xref: /netbsd-src/sys/kern/vfs_vnops.c (revision 3816d47b2c42fcd6e549e3407f842a5b1a1d23ad)
1 /*	$NetBSD: vfs_vnops.c,v 1.169 2010/01/08 11:35:10 pooka Exp $	*/
2 
3 /*-
4  * Copyright (c) 2009 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Andrew Doran.
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  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1982, 1986, 1989, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  * (c) UNIX System Laboratories, Inc.
36  * All or some portions of this file are derived from material licensed
37  * to the University of California by American Telephone and Telegraph
38  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
39  * the permission of UNIX System Laboratories, Inc.
40  *
41  * Redistribution and use in source and binary forms, with or without
42  * modification, are permitted provided that the following conditions
43  * are met:
44  * 1. Redistributions of source code must retain the above copyright
45  *    notice, this list of conditions and the following disclaimer.
46  * 2. Redistributions in binary form must reproduce the above copyright
47  *    notice, this list of conditions and the following disclaimer in the
48  *    documentation and/or other materials provided with the distribution.
49  * 3. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)vfs_vnops.c	8.14 (Berkeley) 6/15/95
66  */
67 
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.169 2010/01/08 11:35:10 pooka Exp $");
70 
71 #include "veriexec.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/kernel.h>
76 #include <sys/file.h>
77 #include <sys/stat.h>
78 #include <sys/buf.h>
79 #include <sys/proc.h>
80 #include <sys/mount.h>
81 #include <sys/namei.h>
82 #include <sys/vnode.h>
83 #include <sys/ioctl.h>
84 #include <sys/tty.h>
85 #include <sys/poll.h>
86 #include <sys/kauth.h>
87 #include <sys/syslog.h>
88 #include <sys/fstrans.h>
89 #include <sys/atomic.h>
90 #include <sys/filedesc.h>
91 #include <sys/wapbl.h>
92 
93 #include <miscfs/specfs/specdev.h>
94 
95 #include <uvm/uvm_extern.h>
96 #include <uvm/uvm_readahead.h>
97 
98 #ifdef UNION
99 #include <fs/union/union.h>
100 #endif
101 
102 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct lwp *);
103 
104 #include <sys/verified_exec.h>
105 
106 static int vn_read(file_t *fp, off_t *offset, struct uio *uio,
107 	    kauth_cred_t cred, int flags);
108 static int vn_write(file_t *fp, off_t *offset, struct uio *uio,
109 	    kauth_cred_t cred, int flags);
110 static int vn_closefile(file_t *fp);
111 static int vn_poll(file_t *fp, int events);
112 static int vn_fcntl(file_t *fp, u_int com, void *data);
113 static int vn_statfile(file_t *fp, struct stat *sb);
114 static int vn_ioctl(file_t *fp, u_long com, void *data);
115 
116 const struct fileops vnops = {
117 	.fo_read = vn_read,
118 	.fo_write = vn_write,
119 	.fo_ioctl = vn_ioctl,
120 	.fo_fcntl = vn_fcntl,
121 	.fo_poll = vn_poll,
122 	.fo_stat = vn_statfile,
123 	.fo_close = vn_closefile,
124 	.fo_kqfilter = vn_kqfilter,
125 	.fo_restart = fnullop_restart,
126 };
127 
128 /*
129  * Common code for vnode open operations.
130  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
131  */
132 int
133 vn_open(struct nameidata *ndp, int fmode, int cmode)
134 {
135 	struct vnode *vp;
136 	struct lwp *l = curlwp;
137 	kauth_cred_t cred = l->l_cred;
138 	struct vattr va;
139 	int error;
140 	char *path;
141 
142 	ndp->ni_cnd.cn_flags &= TRYEMULROOT | NOCHROOT;
143 
144 	if (fmode & O_CREAT) {
145 		ndp->ni_cnd.cn_nameiop = CREATE;
146 		ndp->ni_cnd.cn_flags |= LOCKPARENT | LOCKLEAF;
147 		if ((fmode & O_EXCL) == 0 &&
148 		    ((fmode & O_NOFOLLOW) == 0))
149 			ndp->ni_cnd.cn_flags |= FOLLOW;
150 	} else {
151 		ndp->ni_cnd.cn_nameiop = LOOKUP;
152 		ndp->ni_cnd.cn_flags |= LOCKLEAF;
153 		if ((fmode & O_NOFOLLOW) == 0)
154 			ndp->ni_cnd.cn_flags |= FOLLOW;
155 	}
156 
157 	VERIEXEC_PATH_GET(ndp->ni_dirp, ndp->ni_segflg, ndp->ni_dirp, path);
158 
159 	error = namei(ndp);
160 	if (error)
161 		goto out;
162 
163 	vp = ndp->ni_vp;
164 
165 #if NVERIEXEC > 0
166 	error = veriexec_openchk(l, ndp->ni_vp, ndp->ni_dirp, fmode);
167 	if (error)
168 		goto bad;
169 #endif /* NVERIEXEC > 0 */
170 
171 	if (fmode & O_CREAT) {
172 		if (ndp->ni_vp == NULL) {
173 			vattr_null(&va);
174 			va.va_type = VREG;
175 			va.va_mode = cmode;
176 			if (fmode & O_EXCL)
177 				 va.va_vaflags |= VA_EXCLUSIVE;
178 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
179 					   &ndp->ni_cnd, &va);
180 			if (error)
181 				goto out;
182 			fmode &= ~O_TRUNC;
183 			vp = ndp->ni_vp;
184 		} else {
185 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
186 			if (ndp->ni_dvp == ndp->ni_vp)
187 				vrele(ndp->ni_dvp);
188 			else
189 				vput(ndp->ni_dvp);
190 			ndp->ni_dvp = NULL;
191 			vp = ndp->ni_vp;
192 			if (fmode & O_EXCL) {
193 				error = EEXIST;
194 				goto bad;
195 			}
196 			fmode &= ~O_CREAT;
197 		}
198 	} else {
199 		vp = ndp->ni_vp;
200 	}
201 	if (vp->v_type == VSOCK) {
202 		error = EOPNOTSUPP;
203 		goto bad;
204 	}
205 	if (ndp->ni_vp->v_type == VLNK) {
206 		error = EFTYPE;
207 		goto bad;
208 	}
209 
210 	if ((fmode & O_CREAT) == 0) {
211 		error = vn_openchk(vp, cred, fmode);
212 		if (error != 0)
213 			goto bad;
214 	}
215 
216 	if (fmode & O_TRUNC) {
217 		VOP_UNLOCK(vp, 0);			/* XXX */
218 
219 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
220 		vattr_null(&va);
221 		va.va_size = 0;
222 		error = VOP_SETATTR(vp, &va, cred);
223 		if (error != 0)
224 			goto bad;
225 	}
226 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0)
227 		goto bad;
228 	if (fmode & FWRITE) {
229 		mutex_enter(&vp->v_interlock);
230 		vp->v_writecount++;
231 		mutex_exit(&vp->v_interlock);
232 	}
233 
234 bad:
235 	if (error)
236 		vput(vp);
237 out:
238 	VERIEXEC_PATH_PUT(path);
239 	return (error);
240 }
241 
242 /*
243  * Check for write permissions on the specified vnode.
244  * Prototype text segments cannot be written.
245  */
246 int
247 vn_writechk(struct vnode *vp)
248 {
249 
250 	/*
251 	 * If the vnode is in use as a process's text,
252 	 * we can't allow writing.
253 	 */
254 	if (vp->v_iflag & VI_TEXT)
255 		return (ETXTBSY);
256 	return (0);
257 }
258 
259 int
260 vn_openchk(struct vnode *vp, kauth_cred_t cred, int fflags)
261 {
262 	int permbits = 0;
263 	int error;
264 
265 	if ((fflags & FREAD) != 0) {
266 		permbits = VREAD;
267 	}
268 	if ((fflags & (FWRITE | O_TRUNC)) != 0) {
269 		permbits |= VWRITE;
270 		if (vp->v_type == VDIR) {
271 			error = EISDIR;
272 			goto bad;
273 		}
274 		error = vn_writechk(vp);
275 		if (error != 0)
276 			goto bad;
277 	}
278 	error = VOP_ACCESS(vp, permbits, cred);
279 bad:
280 	return error;
281 }
282 
283 /*
284  * Mark a vnode as having executable mappings.
285  */
286 void
287 vn_markexec(struct vnode *vp)
288 {
289 
290 	if ((vp->v_iflag & VI_EXECMAP) != 0) {
291 		/* Safe unlocked, as long as caller holds a reference. */
292 		return;
293 	}
294 
295 	mutex_enter(&vp->v_interlock);
296 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
297 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
298 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
299 		vp->v_iflag |= VI_EXECMAP;
300 	}
301 	mutex_exit(&vp->v_interlock);
302 }
303 
304 /*
305  * Mark a vnode as being the text of a process.
306  * Fail if the vnode is currently writable.
307  */
308 int
309 vn_marktext(struct vnode *vp)
310 {
311 
312 	if ((vp->v_iflag & (VI_TEXT|VI_EXECMAP)) == (VI_TEXT|VI_EXECMAP)) {
313 		/* Safe unlocked, as long as caller holds a reference. */
314 		return (0);
315 	}
316 
317 	mutex_enter(&vp->v_interlock);
318 	if (vp->v_writecount != 0) {
319 		KASSERT((vp->v_iflag & VI_TEXT) == 0);
320 		mutex_exit(&vp->v_interlock);
321 		return (ETXTBSY);
322 	}
323 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
324 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
325 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
326 	}
327 	vp->v_iflag |= (VI_TEXT | VI_EXECMAP);
328 	mutex_exit(&vp->v_interlock);
329 	return (0);
330 }
331 
332 /*
333  * Vnode close call
334  *
335  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
336  */
337 int
338 vn_close(struct vnode *vp, int flags, kauth_cred_t cred)
339 {
340 	int error;
341 
342 	if (flags & FWRITE) {
343 		mutex_enter(&vp->v_interlock);
344 		vp->v_writecount--;
345 		mutex_exit(&vp->v_interlock);
346 	}
347 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
348 	error = VOP_CLOSE(vp, flags, cred);
349 	vput(vp);
350 	return (error);
351 }
352 
353 /*
354  * Package up an I/O request on a vnode into a uio and do it.
355  */
356 int
357 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset,
358     enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid,
359     struct lwp *l)
360 {
361 	struct uio auio;
362 	struct iovec aiov;
363 	int error;
364 
365 	if ((ioflg & IO_NODELOCKED) == 0) {
366 		if (rw == UIO_READ) {
367 			vn_lock(vp, LK_SHARED | LK_RETRY);
368 		} else /* UIO_WRITE */ {
369 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
370 		}
371 	}
372 	auio.uio_iov = &aiov;
373 	auio.uio_iovcnt = 1;
374 	aiov.iov_base = base;
375 	aiov.iov_len = len;
376 	auio.uio_resid = len;
377 	auio.uio_offset = offset;
378 	auio.uio_rw = rw;
379 	if (segflg == UIO_SYSSPACE) {
380 		UIO_SETUP_SYSSPACE(&auio);
381 	} else {
382 		auio.uio_vmspace = l->l_proc->p_vmspace;
383 	}
384 	if (rw == UIO_READ) {
385 		error = VOP_READ(vp, &auio, ioflg, cred);
386 	} else {
387 		error = VOP_WRITE(vp, &auio, ioflg, cred);
388 	}
389 	if (aresid)
390 		*aresid = auio.uio_resid;
391 	else
392 		if (auio.uio_resid && error == 0)
393 			error = EIO;
394 	if ((ioflg & IO_NODELOCKED) == 0) {
395 		VOP_UNLOCK(vp, 0);
396 	}
397 	return (error);
398 }
399 
400 int
401 vn_readdir(file_t *fp, char *bf, int segflg, u_int count, int *done,
402     struct lwp *l, off_t **cookies, int *ncookies)
403 {
404 	struct vnode *vp = (struct vnode *)fp->f_data;
405 	struct iovec aiov;
406 	struct uio auio;
407 	int error, eofflag;
408 
409 	/* Limit the size on any kernel buffers used by VOP_READDIR */
410 	count = min(MAXBSIZE, count);
411 
412 unionread:
413 	if (vp->v_type != VDIR)
414 		return (EINVAL);
415 	aiov.iov_base = bf;
416 	aiov.iov_len = count;
417 	auio.uio_iov = &aiov;
418 	auio.uio_iovcnt = 1;
419 	auio.uio_rw = UIO_READ;
420 	if (segflg == UIO_SYSSPACE) {
421 		UIO_SETUP_SYSSPACE(&auio);
422 	} else {
423 		KASSERT(l == curlwp);
424 		auio.uio_vmspace = l->l_proc->p_vmspace;
425 	}
426 	auio.uio_resid = count;
427 	vn_lock(vp, LK_SHARED | LK_RETRY);
428 	auio.uio_offset = fp->f_offset;
429 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
430 		    ncookies);
431 	mutex_enter(&fp->f_lock);
432 	fp->f_offset = auio.uio_offset;
433 	mutex_exit(&fp->f_lock);
434 	VOP_UNLOCK(vp, 0);
435 	if (error)
436 		return (error);
437 
438 	if (count == auio.uio_resid && vn_union_readdir_hook) {
439 		struct vnode *ovp = vp;
440 
441 		error = (*vn_union_readdir_hook)(&vp, fp, l);
442 		if (error)
443 			return (error);
444 		if (vp != ovp)
445 			goto unionread;
446 	}
447 
448 	if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) &&
449 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
450 		struct vnode *tvp = vp;
451 		vp = vp->v_mount->mnt_vnodecovered;
452 		vref(vp);
453 		mutex_enter(&fp->f_lock);
454 		fp->f_data = vp;
455 		fp->f_offset = 0;
456 		mutex_exit(&fp->f_lock);
457 		vrele(tvp);
458 		goto unionread;
459 	}
460 	*done = count - auio.uio_resid;
461 	return error;
462 }
463 
464 /*
465  * File table vnode read routine.
466  */
467 static int
468 vn_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
469     int flags)
470 {
471 	struct vnode *vp = (struct vnode *)fp->f_data;
472 	int count, error, ioflag, fflag;
473 
474 	ioflag = IO_ADV_ENCODE(fp->f_advice);
475 	fflag = fp->f_flag;
476 	if (fflag & FNONBLOCK)
477 		ioflag |= IO_NDELAY;
478 	if ((fflag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
479 		ioflag |= IO_SYNC;
480 	if (fflag & FALTIO)
481 		ioflag |= IO_ALTSEMANTICS;
482 	if (fflag & FDIRECT)
483 		ioflag |= IO_DIRECT;
484 	vn_lock(vp, LK_SHARED | LK_RETRY);
485 	uio->uio_offset = *offset;
486 	count = uio->uio_resid;
487 	error = VOP_READ(vp, uio, ioflag, cred);
488 	if (flags & FOF_UPDATE_OFFSET)
489 		*offset += count - uio->uio_resid;
490 	VOP_UNLOCK(vp, 0);
491 	return (error);
492 }
493 
494 /*
495  * File table vnode write routine.
496  */
497 static int
498 vn_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
499     int flags)
500 {
501 	struct vnode *vp = (struct vnode *)fp->f_data;
502 	int count, error, ioflag, fflag;
503 
504 	ioflag = IO_ADV_ENCODE(fp->f_advice) | IO_UNIT;
505 	fflag = fp->f_flag;
506 	if (vp->v_type == VREG && (fflag & O_APPEND))
507 		ioflag |= IO_APPEND;
508 	if (fflag & FNONBLOCK)
509 		ioflag |= IO_NDELAY;
510 	if (fflag & FFSYNC ||
511 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
512 		ioflag |= IO_SYNC;
513 	else if (fflag & FDSYNC)
514 		ioflag |= IO_DSYNC;
515 	if (fflag & FALTIO)
516 		ioflag |= IO_ALTSEMANTICS;
517 	if (fflag & FDIRECT)
518 		ioflag |= IO_DIRECT;
519 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
520 	uio->uio_offset = *offset;
521 	count = uio->uio_resid;
522 	error = VOP_WRITE(vp, uio, ioflag, cred);
523 	if (flags & FOF_UPDATE_OFFSET) {
524 		if (ioflag & IO_APPEND) {
525 			/*
526 			 * SUSv3 describes behaviour for count = 0 as following:
527 			 * "Before any action ... is taken, and if nbyte is zero
528 			 * and the file is a regular file, the write() function
529 			 * ... in the absence of errors ... shall return zero
530 			 * and have no other results."
531 			 */
532 			if (count)
533 				*offset = uio->uio_offset;
534 		} else
535 			*offset += count - uio->uio_resid;
536 	}
537 	VOP_UNLOCK(vp, 0);
538 	return (error);
539 }
540 
541 /*
542  * File table vnode stat routine.
543  */
544 static int
545 vn_statfile(file_t *fp, struct stat *sb)
546 {
547 	struct vnode *vp = fp->f_data;
548 	int error;
549 
550 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
551 	error = vn_stat(vp, sb);
552 	VOP_UNLOCK(vp, 0);
553 	return error;
554 }
555 
556 int
557 vn_stat(struct vnode *vp, struct stat *sb)
558 {
559 	struct vattr va;
560 	int error;
561 	mode_t mode;
562 
563 	error = VOP_GETATTR(vp, &va, kauth_cred_get());
564 	if (error)
565 		return (error);
566 	/*
567 	 * Copy from vattr table
568 	 */
569 	sb->st_dev = va.va_fsid;
570 	sb->st_ino = va.va_fileid;
571 	mode = va.va_mode;
572 	switch (vp->v_type) {
573 	case VREG:
574 		mode |= S_IFREG;
575 		break;
576 	case VDIR:
577 		mode |= S_IFDIR;
578 		break;
579 	case VBLK:
580 		mode |= S_IFBLK;
581 		break;
582 	case VCHR:
583 		mode |= S_IFCHR;
584 		break;
585 	case VLNK:
586 		mode |= S_IFLNK;
587 		break;
588 	case VSOCK:
589 		mode |= S_IFSOCK;
590 		break;
591 	case VFIFO:
592 		mode |= S_IFIFO;
593 		break;
594 	default:
595 		return (EBADF);
596 	};
597 	sb->st_mode = mode;
598 	sb->st_nlink = va.va_nlink;
599 	sb->st_uid = va.va_uid;
600 	sb->st_gid = va.va_gid;
601 	sb->st_rdev = va.va_rdev;
602 	sb->st_size = va.va_size;
603 	sb->st_atimespec = va.va_atime;
604 	sb->st_mtimespec = va.va_mtime;
605 	sb->st_ctimespec = va.va_ctime;
606 	sb->st_birthtimespec = va.va_birthtime;
607 	sb->st_blksize = va.va_blocksize;
608 	sb->st_flags = va.va_flags;
609 	sb->st_gen = 0;
610 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
611 	return (0);
612 }
613 
614 /*
615  * File table vnode fcntl routine.
616  */
617 static int
618 vn_fcntl(file_t *fp, u_int com, void *data)
619 {
620 	struct vnode *vp = fp->f_data;
621 	int error;
622 
623 	error = VOP_FCNTL(vp, com, data, fp->f_flag, kauth_cred_get());
624 	return (error);
625 }
626 
627 /*
628  * File table vnode ioctl routine.
629  */
630 static int
631 vn_ioctl(file_t *fp, u_long com, void *data)
632 {
633 	struct vnode *vp = fp->f_data, *ovp;
634 	struct vattr vattr;
635 	int error;
636 
637 	switch (vp->v_type) {
638 
639 	case VREG:
640 	case VDIR:
641 		if (com == FIONREAD) {
642 			error = VOP_GETATTR(vp, &vattr,
643 			    kauth_cred_get());
644 			if (error)
645 				return (error);
646 			*(int *)data = vattr.va_size - fp->f_offset;
647 			return (0);
648 		}
649 		if ((com == FIONWRITE) || (com == FIONSPACE)) {
650 			/*
651 			 * Files don't have send queues, so there never
652 			 * are any bytes in them, nor is there any
653 			 * open space in them.
654 			 */
655 			*(int *)data = 0;
656 			return (0);
657 		}
658 		if (com == FIOGETBMAP) {
659 			daddr_t *block;
660 
661 			if (*(daddr_t *)data < 0)
662 				return (EINVAL);
663 			block = (daddr_t *)data;
664 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
665 		}
666 		if (com == OFIOGETBMAP) {
667 			daddr_t ibn, obn;
668 
669 			if (*(int32_t *)data < 0)
670 				return (EINVAL);
671 			ibn = (daddr_t)*(int32_t *)data;
672 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
673 			*(int32_t *)data = (int32_t)obn;
674 			return error;
675 		}
676 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
677 			return (0);			/* XXX */
678 		/* fall into ... */
679 	case VFIFO:
680 	case VCHR:
681 	case VBLK:
682 		error = VOP_IOCTL(vp, com, data, fp->f_flag,
683 		    kauth_cred_get());
684 		if (error == 0 && com == TIOCSCTTY) {
685 			vref(vp);
686 			mutex_enter(proc_lock);
687 			ovp = curproc->p_session->s_ttyvp;
688 			curproc->p_session->s_ttyvp = vp;
689 			mutex_exit(proc_lock);
690 			if (ovp != NULL)
691 				vrele(ovp);
692 		}
693 		return (error);
694 
695 	default:
696 		return (EPASSTHROUGH);
697 	}
698 }
699 
700 /*
701  * File table vnode poll routine.
702  */
703 static int
704 vn_poll(file_t *fp, int events)
705 {
706 
707 	return (VOP_POLL(fp->f_data, events));
708 }
709 
710 /*
711  * File table vnode kqfilter routine.
712  */
713 int
714 vn_kqfilter(file_t *fp, struct knote *kn)
715 {
716 
717 	return (VOP_KQFILTER(fp->f_data, kn));
718 }
719 
720 /*
721  * Check that the vnode is still valid, and if so
722  * acquire requested lock.
723  */
724 int
725 vn_lock(struct vnode *vp, int flags)
726 {
727 	int error;
728 
729 #if 0
730 	KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0
731 	    || (vp->v_iflag & VI_ONWORKLST) != 0);
732 #endif
733 	KASSERT((flags &
734 	    ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT|LK_RETRY|
735 	    LK_CANRECURSE))
736 	    == 0);
737 
738 #ifdef DIAGNOSTIC
739 	if (wapbl_vphaswapbl(vp))
740 		WAPBL_JUNLOCK_ASSERT(wapbl_vptomp(vp));
741 #endif
742 
743 	do {
744 		/*
745 		 * XXX PR 37706 forced unmount of file systems is unsafe.
746 		 * Race between vclean() and this the remaining problem.
747 		 */
748 		if (vp->v_iflag & VI_XLOCK) {
749 			if ((flags & LK_INTERLOCK) == 0) {
750 				mutex_enter(&vp->v_interlock);
751 			}
752 			flags &= ~LK_INTERLOCK;
753 			if (flags & LK_NOWAIT) {
754 				mutex_exit(&vp->v_interlock);
755 				return EBUSY;
756 			}
757 			vwait(vp, VI_XLOCK);
758 			mutex_exit(&vp->v_interlock);
759 			error = ENOENT;
760 		} else {
761 			if ((flags & LK_INTERLOCK) != 0) {
762 				mutex_exit(&vp->v_interlock);
763 			}
764 			flags &= ~LK_INTERLOCK;
765 			error = VOP_LOCK(vp, (flags & ~LK_RETRY));
766 			if (error == 0 || error == EDEADLK || error == EBUSY)
767 				return (error);
768 		}
769 	} while (flags & LK_RETRY);
770 	return (error);
771 }
772 
773 /*
774  * File table vnode close routine.
775  */
776 static int
777 vn_closefile(file_t *fp)
778 {
779 
780 	return vn_close(fp->f_data, fp->f_flag, fp->f_cred);
781 }
782 
783 /*
784  * Enable LK_CANRECURSE on lock. Return prior status.
785  */
786 u_int
787 vn_setrecurse(struct vnode *vp)
788 {
789 	struct vnlock *lkp;
790 
791 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
792 	atomic_inc_uint(&lkp->vl_canrecurse);
793 
794 	return 0;
795 }
796 
797 /*
798  * Called when done with locksetrecurse.
799  */
800 void
801 vn_restorerecurse(struct vnode *vp, u_int flags)
802 {
803 	struct vnlock *lkp;
804 
805 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
806 	atomic_dec_uint(&lkp->vl_canrecurse);
807 }
808 
809 /*
810  * Simplified in-kernel wrapper calls for extended attribute access.
811  * Both calls pass in a NULL credential, authorizing a "kernel" access.
812  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
813  */
814 int
815 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
816     const char *attrname, size_t *buflen, void *bf, struct lwp *l)
817 {
818 	struct uio auio;
819 	struct iovec aiov;
820 	int error;
821 
822 	aiov.iov_len = *buflen;
823 	aiov.iov_base = bf;
824 
825 	auio.uio_iov = &aiov;
826 	auio.uio_iovcnt = 1;
827 	auio.uio_rw = UIO_READ;
828 	auio.uio_offset = 0;
829 	auio.uio_resid = *buflen;
830 	UIO_SETUP_SYSSPACE(&auio);
831 
832 	if ((ioflg & IO_NODELOCKED) == 0)
833 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
834 
835 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL);
836 
837 	if ((ioflg & IO_NODELOCKED) == 0)
838 		VOP_UNLOCK(vp, 0);
839 
840 	if (error == 0)
841 		*buflen = *buflen - auio.uio_resid;
842 
843 	return (error);
844 }
845 
846 /*
847  * XXX Failure mode if partially written?
848  */
849 int
850 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
851     const char *attrname, size_t buflen, const void *bf, struct lwp *l)
852 {
853 	struct uio auio;
854 	struct iovec aiov;
855 	int error;
856 
857 	aiov.iov_len = buflen;
858 	aiov.iov_base = __UNCONST(bf);		/* XXXUNCONST kills const */
859 
860 	auio.uio_iov = &aiov;
861 	auio.uio_iovcnt = 1;
862 	auio.uio_rw = UIO_WRITE;
863 	auio.uio_offset = 0;
864 	auio.uio_resid = buflen;
865 	UIO_SETUP_SYSSPACE(&auio);
866 
867 	if ((ioflg & IO_NODELOCKED) == 0) {
868 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
869 	}
870 
871 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL);
872 
873 	if ((ioflg & IO_NODELOCKED) == 0) {
874 		VOP_UNLOCK(vp, 0);
875 	}
876 
877 	return (error);
878 }
879 
880 int
881 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
882     const char *attrname, struct lwp *l)
883 {
884 	int error;
885 
886 	if ((ioflg & IO_NODELOCKED) == 0) {
887 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
888 	}
889 
890 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL);
891 	if (error == EOPNOTSUPP)
892 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL);
893 
894 	if ((ioflg & IO_NODELOCKED) == 0) {
895 		VOP_UNLOCK(vp, 0);
896 	}
897 
898 	return (error);
899 }
900 
901 void
902 vn_ra_allocctx(struct vnode *vp)
903 {
904 	struct uvm_ractx *ra = NULL;
905 
906 	KASSERT(mutex_owned(&vp->v_interlock));
907 
908 	if (vp->v_type != VREG) {
909 		return;
910 	}
911 	if (vp->v_ractx != NULL) {
912 		return;
913 	}
914 	if (vp->v_ractx == NULL) {
915 		mutex_exit(&vp->v_interlock);
916 		ra = uvm_ra_allocctx();
917 		mutex_enter(&vp->v_interlock);
918 		if (ra != NULL && vp->v_ractx == NULL) {
919 			vp->v_ractx = ra;
920 			ra = NULL;
921 		}
922 	}
923 	if (ra != NULL) {
924 		uvm_ra_freectx(ra);
925 	}
926 }
927