Home | History | Annotate | Line # | Download | only in fdesc
fdesc_vnops.c revision 1.77.2.3
      1 /*	$NetBSD: fdesc_vnops.c,v 1.77.2.3 2004/08/24 17:57:39 skrll 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 1.77.2.3 2004/08/24 17:57:39 skrll Exp $");
     45 
     46 #include <sys/param.h>
     47 #include <sys/systm.h>
     48 #include <sys/time.h>
     49 #include <sys/proc.h>
     50 #include <sys/kernel.h>	/* boottime */
     51 #include <sys/resourcevar.h>
     52 #include <sys/socketvar.h>
     53 #include <sys/filedesc.h>
     54 #include <sys/vnode.h>
     55 #include <sys/malloc.h>
     56 #include <sys/conf.h>
     57 #include <sys/file.h>
     58 #include <sys/stat.h>
     59 #include <sys/mount.h>
     60 #include <sys/namei.h>
     61 #include <sys/buf.h>
     62 #include <sys/dirent.h>
     63 #include <sys/tty.h>
     64 
     65 #include <miscfs/fdesc/fdesc.h>
     66 #include <miscfs/genfs/genfs.h>
     67 
     68 #define cttyvp(p) ((p)->p_flag & P_CONTROLT ? (p)->p_session->s_ttyvp : NULL)
     69 
     70 #define FDL_WANT	0x01
     71 #define FDL_LOCKED	0x02
     72 static int fdcache_lock;
     73 
     74 dev_t devctty;
     75 
     76 #if (FD_STDIN != FD_STDOUT-1) || (FD_STDOUT != FD_STDERR-1)
     77 FD_STDIN, FD_STDOUT, FD_STDERR must be a sequence n, n+1, n+2
     78 #endif
     79 
     80 #define	NFDCACHE 4
     81 
     82 #define FD_NHASH(ix) \
     83 	(&fdhashtbl[(ix) & fdhash])
     84 LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
     85 u_long fdhash;
     86 
     87 int	fdesc_lookup	__P((void *));
     88 #define	fdesc_create	genfs_eopnotsupp
     89 #define	fdesc_mknod	genfs_eopnotsupp
     90 int	fdesc_open	__P((void *));
     91 #define	fdesc_close	genfs_nullop
     92 #define	fdesc_access	genfs_nullop
     93 int	fdesc_getattr	__P((void *));
     94 int	fdesc_setattr	__P((void *));
     95 int	fdesc_read	__P((void *));
     96 int	fdesc_write	__P((void *));
     97 int	fdesc_ioctl	__P((void *));
     98 int	fdesc_poll	__P((void *));
     99 int	fdesc_kqfilter	__P((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	__P((void *));
    106 #define	fdesc_rename	genfs_eopnotsupp
    107 #define	fdesc_mkdir	genfs_eopnotsupp
    108 #define	fdesc_rmdir	genfs_eopnotsupp
    109 int	fdesc_symlink	__P((void *));
    110 int	fdesc_readdir	__P((void *));
    111 int	fdesc_readlink	__P((void *));
    112 #define	fdesc_abortop	genfs_abortop
    113 int	fdesc_inactive	__P((void *));
    114 int	fdesc_reclaim	__P((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	__P((void *));
    120 int	fdesc_pathconf	__P((void *));
    121 #define	fdesc_islocked	genfs_islocked
    122 #define	fdesc_advlock	genfs_einval
    123 #define	fdesc_blkatoff	genfs_eopnotsupp
    124 #define	fdesc_valloc	genfs_eopnotsupp
    125 #define	fdesc_vfree	genfs_nullop
    126 #define	fdesc_truncate	genfs_eopnotsupp
    127 #define	fdesc_update	genfs_nullop
    128 #define	fdesc_bwrite	genfs_eopnotsupp
    129 #define fdesc_revoke	genfs_revoke
    130 #define fdesc_putpages	genfs_null_putpages
    131 
    132 static int fdesc_attr __P((int, struct vattr *, struct ucred *, struct lwp *));
    133 
    134 int (**fdesc_vnodeop_p) __P((void *));
    135 const struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = {
    136 	{ &vop_default_desc, vn_default_error },
    137 	{ &vop_lookup_desc, fdesc_lookup },		/* lookup */
    138 	{ &vop_create_desc, fdesc_create },		/* create */
    139 	{ &vop_mknod_desc, fdesc_mknod },		/* mknod */
    140 	{ &vop_open_desc, fdesc_open },			/* open */
    141 	{ &vop_close_desc, fdesc_close },		/* close */
    142 	{ &vop_access_desc, fdesc_access },		/* access */
    143 	{ &vop_getattr_desc, fdesc_getattr },		/* getattr */
    144 	{ &vop_setattr_desc, fdesc_setattr },		/* setattr */
    145 	{ &vop_read_desc, fdesc_read },			/* read */
    146 	{ &vop_write_desc, fdesc_write },		/* write */
    147 	{ &vop_ioctl_desc, fdesc_ioctl },		/* ioctl */
    148 	{ &vop_fcntl_desc, fdesc_fcntl },		/* fcntl */
    149 	{ &vop_poll_desc, fdesc_poll },			/* poll */
    150 	{ &vop_kqfilter_desc, fdesc_kqfilter },		/* kqfilter */
    151 	{ &vop_revoke_desc, fdesc_revoke },		/* revoke */
    152 	{ &vop_mmap_desc, fdesc_mmap },			/* mmap */
    153 	{ &vop_fsync_desc, fdesc_fsync },		/* fsync */
    154 	{ &vop_seek_desc, fdesc_seek },			/* seek */
    155 	{ &vop_remove_desc, fdesc_remove },		/* remove */
    156 	{ &vop_link_desc, fdesc_link },			/* link */
    157 	{ &vop_rename_desc, fdesc_rename },		/* rename */
    158 	{ &vop_mkdir_desc, fdesc_mkdir },		/* mkdir */
    159 	{ &vop_rmdir_desc, fdesc_rmdir },		/* rmdir */
    160 	{ &vop_symlink_desc, fdesc_symlink },		/* symlink */
    161 	{ &vop_readdir_desc, fdesc_readdir },		/* readdir */
    162 	{ &vop_readlink_desc, fdesc_readlink },		/* readlink */
    163 	{ &vop_abortop_desc, fdesc_abortop },		/* abortop */
    164 	{ &vop_inactive_desc, fdesc_inactive },		/* inactive */
    165 	{ &vop_reclaim_desc, fdesc_reclaim },		/* reclaim */
    166 	{ &vop_lock_desc, fdesc_lock },			/* lock */
    167 	{ &vop_unlock_desc, fdesc_unlock },		/* unlock */
    168 	{ &vop_bmap_desc, fdesc_bmap },			/* bmap */
    169 	{ &vop_strategy_desc, fdesc_strategy },		/* strategy */
    170 	{ &vop_print_desc, fdesc_print },		/* print */
    171 	{ &vop_islocked_desc, fdesc_islocked },		/* islocked */
    172 	{ &vop_pathconf_desc, fdesc_pathconf },		/* pathconf */
    173 	{ &vop_advlock_desc, fdesc_advlock },		/* advlock */
    174 	{ &vop_blkatoff_desc, fdesc_blkatoff },		/* blkatoff */
    175 	{ &vop_valloc_desc, fdesc_valloc },		/* valloc */
    176 	{ &vop_vfree_desc, fdesc_vfree },		/* vfree */
    177 	{ &vop_truncate_desc, fdesc_truncate },		/* truncate */
    178 	{ &vop_update_desc, fdesc_update },		/* update */
    179 	{ &vop_bwrite_desc, fdesc_bwrite },		/* bwrite */
    180 	{ &vop_putpages_desc, fdesc_putpages },		/* putpages */
    181 	{ NULL, NULL }
    182 };
    183 
    184 const struct vnodeopv_desc fdesc_vnodeop_opv_desc =
    185 	{ &fdesc_vnodeop_p, fdesc_vnodeop_entries };
    186 
    187 extern const struct cdevsw ctty_cdevsw;
    188 
    189 /*
    190  * Initialise cache headers
    191  */
    192 void
    193 fdesc_init()
    194 {
    195 	int cttymajor;
    196 
    197 	/* locate the major number */
    198 	cttymajor = cdevsw_lookup_major(&ctty_cdevsw);
    199 	devctty = makedev(cttymajor, 0);
    200 	fdhashtbl = hashinit(NFDCACHE, HASH_LIST, M_CACHE, M_NOWAIT, &fdhash);
    201 }
    202 
    203 /*
    204  * Free hash table.
    205  */
    206 void
    207 fdesc_done()
    208 {
    209 	hashdone(fdhashtbl, M_CACHE);
    210 }
    211 
    212 /*
    213  * Return a locked vnode of the correct type.
    214  */
    215 int
    216 fdesc_allocvp(ftype, ix, mp, vpp)
    217 	fdntype ftype;
    218 	int ix;
    219 	struct mount *mp;
    220 	struct vnode **vpp;
    221 {
    222 	struct fdhashhead *fc;
    223 	struct fdescnode *fd;
    224 	int error = 0;
    225 
    226 	fc = FD_NHASH(ix);
    227 loop:
    228 	for (fd = fc->lh_first; fd != 0; fd = fd->fd_hash.le_next) {
    229 		if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
    230 			if (vget(fd->fd_vnode, LK_EXCLUSIVE))
    231 				goto loop;
    232 			*vpp = fd->fd_vnode;
    233 			return (error);
    234 		}
    235 	}
    236 
    237 	/*
    238 	 * otherwise lock the array while we call getnewvnode
    239 	 * since that can block.
    240 	 */
    241 	if (fdcache_lock & FDL_LOCKED) {
    242 		fdcache_lock |= FDL_WANT;
    243 		(void) tsleep(&fdcache_lock, PINOD, "fdcache", 0);
    244 		goto loop;
    245 	}
    246 	fdcache_lock |= FDL_LOCKED;
    247 
    248 	error = getnewvnode(VT_FDESC, mp, fdesc_vnodeop_p, vpp);
    249 	if (error)
    250 		goto out;
    251 	MALLOC(fd, void *, sizeof(struct fdescnode), M_TEMP, M_WAITOK);
    252 	(*vpp)->v_data = fd;
    253 	fd->fd_vnode = *vpp;
    254 	fd->fd_type = ftype;
    255 	fd->fd_fd = -1;
    256 	fd->fd_link = 0;
    257 	fd->fd_ix = ix;
    258 	VOP_LOCK(*vpp, LK_EXCLUSIVE);
    259 	LIST_INSERT_HEAD(fc, fd, fd_hash);
    260 
    261 out:;
    262 	fdcache_lock &= ~FDL_LOCKED;
    263 
    264 	if (fdcache_lock & FDL_WANT) {
    265 		fdcache_lock &= ~FDL_WANT;
    266 		wakeup(&fdcache_lock);
    267 	}
    268 
    269 	return (error);
    270 }
    271 
    272 /*
    273  * vp is the current namei directory
    274  * ndp is the name to locate in that directory...
    275  */
    276 int
    277 fdesc_lookup(v)
    278 	void *v;
    279 {
    280 	struct vop_lookup_args /* {
    281 		struct vnode * a_dvp;
    282 		struct vnode ** a_vpp;
    283 		struct componentname * a_cnp;
    284 	} */ *ap = v;
    285 	struct vnode **vpp = ap->a_vpp;
    286 	struct vnode *dvp = ap->a_dvp;
    287 	struct componentname *cnp = ap->a_cnp;
    288 	struct lwp *l = cnp->cn_lwp;
    289 	const char *pname = cnp->cn_nameptr;
    290 	struct proc *p;
    291 	int nfiles;
    292 	unsigned fd = 0;
    293 	int error;
    294 	struct vnode *fvp;
    295 	char *ln;
    296 
    297 	if (cnp->cn_namelen == 1 && *pname == '.') {
    298 		*vpp = dvp;
    299 		VREF(dvp);
    300 		return (0);
    301 	}
    302 
    303 	p = l->l_proc;
    304 	nfiles = p->p_fd->fd_nfiles;
    305 
    306 	switch (VTOFDESC(dvp)->fd_type) {
    307 	default:
    308 	case Flink:
    309 	case Fdesc:
    310 	case Fctty:
    311 		error = ENOTDIR;
    312 		goto bad;
    313 
    314 	case Froot:
    315 		if (cnp->cn_namelen == 2 && memcmp(pname, "fd", 2) == 0) {
    316 			error = fdesc_allocvp(Fdevfd, FD_DEVFD, dvp->v_mount, &fvp);
    317 			if (error)
    318 				goto bad;
    319 			*vpp = fvp;
    320 			fvp->v_type = VDIR;
    321 			goto good;
    322 		}
    323 
    324 		if (cnp->cn_namelen == 3 && memcmp(pname, "tty", 3) == 0) {
    325 			struct vnode *ttyvp = cttyvp(p);
    326 			if (ttyvp == NULL) {
    327 				error = ENXIO;
    328 				goto bad;
    329 			}
    330 			error = fdesc_allocvp(Fctty, FD_CTTY, dvp->v_mount, &fvp);
    331 			if (error)
    332 				goto bad;
    333 			*vpp = fvp;
    334 			fvp->v_type = VCHR;
    335 			goto good;
    336 		}
    337 
    338 		ln = 0;
    339 		switch (cnp->cn_namelen) {
    340 		case 5:
    341 			if (memcmp(pname, "stdin", 5) == 0) {
    342 				ln = "fd/0";
    343 				fd = FD_STDIN;
    344 			}
    345 			break;
    346 		case 6:
    347 			if (memcmp(pname, "stdout", 6) == 0) {
    348 				ln = "fd/1";
    349 				fd = FD_STDOUT;
    350 			} else
    351 			if (memcmp(pname, "stderr", 6) == 0) {
    352 				ln = "fd/2";
    353 				fd = FD_STDERR;
    354 			}
    355 			break;
    356 		}
    357 
    358 		if (ln) {
    359 			error = fdesc_allocvp(Flink, fd, dvp->v_mount, &fvp);
    360 			if (error)
    361 				goto bad;
    362 			VTOFDESC(fvp)->fd_link = ln;
    363 			*vpp = fvp;
    364 			fvp->v_type = VLNK;
    365 			goto good;
    366 		} else {
    367 			error = ENOENT;
    368 			goto bad;
    369 		}
    370 
    371 		/* FALL THROUGH */
    372 
    373 	case Fdevfd:
    374 		if (cnp->cn_namelen == 2 && memcmp(pname, "..", 2) == 0) {
    375 			VOP_UNLOCK(dvp, 0);
    376 			cnp->cn_flags |= PDIRUNLOCK;
    377 			error = fdesc_root(dvp->v_mount, vpp);
    378 			if (error)
    379 				goto bad;
    380 			/*
    381 			 * If we're at the last component and need the
    382 			 * parent locked, undo the unlock above.
    383 			 */
    384 			if (((~cnp->cn_flags & (ISLASTCN | LOCKPARENT)) == 0) &&
    385 				   ((error = vn_lock(dvp, LK_EXCLUSIVE)) == 0))
    386 				cnp->cn_flags &= ~PDIRUNLOCK;
    387 			return (error);
    388 		}
    389 
    390 		fd = 0;
    391 		while (*pname >= '0' && *pname <= '9') {
    392 			fd = 10 * fd + *pname++ - '0';
    393 			if (fd >= nfiles)
    394 				break;
    395 		}
    396 
    397 		if (*pname != '\0') {
    398 			error = ENOENT;
    399 			goto bad;
    400 		}
    401 
    402 		if (fd >= nfiles || p->p_fd->fd_ofiles[fd] == NULL ||
    403 		    FILE_IS_USABLE(p->p_fd->fd_ofiles[fd]) == 0) {
    404 			error = EBADF;
    405 			goto bad;
    406 		}
    407 
    408 		error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp);
    409 		if (error)
    410 			goto bad;
    411 		VTOFDESC(fvp)->fd_fd = fd;
    412 		*vpp = fvp;
    413 		goto good;
    414 	}
    415 
    416 bad:;
    417 	*vpp = NULL;
    418 	return (error);
    419 
    420 good:;
    421 	/*
    422 	 * As "." was special cased above, we now unlock the parent if we're
    423 	 * suppoed to. We're only supposed to not unlock if this is the
    424 	 * last component, and the caller requested LOCKPARENT. So if either
    425 	 * condition is false, unlock.
    426 	 */
    427 	if (((~cnp->cn_flags) & (ISLASTCN | LOCKPARENT)) != 0) {
    428 		VOP_UNLOCK(dvp, 0);
    429 		cnp->cn_flags |= PDIRUNLOCK;
    430 	}
    431 	return (0);
    432 }
    433 
    434 int
    435 fdesc_open(v)
    436 	void *v;
    437 {
    438 	struct vop_open_args /* {
    439 		struct vnode *a_vp;
    440 		int  a_mode;
    441 		struct ucred *a_cred;
    442 		struct lwp *a_l;
    443 	} */ *ap = v;
    444 	struct vnode *vp = ap->a_vp;
    445 
    446 	switch (VTOFDESC(vp)->fd_type) {
    447 	case Fdesc:
    448 		/*
    449 		 * XXX Kludge: set dupfd to contain the value of the
    450 		 * the file descriptor being sought for duplication. The error
    451 		 * return ensures that the vnode for this device will be
    452 		 * released by vn_open. Open will detect this special error and
    453 		 * take the actions in dupfdopen.  Other callers of vn_open or
    454 		 * VOP_OPEN will simply report the error.
    455 		 */
    456 		curlwp->l_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
    457 		return (ENODEV);
    458 
    459 	case Fctty:
    460 		return ((*ctty_cdevsw.d_open)(devctty, ap->a_mode, 0, ap->a_l));
    461 	case Froot:
    462 	case Fdevfd:
    463 	case Flink:
    464 		break;
    465 	}
    466 
    467 	return (0);
    468 }
    469 
    470 static int
    471 fdesc_attr(fd, vap, cred, l)
    472 	int fd;
    473 	struct vattr *vap;
    474 	struct ucred *cred;
    475 	struct lwp *l;
    476 {
    477 	struct proc *p = l->l_proc;
    478 	struct filedesc *fdp = p->p_fd;
    479 	struct file *fp;
    480 	struct stat stb;
    481 	int error;
    482 
    483 	if ((fp = fd_getfile(fdp, fd)) == NULL)
    484 		return (EBADF);
    485 
    486 	switch (fp->f_type) {
    487 	case DTYPE_VNODE:
    488 		simple_unlock(&fp->f_slock);
    489 		error = VOP_GETATTR((struct vnode *) fp->f_data, vap, cred, l);
    490 		if (error == 0 && vap->va_type == VDIR) {
    491 			/*
    492 			 * directories can cause loops in the namespace,
    493 			 * so turn off the 'x' bits to avoid trouble.
    494 			 */
    495 			vap->va_mode &= ~(S_IXUSR|S_IXGRP|S_IXOTH);
    496 		}
    497 		break;
    498 
    499 	default:
    500 		FILE_USE(fp);
    501 		memset(&stb, 0, sizeof(stb));
    502 		error = (*fp->f_ops->fo_stat)(fp, &stb, l);
    503 		FILE_UNUSE(fp, l);
    504 		if (error)
    505 			break;
    506 
    507 		vattr_null(vap);
    508 		switch(fp->f_type) {
    509 		case DTYPE_SOCKET:
    510 			vap->va_type = VSOCK;
    511 			break;
    512 		case DTYPE_PIPE:
    513 			vap->va_type = VFIFO;
    514 			break;
    515 		default:
    516 			/* use VNON perhaps? */
    517 			vap->va_type = VBAD;
    518 			break;
    519 		}
    520 		vap->va_mode = stb.st_mode;
    521 		vap->va_nlink = stb.st_nlink;
    522 		vap->va_uid = stb.st_uid;
    523 		vap->va_gid = stb.st_gid;
    524 		vap->va_fsid = stb.st_dev;
    525 		vap->va_fileid = stb.st_ino;
    526 		vap->va_size = stb.st_size;
    527 		vap->va_blocksize = stb.st_blksize;
    528 		vap->va_atime = stb.st_atimespec;
    529 		vap->va_mtime = stb.st_mtimespec;
    530 		vap->va_ctime = stb.st_ctimespec;
    531 		vap->va_gen = stb.st_gen;
    532 		vap->va_flags = stb.st_flags;
    533 		vap->va_rdev = stb.st_rdev;
    534 		vap->va_bytes = stb.st_blocks * stb.st_blksize;
    535 		break;
    536 	}
    537 
    538 	return (error);
    539 }
    540 
    541 int
    542 fdesc_getattr(v)
    543 	void *v;
    544 {
    545 	struct vop_getattr_args /* {
    546 		struct vnode *a_vp;
    547 		struct vattr *a_vap;
    548 		struct ucred *a_cred;
    549 		struct lwp *a_l;
    550 	} */ *ap = v;
    551 	struct vnode *vp = ap->a_vp;
    552 	struct vattr *vap = ap->a_vap;
    553 	unsigned fd;
    554 	int error = 0;
    555 
    556 	switch (VTOFDESC(vp)->fd_type) {
    557 	case Froot:
    558 	case Fdevfd:
    559 	case Flink:
    560 	case Fctty:
    561 		VATTR_NULL(vap);
    562 		vap->va_fileid = VTOFDESC(vp)->fd_ix;
    563 
    564 #define R_ALL (S_IRUSR|S_IRGRP|S_IROTH)
    565 #define W_ALL (S_IWUSR|S_IWGRP|S_IWOTH)
    566 #define X_ALL (S_IXUSR|S_IXGRP|S_IXOTH)
    567 
    568 		switch (VTOFDESC(vp)->fd_type) {
    569 		case Flink:
    570 			vap->va_mode = R_ALL|X_ALL;
    571 			vap->va_type = VLNK;
    572 			vap->va_rdev = 0;
    573 			vap->va_nlink = 1;
    574 			vap->va_size = strlen(VTOFDESC(vp)->fd_link);
    575 			break;
    576 
    577 		case Fctty:
    578 			vap->va_mode = R_ALL|W_ALL;
    579 			vap->va_type = VCHR;
    580 			vap->va_rdev = devctty;
    581 			vap->va_nlink = 1;
    582 			vap->va_size = 0;
    583 			break;
    584 
    585 		default:
    586 			vap->va_mode = R_ALL|X_ALL;
    587 			vap->va_type = VDIR;
    588 			vap->va_rdev = 0;
    589 			vap->va_nlink = 2;
    590 			vap->va_size = DEV_BSIZE;
    591 			break;
    592 		}
    593 		vap->va_uid = 0;
    594 		vap->va_gid = 0;
    595 		vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0];
    596 		vap->va_blocksize = DEV_BSIZE;
    597 		vap->va_atime.tv_sec = boottime.tv_sec;
    598 		vap->va_atime.tv_nsec = 0;
    599 		vap->va_mtime = vap->va_atime;
    600 		vap->va_ctime = vap->va_mtime;
    601 		vap->va_gen = 0;
    602 		vap->va_flags = 0;
    603 		vap->va_bytes = 0;
    604 		break;
    605 
    606 	case Fdesc:
    607 		fd = VTOFDESC(vp)->fd_fd;
    608 		error = fdesc_attr(fd, vap, ap->a_cred, ap->a_l);
    609 		break;
    610 
    611 	default:
    612 		panic("fdesc_getattr");
    613 		break;
    614 	}
    615 
    616 	if (error == 0)
    617 		vp->v_type = vap->va_type;
    618 
    619 	return (error);
    620 }
    621 
    622 int
    623 fdesc_setattr(v)
    624 	void *v;
    625 {
    626 	struct vop_setattr_args /* {
    627 		struct vnode *a_vp;
    628 		struct vattr *a_vap;
    629 		struct ucred *a_cred;
    630 		struct lwp *a_l;
    631 	} */ *ap = v;
    632 	struct filedesc *fdp = ap->a_l->l_proc->p_fd;
    633 	struct file *fp;
    634 	unsigned fd;
    635 
    636 	/*
    637 	 * Can't mess with the root vnode
    638 	 */
    639 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    640 	case Fdesc:
    641 		break;
    642 
    643 	case Fctty:
    644 		return (0);
    645 
    646 	default:
    647 		return (EACCES);
    648 	}
    649 
    650 	fd = VTOFDESC(ap->a_vp)->fd_fd;
    651 	if ((fp = fd_getfile(fdp, fd)) == NULL)
    652 		return (EBADF);
    653 
    654 	/*
    655 	 * XXX: Can't reasonably set the attr's on any types currently.
    656 	 *      On vnode's this will cause truncation and socket/pipes make
    657 	 *      no sense.
    658 	 */
    659 	simple_unlock(&fp->f_slock);
    660 	return (0);
    661 }
    662 
    663 #define UIO_MX 32
    664 
    665 struct fdesc_target {
    666 	ino_t ft_fileno;
    667 	u_char ft_type;
    668 	u_char ft_namlen;
    669 	char *ft_name;
    670 } fdesc_targets[] = {
    671 /* NOTE: The name must be less than UIO_MX-16 chars in length */
    672 #define N(s) sizeof(s)-1, s
    673 	{ FD_DEVFD,  DT_DIR,     N("fd")     },
    674 	{ FD_STDIN,  DT_LNK,     N("stdin")  },
    675 	{ FD_STDOUT, DT_LNK,     N("stdout") },
    676 	{ FD_STDERR, DT_LNK,     N("stderr") },
    677 	{ FD_CTTY,   DT_UNKNOWN, N("tty")    },
    678 #undef N
    679 };
    680 static int nfdesc_targets = sizeof(fdesc_targets) / sizeof(fdesc_targets[0]);
    681 
    682 int
    683 fdesc_readdir(v)
    684 	void *v;
    685 {
    686 	struct vop_readdir_args /* {
    687 		struct vnode *a_vp;
    688 		struct uio *a_uio;
    689 		struct ucred *a_cred;
    690 		int *a_eofflag;
    691 		off_t **a_cookies;
    692 		int *a_ncookies;
    693 	} */ *ap = v;
    694 	struct uio *uio = ap->a_uio;
    695 	struct dirent d;
    696 	struct filedesc *fdp;
    697 	off_t i;
    698 	int error;
    699 	off_t *cookies = NULL;
    700 	int ncookies = 0;
    701 
    702 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    703 	case Fctty:
    704 		return (0);
    705 
    706 	case Fdesc:
    707 		return (ENOTDIR);
    708 
    709 	default:
    710 		break;
    711 	}
    712 
    713 	fdp = uio->uio_lwp->l_proc->p_fd;
    714 
    715 	if (uio->uio_resid < UIO_MX)
    716 		return (EINVAL);
    717 	if (uio->uio_offset < 0)
    718 		return (EINVAL);
    719 
    720 	error = 0;
    721 	i = uio->uio_offset;
    722 	memset(&d, 0, UIO_MX);
    723 	d.d_reclen = UIO_MX;
    724 	if (ap->a_ncookies)
    725 		ncookies = (uio->uio_resid / UIO_MX);
    726 
    727 	if (VTOFDESC(ap->a_vp)->fd_type == Froot) {
    728 		struct fdesc_target *ft;
    729 
    730 		if (i >= nfdesc_targets)
    731 			return 0;
    732 
    733 		if (ap->a_ncookies) {
    734 			ncookies = min(ncookies, (nfdesc_targets - i));
    735 			cookies = malloc(ncookies * sizeof(off_t),
    736 			    M_TEMP, M_WAITOK);
    737 			*ap->a_cookies = cookies;
    738 			*ap->a_ncookies = ncookies;
    739 		}
    740 
    741 		for (ft = &fdesc_targets[i];
    742 		     uio->uio_resid >= UIO_MX && i < nfdesc_targets; ft++, i++) {
    743 			switch (ft->ft_fileno) {
    744 			case FD_CTTY:
    745 				if (cttyvp(uio->uio_lwp->l_proc) == NULL)
    746 					continue;
    747 				break;
    748 
    749 			case FD_STDIN:
    750 			case FD_STDOUT:
    751 			case FD_STDERR:
    752 				if ((ft->ft_fileno - FD_STDIN) >= fdp->fd_nfiles)
    753 					continue;
    754 				if (fdp->fd_ofiles[ft->ft_fileno - FD_STDIN] == NULL
    755 				    || FILE_IS_USABLE(fdp->fd_ofiles[ft->ft_fileno - FD_STDIN]) == 0)
    756 					continue;
    757 				break;
    758 			}
    759 
    760 			d.d_fileno = ft->ft_fileno;
    761 			d.d_namlen = ft->ft_namlen;
    762 			memcpy(d.d_name, ft->ft_name, ft->ft_namlen + 1);
    763 			d.d_type = ft->ft_type;
    764 
    765 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
    766 				break;
    767 			if (cookies)
    768 				*cookies++ = i + 1;
    769 		}
    770 	} else {
    771 		if (ap->a_ncookies) {
    772 			ncookies = min(ncookies, (fdp->fd_nfiles + 2));
    773 			cookies = malloc(ncookies * sizeof(off_t),
    774 			    M_TEMP, M_WAITOK);
    775 			*ap->a_cookies = cookies;
    776 			*ap->a_ncookies = ncookies;
    777 		}
    778 		for (; i - 2 < fdp->fd_nfiles && uio->uio_resid >= UIO_MX;
    779 		     i++) {
    780 			switch (i) {
    781 			case 0:
    782 			case 1:
    783 				d.d_fileno = FD_ROOT;		/* XXX */
    784 				d.d_namlen = i + 1;
    785 				memcpy(d.d_name, "..", d.d_namlen);
    786 				d.d_name[i + 1] = '\0';
    787 				d.d_type = DT_DIR;
    788 				break;
    789 
    790 			default:
    791 				if (fdp->fd_ofiles[i - 2] == NULL ||
    792 				    FILE_IS_USABLE(fdp->fd_ofiles[i - 2]) == 0)
    793 					continue;
    794 				d.d_fileno = i - 2 + FD_STDIN;
    795 				d.d_namlen = sprintf(d.d_name, "%d", (int) i - 2);
    796 				d.d_type = DT_UNKNOWN;
    797 				break;
    798 			}
    799 
    800 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
    801 				break;
    802 			if (cookies)
    803 				*cookies++ = i + 1;
    804 		}
    805 	}
    806 
    807 	if (ap->a_ncookies && error) {
    808 		free(*ap->a_cookies, M_TEMP);
    809 		*ap->a_ncookies = 0;
    810 		*ap->a_cookies = NULL;
    811 	}
    812 
    813 	uio->uio_offset = i;
    814 	return (error);
    815 }
    816 
    817 int
    818 fdesc_readlink(v)
    819 	void *v;
    820 {
    821 	struct vop_readlink_args /* {
    822 		struct vnode *a_vp;
    823 		struct uio *a_uio;
    824 		struct ucred *a_cred;
    825 	} */ *ap = v;
    826 	struct vnode *vp = ap->a_vp;
    827 	int error;
    828 
    829 	if (vp->v_type != VLNK)
    830 		return (EPERM);
    831 
    832 	if (VTOFDESC(vp)->fd_type == Flink) {
    833 		char *ln = VTOFDESC(vp)->fd_link;
    834 		error = uiomove(ln, strlen(ln), ap->a_uio);
    835 	} else {
    836 		error = EOPNOTSUPP;
    837 	}
    838 
    839 	return (error);
    840 }
    841 
    842 int
    843 fdesc_read(v)
    844 	void *v;
    845 {
    846 	struct vop_read_args /* {
    847 		struct vnode *a_vp;
    848 		struct uio *a_uio;
    849 		int  a_ioflag;
    850 		struct ucred *a_cred;
    851 	} */ *ap = v;
    852 	int error = EOPNOTSUPP;
    853 	struct vnode *vp = ap->a_vp;
    854 
    855 	switch (VTOFDESC(vp)->fd_type) {
    856 	case Fctty:
    857 		VOP_UNLOCK(vp, 0);
    858 		error = (*ctty_cdevsw.d_read)(devctty, ap->a_uio, ap->a_ioflag);
    859 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    860 		break;
    861 
    862 	default:
    863 		error = EOPNOTSUPP;
    864 		break;
    865 	}
    866 
    867 	return (error);
    868 }
    869 
    870 int
    871 fdesc_write(v)
    872 	void *v;
    873 {
    874 	struct vop_write_args /* {
    875 		struct vnode *a_vp;
    876 		struct uio *a_uio;
    877 		int  a_ioflag;
    878 		struct ucred *a_cred;
    879 	} */ *ap = v;
    880 	int error = EOPNOTSUPP;
    881 	struct vnode *vp = ap->a_vp;
    882 
    883 	switch (VTOFDESC(vp)->fd_type) {
    884 	case Fctty:
    885 		VOP_UNLOCK(vp, 0);
    886 		error = (*ctty_cdevsw.d_write)(devctty, ap->a_uio,
    887 					       ap->a_ioflag);
    888 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    889 		break;
    890 
    891 	default:
    892 		error = EOPNOTSUPP;
    893 		break;
    894 	}
    895 
    896 	return (error);
    897 }
    898 
    899 int
    900 fdesc_ioctl(v)
    901 	void *v;
    902 {
    903 	struct vop_ioctl_args /* {
    904 		struct vnode *a_vp;
    905 		u_long a_command;
    906 		void *a_data;
    907 		int  a_fflag;
    908 		struct ucred *a_cred;
    909 		struct lwp *a_l;
    910 	} */ *ap = v;
    911 	int error = EOPNOTSUPP;
    912 
    913 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    914 	case Fctty:
    915 		error = (*ctty_cdevsw.d_ioctl)(devctty, ap->a_command,
    916 					       ap->a_data, ap->a_fflag,
    917 					       ap->a_l);
    918 		break;
    919 
    920 	default:
    921 		error = EOPNOTSUPP;
    922 		break;
    923 	}
    924 
    925 	return (error);
    926 }
    927 
    928 int
    929 fdesc_poll(v)
    930 	void *v;
    931 {
    932 	struct vop_poll_args /* {
    933 		struct vnode *a_vp;
    934 		int a_events;
    935 		struct lwp *a_l;
    936 	} */ *ap = v;
    937 	int revents;
    938 
    939 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    940 	case Fctty:
    941 		revents = (*ctty_cdevsw.d_poll)(devctty, ap->a_events, ap->a_l);
    942 		break;
    943 
    944 	default:
    945 		revents = genfs_poll(v);
    946 		break;
    947 	}
    948 
    949 	return (revents);
    950 }
    951 
    952 int
    953 fdesc_kqfilter(v)
    954 	void *v;
    955 {
    956 	struct vop_kqfilter_args /* {
    957 		struct vnode *a_vp;
    958 		struct knote *a_kn;
    959 	} */ *ap = v;
    960 	int error;
    961 	struct proc *p;
    962 	struct lwp *l;
    963 	struct file *fp;
    964 
    965 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    966 	case Fctty:
    967 		error = (*ctty_cdevsw.d_kqfilter)(devctty, ap->a_kn);
    968 		break;
    969 
    970 	case Fdesc:
    971 		/* just invoke kqfilter for the underlying descriptor */
    972 		l = curlwp;	/* XXX hopefully ok to use curproc here */
    973 		p = l->l_proc;
    974 		if ((fp = fd_getfile(p->p_fd, VTOFDESC(ap->a_vp)->fd_fd)) == NULL)
    975 			return (1);
    976 
    977 		FILE_USE(fp);
    978 		error = (*fp->f_ops->fo_kqfilter)(fp, ap->a_kn);
    979 		FILE_UNUSE(fp, l);
    980 		break;
    981 
    982 	default:
    983 		return (genfs_kqfilter(v));
    984 	}
    985 
    986 	return (error);
    987 }
    988 
    989 int
    990 fdesc_inactive(v)
    991 	void *v;
    992 {
    993 	struct vop_inactive_args /* {
    994 		struct vnode *a_vp;
    995 		struct lwp *a_l;
    996 	} */ *ap = v;
    997 	struct vnode *vp = ap->a_vp;
    998 
    999 	/*
   1000 	 * Clear out the v_type field to avoid
   1001 	 * nasty things happening in vgone().
   1002 	 */
   1003 	VOP_UNLOCK(vp, 0);
   1004 	vp->v_type = VNON;
   1005 	return (0);
   1006 }
   1007 
   1008 int
   1009 fdesc_reclaim(v)
   1010 	void *v;
   1011 {
   1012 	struct vop_reclaim_args /* {
   1013 		struct vnode *a_vp;
   1014 	} */ *ap = v;
   1015 	struct vnode *vp = ap->a_vp;
   1016 	struct fdescnode *fd = VTOFDESC(vp);
   1017 
   1018 	LIST_REMOVE(fd, fd_hash);
   1019 	FREE(vp->v_data, M_TEMP);
   1020 	vp->v_data = 0;
   1021 
   1022 	return (0);
   1023 }
   1024 
   1025 /*
   1026  * Return POSIX pathconf information applicable to special devices.
   1027  */
   1028 int
   1029 fdesc_pathconf(v)
   1030 	void *v;
   1031 {
   1032 	struct vop_pathconf_args /* {
   1033 		struct vnode *a_vp;
   1034 		int a_name;
   1035 		register_t *a_retval;
   1036 	} */ *ap = v;
   1037 
   1038 	switch (ap->a_name) {
   1039 	case _PC_LINK_MAX:
   1040 		*ap->a_retval = LINK_MAX;
   1041 		return (0);
   1042 	case _PC_MAX_CANON:
   1043 		*ap->a_retval = MAX_CANON;
   1044 		return (0);
   1045 	case _PC_MAX_INPUT:
   1046 		*ap->a_retval = MAX_INPUT;
   1047 		return (0);
   1048 	case _PC_PIPE_BUF:
   1049 		*ap->a_retval = PIPE_BUF;
   1050 		return (0);
   1051 	case _PC_CHOWN_RESTRICTED:
   1052 		*ap->a_retval = 1;
   1053 		return (0);
   1054 	case _PC_VDISABLE:
   1055 		*ap->a_retval = _POSIX_VDISABLE;
   1056 		return (0);
   1057 	case _PC_SYNC_IO:
   1058 		*ap->a_retval = 1;
   1059 		return (0);
   1060 	default:
   1061 		return (EINVAL);
   1062 	}
   1063 	/* NOTREACHED */
   1064 }
   1065 
   1066 /*
   1067  * Print out the contents of a /dev/fd vnode.
   1068  */
   1069 /* ARGSUSED */
   1070 int
   1071 fdesc_print(v)
   1072 	void *v;
   1073 {
   1074 	printf("tag VT_NON, fdesc vnode\n");
   1075 	return (0);
   1076 }
   1077 
   1078 int
   1079 fdesc_link(v)
   1080 	void *v;
   1081 {
   1082 	struct vop_link_args /* {
   1083 		struct vnode *a_dvp;
   1084 		struct vnode *a_vp;
   1085 		struct componentname *a_cnp;
   1086 	} */ *ap = v;
   1087 
   1088 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
   1089 	vput(ap->a_dvp);
   1090 	return (EROFS);
   1091 }
   1092 
   1093 int
   1094 fdesc_symlink(v)
   1095 	void *v;
   1096 {
   1097 	struct vop_symlink_args /* {
   1098 		struct vnode *a_dvp;
   1099 		struct vnode **a_vpp;
   1100 		struct componentname *a_cnp;
   1101 		struct vattr *a_vap;
   1102 		char *a_target;
   1103 	} */ *ap = v;
   1104 
   1105 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
   1106 	vput(ap->a_dvp);
   1107 	return (EROFS);
   1108 }
   1109