_nlink;
482 		vap->va_uid = stb.st_uid;
483 		vap->va_gid = stb.st_gid;
484 		vap->va_fsid = stb.st_dev;
485 		vap->va_fileid = stb.st_ino;
486 		vap->va_size = stb.st_size;
487 		vap->va_blocksize = stb.st_blksize;
488 		vap->va_atime = stb.st_atimespec;
489 		vap->va_mtime = stb.st_mtimespec;
490 		vap->va_ctime = stb.st_ctimespec;
491 		vap->va_gen = stb.st_gen;
492 		vap->va_flags = stb.st_flags;
493 		vap->va_rdev = stb.st_rdev;
494 		vap->va_bytes = stb.st_blocks * stb.st_blksize;
495 		break;
496 	}
497 
498 	fd_putfile(fd);
499 	return (error);
500 }
501 
502 int
503 fdesc_getattr(void *v)
504 {
505 	struct vop_getattr_args /* {
506 		struct vnode *a_vp;
507 		struct vattr *a_vap;
508 		kauth_cred_t a_cred;
509 		struct lwp *a_l;
510 	} */ *ap = v;
511 	struct vnode *vp = ap->a_vp;
512 	struct vattr *vap = ap->a_vap;
513 	unsigned fd;
514 	int error = 0;
515 
516 	switch (VTOFDESC(vp)->fd_type) {
517 	case Froot:
518 	case Fdevfd:
519 	case Flink:
520 	case Fctty:
521 		vattr_null(vap);
522 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
523 
524 #define R_ALL (S_IRUSR|S_IRGRP|S_IROTH)
525 #define W_ALL (S_IWUSR|S_IWGRP|S_IWOTH)
526 #define X_ALL (S_IXUSR|S_IXGRP|S_IXOTH)
527 
528 		switch (VTOFDESC(vp)->fd_type) {
529 		case Flink:
530 			vap->va_mode = R_ALL|X_ALL;
531 			vap->va_type = VLNK;
532 			vap->va_rdev = 0;
533 			vap->va_nlink = 1;
534 			vap->va_size = strlen(VTOFDESC(vp)->fd_link);
535 			break;
536 
537 		case Fctty:
538 			vap->va_mode = R_ALL|W_ALL;
539 			vap->va_type = VCHR;
540 			vap->va_rdev = devctty;
541 			vap->va_nlink = 1;
542 			vap->va_size = 0;
543 			break;
544 
545 		default:
546 			vap->va_mode = R_ALL|X_ALL;
547 			vap->va_type = VDIR;
548 			vap->va_rdev = 0;
549 			vap->va_nlink = 2;
550 			vap->va_size = DEV_BSIZE;
551 			break;
552 		}
553 		vap->va_uid = 0;
554 		vap->va_gid = 0;
555 		vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0];
556 		vap->va_blocksize = DEV_BSIZE;
557 		vap->va_atime.tv_sec = boottime.tv_sec;
558 		vap->va_atime.tv_nsec = 0;
559 		vap->va_mtime = vap->va_atime;
560 		vap->va_ctime = vap->va_mtime;
561 		vap->va_gen = 0;
562 		vap->va_flags = 0;
563 		vap->va_bytes = 0;
564 		break;
565 
566 	case Fdesc:
567 		fd = VTOFDESC(vp)->fd_fd;
568 		error = fdesc_attr(fd, vap, ap->a_cred);
569 		break;
570 
571 	default:
572 		panic("fdesc_getattr");
573 		break;
574 	}
575 
576 	if (error == 0)
577 		vp->v_type = vap->va_type;
578 
579 	return (error);
580 }
581 
582 int
583 fdesc_setattr(void *v)
584 {
585 	struct vop_setattr_args /* {
586 		struct vnode *a_vp;
587 		struct vattr *a_vap;
588 		kauth_cred_t a_cred;
589 	} */ *ap = v;
590 	file_t *fp;
591 	unsigned fd;
592 
593 	/*
594 	 * Can't mess with the root vnode
595 	 */
596 	switch (VTOFDESC(ap->a_vp)->fd_type) {
597 	case Fdesc:
598 		break;
599 
600 	case Fctty:
601 		return (0);
602 
603 	default:
604 		return (EACCES);
605 	}
606 
607 	fd = VTOFDESC(ap->a_vp)->fd_fd;
608 	if ((fp = fd_getfile(fd)) == NULL)
609 		return (EBADF);
610 
611 	/*
612 	 * XXX: Can't reasonably set the attr's on any types currently.
613 	 *      On vnode's this will cause truncation and socket/pipes make
614 	 *      no sense.
615 	 */
616 	fd_putfile(fd);
617 	return (0);
618 }
619 
620 
621 struct fdesc_target {
622 	ino_t ft_fileno;
623 	u_char ft_type;
624 	u_char ft_namlen;
625 	const char *ft_name;
626 } fdesc_targets[] = {
627 #define N(s) sizeof(s)-1, s
628 	{ FD_DEVFD,  DT_DIR,     N("fd")     },
629 	{ FD_STDIN,  DT_LNK,     N("stdin")  },
630 	{ FD_STDOUT, DT_LNK,     N("stdout") },
631 	{ FD_STDERR, DT_LNK,     N("stderr") },
632 	{ FD_CTTY,   DT_UNKNOWN, N("tty")    },
633 #undef N
634 #define UIO_MX _DIRENT_RECLEN((struct dirent *)NULL, sizeof("stderr") - 1)
635 };
636 static int nfdesc_targets = sizeof(fdesc_targets) / sizeof(fdesc_targets[0]);
637 
638 int
639 fdesc_readdir(void *v)
640 {
641 	struct vop_readdir_args /* {
642 		struct vnode *a_vp;
643 		struct uio *a_uio;
644 		kauth_cred_t a_cred;
645 		int *a_eofflag;
646 		off_t **a_cookies;
647 		int *a_ncookies;
648 	} */ *ap = v;
649 	struct uio *uio = ap->a_uio;
650 	struct dirent d;
651 	off_t i;
652 	int j;
653 	int error;
654 	off_t *cookies = NULL;
655 	int ncookies;
656 	fdtab_t *dt;
657 
658 	switch (VTOFDESC(ap->a_vp)->fd_type) {
659 	case Fctty:
660 		return 0;
661 
662 	case Fdesc:
663 		return ENOTDIR;
664 
665 	default:
666 		break;
667 	}
668 
669 	dt = curlwp->l_fd->fd_dt;
670 
671 	if (uio->uio_resid < UIO_MX)
672 		return EINVAL;
673 	if (uio->uio_offset < 0)
674 		return EINVAL;
675 
676 	error = 0;
677 	i = uio->uio_offset;
678 	(void)memset(&d, 0, UIO_MX);
679 	d.d_reclen = UIO_MX;
680 	if (ap->a_ncookies)
681 		ncookies = uio->uio_resid / UIO_MX;
682 	else
683 		ncookies = 0;
684 
685 	if (VTOFDESC(ap->a_vp)->fd_type == Froot) {
686 		struct fdesc_target *ft;
687 
688 		if (i >= nfdesc_targets)
689 			return 0;
690 
691 		if (ap->a_ncookies) {
692 			ncookies = min(ncookies, (nfdesc_targets - i));
693 			cookies = malloc(ncookies * sizeof(off_t),
694 			    M_TEMP, M_WAITOK);
695 			*ap->a_cookies = cookies;
696 			*ap->a_ncookies = ncookies;
697 		}
698 
699 		for (ft = &fdesc_targets[i]; uio->uio_resid >= UIO_MX &&
700 		    i < nfdesc_targets; ft++, i++) {
701 			switch (ft->ft_fileno) {
702 			case FD_CTTY:
703 				if (cttyvp(curproc) == NULL)
704 					continue;
705 				break;
706 
707 			case FD_STDIN:
708 			case FD_STDOUT:
709 			case FD_STDERR:
710 				if ((ft->ft_fileno - FD_STDIN) >=
711 				    dt->dt_nfiles)
712 					continue;
713 				if (dt->dt_ff[ft->ft_fileno - FD_STDIN]
714 				    == NULL || dt->dt_ff[ft->ft_fileno -
715 				    FD_STDIN]->ff_file == NULL)
716 					continue;
717 				break;
718 			}
719 
720 			d.d_fileno = ft->ft_fileno;
721 			d.d_namlen = ft->ft_namlen;
722 			(void)memcpy(d.d_name, ft->ft_name, ft->ft_namlen + 1);
723 			d.d_type = ft->ft_type;
724 
725 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
726 				break;
727 			if (cookies)
728 				*cookies++ = i + 1;
729 		}
730 	} else {
731 		membar_consumer();
732 		if (ap->a_ncookies) {
733 			ncookies = min(ncookies, dt->dt_nfiles + 2);
734 			cookies = malloc(ncookies * sizeof(off_t),
735 			    M_TEMP, M_WAITOK);
736 			*ap->a_cookies = cookies;
737 			*ap->a_ncookies = ncookies;
738 		}
739 		for (; i - 2 < dt->dt_nfiles && uio->uio_resid >= UIO_MX; i++) {
740 			switch (i) {
741 			case 0:
742 			case 1:
743 				d.d_fileno = FD_ROOT;		/* XXX */
744 				d.d_namlen = i + 1;
745 				(void)memcpy(d.d_name, "..", d.d_namlen);
746 				d.d_name[i + 1] = '\0';
747 				d.d_type = DT_DIR;
748 				break;
749 
750 			default:
751 				j = (int)i - 2;
752 				if (dt->dt_ff[j] == NULL ||
753 				    dt->dt_ff[j]->ff_file == NULL)
754 					continue;
755 				d.d_fileno = j + FD_STDIN;
756 				d.d_namlen = sprintf(d.d_name, "%d", j);
757 				d.d_type = DT_UNKNOWN;
758 				break;
759 			}
760 
761 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
762 				break;
763 			if (cookies)
764 				*cookies++ = i + 1;
765 		}
766 	}
767 
768 	if (ap->a_ncookies && error) {
769 		free(*ap->a_cookies, M_TEMP);
770 		*ap->a_ncookies = 0;
771 		*ap->a_cookies = NULL;
772 	}
773 
774 	uio->uio_offset = i;
775 	return error;
776 }
777 
778 int
779 fdesc_readlink(void *v)
780 {
781 	struct vop_readlink_args /* {
782 		struct vnode *a_vp;
783 		struct uio *a_uio;
784 		kauth_cred_t a_cred;
785 	} */ *ap = v;
786 	struct vnode *vp = ap->a_vp;
787 	int error;
788 
789 	if (vp->v_type != VLNK)
790 		return (EPERM);
791 
792 	if (VTOFDESC(vp)->fd_type == Flink) {
793 		char *ln = VTOFDESC(vp)->fd_link;
794 		error = uiomove(ln, strlen(ln), ap->a_uio);
795 	} else {
796 		error = EOPNOTSUPP;
797 	}
798 
799 	return (error);
800 }
801 
802 int
803 fdesc_read(void *v)
804 {
805 	struct vop_read_args /* {
806 		struct vnode *a_vp;
807 		struct uio *a_uio;
808 		int  a_ioflag;
809 		kauth_cred_t a_cred;
810 	} */ *ap = v;
811 	int error = EOPNOTSUPP;
812 	struct vnode *vp = ap->a_vp;
813 
814 	switch (VTOFDESC(vp)->fd_type) {
815 	case Fctty:
816 		VOP_UNLOCK(vp);
817 		error = cdev_read(devctty, ap->a_uio, ap->a_ioflag);
818 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
819 		break;
820 
821 	default:
822 		error = EOPNOTSUPP;
823 		break;
824 	}
825 
826 	return (error);
827 }
828 
829 int
830 fdesc_write(void *v)
831 {
832 	struct vop_write_args /* {
833 		struct vnode *a_vp;
834 		struct uio *a_uio;
835 		int  a_ioflag;
836 		kauth_cred_t a_cred;
837 	} */ *ap = v;
838 	int error = EOPNOTSUPP;
839 	struct vnode *vp = ap->a_vp;
840 
841 	switch (VTOFDESC(vp)->fd_type) {
842 	case Fctty:
843 		VOP_UNLOCK(vp);
844 		error = cdev_write(devctty, ap->a_uio, ap->a_ioflag);
845 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
846 		break;
847 
848 	default:
849 		error = EOPNOTSUPP;
850 		break;
851 	}
852 
853 	return (error);
854 }
855 
856 int
857 fdesc_ioctl(void *v)
858 {
859 	struct vop_ioctl_args /* {
860 		struct vnode *a_vp;
861 		u_long a_command;
862 		void *a_data;
863 		int  a_fflag;
864 		kauth_cred_t a_cred;
865 	} */ *ap = v;
866 	int error = EOPNOTSUPP;
867 
868 	switch (VTOFDESC(ap->a_vp)->fd_type) {
869 	case Fctty:
870 		error = cdev_ioctl(devctty, ap->a_command, ap->a_data,
871 		    ap->a_fflag, curlwp);
872 		break;
873 
874 	default:
875 		error = EOPNOTSUPP;
876 		break;
877 	}
878 
879 	return (error);
880 }
881 
882 int
883 fdesc_poll(void *v)
884 {
885 	struct vop_poll_args /* {
886 		struct vnode *a_vp;
887 		int a_events;
888 	} */ *ap = v;
889 	int revents;
890 
891 	switch (VTOFDESC(ap->a_vp)->fd_type) {
892 	case Fctty:
893 		revents = cdev_poll(devctty, ap->a_events, curlwp);
894 		break;
895 
896 	default:
897 		revents = genfs_poll(v);
898 		break;
899 	}
900 
901 	return (revents);
902 }
903 
904 int
905 fdesc_kqfilter(void *v)
906 {
907 	struct vop_kqfilter_args /* {
908 		struct vnode *a_vp;
909 		struct knote *a_kn;
910 	} */ *ap = v;
911 	int error, fd;
912 	file_t *fp;
913 
914 	switch (VTOFDESC(ap->a_vp)->fd_type) {
915 	case Fctty:
916 		error = cdev_kqfilter(devctty, ap->a_kn);
917 		break;
918 
919 	case Fdesc:
920 		/* just invoke kqfilter for the underlying descriptor */
921 		fd = VTOFDESC(ap->a_vp)->fd_fd;
922 		if ((fp = fd_getfile(fd)) == NULL)
923 			return (1);
924 		error = (*fp->f_ops->fo_kqfilter)(fp, ap->a_kn);
925 		fd_putfile(fd);
926 		break;
927 
928 	default:
929 		return (genfs_kqfilter(v));
930 	}
931 
932 	return (error);
933 }
934 
935 int
936 fdesc_inactive(void *v)
937 {
938 	struct vop_inactive_args /* {
939 		struct vnode *a_vp;
940 	} */ *ap = v;
941 	struct vnode *vp = ap->a_vp;
942 
943 	/*
944 	 * Clear out the v_type field to avoid
945 	 * nasty things happening in vgone().
946 	 */
947 	VOP_UNLOCK(vp);
948 	vp->v_type = VNON;
949 	return (0);
950 }
951 
952 int
953 fdesc_reclaim(void *v)
954 {
955 	struct vop_reclaim_args /* {
956 		struct vnode *a_vp;
957 	} */ *ap = v;
958 	struct vnode *vp = ap->a_vp;
959 	struct fdescnode *fd = VTOFDESC(vp);
960 
961 	mutex_enter(&fdcache_lock);
962 	LIST_REMOVE(fd, fd_hash);
963 	free(vp->v_data, M_TEMP);
964 	vp->v_data = 0;
965 	mutex_exit(&fdcache_lock);
966 
967 	return (0);
968 }
969 
970 /*
971  * Return POSIX pathconf information applicable to special devices.
972  */
973 int
974 fdesc_pathconf(void *v)
975 {
976 	struct vop_pathconf_args /* {
977 		struct vnode *a_vp;
978 		int a_name;
979 		register_t *a_retval;
980 	} */ *ap = v;
981 
982 	switch (ap->a_name) {
983 	case _PC_LINK_MAX:
984 		*ap->a_retval = LINK_MAX;
985 		return (0);
986 	case _PC_MAX_CANON:
987 		*ap->a_retval = MAX_CANON;
988 		return (0);
989 	case _PC_MAX_INPUT:
990 		*ap->a_retval = MAX_INPUT;
991 		return (0);
992 	case _PC_PIPE_BUF:
993 		*ap->a_retval = PIPE_BUF;
994 		return (0);
995 	case _PC_CHOWN_RESTRICTED:
996 		*ap->a_retval = 1;
997 		return (0);
998 	case _PC_VDISABLE:
999 		*ap->a_retval = _POSIX_VDISABLE;
1000 		return (0);
1001 	case _PC_SYNC_IO:
1002 		*ap->a_retval = 1;
1003 		return (0);
1004 	default:
1005 		return (EINVAL);
1006 	}
1007 	/* NOTREACHED */
1008 }
1009 
1010 /*
1011  * Print out the contents of a /dev/fd vnode.
1012  */
1013 /* ARGSUSED */
1014 int
1015 fdesc_print(void *v)
1016 {
1017 	printf("tag VT_NON, fdesc vnode\n");
1018 	return (0);
1019 }
1020 
1021 int
1022 fdesc_link(void *v)
1023 {
1024 	struct vop_link_args /* {
1025 		struct vnode *a_dvp;
1026 		struct vnode *a_vp;
1027 		struct componentname *a_cnp;
1028 	} */ *ap = v;
1029 
1030 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1031 	vput(ap->a_dvp);
1032 	return (EROFS);
1033 }
1034 
1035 int
1036 fdesc_symlink(void *v)
1037 {
1038 	struct vop_symlink_args /* {
1039 		struct vnode *a_dvp;
1040 		struct vnode **a_vpp;
1041 		struct componentname *a_cnp;
1042 		struct vattr *a_vap;
1043 		char *a_target;
1044 	} */ *ap = v;
1045 
1046 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1047 	vput(ap->a_dvp);
1048 	return (EROFS);
1049 }
1050