Home | History | Annotate | Line # | Download | only in fdesc
fdesc_vnops.c revision 1.134.6.2
      1 /*	$NetBSD: fdesc_vnops.c,v 1.134.6.2 2021/08/01 22:42:39 thorpej 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 1.134.6.2 2021/08/01 22:42:39 thorpej 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/resourcevar.h>
     51 #include <sys/socketvar.h>
     52 #include <sys/filedesc.h>
     53 #include <sys/vnode.h>
     54 #include <sys/malloc.h>
     55 #include <sys/conf.h>
     56 #include <sys/file.h>
     57 #include <sys/stat.h>
     58 #include <sys/mount.h>
     59 #include <sys/namei.h>
     60 #include <sys/buf.h>
     61 #include <sys/dirent.h>
     62 #include <sys/tty.h>
     63 #include <sys/kauth.h>
     64 #include <sys/atomic.h>
     65 
     66 #include <miscfs/fdesc/fdesc.h>
     67 #include <miscfs/genfs/genfs.h>
     68 
     69 #define cttyvp(p) ((p)->p_lflag & PL_CONTROLT ? (p)->p_session->s_ttyvp : NULL)
     70 
     71 dev_t devctty;
     72 
     73 #if (FD_STDIN != FD_STDOUT-1) || (FD_STDOUT != FD_STDERR-1)
     74 FD_STDIN, FD_STDOUT, FD_STDERR must be a sequence n, n+1, n+2
     75 #endif
     76 
     77 int	fdesc_lookup(void *);
     78 #define	fdesc_create	genfs_eopnotsupp
     79 #define	fdesc_mknod	genfs_eopnotsupp
     80 int	fdesc_open(void *);
     81 #define	fdesc_close	genfs_nullop
     82 #define	fdesc_access	genfs_nullop
     83 int	fdesc_getattr(void *);
     84 int	fdesc_setattr(void *);
     85 int	fdesc_read(void *);
     86 int	fdesc_write(void *);
     87 int	fdesc_ioctl(void *);
     88 int	fdesc_poll(void *);
     89 int	fdesc_kqfilter(void *);
     90 #define	fdesc_mmap	genfs_eopnotsupp
     91 #define	fdesc_fcntl	genfs_fcntl
     92 #define	fdesc_fsync	genfs_nullop
     93 #define	fdesc_seek	genfs_seek
     94 #define	fdesc_remove	genfs_eopnotsupp
     95 int	fdesc_link(void *);
     96 #define	fdesc_rename	genfs_eopnotsupp
     97 #define	fdesc_mkdir	genfs_eopnotsupp
     98 #define	fdesc_rmdir	genfs_eopnotsupp
     99 int	fdesc_symlink(void *);
    100 int	fdesc_readdir(void *);
    101 int	fdesc_readlink(void *);
    102 #define	fdesc_abortop	genfs_abortop
    103 int	fdesc_inactive(void *);
    104 int	fdesc_reclaim(void *);
    105 #define	fdesc_lock	genfs_lock
    106 #define	fdesc_unlock	genfs_unlock
    107 #define	fdesc_bmap	genfs_eopnotsupp
    108 #define	fdesc_strategy	genfs_badop
    109 int	fdesc_print(void *);
    110 int	fdesc_pathconf(void *);
    111 #define	fdesc_islocked	genfs_islocked
    112 #define	fdesc_advlock	genfs_einval
    113 #define	fdesc_bwrite	genfs_eopnotsupp
    114 #define fdesc_revoke	genfs_revoke
    115 #define fdesc_putpages	genfs_null_putpages
    116 
    117 static int fdesc_attr(int, struct vattr *, kauth_cred_t);
    118 
    119 int (**fdesc_vnodeop_p)(void *);
    120 const struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = {
    121 	{ &vop_default_desc, vn_default_error },
    122 	{ &vop_parsepath_desc, genfs_parsepath },	/* parsepath */
    123 	{ &vop_lookup_desc, fdesc_lookup },		/* lookup */
    124 	{ &vop_create_desc, fdesc_create },		/* create */
    125 	{ &vop_mknod_desc, fdesc_mknod },		/* mknod */
    126 	{ &vop_open_desc, fdesc_open },			/* open */
    127 	{ &vop_close_desc, fdesc_close },		/* close */
    128 	{ &vop_access_desc, fdesc_access },		/* access */
    129 	{ &vop_accessx_desc, genfs_accessx },		/* accessx */
    130 	{ &vop_getattr_desc, fdesc_getattr },		/* getattr */
    131 	{ &vop_setattr_desc, fdesc_setattr },		/* setattr */
    132 	{ &vop_read_desc, fdesc_read },			/* read */
    133 	{ &vop_write_desc, fdesc_write },		/* write */
    134 	{ &vop_fallocate_desc, genfs_eopnotsupp },	/* fallocate */
    135 	{ &vop_fdiscard_desc, genfs_eopnotsupp },	/* fdiscard */
    136 	{ &vop_ioctl_desc, fdesc_ioctl },		/* ioctl */
    137 	{ &vop_fcntl_desc, fdesc_fcntl },		/* fcntl */
    138 	{ &vop_poll_desc, fdesc_poll },			/* poll */
    139 	{ &vop_kqfilter_desc, fdesc_kqfilter },		/* kqfilter */
    140 	{ &vop_revoke_desc, fdesc_revoke },		/* revoke */
    141 	{ &vop_mmap_desc, fdesc_mmap },			/* mmap */
    142 	{ &vop_fsync_desc, fdesc_fsync },		/* fsync */
    143 	{ &vop_seek_desc, fdesc_seek },			/* seek */
    144 	{ &vop_remove_desc, fdesc_remove },		/* remove */
    145 	{ &vop_link_desc, fdesc_link },			/* link */
    146 	{ &vop_rename_desc, fdesc_rename },		/* rename */
    147 	{ &vop_mkdir_desc, fdesc_mkdir },		/* mkdir */
    148 	{ &vop_rmdir_desc, fdesc_rmdir },		/* rmdir */
    149 	{ &vop_symlink_desc, fdesc_symlink },		/* symlink */
    150 	{ &vop_readdir_desc, fdesc_readdir },		/* readdir */
    151 	{ &vop_readlink_desc, fdesc_readlink },		/* readlink */
    152 	{ &vop_abortop_desc, fdesc_abortop },		/* abortop */
    153 	{ &vop_inactive_desc, fdesc_inactive },		/* inactive */
    154 	{ &vop_reclaim_desc, fdesc_reclaim },		/* reclaim */
    155 	{ &vop_lock_desc, fdesc_lock },			/* lock */
    156 	{ &vop_unlock_desc, fdesc_unlock },		/* unlock */
    157 	{ &vop_bmap_desc, fdesc_bmap },			/* bmap */
    158 	{ &vop_strategy_desc, fdesc_strategy },		/* strategy */
    159 	{ &vop_print_desc, fdesc_print },		/* print */
    160 	{ &vop_islocked_desc, fdesc_islocked },		/* islocked */
    161 	{ &vop_pathconf_desc, fdesc_pathconf },		/* pathconf */
    162 	{ &vop_advlock_desc, fdesc_advlock },		/* advlock */
    163 	{ &vop_bwrite_desc, fdesc_bwrite },		/* bwrite */
    164 	{ &vop_putpages_desc, fdesc_putpages },		/* putpages */
    165 	{ NULL, NULL }
    166 };
    167 
    168 const struct vnodeopv_desc fdesc_vnodeop_opv_desc =
    169 	{ &fdesc_vnodeop_p, fdesc_vnodeop_entries };
    170 
    171 /*
    172  * Initialise cache headers
    173  */
    174 void
    175 fdesc_init(void)
    176 {
    177 	int cttymajor;
    178 
    179 	/* locate the major number */
    180 	cttymajor = devsw_name2chr("ctty", NULL, 0);
    181 	devctty = makedev(cttymajor, 0);
    182 }
    183 
    184 void
    185 fdesc_done(void)
    186 {
    187 }
    188 
    189 /*
    190  * vp is the current namei directory
    191  * ndp is the name to locate in that directory...
    192  */
    193 int
    194 fdesc_lookup(void *v)
    195 {
    196 	struct vop_lookup_v2_args /* {
    197 		struct vnode * a_dvp;
    198 		struct vnode ** a_vpp;
    199 		struct componentname * a_cnp;
    200 	} */ *ap = v;
    201 	struct vnode **vpp = ap->a_vpp;
    202 	struct vnode *dvp = ap->a_dvp;
    203 	struct componentname *cnp = ap->a_cnp;
    204 	struct lwp *l = curlwp;
    205 	const char *pname = cnp->cn_nameptr;
    206 	struct proc *p = l->l_proc;
    207 	unsigned fd = 0;
    208 	int error, ix = -1;
    209 	fdtab_t *dt;
    210 
    211 	dt = atomic_load_consume(&curlwp->l_fd->fd_dt);
    212 
    213 	if (cnp->cn_namelen == 1 && *pname == '.') {
    214 		*vpp = dvp;
    215 		vref(dvp);
    216 		return (0);
    217 	}
    218 
    219 	switch (VTOFDESC(dvp)->fd_type) {
    220 	default:
    221 	case Flink:
    222 	case Fdesc:
    223 	case Fctty:
    224 		error = ENOTDIR;
    225 		goto bad;
    226 
    227 	case Froot:
    228 		if (cnp->cn_namelen == 2 && memcmp(pname, "fd", 2) == 0) {
    229 			ix = FD_DEVFD;
    230 			goto good;
    231 		}
    232 
    233 		if (cnp->cn_namelen == 3 && memcmp(pname, "tty", 3) == 0) {
    234 			struct vnode *ttyvp = cttyvp(p);
    235 			if (ttyvp == NULL) {
    236 				error = ENXIO;
    237 				goto bad;
    238 			}
    239 			ix = FD_CTTY;
    240 			goto good;
    241 		}
    242 
    243 		switch (cnp->cn_namelen) {
    244 		case 5:
    245 			if (memcmp(pname, "stdin", 5) == 0) {
    246 				ix = FD_STDIN;
    247 				goto good;
    248 			}
    249 			break;
    250 		case 6:
    251 			if (memcmp(pname, "stdout", 6) == 0) {
    252 				ix = FD_STDOUT;
    253 				goto good;
    254 			} else if (memcmp(pname, "stderr", 6) == 0) {
    255 				ix = FD_STDERR;
    256 				goto good;
    257 			}
    258 			break;
    259 		}
    260 
    261 		error = ENOENT;
    262 		goto bad;
    263 
    264 	case Fdevfd:
    265 		if (cnp->cn_namelen == 2 && memcmp(pname, "..", 2) == 0) {
    266 			ix = FD_ROOT;
    267 			goto good;
    268 		}
    269 
    270 		fd = 0;
    271 		while (*pname >= '0' && *pname <= '9') {
    272 			fd = 10 * fd + *pname++ - '0';
    273 			if (fd >= dt->dt_nfiles)
    274 				break;
    275 		}
    276 
    277 		if (*pname != '\0') {
    278 			error = ENOENT;
    279 			goto bad;
    280 		}
    281 
    282 		if (fd >= dt->dt_nfiles || dt->dt_ff[fd] == NULL ||
    283 		    dt->dt_ff[fd]->ff_file == NULL) {
    284 			error = EBADF;
    285 			goto bad;
    286 		}
    287 
    288 		ix = FD_DESC + fd;
    289 		goto good;
    290 	}
    291 
    292 bad:
    293 	*vpp = NULL;
    294 	return error;
    295 
    296 good:
    297 	KASSERT(ix != -1);
    298 	error = vcache_get(dvp->v_mount, &ix, sizeof(ix), vpp);
    299 	if (error)
    300 		return error;
    301 
    302 	/*
    303 	 * Prevent returning VNON nodes.
    304 	 * Operation fdesc_inactive() will reset the type to VNON.
    305 	 */
    306 	if (ix == FD_CTTY)
    307 		(*vpp)->v_type = VCHR;
    308 	else if (ix >= FD_DESC)
    309 		(*vpp)->v_type = VREG;
    310 	KASSERT((*vpp)->v_type != VNON);
    311 
    312 	return 0;
    313 }
    314 
    315 int
    316 fdesc_open(void *v)
    317 {
    318 	struct vop_open_args /* {
    319 		struct vnode *a_vp;
    320 		int  a_mode;
    321 		kauth_cred_t a_cred;
    322 	} */ *ap = v;
    323 	struct vnode *vp = ap->a_vp;
    324 
    325 	switch (VTOFDESC(vp)->fd_type) {
    326 	case Fdesc:
    327 		/*
    328 		 * XXX Kludge: set dupfd to contain the value of the
    329 		 * the file descriptor being sought for duplication.
    330 		 * The error return ensures that the vnode for this
    331 		 * device will be released by vn_open. vn_open will
    332 		 * then detect this special error and take the actions
    333 		 * in fd_dupopen. Other callers of vn_open or VOP_OPEN
    334 		 * not prepared to deal with this situation will
    335 		 * report a real error.
    336 		 */
    337 		curlwp->l_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
    338 		return EDUPFD;
    339 
    340 	case Fctty:
    341 		return cdev_open(devctty, ap->a_mode, 0, curlwp);
    342 	case Froot:
    343 	case Fdevfd:
    344 	case Flink:
    345 		break;
    346 	}
    347 
    348 	return (0);
    349 }
    350 
    351 static int
    352 fdesc_attr(int fd, struct vattr *vap, kauth_cred_t cred)
    353 {
    354 	file_t *fp;
    355 	struct stat stb;
    356 	int error;
    357 
    358 	if ((fp = fd_getfile(fd)) == NULL)
    359 		return (EBADF);
    360 
    361 	switch (fp->f_type) {
    362 	case DTYPE_VNODE:
    363 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
    364 		error = VOP_GETATTR(fp->f_vnode, vap, cred);
    365 		VOP_UNLOCK(fp->f_vnode);
    366 		if (error == 0 && vap->va_type == VDIR) {
    367 			/*
    368 			 * directories can cause loops in the namespace,
    369 			 * so turn off the 'x' bits to avoid trouble.
    370 			 */
    371 			vap->va_mode &= ~(S_IXUSR|S_IXGRP|S_IXOTH);
    372 		}
    373 		break;
    374 
    375 	default:
    376 		memset(&stb, 0, sizeof(stb));
    377 		error = (*fp->f_ops->fo_stat)(fp, &stb);
    378 		if (error)
    379 			break;
    380 
    381 		vattr_null(vap);
    382 		switch(fp->f_type) {
    383 		case DTYPE_SOCKET:
    384 			vap->va_type = VSOCK;
    385 			break;
    386 		case DTYPE_PIPE:
    387 			vap->va_type = VFIFO;
    388 			break;
    389 		default:
    390 			/* use VNON perhaps? */
    391 			vap->va_type = VBAD;
    392 			break;
    393 		}
    394 		vap->va_mode = stb.st_mode;
    395 		vap->va_nlink = stb.st_nlink;
    396 		vap->va_uid = stb.st_uid;
    397 		vap->va_gid = stb.st_gid;
    398 		vap->va_fsid = stb.st_dev;
    399 		vap->va_fileid = stb.st_ino;
    400 		vap->va_size = stb.st_size;
    401 		vap->va_blocksize = stb.st_blksize;
    402 		vap->va_atime = stb.st_atimespec;
    403 		vap->va_mtime = stb.st_mtimespec;
    404 		vap->va_ctime = stb.st_ctimespec;
    405 		vap->va_gen = stb.st_gen;
    406 		vap->va_flags = stb.st_flags;
    407 		vap->va_rdev = stb.st_rdev;
    408 		vap->va_bytes = stb.st_blocks * stb.st_blksize;
    409 		break;
    410 	}
    411 
    412 	fd_putfile(fd);
    413 	return (error);
    414 }
    415 
    416 int
    417 fdesc_getattr(void *v)
    418 {
    419 	struct vop_getattr_args /* {
    420 		struct vnode *a_vp;
    421 		struct vattr *a_vap;
    422 		kauth_cred_t a_cred;
    423 		struct lwp *a_l;
    424 	} */ *ap = v;
    425 	struct vnode *vp = ap->a_vp;
    426 	struct vattr *vap = ap->a_vap;
    427 	unsigned fd;
    428 	int error = 0;
    429 	struct timeval tv;
    430 
    431 	switch (VTOFDESC(vp)->fd_type) {
    432 	case Froot:
    433 	case Fdevfd:
    434 	case Flink:
    435 	case Fctty:
    436 		vattr_null(vap);
    437 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
    438 
    439 #define R_ALL (S_IRUSR|S_IRGRP|S_IROTH)
    440 #define W_ALL (S_IWUSR|S_IWGRP|S_IWOTH)
    441 #define X_ALL (S_IXUSR|S_IXGRP|S_IXOTH)
    442 
    443 		switch (VTOFDESC(vp)->fd_type) {
    444 		case Flink:
    445 			vap->va_mode = R_ALL|X_ALL;
    446 			vap->va_type = VLNK;
    447 			vap->va_rdev = 0;
    448 			vap->va_nlink = 1;
    449 			vap->va_size = strlen(VTOFDESC(vp)->fd_link);
    450 			break;
    451 
    452 		case Fctty:
    453 			vap->va_mode = R_ALL|W_ALL;
    454 			vap->va_type = VCHR;
    455 			vap->va_rdev = devctty;
    456 			vap->va_nlink = 1;
    457 			vap->va_size = 0;
    458 			break;
    459 
    460 		default:
    461 			vap->va_mode = R_ALL|X_ALL;
    462 			vap->va_type = VDIR;
    463 			vap->va_rdev = 0;
    464 			vap->va_nlink = 2;
    465 			vap->va_size = DEV_BSIZE;
    466 			break;
    467 		}
    468 		vap->va_uid = 0;
    469 		vap->va_gid = 0;
    470 		vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0];
    471 		vap->va_blocksize = DEV_BSIZE;
    472 		getmicroboottime(&tv);
    473 		vap->va_atime.tv_sec = tv.tv_sec;
    474 		vap->va_atime.tv_nsec = 0;
    475 		vap->va_mtime = vap->va_atime;
    476 		vap->va_ctime = vap->va_mtime;
    477 		vap->va_gen = 0;
    478 		vap->va_flags = 0;
    479 		vap->va_bytes = 0;
    480 		break;
    481 
    482 	case Fdesc:
    483 		fd = VTOFDESC(vp)->fd_fd;
    484 		error = fdesc_attr(fd, vap, ap->a_cred);
    485 		break;
    486 
    487 	default:
    488 		panic("fdesc_getattr");
    489 		break;
    490 	}
    491 
    492 	if (error == 0)
    493 		vp->v_type = vap->va_type;
    494 
    495 	return (error);
    496 }
    497 
    498 int
    499 fdesc_setattr(void *v)
    500 {
    501 	struct vop_setattr_args /* {
    502 		struct vnode *a_vp;
    503 		struct vattr *a_vap;
    504 		kauth_cred_t a_cred;
    505 	} */ *ap = v;
    506 	file_t *fp;
    507 	unsigned fd;
    508 
    509 	/*
    510 	 * Can't mess with the root vnode
    511 	 */
    512 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    513 	case Fdesc:
    514 		break;
    515 
    516 	case Fctty:
    517 		return (0);
    518 
    519 	default:
    520 		return (EACCES);
    521 	}
    522 
    523 	fd = VTOFDESC(ap->a_vp)->fd_fd;
    524 	if ((fp = fd_getfile(fd)) == NULL)
    525 		return (EBADF);
    526 
    527 	/*
    528 	 * XXX: Can't reasonably set the attr's on any types currently.
    529 	 *      On vnode's this will cause truncation and socket/pipes make
    530 	 *      no sense.
    531 	 */
    532 	fd_putfile(fd);
    533 	return (0);
    534 }
    535 
    536 
    537 struct fdesc_target {
    538 	ino_t ft_fileno;
    539 	u_char ft_type;
    540 	u_char ft_namlen;
    541 	const char *ft_name;
    542 } fdesc_targets[] = {
    543 #define N(s) sizeof(s)-1, s
    544 	{ FD_DEVFD,  DT_DIR,     N("fd")     },
    545 	{ FD_STDIN,  DT_LNK,     N("stdin")  },
    546 	{ FD_STDOUT, DT_LNK,     N("stdout") },
    547 	{ FD_STDERR, DT_LNK,     N("stderr") },
    548 	{ FD_CTTY,   DT_UNKNOWN, N("tty")    },
    549 #undef N
    550 #define UIO_MX _DIRENT_RECLEN((struct dirent *)NULL, sizeof("stderr") - 1)
    551 };
    552 static int nfdesc_targets = sizeof(fdesc_targets) / sizeof(fdesc_targets[0]);
    553 
    554 int
    555 fdesc_readdir(void *v)
    556 {
    557 	struct vop_readdir_args /* {
    558 		struct vnode *a_vp;
    559 		struct uio *a_uio;
    560 		kauth_cred_t a_cred;
    561 		int *a_eofflag;
    562 		off_t **a_cookies;
    563 		int *a_ncookies;
    564 	} */ *ap = v;
    565 	struct uio *uio = ap->a_uio;
    566 	struct dirent d;
    567 	off_t i;
    568 	int j;
    569 	int error;
    570 	off_t *cookies = NULL;
    571 	int ncookies;
    572 	fdtab_t *dt;
    573 
    574 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    575 	case Fctty:
    576 		return 0;
    577 
    578 	case Fdesc:
    579 		return ENOTDIR;
    580 
    581 	default:
    582 		break;
    583 	}
    584 
    585 	dt = atomic_load_consume(&curlwp->l_fd->fd_dt);
    586 
    587 	if (uio->uio_resid < UIO_MX)
    588 		return EINVAL;
    589 	if (uio->uio_offset < 0)
    590 		return EINVAL;
    591 
    592 	error = 0;
    593 	i = uio->uio_offset;
    594 	(void)memset(&d, 0, UIO_MX);
    595 	d.d_reclen = UIO_MX;
    596 	if (ap->a_ncookies)
    597 		ncookies = uio->uio_resid / UIO_MX;
    598 	else
    599 		ncookies = 0;
    600 
    601 	if (VTOFDESC(ap->a_vp)->fd_type == Froot) {
    602 		struct fdesc_target *ft;
    603 
    604 		if (i >= nfdesc_targets)
    605 			return 0;
    606 
    607 		if (ap->a_ncookies) {
    608 			ncookies = uimin(ncookies, (nfdesc_targets - i));
    609 			cookies = malloc(ncookies * sizeof(off_t),
    610 			    M_TEMP, M_WAITOK);
    611 			*ap->a_cookies = cookies;
    612 			*ap->a_ncookies = ncookies;
    613 		}
    614 
    615 		for (ft = &fdesc_targets[i]; uio->uio_resid >= UIO_MX &&
    616 		    i < nfdesc_targets; ft++, i++) {
    617 			switch (ft->ft_fileno) {
    618 			case FD_CTTY:
    619 				if (cttyvp(curproc) == NULL)
    620 					continue;
    621 				break;
    622 
    623 			case FD_STDIN:
    624 			case FD_STDOUT:
    625 			case FD_STDERR:
    626 				if ((ft->ft_fileno - FD_STDIN) >=
    627 				    dt->dt_nfiles)
    628 					continue;
    629 				if (dt->dt_ff[ft->ft_fileno - FD_STDIN]
    630 				    == NULL || dt->dt_ff[ft->ft_fileno -
    631 				    FD_STDIN]->ff_file == NULL)
    632 					continue;
    633 				break;
    634 			}
    635 
    636 			d.d_fileno = ft->ft_fileno;
    637 			d.d_namlen = ft->ft_namlen;
    638 			(void)memcpy(d.d_name, ft->ft_name, ft->ft_namlen + 1);
    639 			d.d_type = ft->ft_type;
    640 
    641 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
    642 				break;
    643 			if (cookies)
    644 				*cookies++ = i + 1;
    645 		}
    646 	} else {
    647 		if (ap->a_ncookies) {
    648 			ncookies = uimin(ncookies, dt->dt_nfiles + 2);
    649 			cookies = malloc(ncookies * sizeof(off_t),
    650 			    M_TEMP, M_WAITOK);
    651 			*ap->a_cookies = cookies;
    652 			*ap->a_ncookies = ncookies;
    653 		}
    654 		for (; i - 2 < dt->dt_nfiles && uio->uio_resid >= UIO_MX; i++) {
    655 			switch (i) {
    656 			case 0:
    657 			case 1:
    658 				d.d_fileno = FD_ROOT;		/* XXX */
    659 				d.d_namlen = i + 1;
    660 				(void)memcpy(d.d_name, "..", d.d_namlen);
    661 				d.d_name[i + 1] = '\0';
    662 				d.d_type = DT_DIR;
    663 				break;
    664 
    665 			default:
    666 				j = (int)i - 2;
    667 				if (dt->dt_ff[j] == NULL ||
    668 				    dt->dt_ff[j]->ff_file == NULL)
    669 					continue;
    670 				d.d_fileno = j + FD_STDIN;
    671 				d.d_namlen = snprintf(d.d_name,
    672 				    sizeof(d.d_name), "%d", j);
    673 				d.d_type = DT_UNKNOWN;
    674 				break;
    675 			}
    676 
    677 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
    678 				break;
    679 			if (cookies)
    680 				*cookies++ = i + 1;
    681 		}
    682 	}
    683 
    684 	if (ap->a_ncookies && error) {
    685 		free(*ap->a_cookies, M_TEMP);
    686 		*ap->a_ncookies = 0;
    687 		*ap->a_cookies = NULL;
    688 	}
    689 
    690 	uio->uio_offset = i;
    691 	return error;
    692 }
    693 
    694 int
    695 fdesc_readlink(void *v)
    696 {
    697 	struct vop_readlink_args /* {
    698 		struct vnode *a_vp;
    699 		struct uio *a_uio;
    700 		kauth_cred_t a_cred;
    701 	} */ *ap = v;
    702 	struct vnode *vp = ap->a_vp;
    703 	int error;
    704 
    705 	if (vp->v_type != VLNK)
    706 		return (EPERM);
    707 
    708 	if (VTOFDESC(vp)->fd_type == Flink) {
    709 		const char *ln = VTOFDESC(vp)->fd_link;
    710 		error = uiomove(__UNCONST(ln), strlen(ln), ap->a_uio);
    711 	} else {
    712 		error = EOPNOTSUPP;
    713 	}
    714 
    715 	return (error);
    716 }
    717 
    718 int
    719 fdesc_read(void *v)
    720 {
    721 	struct vop_read_args /* {
    722 		struct vnode *a_vp;
    723 		struct uio *a_uio;
    724 		int  a_ioflag;
    725 		kauth_cred_t a_cred;
    726 	} */ *ap = v;
    727 	int error = EOPNOTSUPP;
    728 	struct vnode *vp = ap->a_vp;
    729 
    730 	switch (VTOFDESC(vp)->fd_type) {
    731 	case Fctty:
    732 		VOP_UNLOCK(vp);
    733 		error = cdev_read(devctty, ap->a_uio, ap->a_ioflag);
    734 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    735 		break;
    736 
    737 	default:
    738 		error = EOPNOTSUPP;
    739 		break;
    740 	}
    741 
    742 	return (error);
    743 }
    744 
    745 int
    746 fdesc_write(void *v)
    747 {
    748 	struct vop_write_args /* {
    749 		struct vnode *a_vp;
    750 		struct uio *a_uio;
    751 		int  a_ioflag;
    752 		kauth_cred_t a_cred;
    753 	} */ *ap = v;
    754 	int error = EOPNOTSUPP;
    755 	struct vnode *vp = ap->a_vp;
    756 
    757 	switch (VTOFDESC(vp)->fd_type) {
    758 	case Fctty:
    759 		VOP_UNLOCK(vp);
    760 		error = cdev_write(devctty, ap->a_uio, ap->a_ioflag);
    761 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    762 		break;
    763 
    764 	default:
    765 		error = EOPNOTSUPP;
    766 		break;
    767 	}
    768 
    769 	return (error);
    770 }
    771 
    772 int
    773 fdesc_ioctl(void *v)
    774 {
    775 	struct vop_ioctl_args /* {
    776 		struct vnode *a_vp;
    777 		u_long a_command;
    778 		void *a_data;
    779 		int  a_fflag;
    780 		kauth_cred_t a_cred;
    781 	} */ *ap = v;
    782 	int error = EOPNOTSUPP;
    783 
    784 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    785 	case Fctty:
    786 		error = cdev_ioctl(devctty, ap->a_command, ap->a_data,
    787 		    ap->a_fflag, curlwp);
    788 		break;
    789 
    790 	default:
    791 		error = EOPNOTSUPP;
    792 		break;
    793 	}
    794 
    795 	return (error);
    796 }
    797 
    798 int
    799 fdesc_poll(void *v)
    800 {
    801 	struct vop_poll_args /* {
    802 		struct vnode *a_vp;
    803 		int a_events;
    804 	} */ *ap = v;
    805 	int revents;
    806 
    807 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    808 	case Fctty:
    809 		revents = cdev_poll(devctty, ap->a_events, curlwp);
    810 		break;
    811 
    812 	default:
    813 		revents = genfs_poll(v);
    814 		break;
    815 	}
    816 
    817 	return (revents);
    818 }
    819 
    820 int
    821 fdesc_kqfilter(void *v)
    822 {
    823 	struct vop_kqfilter_args /* {
    824 		struct vnode *a_vp;
    825 		struct knote *a_kn;
    826 	} */ *ap = v;
    827 	int error, fd;
    828 	file_t *fp;
    829 
    830 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    831 	case Fctty:
    832 		error = cdev_kqfilter(devctty, ap->a_kn);
    833 		break;
    834 
    835 	case Fdesc:
    836 		/* just invoke kqfilter for the underlying descriptor */
    837 		fd = VTOFDESC(ap->a_vp)->fd_fd;
    838 		if ((fp = fd_getfile(fd)) == NULL)
    839 			return (1);
    840 		error = (*fp->f_ops->fo_kqfilter)(fp, ap->a_kn);
    841 		fd_putfile(fd);
    842 		break;
    843 
    844 	default:
    845 		return (genfs_kqfilter(v));
    846 	}
    847 
    848 	return (error);
    849 }
    850 
    851 int
    852 fdesc_inactive(void *v)
    853 {
    854 	struct vop_inactive_v2_args /* {
    855 		struct vnode *a_vp;
    856 	} */ *ap = v;
    857 	struct vnode *vp = ap->a_vp;
    858 	struct fdescnode *fd = VTOFDESC(vp);
    859 
    860 	/*
    861 	 * Clear out the v_type field to avoid
    862 	 * nasty things happening on reclaim.
    863 	 */
    864 	if (fd->fd_type == Fctty || fd->fd_type == Fdesc)
    865 		vp->v_type = VNON;
    866 
    867 	return (0);
    868 }
    869 
    870 int
    871 fdesc_reclaim(void *v)
    872 {
    873 	struct vop_reclaim_v2_args /* {
    874 		struct vnode *a_vp;
    875 	} */ *ap = v;
    876 	struct vnode *vp = ap->a_vp;
    877 	struct fdescnode *fd = VTOFDESC(vp);
    878 
    879 	VOP_UNLOCK(vp);
    880 
    881 	vp->v_data = NULL;
    882 	kmem_free(fd, sizeof(struct fdescnode));
    883 
    884 	return (0);
    885 }
    886 
    887 /*
    888  * Return POSIX pathconf information applicable to special devices.
    889  */
    890 int
    891 fdesc_pathconf(void *v)
    892 {
    893 	struct vop_pathconf_args /* {
    894 		struct vnode *a_vp;
    895 		int a_name;
    896 		register_t *a_retval;
    897 	} */ *ap = v;
    898 
    899 	switch (ap->a_name) {
    900 	case _PC_LINK_MAX:
    901 		*ap->a_retval = LINK_MAX;
    902 		return (0);
    903 	case _PC_MAX_CANON:
    904 		*ap->a_retval = MAX_CANON;
    905 		return (0);
    906 	case _PC_MAX_INPUT:
    907 		*ap->a_retval = MAX_INPUT;
    908 		return (0);
    909 	case _PC_PIPE_BUF:
    910 		*ap->a_retval = PIPE_BUF;
    911 		return (0);
    912 	case _PC_CHOWN_RESTRICTED:
    913 		*ap->a_retval = 1;
    914 		return (0);
    915 	case _PC_VDISABLE:
    916 		*ap->a_retval = _POSIX_VDISABLE;
    917 		return (0);
    918 	case _PC_SYNC_IO:
    919 		*ap->a_retval = 1;
    920 		return (0);
    921 	default:
    922 		return genfs_pathconf(ap);
    923 	}
    924 	/* NOTREACHED */
    925 }
    926 
    927 /*
    928  * Print out the contents of a /dev/fd vnode.
    929  */
    930 /* ARGSUSED */
    931 int
    932 fdesc_print(void *v)
    933 {
    934 	printf("tag VT_NON, fdesc vnode\n");
    935 	return (0);
    936 }
    937 
    938 int
    939 fdesc_link(void *v)
    940 {
    941 	struct vop_link_v2_args /* {
    942 		struct vnode *a_dvp;
    943 		struct vnode *a_vp;
    944 		struct componentname *a_cnp;
    945 	} */ *ap = v;
    946 
    947 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
    948 	return (EROFS);
    949 }
    950 
    951 int
    952 fdesc_symlink(void *v)
    953 {
    954 	struct vop_symlink_v3_args /* {
    955 		struct vnode *a_dvp;
    956 		struct vnode **a_vpp;
    957 		struct componentname *a_cnp;
    958 		struct vattr *a_vap;
    959 		char *a_target;
    960 	} */ *ap = v;
    961 
    962 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
    963 	return (EROFS);
    964 }
    965