Home | History | Annotate | Line # | Download | only in fdesc
fdesc_vnops.c revision 1.103.8.2
      1 /*	$NetBSD: fdesc_vnops.c,v 1.103.8.2 2009/04/28 07:37:14 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 1.103.8.2 2009/04/28 07:37:14 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 #include <sys/kauth.h>
     65 #include <sys/atomic.h>
     66 
     67 #include <miscfs/fdesc/fdesc.h>
     68 #include <miscfs/genfs/genfs.h>
     69 
     70 #define cttyvp(p) ((p)->p_lflag & PL_CONTROLT ? (p)->p_session->s_ttyvp : NULL)
     71 
     72 #define FDL_WANT	0x01
     73 #define FDL_LOCKED	0x02
     74 static int fdcache_lock;
     75 
     76 dev_t devctty;
     77 
     78 #if (FD_STDIN != FD_STDOUT-1) || (FD_STDOUT != FD_STDERR-1)
     79 FD_STDIN, FD_STDOUT, FD_STDERR must be a sequence n, n+1, n+2
     80 #endif
     81 
     82 #define	NFDCACHE 4
     83 
     84 #define FD_NHASH(ix) \
     85 	(&fdhashtbl[(ix) & fdhash])
     86 LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
     87 u_long fdhash;
     88 
     89 int	fdesc_lookup(void *);
     90 #define	fdesc_create	genfs_eopnotsupp
     91 #define	fdesc_mknod	genfs_eopnotsupp
     92 int	fdesc_open(void *);
     93 #define	fdesc_close	genfs_nullop
     94 #define	fdesc_access	genfs_nullop
     95 int	fdesc_getattr(void *);
     96 int	fdesc_setattr(void *);
     97 int	fdesc_read(void *);
     98 int	fdesc_write(void *);
     99 int	fdesc_ioctl(void *);
    100 int	fdesc_poll(void *);
    101 int	fdesc_kqfilter(void *);
    102 #define	fdesc_mmap	genfs_eopnotsupp
    103 #define	fdesc_fcntl	genfs_fcntl
    104 #define	fdesc_fsync	genfs_nullop
    105 #define	fdesc_seek	genfs_seek
    106 #define	fdesc_remove	genfs_eopnotsupp
    107 int	fdesc_link(void *);
    108 #define	fdesc_rename	genfs_eopnotsupp
    109 #define	fdesc_mkdir	genfs_eopnotsupp
    110 #define	fdesc_rmdir	genfs_eopnotsupp
    111 int	fdesc_symlink(void *);
    112 int	fdesc_readdir(void *);
    113 int	fdesc_readlink(void *);
    114 #define	fdesc_abortop	genfs_abortop
    115 int	fdesc_inactive(void *);
    116 int	fdesc_reclaim(void *);
    117 #define	fdesc_lock	genfs_lock
    118 #define	fdesc_unlock	genfs_unlock
    119 #define	fdesc_bmap	genfs_badop
    120 #define	fdesc_strategy	genfs_badop
    121 int	fdesc_print(void *);
    122 int	fdesc_pathconf(void *);
    123 #define	fdesc_islocked	genfs_islocked
    124 #define	fdesc_advlock	genfs_einval
    125 #define	fdesc_bwrite	genfs_eopnotsupp
    126 #define fdesc_revoke	genfs_revoke
    127 #define fdesc_putpages	genfs_null_putpages
    128 
    129 static int fdesc_attr(int, struct vattr *, kauth_cred_t);
    130 
    131 int (**fdesc_vnodeop_p)(void *);
    132 const struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = {
    133 	{ &vop_default_desc, vn_default_error },
    134 	{ &vop_lookup_desc, fdesc_lookup },		/* lookup */
    135 	{ &vop_create_desc, fdesc_create },		/* create */
    136 	{ &vop_mknod_desc, fdesc_mknod },		/* mknod */
    137 	{ &vop_open_desc, fdesc_open },			/* open */
    138 	{ &vop_close_desc, fdesc_close },		/* close */
    139 	{ &vop_access_desc, fdesc_access },		/* access */
    140 	{ &vop_getattr_desc, fdesc_getattr },		/* getattr */
    141 	{ &vop_setattr_desc, fdesc_setattr },		/* setattr */
    142 	{ &vop_read_desc, fdesc_read },			/* read */
    143 	{ &vop_write_desc, fdesc_write },		/* write */
    144 	{ &vop_ioctl_desc, fdesc_ioctl },		/* ioctl */
    145 	{ &vop_fcntl_desc, fdesc_fcntl },		/* fcntl */
    146 	{ &vop_poll_desc, fdesc_poll },			/* poll */
    147 	{ &vop_kqfilter_desc, fdesc_kqfilter },		/* kqfilter */
    148 	{ &vop_revoke_desc, fdesc_revoke },		/* revoke */
    149 	{ &vop_mmap_desc, fdesc_mmap },			/* mmap */
    150 	{ &vop_fsync_desc, fdesc_fsync },		/* fsync */
    151 	{ &vop_seek_desc, fdesc_seek },			/* seek */
    152 	{ &vop_remove_desc, fdesc_remove },		/* remove */
    153 	{ &vop_link_desc, fdesc_link },			/* link */
    154 	{ &vop_rename_desc, fdesc_rename },		/* rename */
    155 	{ &vop_mkdir_desc, fdesc_mkdir },		/* mkdir */
    156 	{ &vop_rmdir_desc, fdesc_rmdir },		/* rmdir */
    157 	{ &vop_symlink_desc, fdesc_symlink },		/* symlink */
    158 	{ &vop_readdir_desc, fdesc_readdir },		/* readdir */
    159 	{ &vop_readlink_desc, fdesc_readlink },		/* readlink */
    160 	{ &vop_abortop_desc, fdesc_abortop },		/* abortop */
    161 	{ &vop_inactive_desc, fdesc_inactive },		/* inactive */
    162 	{ &vop_reclaim_desc, fdesc_reclaim },		/* reclaim */
    163 	{ &vop_lock_desc, fdesc_lock },			/* lock */
    164 	{ &vop_unlock_desc, fdesc_unlock },		/* unlock */
    165 	{ &vop_bmap_desc, fdesc_bmap },			/* bmap */
    166 	{ &vop_strategy_desc, fdesc_strategy },		/* strategy */
    167 	{ &vop_print_desc, fdesc_print },		/* print */
    168 	{ &vop_islocked_desc, fdesc_islocked },		/* islocked */
    169 	{ &vop_pathconf_desc, fdesc_pathconf },		/* pathconf */
    170 	{ &vop_advlock_desc, fdesc_advlock },		/* advlock */
    171 	{ &vop_bwrite_desc, fdesc_bwrite },		/* bwrite */
    172 	{ &vop_putpages_desc, fdesc_putpages },		/* putpages */
    173 	{ NULL, NULL }
    174 };
    175 
    176 const struct vnodeopv_desc fdesc_vnodeop_opv_desc =
    177 	{ &fdesc_vnodeop_p, fdesc_vnodeop_entries };
    178 
    179 extern const struct cdevsw ctty_cdevsw;
    180 
    181 /*
    182  * Initialise cache headers
    183  */
    184 void
    185 fdesc_init(void)
    186 {
    187 	int cttymajor;
    188 
    189 	/* locate the major number */
    190 	cttymajor = cdevsw_lookup_major(&ctty_cdevsw);
    191 	devctty = makedev(cttymajor, 0);
    192 	fdhashtbl = hashinit(NFDCACHE, HASH_LIST, true, &fdhash);
    193 }
    194 
    195 /*
    196  * Free hash table.
    197  */
    198 void
    199 fdesc_done(void)
    200 {
    201 	hashdone(fdhashtbl, HASH_LIST, fdhash);
    202 }
    203 
    204 /*
    205  * Return a locked vnode of the correct type.
    206  */
    207 int
    208 fdesc_allocvp(fdntype ftype, int ix, struct mount *mp, struct vnode **vpp)
    209 {
    210 	struct fdhashhead *fc;
    211 	struct fdescnode *fd;
    212 	int error = 0;
    213 
    214 	fc = FD_NHASH(ix);
    215 loop:
    216 	for (fd = fc->lh_first; fd != 0; fd = fd->fd_hash.le_next) {
    217 		if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
    218 			if (vget(fd->fd_vnode, LK_EXCLUSIVE))
    219 				goto loop;
    220 			*vpp = fd->fd_vnode;
    221 			return (error);
    222 		}
    223 	}
    224 
    225 	/*
    226 	 * otherwise lock the array while we call getnewvnode
    227 	 * since that can block.
    228 	 */
    229 	if (fdcache_lock & FDL_LOCKED) {
    230 		fdcache_lock |= FDL_WANT;
    231 		(void) tsleep(&fdcache_lock, PINOD, "fdcache", 0);
    232 		goto loop;
    233 	}
    234 	fdcache_lock |= FDL_LOCKED;
    235 
    236 	error = getnewvnode(VT_FDESC, mp, fdesc_vnodeop_p, vpp);
    237 	if (error)
    238 		goto out;
    239 	fd = malloc(sizeof(struct fdescnode), M_TEMP, M_WAITOK);
    240 	(*vpp)->v_data = fd;
    241 	fd->fd_vnode = *vpp;
    242 	fd->fd_type = ftype;
    243 	fd->fd_fd = -1;
    244 	fd->fd_link = 0;
    245 	fd->fd_ix = ix;
    246 	uvm_vnp_setsize(*vpp, 0);
    247 	VOP_LOCK(*vpp, LK_EXCLUSIVE);
    248 	LIST_INSERT_HEAD(fc, fd, fd_hash);
    249 
    250 out:;
    251 	fdcache_lock &= ~FDL_LOCKED;
    252 
    253 	if (fdcache_lock & FDL_WANT) {
    254 		fdcache_lock &= ~FDL_WANT;
    255 		wakeup(&fdcache_lock);
    256 	}
    257 
    258 	return (error);
    259 }
    260 
    261 /*
    262  * vp is the current namei directory
    263  * ndp is the name to locate in that directory...
    264  */
    265 int
    266 fdesc_lookup(void *v)
    267 {
    268 	struct vop_lookup_args /* {
    269 		struct vnode * a_dvp;
    270 		struct vnode ** a_vpp;
    271 		struct componentname * a_cnp;
    272 	} */ *ap = v;
    273 	struct vnode **vpp = ap->a_vpp;
    274 	struct vnode *dvp = ap->a_dvp;
    275 	struct componentname *cnp = ap->a_cnp;
    276 	struct lwp *l = curlwp;
    277 	const char *pname = cnp->cn_nameptr;
    278 	struct proc *p = l->l_proc;
    279 	filedesc_t *fdp = p->p_fd;
    280 	int numfiles = fdp->fd_nfiles;
    281 	unsigned fd = 0;
    282 	int error;
    283 	struct vnode *fvp;
    284 	const char *ln;
    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, 0);
    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 >= numfiles)
    374 				break;
    375 		}
    376 
    377 		if (*pname != '\0') {
    378 			error = ENOENT;
    379 			goto bad;
    380 		}
    381 
    382 		mutex_enter(&fdp->fd_lock);
    383 		if (fd >= numfiles ||fdp->fd_ofiles[fd] == NULL ||
    384 		    fdp->fd_ofiles[fd]->ff_file == NULL) {
    385 			mutex_exit(&fdp->fd_lock);
    386 			error = EBADF;
    387 			goto bad;
    388 		}
    389 		mutex_exit(&fdp->fd_lock);
    390 
    391 		error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp);
    392 		if (error)
    393 			goto bad;
    394 		VTOFDESC(fvp)->fd_fd = fd;
    395 		*vpp = fvp;
    396 		goto good;
    397 	}
    398 
    399 bad:
    400 	*vpp = NULL;
    401 	return (error);
    402 
    403 good:
    404 	return (0);
    405 }
    406 
    407 int
    408 fdesc_open(void *v)
    409 {
    410 	struct vop_open_args /* {
    411 		struct vnode *a_vp;
    412 		int  a_mode;
    413 		kauth_cred_t a_cred;
    414 	} */ *ap = v;
    415 	struct vnode *vp = ap->a_vp;
    416 
    417 	switch (VTOFDESC(vp)->fd_type) {
    418 	case Fdesc:
    419 		/*
    420 		 * XXX Kludge: set dupfd to contain the value of the
    421 		 * the file descriptor being sought for duplication. The error
    422 		 * return ensures that the vnode for this device will be
    423 		 * released by vn_open. Open will detect this special error and
    424 		 * take the actions in dupfdopen.  Other callers of vn_open or
    425 		 * VOP_OPEN will simply report the error.
    426 		 */
    427 		curlwp->l_dupfd = VTOFDESC(vp)->fd_fd;	/* XXX */
    428 		return EDUPFD;
    429 
    430 	case Fctty:
    431 		return cdev_open(devctty, ap->a_mode, 0, curlwp);
    432 	case Froot:
    433 	case Fdevfd:
    434 	case Flink:
    435 		break;
    436 	}
    437 
    438 	return (0);
    439 }
    440 
    441 static int
    442 fdesc_attr(int fd, struct vattr *vap, kauth_cred_t cred)
    443 {
    444 	file_t *fp;
    445 	struct stat stb;
    446 	int error;
    447 
    448 	if ((fp = fd_getfile(fd)) == NULL)
    449 		return (EBADF);
    450 
    451 	switch (fp->f_type) {
    452 	case DTYPE_VNODE:
    453 		error = VOP_GETATTR((struct vnode *) fp->f_data, vap, cred);
    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 	filedesc_t *fdp;
    654 	off_t i;
    655 	int j;
    656 	int error;
    657 	off_t *cookies = NULL;
    658 	int ncookies;
    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 	fdp = curproc->p_fd;
    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 (fdp == NULL)
    713 					continue;
    714 				if ((ft->ft_fileno - FD_STDIN) >=
    715 				    fdp->fd_nfiles)
    716 					continue;
    717 				membar_consumer();
    718 				if (fdp->fd_ofiles[ft->ft_fileno - FD_STDIN]
    719 				    == NULL || fdp->fd_ofiles[ft->ft_fileno -
    720 				    FD_STDIN]->ff_file == NULL)
    721 					continue;
    722 				break;
    723 			}
    724 
    725 			d.d_fileno = ft->ft_fileno;
    726 			d.d_namlen = ft->ft_namlen;
    727 			(void)memcpy(d.d_name, ft->ft_name, ft->ft_namlen + 1);
    728 			d.d_type = ft->ft_type;
    729 
    730 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
    731 				break;
    732 			if (cookies)
    733 				*cookies++ = i + 1;
    734 		}
    735 	} else {
    736 		int nfdp = fdp ? fdp->fd_nfiles : 0;
    737 		membar_consumer();
    738 		if (ap->a_ncookies) {
    739 			ncookies = min(ncookies, nfdp + 2);
    740 			cookies = malloc(ncookies * sizeof(off_t),
    741 			    M_TEMP, M_WAITOK);
    742 			*ap->a_cookies = cookies;
    743 			*ap->a_ncookies = ncookies;
    744 		}
    745 		for (; i - 2 < nfdp && uio->uio_resid >= UIO_MX; i++) {
    746 			switch (i) {
    747 			case 0:
    748 			case 1:
    749 				d.d_fileno = FD_ROOT;		/* XXX */
    750 				d.d_namlen = i + 1;
    751 				(void)memcpy(d.d_name, "..", d.d_namlen);
    752 				d.d_name[i + 1] = '\0';
    753 				d.d_type = DT_DIR;
    754 				break;
    755 
    756 			default:
    757 				KASSERT(fdp != NULL);
    758 				j = (int)i - 2;
    759 				if (fdp == NULL || fdp->fd_ofiles[j] == NULL ||
    760 				    fdp->fd_ofiles[j]->ff_file == NULL)
    761 					continue;
    762 				d.d_fileno = j + FD_STDIN;
    763 				d.d_namlen = sprintf(d.d_name, "%d", j);
    764 				d.d_type = DT_UNKNOWN;
    765 				break;
    766 			}
    767 
    768 			if ((error = uiomove(&d, UIO_MX, uio)) != 0)
    769 				break;
    770 			if (cookies)
    771 				*cookies++ = i + 1;
    772 		}
    773 	}
    774 
    775 	if (ap->a_ncookies && error) {
    776 		free(*ap->a_cookies, M_TEMP);
    777 		*ap->a_ncookies = 0;
    778 		*ap->a_cookies = NULL;
    779 	}
    780 
    781 	uio->uio_offset = i;
    782 	return error;
    783 }
    784 
    785 int
    786 fdesc_readlink(void *v)
    787 {
    788 	struct vop_readlink_args /* {
    789 		struct vnode *a_vp;
    790 		struct uio *a_uio;
    791 		kauth_cred_t a_cred;
    792 	} */ *ap = v;
    793 	struct vnode *vp = ap->a_vp;
    794 	int error;
    795 
    796 	if (vp->v_type != VLNK)
    797 		return (EPERM);
    798 
    799 	if (VTOFDESC(vp)->fd_type == Flink) {
    800 		char *ln = VTOFDESC(vp)->fd_link;
    801 		error = uiomove(ln, strlen(ln), ap->a_uio);
    802 	} else {
    803 		error = EOPNOTSUPP;
    804 	}
    805 
    806 	return (error);
    807 }
    808 
    809 int
    810 fdesc_read(void *v)
    811 {
    812 	struct vop_read_args /* {
    813 		struct vnode *a_vp;
    814 		struct uio *a_uio;
    815 		int  a_ioflag;
    816 		kauth_cred_t a_cred;
    817 	} */ *ap = v;
    818 	int error = EOPNOTSUPP;
    819 	struct vnode *vp = ap->a_vp;
    820 
    821 	switch (VTOFDESC(vp)->fd_type) {
    822 	case Fctty:
    823 		VOP_UNLOCK(vp, 0);
    824 		error = cdev_read(devctty, ap->a_uio, ap->a_ioflag);
    825 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    826 		break;
    827 
    828 	default:
    829 		error = EOPNOTSUPP;
    830 		break;
    831 	}
    832 
    833 	return (error);
    834 }
    835 
    836 int
    837 fdesc_write(void *v)
    838 {
    839 	struct vop_write_args /* {
    840 		struct vnode *a_vp;
    841 		struct uio *a_uio;
    842 		int  a_ioflag;
    843 		kauth_cred_t a_cred;
    844 	} */ *ap = v;
    845 	int error = EOPNOTSUPP;
    846 	struct vnode *vp = ap->a_vp;
    847 
    848 	switch (VTOFDESC(vp)->fd_type) {
    849 	case Fctty:
    850 		VOP_UNLOCK(vp, 0);
    851 		error = cdev_write(devctty, ap->a_uio, ap->a_ioflag);
    852 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    853 		break;
    854 
    855 	default:
    856 		error = EOPNOTSUPP;
    857 		break;
    858 	}
    859 
    860 	return (error);
    861 }
    862 
    863 int
    864 fdesc_ioctl(void *v)
    865 {
    866 	struct vop_ioctl_args /* {
    867 		struct vnode *a_vp;
    868 		u_long a_command;
    869 		void *a_data;
    870 		int  a_fflag;
    871 		kauth_cred_t a_cred;
    872 	} */ *ap = v;
    873 	int error = EOPNOTSUPP;
    874 
    875 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    876 	case Fctty:
    877 		error = cdev_ioctl(devctty, ap->a_command, ap->a_data,
    878 		    ap->a_fflag, curlwp);
    879 		break;
    880 
    881 	default:
    882 		error = EOPNOTSUPP;
    883 		break;
    884 	}
    885 
    886 	return (error);
    887 }
    888 
    889 int
    890 fdesc_poll(void *v)
    891 {
    892 	struct vop_poll_args /* {
    893 		struct vnode *a_vp;
    894 		int a_events;
    895 	} */ *ap = v;
    896 	int revents;
    897 
    898 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    899 	case Fctty:
    900 		revents = cdev_poll(devctty, ap->a_events, curlwp);
    901 		break;
    902 
    903 	default:
    904 		revents = genfs_poll(v);
    905 		break;
    906 	}
    907 
    908 	return (revents);
    909 }
    910 
    911 int
    912 fdesc_kqfilter(void *v)
    913 {
    914 	struct vop_kqfilter_args /* {
    915 		struct vnode *a_vp;
    916 		struct knote *a_kn;
    917 	} */ *ap = v;
    918 	int error, fd;
    919 	file_t *fp;
    920 
    921 	switch (VTOFDESC(ap->a_vp)->fd_type) {
    922 	case Fctty:
    923 		error = cdev_kqfilter(devctty, ap->a_kn);
    924 		break;
    925 
    926 	case Fdesc:
    927 		/* just invoke kqfilter for the underlying descriptor */
    928 		fd = VTOFDESC(ap->a_vp)->fd_fd;
    929 		if ((fp = fd_getfile(fd)) == NULL)
    930 			return (1);
    931 		error = (*fp->f_ops->fo_kqfilter)(fp, ap->a_kn);
    932 		fd_putfile(fd);
    933 		break;
    934 
    935 	default:
    936 		return (genfs_kqfilter(v));
    937 	}
    938 
    939 	return (error);
    940 }
    941 
    942 int
    943 fdesc_inactive(void *v)
    944 {
    945 	struct vop_inactive_args /* {
    946 		struct vnode *a_vp;
    947 	} */ *ap = v;
    948 	struct vnode *vp = ap->a_vp;
    949 
    950 	/*
    951 	 * Clear out the v_type field to avoid
    952 	 * nasty things happening in vgone().
    953 	 */
    954 	VOP_UNLOCK(vp, 0);
    955 	vp->v_type = VNON;
    956 	return (0);
    957 }
    958 
    959 int
    960 fdesc_reclaim(void *v)
    961 {
    962 	struct vop_reclaim_args /* {
    963 		struct vnode *a_vp;
    964 	} */ *ap = v;
    965 	struct vnode *vp = ap->a_vp;
    966 	struct fdescnode *fd = VTOFDESC(vp);
    967 
    968 	LIST_REMOVE(fd, fd_hash);
    969 	free(vp->v_data, M_TEMP);
    970 	vp->v_data = 0;
    971 
    972 	return (0);
    973 }
    974 
    975 /*
    976  * Return POSIX pathconf information applicable to special devices.
    977  */
    978 int
    979 fdesc_pathconf(void *v)
    980 {
    981 	struct vop_pathconf_args /* {
    982 		struct vnode *a_vp;
    983 		int a_name;
    984 		register_t *a_retval;
    985 	} */ *ap = v;
    986 
    987 	switch (ap->a_name) {
    988 	case _PC_LINK_MAX:
    989 		*ap->a_retval = LINK_MAX;
    990 		return (0);
    991 	case _PC_MAX_CANON:
    992 		*ap->a_retval = MAX_CANON;
    993 		return (0);
    994 	case _PC_MAX_INPUT:
    995 		*ap->a_retval = MAX_INPUT;
    996 		return (0);
    997 	case _PC_PIPE_BUF:
    998 		*ap->a_retval = PIPE_BUF;
    999 		return (0);
   1000 	case _PC_CHOWN_RESTRICTED:
   1001 		*ap->a_retval = 1;
   1002 		return (0);
   1003 	case _PC_VDISABLE:
   1004 		*ap->a_retval = _POSIX_VDISABLE;
   1005 		return (0);
   1006 	case _PC_SYNC_IO:
   1007 		*ap->a_retval = 1;
   1008 		return (0);
   1009 	default:
   1010 		return (EINVAL);
   1011 	}
   1012 	/* NOTREACHED */
   1013 }
   1014 
   1015 /*
   1016  * Print out the contents of a /dev/fd vnode.
   1017  */
   1018 /* ARGSUSED */
   1019 int
   1020 fdesc_print(void *v)
   1021 {
   1022 	printf("tag VT_NON, fdesc vnode\n");
   1023 	return (0);
   1024 }
   1025 
   1026 int
   1027 fdesc_link(void *v)
   1028 {
   1029 	struct vop_link_args /* {
   1030 		struct vnode *a_dvp;
   1031 		struct vnode *a_vp;
   1032 		struct componentname *a_cnp;
   1033 	} */ *ap = v;
   1034 
   1035 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
   1036 	vput(ap->a_dvp);
   1037 	return (EROFS);
   1038 }
   1039 
   1040 int
   1041 fdesc_symlink(void *v)
   1042 {
   1043 	struct vop_symlink_args /* {
   1044 		struct vnode *a_dvp;
   1045 		struct vnode **a_vpp;
   1046 		struct componentname *a_cnp;
   1047 		struct vattr *a_vap;
   1048 		char *a_target;
   1049 	} */ *ap = v;
   1050 
   1051 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
   1052 	vput(ap->a_dvp);
   1053 	return (EROFS);
   1054 }
   1055