xref: /netbsd-src/sys/kern/vfs_vnops.c (revision b757af438b42b93f8c6571f026d8b8ef3eaf5fc9)
1 /*	$NetBSD: vfs_vnops.c,v 1.183 2011/10/14 09:23:31 hannken 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 1.183 2011/10/14 09:23:31 hannken 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 #include <miscfs/fifofs/fifo.h>
95 
96 #include <uvm/uvm_extern.h>
97 #include <uvm/uvm_readahead.h>
98 
99 #ifdef UNION
100 #include <fs/union/union.h>
101 #endif
102 
103 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct lwp *);
104 
105 #include <sys/verified_exec.h>
106 
107 static int vn_read(file_t *fp, off_t *offset, struct uio *uio,
108 	    kauth_cred_t cred, int flags);
109 static int vn_write(file_t *fp, off_t *offset, struct uio *uio,
110 	    kauth_cred_t cred, int flags);
111 static int vn_closefile(file_t *fp);
112 static int vn_poll(file_t *fp, int events);
113 static int vn_fcntl(file_t *fp, u_int com, void *data);
114 static int vn_statfile(file_t *fp, struct stat *sb);
115 static int vn_ioctl(file_t *fp, u_long com, void *data);
116 
117 const struct fileops vnops = {
118 	.fo_read = vn_read,
119 	.fo_write = vn_write,
120 	.fo_ioctl = vn_ioctl,
121 	.fo_fcntl = vn_fcntl,
122 	.fo_poll = vn_poll,
123 	.fo_stat = vn_statfile,
124 	.fo_close = vn_closefile,
125 	.fo_kqfilter = vn_kqfilter,
126 	.fo_restart = fnullop_restart,
127 };
128 
129 /*
130  * Common code for vnode open operations.
131  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
132  */
133 int
134 vn_open(struct nameidata *ndp, int fmode, int cmode)
135 {
136 	struct vnode *vp;
137 	struct lwp *l = curlwp;
138 	kauth_cred_t cred = l->l_cred;
139 	struct vattr va;
140 	int error;
141 	const char *pathstring;
142 
143 	if ((fmode & (O_CREAT | O_DIRECTORY)) == (O_CREAT | O_DIRECTORY))
144 		return EINVAL;
145 
146 	ndp->ni_cnd.cn_flags &= TRYEMULROOT | NOCHROOT;
147 
148 	if (fmode & O_CREAT) {
149 		ndp->ni_cnd.cn_nameiop = CREATE;
150 		ndp->ni_cnd.cn_flags |= LOCKPARENT | LOCKLEAF;
151 		if ((fmode & O_EXCL) == 0 &&
152 		    ((fmode & O_NOFOLLOW) == 0))
153 			ndp->ni_cnd.cn_flags |= FOLLOW;
154 	} else {
155 		ndp->ni_cnd.cn_nameiop = LOOKUP;
156 		ndp->ni_cnd.cn_flags |= LOCKLEAF;
157 		if ((fmode & O_NOFOLLOW) == 0)
158 			ndp->ni_cnd.cn_flags |= FOLLOW;
159 	}
160 
161 	pathstring = pathbuf_stringcopy_get(ndp->ni_pathbuf);
162 	if (pathstring == NULL) {
163 		return ENOMEM;
164 	}
165 
166 	error = namei(ndp);
167 	if (error)
168 		goto out;
169 
170 	vp = ndp->ni_vp;
171 
172 #if NVERIEXEC > 0
173 	error = veriexec_openchk(l, ndp->ni_vp, pathstring, fmode);
174 	if (error)
175 		goto bad;
176 #endif /* NVERIEXEC > 0 */
177 
178 	if (fmode & O_CREAT) {
179 		if (ndp->ni_vp == NULL) {
180 			vattr_null(&va);
181 			va.va_type = VREG;
182 			va.va_mode = cmode;
183 			if (fmode & O_EXCL)
184 				 va.va_vaflags |= VA_EXCLUSIVE;
185 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
186 					   &ndp->ni_cnd, &va);
187 			if (error)
188 				goto out;
189 			fmode &= ~O_TRUNC;
190 			vp = ndp->ni_vp;
191 		} else {
192 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
193 			if (ndp->ni_dvp == ndp->ni_vp)
194 				vrele(ndp->ni_dvp);
195 			else
196 				vput(ndp->ni_dvp);
197 			ndp->ni_dvp = NULL;
198 			vp = ndp->ni_vp;
199 			if (fmode & O_EXCL) {
200 				error = EEXIST;
201 				goto bad;
202 			}
203 			fmode &= ~O_CREAT;
204 		}
205 	} else {
206 		vp = ndp->ni_vp;
207 	}
208 	if (vp->v_type == VSOCK) {
209 		error = EOPNOTSUPP;
210 		goto bad;
211 	}
212 	if (ndp->ni_vp->v_type == VLNK) {
213 		error = EFTYPE;
214 		goto bad;
215 	}
216 
217 	if ((fmode & O_CREAT) == 0) {
218 		error = vn_openchk(vp, cred, fmode);
219 		if (error != 0)
220 			goto bad;
221 	}
222 
223 	if (fmode & O_TRUNC) {
224 		vattr_null(&va);
225 		va.va_size = 0;
226 		error = VOP_SETATTR(vp, &va, cred);
227 		if (error != 0)
228 			goto bad;
229 	}
230 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0)
231 		goto bad;
232 	if (fmode & FWRITE) {
233 		mutex_enter(vp->v_interlock);
234 		vp->v_writecount++;
235 		mutex_exit(vp->v_interlock);
236 	}
237 
238 bad:
239 	if (error)
240 		vput(vp);
241 out:
242 	pathbuf_stringcopy_put(ndp->ni_pathbuf, pathstring);
243 	return (error);
244 }
245 
246 /*
247  * Check for write permissions on the specified vnode.
248  * Prototype text segments cannot be written.
249  */
250 int
251 vn_writechk(struct vnode *vp)
252 {
253 
254 	/*
255 	 * If the vnode is in use as a process's text,
256 	 * we can't allow writing.
257 	 */
258 	if (vp->v_iflag & VI_TEXT)
259 		return (ETXTBSY);
260 	return (0);
261 }
262 
263 int
264 vn_openchk(struct vnode *vp, kauth_cred_t cred, int fflags)
265 {
266 	int permbits = 0;
267 	int error;
268 
269 	if ((fflags & O_DIRECTORY) != 0 && vp->v_type != VDIR)
270 		return ENOTDIR;
271 
272 	if ((fflags & FREAD) != 0) {
273 		permbits = VREAD;
274 	}
275 	if ((fflags & (FWRITE | O_TRUNC)) != 0) {
276 		permbits |= VWRITE;
277 		if (vp->v_type == VDIR) {
278 			error = EISDIR;
279 			goto bad;
280 		}
281 		error = vn_writechk(vp);
282 		if (error != 0)
283 			goto bad;
284 	}
285 	error = VOP_ACCESS(vp, permbits, cred);
286 bad:
287 	return error;
288 }
289 
290 /*
291  * Mark a vnode as having executable mappings.
292  */
293 void
294 vn_markexec(struct vnode *vp)
295 {
296 
297 	if ((vp->v_iflag & VI_EXECMAP) != 0) {
298 		/* Safe unlocked, as long as caller holds a reference. */
299 		return;
300 	}
301 
302 	mutex_enter(vp->v_interlock);
303 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
304 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
305 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
306 		vp->v_iflag |= VI_EXECMAP;
307 	}
308 	mutex_exit(vp->v_interlock);
309 }
310 
311 /*
312  * Mark a vnode as being the text of a process.
313  * Fail if the vnode is currently writable.
314  */
315 int
316 vn_marktext(struct vnode *vp)
317 {
318 
319 	if ((vp->v_iflag & (VI_TEXT|VI_EXECMAP)) == (VI_TEXT|VI_EXECMAP)) {
320 		/* Safe unlocked, as long as caller holds a reference. */
321 		return (0);
322 	}
323 
324 	mutex_enter(vp->v_interlock);
325 	if (vp->v_writecount != 0) {
326 		KASSERT((vp->v_iflag & VI_TEXT) == 0);
327 		mutex_exit(vp->v_interlock);
328 		return (ETXTBSY);
329 	}
330 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
331 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
332 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
333 	}
334 	vp->v_iflag |= (VI_TEXT | VI_EXECMAP);
335 	mutex_exit(vp->v_interlock);
336 	return (0);
337 }
338 
339 /*
340  * Vnode close call
341  *
342  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
343  */
344 int
345 vn_close(struct vnode *vp, int flags, kauth_cred_t cred)
346 {
347 	int error;
348 
349 	if (flags & FWRITE) {
350 		mutex_enter(vp->v_interlock);
351 		KASSERT(vp->v_writecount > 0);
352 		vp->v_writecount--;
353 		mutex_exit(vp->v_interlock);
354 	}
355 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
356 	error = VOP_CLOSE(vp, flags, cred);
357 	vput(vp);
358 	return (error);
359 }
360 
361 static int
362 enforce_rlimit_fsize(struct vnode *vp, struct uio *uio, int ioflag)
363 {
364 	struct lwp *l = curlwp;
365 	off_t testoff;
366 
367 	if (uio->uio_rw != UIO_WRITE || vp->v_type != VREG)
368 		return 0;
369 
370 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
371 	if (ioflag & IO_APPEND)
372 		testoff = vp->v_size;
373 	else
374 		testoff = uio->uio_offset;
375 
376 	if (testoff + uio->uio_resid >
377 	    l->l_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
378 		mutex_enter(proc_lock);
379 		psignal(l->l_proc, SIGXFSZ);
380 		mutex_exit(proc_lock);
381 		return EFBIG;
382 	}
383 
384 	return 0;
385 }
386 
387 /*
388  * Package up an I/O request on a vnode into a uio and do it.
389  */
390 int
391 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset,
392     enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid,
393     struct lwp *l)
394 {
395 	struct uio auio;
396 	struct iovec aiov;
397 	int error;
398 
399 	if ((ioflg & IO_NODELOCKED) == 0) {
400 		if (rw == UIO_READ) {
401 			vn_lock(vp, LK_SHARED | LK_RETRY);
402 		} else /* UIO_WRITE */ {
403 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
404 		}
405 	}
406 	auio.uio_iov = &aiov;
407 	auio.uio_iovcnt = 1;
408 	aiov.iov_base = base;
409 	aiov.iov_len = len;
410 	auio.uio_resid = len;
411 	auio.uio_offset = offset;
412 	auio.uio_rw = rw;
413 	if (segflg == UIO_SYSSPACE) {
414 		UIO_SETUP_SYSSPACE(&auio);
415 	} else {
416 		auio.uio_vmspace = l->l_proc->p_vmspace;
417 	}
418 
419 	if ((error = enforce_rlimit_fsize(vp, &auio, ioflg)) != 0)
420 		goto out;
421 
422 	if (rw == UIO_READ) {
423 		error = VOP_READ(vp, &auio, ioflg, cred);
424 	} else {
425 		error = VOP_WRITE(vp, &auio, ioflg, cred);
426 	}
427 
428 	if (aresid)
429 		*aresid = auio.uio_resid;
430 	else
431 		if (auio.uio_resid && error == 0)
432 			error = EIO;
433 
434  out:
435 	if ((ioflg & IO_NODELOCKED) == 0) {
436 		VOP_UNLOCK(vp);
437 	}
438 	return (error);
439 }
440 
441 int
442 vn_readdir(file_t *fp, char *bf, int segflg, u_int count, int *done,
443     struct lwp *l, off_t **cookies, int *ncookies)
444 {
445 	struct vnode *vp = (struct vnode *)fp->f_data;
446 	struct iovec aiov;
447 	struct uio auio;
448 	int error, eofflag;
449 
450 	/* Limit the size on any kernel buffers used by VOP_READDIR */
451 	count = min(MAXBSIZE, count);
452 
453 unionread:
454 	if (vp->v_type != VDIR)
455 		return (EINVAL);
456 	aiov.iov_base = bf;
457 	aiov.iov_len = count;
458 	auio.uio_iov = &aiov;
459 	auio.uio_iovcnt = 1;
460 	auio.uio_rw = UIO_READ;
461 	if (segflg == UIO_SYSSPACE) {
462 		UIO_SETUP_SYSSPACE(&auio);
463 	} else {
464 		KASSERT(l == curlwp);
465 		auio.uio_vmspace = l->l_proc->p_vmspace;
466 	}
467 	auio.uio_resid = count;
468 	vn_lock(vp, LK_SHARED | LK_RETRY);
469 	auio.uio_offset = fp->f_offset;
470 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
471 		    ncookies);
472 	mutex_enter(&fp->f_lock);
473 	fp->f_offset = auio.uio_offset;
474 	mutex_exit(&fp->f_lock);
475 	VOP_UNLOCK(vp);
476 	if (error)
477 		return (error);
478 
479 	if (count == auio.uio_resid && vn_union_readdir_hook) {
480 		struct vnode *ovp = vp;
481 
482 		error = (*vn_union_readdir_hook)(&vp, fp, l);
483 		if (error)
484 			return (error);
485 		if (vp != ovp)
486 			goto unionread;
487 	}
488 
489 	if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) &&
490 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
491 		struct vnode *tvp = vp;
492 		vp = vp->v_mount->mnt_vnodecovered;
493 		vref(vp);
494 		mutex_enter(&fp->f_lock);
495 		fp->f_data = vp;
496 		fp->f_offset = 0;
497 		mutex_exit(&fp->f_lock);
498 		vrele(tvp);
499 		goto unionread;
500 	}
501 	*done = count - auio.uio_resid;
502 	return error;
503 }
504 
505 /*
506  * File table vnode read routine.
507  */
508 static int
509 vn_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
510     int flags)
511 {
512 	struct vnode *vp = (struct vnode *)fp->f_data;
513 	int count, error, ioflag, fflag;
514 
515 	ioflag = IO_ADV_ENCODE(fp->f_advice);
516 	fflag = fp->f_flag;
517 	if (fflag & FNONBLOCK)
518 		ioflag |= IO_NDELAY;
519 	if ((fflag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
520 		ioflag |= IO_SYNC;
521 	if (fflag & FALTIO)
522 		ioflag |= IO_ALTSEMANTICS;
523 	if (fflag & FDIRECT)
524 		ioflag |= IO_DIRECT;
525 	vn_lock(vp, LK_SHARED | LK_RETRY);
526 	uio->uio_offset = *offset;
527 	count = uio->uio_resid;
528 	error = VOP_READ(vp, uio, ioflag, cred);
529 	if (flags & FOF_UPDATE_OFFSET)
530 		*offset += count - uio->uio_resid;
531 	VOP_UNLOCK(vp);
532 	return (error);
533 }
534 
535 /*
536  * File table vnode write routine.
537  */
538 static int
539 vn_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
540     int flags)
541 {
542 	struct vnode *vp = (struct vnode *)fp->f_data;
543 	int count, error, ioflag, fflag;
544 
545 	ioflag = IO_ADV_ENCODE(fp->f_advice) | IO_UNIT;
546 	fflag = fp->f_flag;
547 	if (vp->v_type == VREG && (fflag & O_APPEND))
548 		ioflag |= IO_APPEND;
549 	if (fflag & FNONBLOCK)
550 		ioflag |= IO_NDELAY;
551 	if (fflag & FFSYNC ||
552 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
553 		ioflag |= IO_SYNC;
554 	else if (fflag & FDSYNC)
555 		ioflag |= IO_DSYNC;
556 	if (fflag & FALTIO)
557 		ioflag |= IO_ALTSEMANTICS;
558 	if (fflag & FDIRECT)
559 		ioflag |= IO_DIRECT;
560 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
561 	uio->uio_offset = *offset;
562 	count = uio->uio_resid;
563 
564 	if ((error = enforce_rlimit_fsize(vp, uio, ioflag)) != 0)
565 		goto out;
566 
567 	error = VOP_WRITE(vp, uio, ioflag, cred);
568 
569 	if (flags & FOF_UPDATE_OFFSET) {
570 		if (ioflag & IO_APPEND) {
571 			/*
572 			 * SUSv3 describes behaviour for count = 0 as following:
573 			 * "Before any action ... is taken, and if nbyte is zero
574 			 * and the file is a regular file, the write() function
575 			 * ... in the absence of errors ... shall return zero
576 			 * and have no other results."
577 			 */
578 			if (count)
579 				*offset = uio->uio_offset;
580 		} else
581 			*offset += count - uio->uio_resid;
582 	}
583 
584  out:
585 	VOP_UNLOCK(vp);
586 	return (error);
587 }
588 
589 /*
590  * File table vnode stat routine.
591  */
592 static int
593 vn_statfile(file_t *fp, struct stat *sb)
594 {
595 	struct vnode *vp = fp->f_data;
596 	int error;
597 
598 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
599 	error = vn_stat(vp, sb);
600 	VOP_UNLOCK(vp);
601 	return error;
602 }
603 
604 int
605 vn_stat(struct vnode *vp, struct stat *sb)
606 {
607 	struct vattr va;
608 	int error;
609 	mode_t mode;
610 
611 	memset(&va, 0, sizeof(va));
612 	error = VOP_GETATTR(vp, &va, kauth_cred_get());
613 	if (error)
614 		return (error);
615 	/*
616 	 * Copy from vattr table
617 	 */
618 	memset(sb, 0, sizeof(*sb));
619 	sb->st_dev = va.va_fsid;
620 	sb->st_ino = va.va_fileid;
621 	mode = va.va_mode;
622 	switch (vp->v_type) {
623 	case VREG:
624 		mode |= S_IFREG;
625 		break;
626 	case VDIR:
627 		mode |= S_IFDIR;
628 		break;
629 	case VBLK:
630 		mode |= S_IFBLK;
631 		break;
632 	case VCHR:
633 		mode |= S_IFCHR;
634 		break;
635 	case VLNK:
636 		mode |= S_IFLNK;
637 		break;
638 	case VSOCK:
639 		mode |= S_IFSOCK;
640 		break;
641 	case VFIFO:
642 		mode |= S_IFIFO;
643 		break;
644 	default:
645 		return (EBADF);
646 	};
647 	sb->st_mode = mode;
648 	sb->st_nlink = va.va_nlink;
649 	sb->st_uid = va.va_uid;
650 	sb->st_gid = va.va_gid;
651 	sb->st_rdev = va.va_rdev;
652 	sb->st_size = va.va_size;
653 	sb->st_atimespec = va.va_atime;
654 	sb->st_mtimespec = va.va_mtime;
655 	sb->st_ctimespec = va.va_ctime;
656 	sb->st_birthtimespec = va.va_birthtime;
657 	sb->st_blksize = va.va_blocksize;
658 	sb->st_flags = va.va_flags;
659 	sb->st_gen = 0;
660 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
661 	return (0);
662 }
663 
664 /*
665  * File table vnode fcntl routine.
666  */
667 static int
668 vn_fcntl(file_t *fp, u_int com, void *data)
669 {
670 	struct vnode *vp = fp->f_data;
671 	int error;
672 
673 	error = VOP_FCNTL(vp, com, data, fp->f_flag, kauth_cred_get());
674 	return (error);
675 }
676 
677 /*
678  * File table vnode ioctl routine.
679  */
680 static int
681 vn_ioctl(file_t *fp, u_long com, void *data)
682 {
683 	struct vnode *vp = fp->f_data, *ovp;
684 	struct vattr vattr;
685 	int error;
686 
687 	switch (vp->v_type) {
688 
689 	case VREG:
690 	case VDIR:
691 		if (com == FIONREAD) {
692 			vn_lock(vp, LK_SHARED | LK_RETRY);
693 			error = VOP_GETATTR(vp, &vattr, kauth_cred_get());
694 			VOP_UNLOCK(vp);
695 			if (error)
696 				return (error);
697 			*(int *)data = vattr.va_size - fp->f_offset;
698 			return (0);
699 		}
700 		if ((com == FIONWRITE) || (com == FIONSPACE)) {
701 			/*
702 			 * Files don't have send queues, so there never
703 			 * are any bytes in them, nor is there any
704 			 * open space in them.
705 			 */
706 			*(int *)data = 0;
707 			return (0);
708 		}
709 		if (com == FIOGETBMAP) {
710 			daddr_t *block;
711 
712 			if (*(daddr_t *)data < 0)
713 				return (EINVAL);
714 			block = (daddr_t *)data;
715 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
716 		}
717 		if (com == OFIOGETBMAP) {
718 			daddr_t ibn, obn;
719 
720 			if (*(int32_t *)data < 0)
721 				return (EINVAL);
722 			ibn = (daddr_t)*(int32_t *)data;
723 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
724 			*(int32_t *)data = (int32_t)obn;
725 			return error;
726 		}
727 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
728 			return (0);			/* XXX */
729 		/* fall into ... */
730 	case VFIFO:
731 	case VCHR:
732 	case VBLK:
733 		error = VOP_IOCTL(vp, com, data, fp->f_flag,
734 		    kauth_cred_get());
735 		if (error == 0 && com == TIOCSCTTY) {
736 			vref(vp);
737 			mutex_enter(proc_lock);
738 			ovp = curproc->p_session->s_ttyvp;
739 			curproc->p_session->s_ttyvp = vp;
740 			mutex_exit(proc_lock);
741 			if (ovp != NULL)
742 				vrele(ovp);
743 		}
744 		return (error);
745 
746 	default:
747 		return (EPASSTHROUGH);
748 	}
749 }
750 
751 /*
752  * File table vnode poll routine.
753  */
754 static int
755 vn_poll(file_t *fp, int events)
756 {
757 
758 	return (VOP_POLL(fp->f_data, events));
759 }
760 
761 /*
762  * File table vnode kqfilter routine.
763  */
764 int
765 vn_kqfilter(file_t *fp, struct knote *kn)
766 {
767 
768 	return (VOP_KQFILTER(fp->f_data, kn));
769 }
770 
771 /*
772  * Check that the vnode is still valid, and if so
773  * acquire requested lock.
774  */
775 int
776 vn_lock(struct vnode *vp, int flags)
777 {
778 	int error;
779 
780 #if 0
781 	KASSERT(vp->v_usecount > 0 || (vp->v_iflag & VI_ONWORKLST) != 0);
782 #endif
783 	KASSERT((flags & ~(LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT|LK_RETRY)) == 0);
784 	KASSERT(!mutex_owned(vp->v_interlock));
785 
786 #ifdef DIAGNOSTIC
787 	if (wapbl_vphaswapbl(vp))
788 		WAPBL_JUNLOCK_ASSERT(wapbl_vptomp(vp));
789 #endif
790 
791 	do {
792 		/*
793 		 * XXX PR 37706 forced unmount of file systems is unsafe.
794 		 * Race between vclean() and this the remaining problem.
795 		 */
796 		mutex_enter(vp->v_interlock);
797 		if (vp->v_iflag & VI_XLOCK) {
798 			if (flags & LK_NOWAIT) {
799 				mutex_exit(vp->v_interlock);
800 				return EBUSY;
801 			}
802 			vwait(vp, VI_XLOCK);
803 			mutex_exit(vp->v_interlock);
804 			error = ENOENT;
805 		} else {
806 			mutex_exit(vp->v_interlock);
807 			error = VOP_LOCK(vp, (flags & ~LK_RETRY));
808 			if (error == 0 || error == EDEADLK || error == EBUSY)
809 				return (error);
810 		}
811 	} while (flags & LK_RETRY);
812 	return (error);
813 }
814 
815 /*
816  * File table vnode close routine.
817  */
818 static int
819 vn_closefile(file_t *fp)
820 {
821 
822 	return vn_close(fp->f_data, fp->f_flag, fp->f_cred);
823 }
824 
825 /*
826  * Simplified in-kernel wrapper calls for extended attribute access.
827  * Both calls pass in a NULL credential, authorizing a "kernel" access.
828  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
829  */
830 int
831 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
832     const char *attrname, size_t *buflen, void *bf, struct lwp *l)
833 {
834 	struct uio auio;
835 	struct iovec aiov;
836 	int error;
837 
838 	aiov.iov_len = *buflen;
839 	aiov.iov_base = bf;
840 
841 	auio.uio_iov = &aiov;
842 	auio.uio_iovcnt = 1;
843 	auio.uio_rw = UIO_READ;
844 	auio.uio_offset = 0;
845 	auio.uio_resid = *buflen;
846 	UIO_SETUP_SYSSPACE(&auio);
847 
848 	if ((ioflg & IO_NODELOCKED) == 0)
849 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
850 
851 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL);
852 
853 	if ((ioflg & IO_NODELOCKED) == 0)
854 		VOP_UNLOCK(vp);
855 
856 	if (error == 0)
857 		*buflen = *buflen - auio.uio_resid;
858 
859 	return (error);
860 }
861 
862 /*
863  * XXX Failure mode if partially written?
864  */
865 int
866 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
867     const char *attrname, size_t buflen, const void *bf, struct lwp *l)
868 {
869 	struct uio auio;
870 	struct iovec aiov;
871 	int error;
872 
873 	aiov.iov_len = buflen;
874 	aiov.iov_base = __UNCONST(bf);		/* XXXUNCONST kills const */
875 
876 	auio.uio_iov = &aiov;
877 	auio.uio_iovcnt = 1;
878 	auio.uio_rw = UIO_WRITE;
879 	auio.uio_offset = 0;
880 	auio.uio_resid = buflen;
881 	UIO_SETUP_SYSSPACE(&auio);
882 
883 	if ((ioflg & IO_NODELOCKED) == 0) {
884 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
885 	}
886 
887 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL);
888 
889 	if ((ioflg & IO_NODELOCKED) == 0) {
890 		VOP_UNLOCK(vp);
891 	}
892 
893 	return (error);
894 }
895 
896 int
897 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
898     const char *attrname, struct lwp *l)
899 {
900 	int error;
901 
902 	if ((ioflg & IO_NODELOCKED) == 0) {
903 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
904 	}
905 
906 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL);
907 	if (error == EOPNOTSUPP)
908 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL);
909 
910 	if ((ioflg & IO_NODELOCKED) == 0) {
911 		VOP_UNLOCK(vp);
912 	}
913 
914 	return (error);
915 }
916 
917 void
918 vn_ra_allocctx(struct vnode *vp)
919 {
920 	struct uvm_ractx *ra = NULL;
921 
922 	KASSERT(mutex_owned(vp->v_interlock));
923 
924 	if (vp->v_type != VREG) {
925 		return;
926 	}
927 	if (vp->v_ractx != NULL) {
928 		return;
929 	}
930 	if (vp->v_ractx == NULL) {
931 		mutex_exit(vp->v_interlock);
932 		ra = uvm_ra_allocctx();
933 		mutex_enter(vp->v_interlock);
934 		if (ra != NULL && vp->v_ractx == NULL) {
935 			vp->v_ractx = ra;
936 			ra = NULL;
937 		}
938 	}
939 	if (ra != NULL) {
940 		uvm_ra_freectx(ra);
941 	}
942 }
943 
944 int
945 vn_fifo_bypass(void *v)
946 {
947 	struct vop_generic_args *ap = v;
948 
949 	return VOCALL(fifo_vnodeop_p, ap->a_desc->vdesc_offset, v);
950 }
951