xref: /netbsd-src/sys/miscfs/fdesc/fdesc_vnops.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: fdesc_vnops.c,v 1.119 2014/03/20 18:04:05 christos 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 1.119 2014/03/20 18:04:05 christos 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 	error = VOP_LOCK(*vpp, LK_EXCLUSIVE);
253 	KASSERT(error == 0);
254 	LIST_INSERT_HEAD(fc, fd, fd_hash);
255 	mutex_exit(&fdcache_lock);
256 
257 	return 0;
258 }
259 
260 /*
261  * vp is the current namei directory
262  * ndp is the name to locate in that directory...
263  */
264 int
265 fdesc_lookup(void *v)
266 {
267 	struct vop_lookup_v2_args /* {
268 		struct vnode * a_dvp;
269 		struct vnode ** a_vpp;
270 		struct componentname * a_cnp;
271 	} */ *ap = v;
272 	struct vnode **vpp = ap->a_vpp;
273 	struct vnode *dvp = ap->a_dvp;
274 	struct componentname *cnp = ap->a_cnp;
275 	struct lwp *l = curlwp;
276 	const char *pname = cnp->cn_nameptr;
277 	struct proc *p = l->l_proc;
278 	unsigned fd = 0;
279 	int error;
280 	struct vnode *fvp;
281 	const char *ln;
282 	fdtab_t *dt;
283 
284 	dt = curlwp->l_fd->fd_dt;
285 
286 	if (cnp->cn_namelen == 1 && *pname == '.') {
287 		*vpp = dvp;
288 		vref(dvp);
289 		return (0);
290 	}
291 
292 	switch (VTOFDESC(dvp)->fd_type) {
293 	default:
294 	case Flink:
295 	case Fdesc:
296 	case Fctty:
297 		error = ENOTDIR;
298 		goto bad;
299 
300 	case Froot:
301 		if (cnp->cn_namelen == 2 && memcmp(pname, "fd", 2) == 0) {
302 			error = fdesc_allocvp(Fdevfd, FD_DEVFD, dvp->v_mount, &fvp);
303 			if (error)
304 				goto bad;
305 			*vpp = fvp;
306 			fvp->v_type = VDIR;
307 			goto good;
308 		}
309 
310 		if (cnp->cn_namelen == 3 && memcmp(pname, "tty", 3) == 0) {
311 			struct vnode *ttyvp = cttyvp(p);
312 			if (ttyvp == NULL) {
313 				error = ENXIO;
314 				goto bad;
315 			}
316 			error = fdesc_allocvp(Fctty, FD_CTTY, dvp->v_mount, &fvp);
317 			if (error)
318 				goto bad;
319 			*vpp = fvp;
320 			fvp->v_type = VCHR;
321 			goto good;
322 		}
323 
324 		ln = 0;
325 		switch (cnp->cn_namelen) {
326 		case 5:
327 			if (memcmp(pname, "stdin", 5) == 0) {
328 				ln = "fd/0";
329 				fd = FD_STDIN;
330 			}
331 			break;
332 		case 6:
333 			if (memcmp(pname, "stdout", 6) == 0) {
334 				ln = "fd/1";
335 				fd = FD_STDOUT;
336 			} else
337 			if (memcmp(pname, "stderr", 6) == 0) {
338 				ln = "fd/2";
339 				fd = FD_STDERR;
340 			}
341 			break;
342 		}
343 
344 		if (ln) {
345 			error = fdesc_allocvp(Flink, fd, dvp->v_mount, &fvp);
346 			if (error)
347 				goto bad;
348 			/* XXXUNCONST */
349 			VTOFDESC(fvp)->fd_link = __UNCONST(ln);
350 			*vpp = fvp;
351 			fvp->v_type = VLNK;
352 			goto good;
353 		} else {
354 			error = ENOENT;
355 			goto bad;
356 		}
357 
358 		/* FALL THROUGH */
359 
360 	case Fdevfd:
361 		if (cnp->cn_namelen == 2 && memcmp(pname, "..", 2) == 0) {
362 			VOP_UNLOCK(dvp);
363 			error = fdesc_root(dvp->v_mount, vpp);
364 			vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
365 			if (error)
366 				goto bad;
367 			return (error);
368 		}
369 
370 		fd = 0;
371 		while (*pname >= '0' && *pname <= '9') {
372 			fd = 10 * fd + *pname++ - '0';
373 			if (fd >= dt->dt_nfiles)
374 				break;
375 		}
376 
377 		if (*pname != '\0') {
378 			error = ENOENT;
379 			goto bad;
380 		}
381 
382 		if (fd >= dt->dt_nfiles || dt->dt_ff[fd] == NULL ||
383 		    dt->dt_ff[fd]->ff_file == NULL) {
384 			error = EBADF;
385 			goto bad;
386 		}
387 
388 		error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp);
389 		if (error)
390 			goto bad;
391 		VTOFDESC(fvp)->fd_fd = fd;
392 		*vpp = fvp;
393 		goto good;
394 	}
395 
396 bad:
397 	*vpp = NULL;
398 	return (error);
399 
400 good:
401 	VOP_UNLOCK(*vpp);
402 	return (0);
403 }
404 
405 int
406 fdesc_open(void *v)
407 {
408 	struct vop_open_args /* {
409 		struct vnode *a_vp;
410 		int  a_mode;
411 		kauth_cred_t a_cred;
412 	} */ *ap = v;
413 	struct vnode *vp = ap->a_vp;
414 
415 	switch (VTOFDESC(vp)->fd_type) {
416 	case Fdesc:
417 		/*
418 		 * XXX Kludge: set dupfd to contain the value of the
419 		 * the file descriptor being sought for duplication. The error
420 		 * return ensures that the vnode for this device will be
421 		 * released by vn_open. Open will detect this special error and
422 		 * take the actions in dupfdopen.  Other callers of vn_open or
423 		 * VOP_OPEN will simply report the error.
424 		 */
425 		curlwp->l_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
426 		return EDUPFD;
427 
428 	case Fctty:
429 		return cdev_open(devctty, ap->a_mode, 0, curlwp);
430 	case Froot:
431 	case Fdevfd:
432 	case Flink:
433 		break;
434 	}
435 
436 	return (0);
437 }
438 
439 static int
440 fdesc_attr(int fd, struct vattr *vap, kauth_cred_t cred)
441 {
442 	file_t *fp;
443 	struct stat stb;
444 	int error;
445 
446 	if ((fp = fd_getfile(fd)) == NULL)
447 		return (EBADF);
448 
449 	switch (fp->f_type) {
450 	case DTYPE_VNODE:
451 		vn_lock((struct vnode *) fp->f_data, LK_SHARED | LK_RETRY);
452 		error = VOP_GETATTR((struct vnode *) fp->f_data, vap, cred);
453 		VOP_UNLOCK((struct vnode *) fp->f_data);
454 		if (error == 0 && vap->va_type == VDIR) {
455 			/*
456 			 * directories can cause loops in the namespace,
457 			 * so turn off the 'x' bits to avoid trouble.
458 			 */
459 			vap->va_mode &= ~(S_IXUSR|S_IXGRP|S_IXOTH);
460 		}
461 		break;
462 
463 	default:
464 		memset(&stb, 0, sizeof(stb));
465 		error = (*fp->f_ops->fo_stat)(fp, &stb);
466 		if (error)
467 			break;
468 
469 		vattr_null(vap);
470 		switch(fp->f_type) {
471 		case DTYPE_SOCKET:
472 			vap->va_type = VSOCK;
473 			break;
474 		case DTYPE_PIPE:
475 			vap->va_type = VFIFO;
476 			break;
477 		default:
478 			/* use VNON perhaps? */
479 			vap->va_type = VBAD;
480 			break;
481 		}
482 		vap->va_mode = stb.st_mode;
483 		vap->va_nlink = stb.st_nlink;
484 		vap->va_uid = stb.st_uid;
485 		vap->va_gid = stb.st_gid;
486 		vap->va_fsid = stb.st_dev;
487 		vap->va_fileid = stb.st_ino;
488 		vap->va_size = stb.st_size;
489 		vap->va_blocksize = stb.st_blksize;
490 		vap->va_atime = stb.st_atimespec;
491 		vap->va_mtime = stb.st_mtimespec;
492 		vap->va_ctime = stb.st_ctimespec;
493 		vap->va_gen = stb.st_gen;
494 		vap->va_flags = stb.st_flags;
495 		vap->va_rdev = stb.st_rdev;
496 		vap->va_bytes = stb.st_blocks * stb.st_blksize;
497 		break;
498 	}
499 
500 	fd_putfile(fd);
501 	return (error);
502 }
503 
504 int
505 fdesc_getattr(void *v)
506 {
507 	struct vop_getattr_args /* {
508 		struct vnode *a_vp;
509 		struct vattr *a_vap;
510 		kauth_cred_t a_cred;
511 		struct lwp *a_l;
512 	} */ *ap = v;
513 	struct vnode *vp = ap->a_vp;
514 	struct vattr *vap = ap->a_vap;
515 	unsigned fd;
516 	int error = 0;
517 
518 	switch (VTOFDESC(vp)->fd_type) {
519 	case Froot:
520 	case Fdevfd:
521 	case Flink:
522 	case Fctty:
523 		vattr_null(vap);
524 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
525 
526 #define R_ALL (S_IRUSR|S_IRGRP|S_IROTH)
527 #define W_ALL (S_IWUSR|S_IWGRP|S_IWOTH)
528 #define X_ALL (S_IXUSR|S_IXGRP|S_IXOTH)
529 
530 		switch (VTOFDESC(vp)->fd_type) {
531 		case Flink:
532 			vap->va_mode = R_ALL|X_ALL;
533 			vap->va_type = VLNK;
534 			vap->va_rdev = 0;
535 			vap->va_nlink = 1;
536 			vap->va_size = strlen(VTOFDESC(vp)->fd_link);
537 			break;
538 
539 		case Fctty:
540 			vap->va_mode = R_ALL|W_ALL;
541 			vap->va_type = VCHR;
542 			vap->va_rdev = devctty;
543 			vap->va_nlink = 1;
544 			vap->va_size = 0;
545 			break;
546 
547 		default:
548 			vap->va_mode = R_ALL|X_ALL;
549 			vap->va_type = VDIR;
550 			vap->va_rdev = 0;
551 			vap->va_nlink = 2;
552 			vap->va_size = DEV_BSIZE;
553 			break;
554 		}
555 		vap->va_uid = 0;
556 		vap->va_gid = 0;
557 		vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0];
558 		vap->va_blocksize = DEV_BSIZE;
559 		vap->va_atime.tv_sec = boottime.tv_sec;
560 		vap->va_atime.tv_nsec = 0;
561 		vap->va_mtime = vap->va_atime;
562 		vap->va_ctime = vap->va_mtime;
563 		vap->va_gen = 0;
564 		vap->va_flags = 0;
565 		vap->va_bytes = 0;
566 		break;
567 
568 	case Fdesc:
569 		fd = VTOFDESC(vp)->fd_fd;
570 		error = fdesc_attr(fd, vap, ap->a_cred);
571 		break;
572 
573 	default:
574 		panic("fdesc_getattr");
575 		break;
576 	}
577 
578 	if (error == 0)
579 		vp->v_type = vap->va_type;
580 
581 	return (error);
582 }
583 
584 int
585 fdesc_setattr(void *v)
586 {
587 	struct vop_setattr_args /* {
588 		struct vnode *a_vp;
589 		struct vattr *a_vap;
590 		kauth_cred_t a_cred;
591 	} */ *ap = v;
592 	file_t *fp;
593 	unsigned fd;
594 
595 	/*
596 	 * Can't mess with the root vnode
597 	 */
598 	switch (VTOFDESC(ap->a_vp)->fd_type) {
599 	case Fdesc:
600 		break;
601 
602 	case Fctty:
603 		return (0);
604 
605 	default:
606 		return (EACCES);
607 	}
608 
609 	fd = VTOFDESC(ap->a_vp)->fd_fd;
610 	if ((fp = fd_getfile(fd)) == NULL)
611 		return (EBADF);
612 
613 	/*
614 	 * XXX: Can't reasonably set the attr's on any types currently.
615 	 *      On vnode's this will cause truncation and socket/pipes make
616 	 *      no sense.
617 	 */
618 	fd_putfile(fd);
619 	return (0);
620 }
621 
622 
623 struct fdesc_target {
624 	ino_t ft_fileno;
625 	u_char ft_type;
626 	u_char ft_namlen;
627 	const char *ft_name;
628 } fdesc_targets[] = {
629 #define N(s) sizeof(s)-1, s
630 	{ FD_DEVFD,  DT_DIR,     N("fd")     },
631 	{ FD_STDIN,  DT_LNK,     N("stdin")  },
632 	{ FD_STDOUT, DT_LNK,     N("stdout") },
633 	{ FD_STDERR, DT_LNK,     N("stderr") },
634 	{ FD_CTTY,   DT_UNKNOWN, N("tty")    },
635 #undef N
636 #define UIO_MX _DIRENT_RECLEN((struct dirent *)NULL, sizeof("stderr") - 1)
637 };
638 static int nfdesc_targets = sizeof(fdesc_targets) / sizeof(fdesc_targets[0]);
639 
640 int
641 fdesc_readdir(void *v)
642 {
643 	struct vop_readdir_args /* {
644 		struct vnode *a_vp;
645 		struct uio *a_uio;
646 		kauth_cred_t a_cred;
647 		int *a_eofflag;
648 		off_t **a_cookies;
649 		int *a_ncookies;
650 	} */ *ap = v;
651 	struct uio *uio = ap->a_uio;
652 	struct dirent d;
653 	off_t i;
654 	int j;
655 	int error;
656 	off_t *cookies = NULL;
657 	int ncookies;
658 	fdtab_t *dt;
659 
660 	switch (VTOFDESC(ap->a_vp)->fd_type) {
661 	case Fctty:
662 		return 0;
663 
664 	case Fdesc:
665 		return ENOTDIR;
666 
667 	default:
668 		break;
669 	}
670 
671 	dt = curlwp->l_fd->fd_dt;
672 
673 	if (uio->uio_resid < UIO_MX)
674 		return EINVAL;
675 	if (uio->uio_offset < 0)
676 		return EINVAL;
677 
678 	error = 0;
679 	i = uio->uio_offset;
680 	(void)memset(&d, 0, UIO_MX);
681 	d.d_reclen = UIO_MX;
682 	if (ap->a_ncookies)
683 		ncookies = uio->uio_resid / UIO_MX;
684 	else
685 		ncookies = 0;
686 
687 	if (VTOFDESC(ap->a_vp)->fd_type == Froot) {
688 		struct fdesc_target *ft;
689 
690 		if (i >= nfdesc_targets)
691 			return 0;
692 
693 		if (ap->a_ncookies) {
694 			ncookies = min(ncookies, (nfdesc_targets - i));
695 			cookies = malloc(ncookies * sizeof(off_t),
696 			    M_TEMP, M_WAITOK);
697 			*ap->a_cookies = cookies;
698 			*ap->a_ncookies = ncookies;
699 		}
700 
701 		for (ft = &fdesc_targets[i]; uio->uio_resid >= UIO_MX &&
702 		    i < nfdesc_targets; ft++, i++) {
703 			switch (ft->ft_fileno) {
704 			case FD_CTTY:
705 				if (cttyvp(curproc) == NULL)
706 					continue;
707 				break;
708 
709 			case FD_STDIN:
710 			case FD_STDOUT:
711 			case FD_STDERR:
712 				if ((ft->ft_fileno - FD_STDIN) >=
713 				    dt->dt_nfiles)
714 					continue;
715 				if (dt->dt_ff[ft->ft_fileno - FD_STDIN]
716 				    == NULL || dt->dt_ff[ft->ft_fileno -
717 				    FD_STDIN]->ff_file == NULL)
718 					continue;
719 				break;
720 			}
721 
722 			d.d_fileno = ft->ft_fileno;
723 			d.d_namlen = ft->ft_namlen;
724 			(void)memcpy(d.d_name, ft->ft_name, ft->ft_namlen + 1);
725 			d.d_type = ft->ft_type;
726 
727 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
728 				break;
729 			if (cookies)
730 				*cookies++ = i + 1;
731 		}
732 	} else {
733 		membar_consumer();
734 		if (ap->a_ncookies) {
735 			ncookies = min(ncookies, dt->dt_nfiles + 2);
736 			cookies = malloc(ncookies * sizeof(off_t),
737 			    M_TEMP, M_WAITOK);
738 			*ap->a_cookies = cookies;
739 			*ap->a_ncookies = ncookies;
740 		}
741 		for (; i - 2 < dt->dt_nfiles && uio->uio_resid >= UIO_MX; i++) {
742 			switch (i) {
743 			case 0:
744 			case 1:
745 				d.d_fileno = FD_ROOT;		/* XXX */
746 				d.d_namlen = i + 1;
747 				(void)memcpy(d.d_name, "..", d.d_namlen);
748 				d.d_name[i + 1] = '\0';
749 				d.d_type = DT_DIR;
750 				break;
751 
752 			default:
753 				j = (int)i - 2;
754 				if (dt->dt_ff[j] == NULL ||
755 				    dt->dt_ff[j]->ff_file == NULL)
756 					continue;
757 				d.d_fileno = j + FD_STDIN;
758 				d.d_namlen = snprintf(d.d_name,
759 				    sizeof(d.d_name), "%d", j);
760 				d.d_type = DT_UNKNOWN;
761 				break;
762 			}
763 
764 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
765 				break;
766 			if (cookies)
767 				*cookies++ = i + 1;
768 		}
769 	}
770 
771 	if (ap->a_ncookies && error) {
772 		free(*ap->a_cookies, M_TEMP);
773 		*ap->a_ncookies = 0;
774 		*ap->a_cookies = NULL;
775 	}
776 
777 	uio->uio_offset = i;
778 	return error;
779 }
780 
781 int
782 fdesc_readlink(void *v)
783 {
784 	struct vop_readlink_args /* {
785 		struct vnode *a_vp;
786 		struct uio *a_uio;
787 		kauth_cred_t a_cred;
788 	} */ *ap = v;
789 	struct vnode *vp = ap->a_vp;
790 	int error;
791 
792 	if (vp->v_type != VLNK)
793 		return (EPERM);
794 
795 	if (VTOFDESC(vp)->fd_type == Flink) {
796 		char *ln = VTOFDESC(vp)->fd_link;
797 		error = uiomove(ln, strlen(ln), ap->a_uio);
798 	} else {
799 		error = EOPNOTSUPP;
800 	}
801 
802 	return (error);
803 }
804 
805 int
806 fdesc_read(void *v)
807 {
808 	struct vop_read_args /* {
809 		struct vnode *a_vp;
810 		struct uio *a_uio;
811 		int  a_ioflag;
812 		kauth_cred_t a_cred;
813 	} */ *ap = v;
814 	int error = EOPNOTSUPP;
815 	struct vnode *vp = ap->a_vp;
816 
817 	switch (VTOFDESC(vp)->fd_type) {
818 	case Fctty:
819 		VOP_UNLOCK(vp);
820 		error = cdev_read(devctty, ap->a_uio, ap->a_ioflag);
821 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
822 		break;
823 
824 	default:
825 		error = EOPNOTSUPP;
826 		break;
827 	}
828 
829 	return (error);
830 }
831 
832 int
833 fdesc_write(void *v)
834 {
835 	struct vop_write_args /* {
836 		struct vnode *a_vp;
837 		struct uio *a_uio;
838 		int  a_ioflag;
839 		kauth_cred_t a_cred;
840 	} */ *ap = v;
841 	int error = EOPNOTSUPP;
842 	struct vnode *vp = ap->a_vp;
843 
844 	switch (VTOFDESC(vp)->fd_type) {
845 	case Fctty:
846 		VOP_UNLOCK(vp);
847 		error = cdev_write(devctty, ap->a_uio, ap->a_ioflag);
848 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
849 		break;
850 
851 	default:
852 		error = EOPNOTSUPP;
853 		break;
854 	}
855 
856 	return (error);
857 }
858 
859 int
860 fdesc_ioctl(void *v)
861 {
862 	struct vop_ioctl_args /* {
863 		struct vnode *a_vp;
864 		u_long a_command;
865 		void *a_data;
866 		int  a_fflag;
867 		kauth_cred_t a_cred;
868 	} */ *ap = v;
869 	int error = EOPNOTSUPP;
870 
871 	switch (VTOFDESC(ap->a_vp)->fd_type) {
872 	case Fctty:
873 		error = cdev_ioctl(devctty, ap->a_command, ap->a_data,
874 		    ap->a_fflag, curlwp);
875 		break;
876 
877 	default:
878 		error = EOPNOTSUPP;
879 		break;
880 	}
881 
882 	return (error);
883 }
884 
885 int
886 fdesc_poll(void *v)
887 {
888 	struct vop_poll_args /* {
889 		struct vnode *a_vp;
890 		int a_events;
891 	} */ *ap = v;
892 	int revents;
893 
894 	switch (VTOFDESC(ap->a_vp)->fd_type) {
895 	case Fctty:
896 		revents = cdev_poll(devctty, ap->a_events, curlwp);
897 		break;
898 
899 	default:
900 		revents = genfs_poll(v);
901 		break;
902 	}
903 
904 	return (revents);
905 }
906 
907 int
908 fdesc_kqfilter(void *v)
909 {
910 	struct vop_kqfilter_args /* {
911 		struct vnode *a_vp;
912 		struct knote *a_kn;
913 	} */ *ap = v;
914 	int error, fd;
915 	file_t *fp;
916 
917 	switch (VTOFDESC(ap->a_vp)->fd_type) {
918 	case Fctty:
919 		error = cdev_kqfilter(devctty, ap->a_kn);
920 		break;
921 
922 	case Fdesc:
923 		/* just invoke kqfilter for the underlying descriptor */
924 		fd = VTOFDESC(ap->a_vp)->fd_fd;
925 		if ((fp = fd_getfile(fd)) == NULL)
926 			return (1);
927 		error = (*fp->f_ops->fo_kqfilter)(fp, ap->a_kn);
928 		fd_putfile(fd);
929 		break;
930 
931 	default:
932 		return (genfs_kqfilter(v));
933 	}
934 
935 	return (error);
936 }
937 
938 int
939 fdesc_inactive(void *v)
940 {
941 	struct vop_inactive_args /* {
942 		struct vnode *a_vp;
943 	} */ *ap = v;
944 	struct vnode *vp = ap->a_vp;
945 
946 	/*
947 	 * Clear out the v_type field to avoid
948 	 * nasty things happening in vgone().
949 	 */
950 	VOP_UNLOCK(vp);
951 	vp->v_type = VNON;
952 	return (0);
953 }
954 
955 int
956 fdesc_reclaim(void *v)
957 {
958 	struct vop_reclaim_args /* {
959 		struct vnode *a_vp;
960 	} */ *ap = v;
961 	struct vnode *vp = ap->a_vp;
962 	struct fdescnode *fd = VTOFDESC(vp);
963 
964 	mutex_enter(&fdcache_lock);
965 	LIST_REMOVE(fd, fd_hash);
966 	free(vp->v_data, M_TEMP);
967 	vp->v_data = 0;
968 	mutex_exit(&fdcache_lock);
969 
970 	return (0);
971 }
972 
973 /*
974  * Return POSIX pathconf information applicable to special devices.
975  */
976 int
977 fdesc_pathconf(void *v)
978 {
979 	struct vop_pathconf_args /* {
980 		struct vnode *a_vp;
981 		int a_name;
982 		register_t *a_retval;
983 	} */ *ap = v;
984 
985 	switch (ap->a_name) {
986 	case _PC_LINK_MAX:
987 		*ap->a_retval = LINK_MAX;
988 		return (0);
989 	case _PC_MAX_CANON:
990 		*ap->a_retval = MAX_CANON;
991 		return (0);
992 	case _PC_MAX_INPUT:
993 		*ap->a_retval = MAX_INPUT;
994 		return (0);
995 	case _PC_PIPE_BUF:
996 		*ap->a_retval = PIPE_BUF;
997 		return (0);
998 	case _PC_CHOWN_RESTRICTED:
999 		*ap->a_retval = 1;
1000 		return (0);
1001 	case _PC_VDISABLE:
1002 		*ap->a_retval = _POSIX_VDISABLE;
1003 		return (0);
1004 	case _PC_SYNC_IO:
1005 		*ap->a_retval = 1;
1006 		return (0);
1007 	default:
1008 		return (EINVAL);
1009 	}
1010 	/* NOTREACHED */
1011 }
1012 
1013 /*
1014  * Print out the contents of a /dev/fd vnode.
1015  */
1016 /* ARGSUSED */
1017 int
1018 fdesc_print(void *v)
1019 {
1020 	printf("tag VT_NON, fdesc vnode\n");
1021 	return (0);
1022 }
1023 
1024 int
1025 fdesc_link(void *v)
1026 {
1027 	struct vop_link_args /* {
1028 		struct vnode *a_dvp;
1029 		struct vnode *a_vp;
1030 		struct componentname *a_cnp;
1031 	} */ *ap = v;
1032 
1033 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1034 	vput(ap->a_dvp);
1035 	return (EROFS);
1036 }
1037 
1038 int
1039 fdesc_symlink(void *v)
1040 {
1041 	struct vop_symlink_v3_args /* {
1042 		struct vnode *a_dvp;
1043 		struct vnode **a_vpp;
1044 		struct componentname *a_cnp;
1045 		struct vattr *a_vap;
1046 		char *a_target;
1047 	} */ *ap = v;
1048 
1049 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
1050 	return (EROFS);
1051 }
1052