xref: /netbsd-src/sys/kern/vfs_vnops.c (revision b62fc9e20372b08e1785ff6d769312d209fa2005)
1 /*	$NetBSD: vfs_vnops.c,v 1.170 2010/03/29 13:11:32 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 1.170 2010/03/29 13:11:32 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 #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 	char *path;
142 
143 	ndp->ni_cnd.cn_flags &= TRYEMULROOT | NOCHROOT;
144 
145 	if (fmode & O_CREAT) {
146 		ndp->ni_cnd.cn_nameiop = CREATE;
147 		ndp->ni_cnd.cn_flags |= LOCKPARENT | LOCKLEAF;
148 		if ((fmode & O_EXCL) == 0 &&
149 		    ((fmode & O_NOFOLLOW) == 0))
150 			ndp->ni_cnd.cn_flags |= FOLLOW;
151 	} else {
152 		ndp->ni_cnd.cn_nameiop = LOOKUP;
153 		ndp->ni_cnd.cn_flags |= LOCKLEAF;
154 		if ((fmode & O_NOFOLLOW) == 0)
155 			ndp->ni_cnd.cn_flags |= FOLLOW;
156 	}
157 
158 	VERIEXEC_PATH_GET(ndp->ni_dirp, ndp->ni_segflg, ndp->ni_dirp, path);
159 
160 	error = namei(ndp);
161 	if (error)
162 		goto out;
163 
164 	vp = ndp->ni_vp;
165 
166 #if NVERIEXEC > 0
167 	error = veriexec_openchk(l, ndp->ni_vp, ndp->ni_dirp, fmode);
168 	if (error)
169 		goto bad;
170 #endif /* NVERIEXEC > 0 */
171 
172 	if (fmode & O_CREAT) {
173 		if (ndp->ni_vp == NULL) {
174 			vattr_null(&va);
175 			va.va_type = VREG;
176 			va.va_mode = cmode;
177 			if (fmode & O_EXCL)
178 				 va.va_vaflags |= VA_EXCLUSIVE;
179 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
180 					   &ndp->ni_cnd, &va);
181 			if (error)
182 				goto out;
183 			fmode &= ~O_TRUNC;
184 			vp = ndp->ni_vp;
185 		} else {
186 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
187 			if (ndp->ni_dvp == ndp->ni_vp)
188 				vrele(ndp->ni_dvp);
189 			else
190 				vput(ndp->ni_dvp);
191 			ndp->ni_dvp = NULL;
192 			vp = ndp->ni_vp;
193 			if (fmode & O_EXCL) {
194 				error = EEXIST;
195 				goto bad;
196 			}
197 			fmode &= ~O_CREAT;
198 		}
199 	} else {
200 		vp = ndp->ni_vp;
201 	}
202 	if (vp->v_type == VSOCK) {
203 		error = EOPNOTSUPP;
204 		goto bad;
205 	}
206 	if (ndp->ni_vp->v_type == VLNK) {
207 		error = EFTYPE;
208 		goto bad;
209 	}
210 
211 	if ((fmode & O_CREAT) == 0) {
212 		error = vn_openchk(vp, cred, fmode);
213 		if (error != 0)
214 			goto bad;
215 	}
216 
217 	if (fmode & O_TRUNC) {
218 		VOP_UNLOCK(vp, 0);			/* XXX */
219 
220 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
221 		vattr_null(&va);
222 		va.va_size = 0;
223 		error = VOP_SETATTR(vp, &va, cred);
224 		if (error != 0)
225 			goto bad;
226 	}
227 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0)
228 		goto bad;
229 	if (fmode & FWRITE) {
230 		mutex_enter(&vp->v_interlock);
231 		vp->v_writecount++;
232 		mutex_exit(&vp->v_interlock);
233 	}
234 
235 bad:
236 	if (error)
237 		vput(vp);
238 out:
239 	VERIEXEC_PATH_PUT(path);
240 	return (error);
241 }
242 
243 /*
244  * Check for write permissions on the specified vnode.
245  * Prototype text segments cannot be written.
246  */
247 int
248 vn_writechk(struct vnode *vp)
249 {
250 
251 	/*
252 	 * If the vnode is in use as a process's text,
253 	 * we can't allow writing.
254 	 */
255 	if (vp->v_iflag & VI_TEXT)
256 		return (ETXTBSY);
257 	return (0);
258 }
259 
260 int
261 vn_openchk(struct vnode *vp, kauth_cred_t cred, int fflags)
262 {
263 	int permbits = 0;
264 	int error;
265 
266 	if ((fflags & FREAD) != 0) {
267 		permbits = VREAD;
268 	}
269 	if ((fflags & (FWRITE | O_TRUNC)) != 0) {
270 		permbits |= VWRITE;
271 		if (vp->v_type == VDIR) {
272 			error = EISDIR;
273 			goto bad;
274 		}
275 		error = vn_writechk(vp);
276 		if (error != 0)
277 			goto bad;
278 	}
279 	error = VOP_ACCESS(vp, permbits, cred);
280 bad:
281 	return error;
282 }
283 
284 /*
285  * Mark a vnode as having executable mappings.
286  */
287 void
288 vn_markexec(struct vnode *vp)
289 {
290 
291 	if ((vp->v_iflag & VI_EXECMAP) != 0) {
292 		/* Safe unlocked, as long as caller holds a reference. */
293 		return;
294 	}
295 
296 	mutex_enter(&vp->v_interlock);
297 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
298 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
299 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
300 		vp->v_iflag |= VI_EXECMAP;
301 	}
302 	mutex_exit(&vp->v_interlock);
303 }
304 
305 /*
306  * Mark a vnode as being the text of a process.
307  * Fail if the vnode is currently writable.
308  */
309 int
310 vn_marktext(struct vnode *vp)
311 {
312 
313 	if ((vp->v_iflag & (VI_TEXT|VI_EXECMAP)) == (VI_TEXT|VI_EXECMAP)) {
314 		/* Safe unlocked, as long as caller holds a reference. */
315 		return (0);
316 	}
317 
318 	mutex_enter(&vp->v_interlock);
319 	if (vp->v_writecount != 0) {
320 		KASSERT((vp->v_iflag & VI_TEXT) == 0);
321 		mutex_exit(&vp->v_interlock);
322 		return (ETXTBSY);
323 	}
324 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
325 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
326 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
327 	}
328 	vp->v_iflag |= (VI_TEXT | VI_EXECMAP);
329 	mutex_exit(&vp->v_interlock);
330 	return (0);
331 }
332 
333 /*
334  * Vnode close call
335  *
336  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
337  */
338 int
339 vn_close(struct vnode *vp, int flags, kauth_cred_t cred)
340 {
341 	int error;
342 
343 	if (flags & FWRITE) {
344 		mutex_enter(&vp->v_interlock);
345 		vp->v_writecount--;
346 		mutex_exit(&vp->v_interlock);
347 	}
348 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
349 	error = VOP_CLOSE(vp, flags, cred);
350 	vput(vp);
351 	return (error);
352 }
353 
354 /*
355  * Package up an I/O request on a vnode into a uio and do it.
356  */
357 int
358 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset,
359     enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid,
360     struct lwp *l)
361 {
362 	struct uio auio;
363 	struct iovec aiov;
364 	int error;
365 
366 	if ((ioflg & IO_NODELOCKED) == 0) {
367 		if (rw == UIO_READ) {
368 			vn_lock(vp, LK_SHARED | LK_RETRY);
369 		} else /* UIO_WRITE */ {
370 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
371 		}
372 	}
373 	auio.uio_iov = &aiov;
374 	auio.uio_iovcnt = 1;
375 	aiov.iov_base = base;
376 	aiov.iov_len = len;
377 	auio.uio_resid = len;
378 	auio.uio_offset = offset;
379 	auio.uio_rw = rw;
380 	if (segflg == UIO_SYSSPACE) {
381 		UIO_SETUP_SYSSPACE(&auio);
382 	} else {
383 		auio.uio_vmspace = l->l_proc->p_vmspace;
384 	}
385 	if (rw == UIO_READ) {
386 		error = VOP_READ(vp, &auio, ioflg, cred);
387 	} else {
388 		error = VOP_WRITE(vp, &auio, ioflg, cred);
389 	}
390 	if (aresid)
391 		*aresid = auio.uio_resid;
392 	else
393 		if (auio.uio_resid && error == 0)
394 			error = EIO;
395 	if ((ioflg & IO_NODELOCKED) == 0) {
396 		VOP_UNLOCK(vp, 0);
397 	}
398 	return (error);
399 }
400 
401 int
402 vn_readdir(file_t *fp, char *bf, int segflg, u_int count, int *done,
403     struct lwp *l, off_t **cookies, int *ncookies)
404 {
405 	struct vnode *vp = (struct vnode *)fp->f_data;
406 	struct iovec aiov;
407 	struct uio auio;
408 	int error, eofflag;
409 
410 	/* Limit the size on any kernel buffers used by VOP_READDIR */
411 	count = min(MAXBSIZE, count);
412 
413 unionread:
414 	if (vp->v_type != VDIR)
415 		return (EINVAL);
416 	aiov.iov_base = bf;
417 	aiov.iov_len = count;
418 	auio.uio_iov = &aiov;
419 	auio.uio_iovcnt = 1;
420 	auio.uio_rw = UIO_READ;
421 	if (segflg == UIO_SYSSPACE) {
422 		UIO_SETUP_SYSSPACE(&auio);
423 	} else {
424 		KASSERT(l == curlwp);
425 		auio.uio_vmspace = l->l_proc->p_vmspace;
426 	}
427 	auio.uio_resid = count;
428 	vn_lock(vp, LK_SHARED | LK_RETRY);
429 	auio.uio_offset = fp->f_offset;
430 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
431 		    ncookies);
432 	mutex_enter(&fp->f_lock);
433 	fp->f_offset = auio.uio_offset;
434 	mutex_exit(&fp->f_lock);
435 	VOP_UNLOCK(vp, 0);
436 	if (error)
437 		return (error);
438 
439 	if (count == auio.uio_resid && vn_union_readdir_hook) {
440 		struct vnode *ovp = vp;
441 
442 		error = (*vn_union_readdir_hook)(&vp, fp, l);
443 		if (error)
444 			return (error);
445 		if (vp != ovp)
446 			goto unionread;
447 	}
448 
449 	if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) &&
450 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
451 		struct vnode *tvp = vp;
452 		vp = vp->v_mount->mnt_vnodecovered;
453 		vref(vp);
454 		mutex_enter(&fp->f_lock);
455 		fp->f_data = vp;
456 		fp->f_offset = 0;
457 		mutex_exit(&fp->f_lock);
458 		vrele(tvp);
459 		goto unionread;
460 	}
461 	*done = count - auio.uio_resid;
462 	return error;
463 }
464 
465 /*
466  * File table vnode read routine.
467  */
468 static int
469 vn_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
470     int flags)
471 {
472 	struct vnode *vp = (struct vnode *)fp->f_data;
473 	int count, error, ioflag, fflag;
474 
475 	ioflag = IO_ADV_ENCODE(fp->f_advice);
476 	fflag = fp->f_flag;
477 	if (fflag & FNONBLOCK)
478 		ioflag |= IO_NDELAY;
479 	if ((fflag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
480 		ioflag |= IO_SYNC;
481 	if (fflag & FALTIO)
482 		ioflag |= IO_ALTSEMANTICS;
483 	if (fflag & FDIRECT)
484 		ioflag |= IO_DIRECT;
485 	vn_lock(vp, LK_SHARED | LK_RETRY);
486 	uio->uio_offset = *offset;
487 	count = uio->uio_resid;
488 	error = VOP_READ(vp, uio, ioflag, cred);
489 	if (flags & FOF_UPDATE_OFFSET)
490 		*offset += count - uio->uio_resid;
491 	VOP_UNLOCK(vp, 0);
492 	return (error);
493 }
494 
495 /*
496  * File table vnode write routine.
497  */
498 static int
499 vn_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
500     int flags)
501 {
502 	struct vnode *vp = (struct vnode *)fp->f_data;
503 	int count, error, ioflag, fflag;
504 
505 	ioflag = IO_ADV_ENCODE(fp->f_advice) | IO_UNIT;
506 	fflag = fp->f_flag;
507 	if (vp->v_type == VREG && (fflag & O_APPEND))
508 		ioflag |= IO_APPEND;
509 	if (fflag & FNONBLOCK)
510 		ioflag |= IO_NDELAY;
511 	if (fflag & FFSYNC ||
512 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
513 		ioflag |= IO_SYNC;
514 	else if (fflag & FDSYNC)
515 		ioflag |= IO_DSYNC;
516 	if (fflag & FALTIO)
517 		ioflag |= IO_ALTSEMANTICS;
518 	if (fflag & FDIRECT)
519 		ioflag |= IO_DIRECT;
520 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
521 	uio->uio_offset = *offset;
522 	count = uio->uio_resid;
523 	error = VOP_WRITE(vp, uio, ioflag, cred);
524 	if (flags & FOF_UPDATE_OFFSET) {
525 		if (ioflag & IO_APPEND) {
526 			/*
527 			 * SUSv3 describes behaviour for count = 0 as following:
528 			 * "Before any action ... is taken, and if nbyte is zero
529 			 * and the file is a regular file, the write() function
530 			 * ... in the absence of errors ... shall return zero
531 			 * and have no other results."
532 			 */
533 			if (count)
534 				*offset = uio->uio_offset;
535 		} else
536 			*offset += count - uio->uio_resid;
537 	}
538 	VOP_UNLOCK(vp, 0);
539 	return (error);
540 }
541 
542 /*
543  * File table vnode stat routine.
544  */
545 static int
546 vn_statfile(file_t *fp, struct stat *sb)
547 {
548 	struct vnode *vp = fp->f_data;
549 	int error;
550 
551 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
552 	error = vn_stat(vp, sb);
553 	VOP_UNLOCK(vp, 0);
554 	return error;
555 }
556 
557 int
558 vn_stat(struct vnode *vp, struct stat *sb)
559 {
560 	struct vattr va;
561 	int error;
562 	mode_t mode;
563 
564 	error = VOP_GETATTR(vp, &va, kauth_cred_get());
565 	if (error)
566 		return (error);
567 	/*
568 	 * Copy from vattr table
569 	 */
570 	sb->st_dev = va.va_fsid;
571 	sb->st_ino = va.va_fileid;
572 	mode = va.va_mode;
573 	switch (vp->v_type) {
574 	case VREG:
575 		mode |= S_IFREG;
576 		break;
577 	case VDIR:
578 		mode |= S_IFDIR;
579 		break;
580 	case VBLK:
581 		mode |= S_IFBLK;
582 		break;
583 	case VCHR:
584 		mode |= S_IFCHR;
585 		break;
586 	case VLNK:
587 		mode |= S_IFLNK;
588 		break;
589 	case VSOCK:
590 		mode |= S_IFSOCK;
591 		break;
592 	case VFIFO:
593 		mode |= S_IFIFO;
594 		break;
595 	default:
596 		return (EBADF);
597 	};
598 	sb->st_mode = mode;
599 	sb->st_nlink = va.va_nlink;
600 	sb->st_uid = va.va_uid;
601 	sb->st_gid = va.va_gid;
602 	sb->st_rdev = va.va_rdev;
603 	sb->st_size = va.va_size;
604 	sb->st_atimespec = va.va_atime;
605 	sb->st_mtimespec = va.va_mtime;
606 	sb->st_ctimespec = va.va_ctime;
607 	sb->st_birthtimespec = va.va_birthtime;
608 	sb->st_blksize = va.va_blocksize;
609 	sb->st_flags = va.va_flags;
610 	sb->st_gen = 0;
611 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
612 	return (0);
613 }
614 
615 /*
616  * File table vnode fcntl routine.
617  */
618 static int
619 vn_fcntl(file_t *fp, u_int com, void *data)
620 {
621 	struct vnode *vp = fp->f_data;
622 	int error;
623 
624 	error = VOP_FCNTL(vp, com, data, fp->f_flag, kauth_cred_get());
625 	return (error);
626 }
627 
628 /*
629  * File table vnode ioctl routine.
630  */
631 static int
632 vn_ioctl(file_t *fp, u_long com, void *data)
633 {
634 	struct vnode *vp = fp->f_data, *ovp;
635 	struct vattr vattr;
636 	int error;
637 
638 	switch (vp->v_type) {
639 
640 	case VREG:
641 	case VDIR:
642 		if (com == FIONREAD) {
643 			error = VOP_GETATTR(vp, &vattr,
644 			    kauth_cred_get());
645 			if (error)
646 				return (error);
647 			*(int *)data = vattr.va_size - fp->f_offset;
648 			return (0);
649 		}
650 		if ((com == FIONWRITE) || (com == FIONSPACE)) {
651 			/*
652 			 * Files don't have send queues, so there never
653 			 * are any bytes in them, nor is there any
654 			 * open space in them.
655 			 */
656 			*(int *)data = 0;
657 			return (0);
658 		}
659 		if (com == FIOGETBMAP) {
660 			daddr_t *block;
661 
662 			if (*(daddr_t *)data < 0)
663 				return (EINVAL);
664 			block = (daddr_t *)data;
665 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
666 		}
667 		if (com == OFIOGETBMAP) {
668 			daddr_t ibn, obn;
669 
670 			if (*(int32_t *)data < 0)
671 				return (EINVAL);
672 			ibn = (daddr_t)*(int32_t *)data;
673 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
674 			*(int32_t *)data = (int32_t)obn;
675 			return error;
676 		}
677 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
678 			return (0);			/* XXX */
679 		/* fall into ... */
680 	case VFIFO:
681 	case VCHR:
682 	case VBLK:
683 		error = VOP_IOCTL(vp, com, data, fp->f_flag,
684 		    kauth_cred_get());
685 		if (error == 0 && com == TIOCSCTTY) {
686 			vref(vp);
687 			mutex_enter(proc_lock);
688 			ovp = curproc->p_session->s_ttyvp;
689 			curproc->p_session->s_ttyvp = vp;
690 			mutex_exit(proc_lock);
691 			if (ovp != NULL)
692 				vrele(ovp);
693 		}
694 		return (error);
695 
696 	default:
697 		return (EPASSTHROUGH);
698 	}
699 }
700 
701 /*
702  * File table vnode poll routine.
703  */
704 static int
705 vn_poll(file_t *fp, int events)
706 {
707 
708 	return (VOP_POLL(fp->f_data, events));
709 }
710 
711 /*
712  * File table vnode kqfilter routine.
713  */
714 int
715 vn_kqfilter(file_t *fp, struct knote *kn)
716 {
717 
718 	return (VOP_KQFILTER(fp->f_data, kn));
719 }
720 
721 /*
722  * Check that the vnode is still valid, and if so
723  * acquire requested lock.
724  */
725 int
726 vn_lock(struct vnode *vp, int flags)
727 {
728 	int error;
729 
730 #if 0
731 	KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0
732 	    || (vp->v_iflag & VI_ONWORKLST) != 0);
733 #endif
734 	KASSERT((flags &
735 	    ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT|LK_RETRY|
736 	    LK_CANRECURSE))
737 	    == 0);
738 
739 #ifdef DIAGNOSTIC
740 	if (wapbl_vphaswapbl(vp))
741 		WAPBL_JUNLOCK_ASSERT(wapbl_vptomp(vp));
742 #endif
743 
744 	do {
745 		/*
746 		 * XXX PR 37706 forced unmount of file systems is unsafe.
747 		 * Race between vclean() and this the remaining problem.
748 		 */
749 		if (vp->v_iflag & VI_XLOCK) {
750 			if ((flags & LK_INTERLOCK) == 0) {
751 				mutex_enter(&vp->v_interlock);
752 			}
753 			flags &= ~LK_INTERLOCK;
754 			if (flags & LK_NOWAIT) {
755 				mutex_exit(&vp->v_interlock);
756 				return EBUSY;
757 			}
758 			vwait(vp, VI_XLOCK);
759 			mutex_exit(&vp->v_interlock);
760 			error = ENOENT;
761 		} else {
762 			if ((flags & LK_INTERLOCK) != 0) {
763 				mutex_exit(&vp->v_interlock);
764 			}
765 			flags &= ~LK_INTERLOCK;
766 			error = VOP_LOCK(vp, (flags & ~LK_RETRY));
767 			if (error == 0 || error == EDEADLK || error == EBUSY)
768 				return (error);
769 		}
770 	} while (flags & LK_RETRY);
771 	return (error);
772 }
773 
774 /*
775  * File table vnode close routine.
776  */
777 static int
778 vn_closefile(file_t *fp)
779 {
780 
781 	return vn_close(fp->f_data, fp->f_flag, fp->f_cred);
782 }
783 
784 /*
785  * Enable LK_CANRECURSE on lock. Return prior status.
786  */
787 u_int
788 vn_setrecurse(struct vnode *vp)
789 {
790 	struct vnlock *lkp;
791 
792 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
793 	atomic_inc_uint(&lkp->vl_canrecurse);
794 
795 	return 0;
796 }
797 
798 /*
799  * Called when done with locksetrecurse.
800  */
801 void
802 vn_restorerecurse(struct vnode *vp, u_int flags)
803 {
804 	struct vnlock *lkp;
805 
806 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
807 	atomic_dec_uint(&lkp->vl_canrecurse);
808 }
809 
810 /*
811  * Simplified in-kernel wrapper calls for extended attribute access.
812  * Both calls pass in a NULL credential, authorizing a "kernel" access.
813  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
814  */
815 int
816 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
817     const char *attrname, size_t *buflen, void *bf, struct lwp *l)
818 {
819 	struct uio auio;
820 	struct iovec aiov;
821 	int error;
822 
823 	aiov.iov_len = *buflen;
824 	aiov.iov_base = bf;
825 
826 	auio.uio_iov = &aiov;
827 	auio.uio_iovcnt = 1;
828 	auio.uio_rw = UIO_READ;
829 	auio.uio_offset = 0;
830 	auio.uio_resid = *buflen;
831 	UIO_SETUP_SYSSPACE(&auio);
832 
833 	if ((ioflg & IO_NODELOCKED) == 0)
834 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
835 
836 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL);
837 
838 	if ((ioflg & IO_NODELOCKED) == 0)
839 		VOP_UNLOCK(vp, 0);
840 
841 	if (error == 0)
842 		*buflen = *buflen - auio.uio_resid;
843 
844 	return (error);
845 }
846 
847 /*
848  * XXX Failure mode if partially written?
849  */
850 int
851 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
852     const char *attrname, size_t buflen, const void *bf, struct lwp *l)
853 {
854 	struct uio auio;
855 	struct iovec aiov;
856 	int error;
857 
858 	aiov.iov_len = buflen;
859 	aiov.iov_base = __UNCONST(bf);		/* XXXUNCONST kills const */
860 
861 	auio.uio_iov = &aiov;
862 	auio.uio_iovcnt = 1;
863 	auio.uio_rw = UIO_WRITE;
864 	auio.uio_offset = 0;
865 	auio.uio_resid = buflen;
866 	UIO_SETUP_SYSSPACE(&auio);
867 
868 	if ((ioflg & IO_NODELOCKED) == 0) {
869 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
870 	}
871 
872 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL);
873 
874 	if ((ioflg & IO_NODELOCKED) == 0) {
875 		VOP_UNLOCK(vp, 0);
876 	}
877 
878 	return (error);
879 }
880 
881 int
882 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
883     const char *attrname, struct lwp *l)
884 {
885 	int error;
886 
887 	if ((ioflg & IO_NODELOCKED) == 0) {
888 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
889 	}
890 
891 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL);
892 	if (error == EOPNOTSUPP)
893 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL);
894 
895 	if ((ioflg & IO_NODELOCKED) == 0) {
896 		VOP_UNLOCK(vp, 0);
897 	}
898 
899 	return (error);
900 }
901 
902 void
903 vn_ra_allocctx(struct vnode *vp)
904 {
905 	struct uvm_ractx *ra = NULL;
906 
907 	KASSERT(mutex_owned(&vp->v_interlock));
908 
909 	if (vp->v_type != VREG) {
910 		return;
911 	}
912 	if (vp->v_ractx != NULL) {
913 		return;
914 	}
915 	if (vp->v_ractx == NULL) {
916 		mutex_exit(&vp->v_interlock);
917 		ra = uvm_ra_allocctx();
918 		mutex_enter(&vp->v_interlock);
919 		if (ra != NULL && vp->v_ractx == NULL) {
920 			vp->v_ractx = ra;
921 			ra = NULL;
922 		}
923 	}
924 	if (ra != NULL) {
925 		uvm_ra_freectx(ra);
926 	}
927 }
928 
929 int
930 vn_fifo_bypass(void *v)
931 {
932 	struct vop_generic_args *ap = v;
933 
934 	return VOCALL(fifo_vnodeop_p, ap->a_desc->vdesc_offset, v);
935 }
936