xref: /netbsd-src/sys/kern/vfs_vnops.c (revision cd22f25e6f6d1cc1f197fe8c5468a80f51d1c4e1)
1 /*	$NetBSD: vfs_vnops.c,v 1.156 2008/04/24 15:35:30 ad Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_vnops.c	8.14 (Berkeley) 6/15/95
37  */
38 
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.156 2008/04/24 15:35:30 ad Exp $");
41 
42 #include "fs_union.h"
43 #include "veriexec.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/kernel.h>
48 #include <sys/file.h>
49 #include <sys/stat.h>
50 #include <sys/buf.h>
51 #include <sys/proc.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/namei.h>
55 #include <sys/vnode.h>
56 #include <sys/ioctl.h>
57 #include <sys/tty.h>
58 #include <sys/poll.h>
59 #include <sys/kauth.h>
60 #include <sys/syslog.h>
61 #include <sys/fstrans.h>
62 #include <sys/atomic.h>
63 #include <sys/filedesc.h>
64 
65 #include <miscfs/specfs/specdev.h>
66 
67 #include <uvm/uvm_extern.h>
68 #include <uvm/uvm_readahead.h>
69 
70 #ifdef UNION
71 #include <fs/union/union.h>
72 #endif
73 
74 #if defined(LKM) || defined(UNION)
75 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct lwp *);
76 #endif
77 
78 #include <sys/verified_exec.h>
79 
80 static int vn_read(file_t *fp, off_t *offset, struct uio *uio,
81 	    kauth_cred_t cred, int flags);
82 static int vn_write(file_t *fp, off_t *offset, struct uio *uio,
83 	    kauth_cred_t cred, int flags);
84 static int vn_closefile(file_t *fp);
85 static int vn_poll(file_t *fp, int events);
86 static int vn_fcntl(file_t *fp, u_int com, void *data);
87 static int vn_statfile(file_t *fp, struct stat *sb);
88 static int vn_ioctl(file_t *fp, u_long com, void *data);
89 
90 const struct fileops vnops = {
91 	vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll,
92 	vn_statfile, vn_closefile, vn_kqfilter
93 };
94 
95 /*
96  * Common code for vnode open operations.
97  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
98  */
99 int
100 vn_open(struct nameidata *ndp, int fmode, int cmode)
101 {
102 	struct vnode *vp;
103 	struct lwp *l = curlwp;
104 	kauth_cred_t cred = l->l_cred;
105 	struct vattr va;
106 	int error;
107 	char *path;
108 
109 	ndp->ni_cnd.cn_flags &= TRYEMULROOT;
110 
111 	if (fmode & O_CREAT) {
112 		ndp->ni_cnd.cn_nameiop = CREATE;
113 		ndp->ni_cnd.cn_flags |= LOCKPARENT | LOCKLEAF;
114 		if ((fmode & O_EXCL) == 0 &&
115 		    ((fmode & O_NOFOLLOW) == 0))
116 			ndp->ni_cnd.cn_flags |= FOLLOW;
117 	} else {
118 		ndp->ni_cnd.cn_nameiop = LOOKUP;
119 		ndp->ni_cnd.cn_flags |= LOCKLEAF;
120 		if ((fmode & O_NOFOLLOW) == 0)
121 			ndp->ni_cnd.cn_flags |= FOLLOW;
122 	}
123 
124 	VERIEXEC_PATH_GET(ndp->ni_dirp, ndp->ni_segflg, ndp->ni_dirp, path);
125 
126 	error = namei(ndp);
127 	if (error)
128 		goto out;
129 
130 	vp = ndp->ni_vp;
131 
132 #if NVERIEXEC > 0
133 	error = veriexec_openchk(l, ndp->ni_vp, ndp->ni_dirp, fmode);
134 	if (error)
135 		goto bad;
136 #endif /* NVERIEXEC > 0 */
137 
138 	if (fmode & O_CREAT) {
139 		if (ndp->ni_vp == NULL) {
140 			VATTR_NULL(&va);
141 			va.va_type = VREG;
142 			va.va_mode = cmode;
143 			if (fmode & O_EXCL)
144 				 va.va_vaflags |= VA_EXCLUSIVE;
145 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
146 					   &ndp->ni_cnd, &va);
147 			if (error)
148 				goto out;
149 			fmode &= ~O_TRUNC;
150 			vp = ndp->ni_vp;
151 		} else {
152 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
153 			if (ndp->ni_dvp == ndp->ni_vp)
154 				vrele(ndp->ni_dvp);
155 			else
156 				vput(ndp->ni_dvp);
157 			ndp->ni_dvp = NULL;
158 			vp = ndp->ni_vp;
159 			if (fmode & O_EXCL) {
160 				error = EEXIST;
161 				goto bad;
162 			}
163 			fmode &= ~O_CREAT;
164 		}
165 	} else {
166 		vp = ndp->ni_vp;
167 	}
168 	if (vp->v_type == VSOCK) {
169 		error = EOPNOTSUPP;
170 		goto bad;
171 	}
172 	if (ndp->ni_vp->v_type == VLNK) {
173 		error = EFTYPE;
174 		goto bad;
175 	}
176 
177 	if ((fmode & O_CREAT) == 0) {
178 		error = vn_openchk(vp, cred, fmode);
179 		if (error != 0)
180 			goto bad;
181 	}
182 
183 	if (fmode & O_TRUNC) {
184 		VOP_UNLOCK(vp, 0);			/* XXX */
185 
186 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
187 		VATTR_NULL(&va);
188 		va.va_size = 0;
189 		error = VOP_SETATTR(vp, &va, cred);
190 		if (error != 0)
191 			goto bad;
192 	}
193 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0)
194 		goto bad;
195 	if (fmode & FWRITE) {
196 		mutex_enter(&vp->v_interlock);
197 		vp->v_writecount++;
198 		mutex_exit(&vp->v_interlock);
199 	}
200 
201 bad:
202 	if (error)
203 		vput(vp);
204 out:
205 	VERIEXEC_PATH_PUT(path);
206 	return (error);
207 }
208 
209 /*
210  * Check for write permissions on the specified vnode.
211  * Prototype text segments cannot be written.
212  */
213 int
214 vn_writechk(struct vnode *vp)
215 {
216 
217 	/*
218 	 * If the vnode is in use as a process's text,
219 	 * we can't allow writing.
220 	 */
221 	if (vp->v_iflag & VI_TEXT)
222 		return (ETXTBSY);
223 	return (0);
224 }
225 
226 int
227 vn_openchk(struct vnode *vp, kauth_cred_t cred, int fflags)
228 {
229 	int permbits = 0;
230 	int error;
231 
232 	if ((fflags & FREAD) != 0) {
233 		permbits = VREAD;
234 	}
235 	if ((fflags & (FWRITE | O_TRUNC)) != 0) {
236 		permbits |= VWRITE;
237 		if (vp->v_type == VDIR) {
238 			error = EISDIR;
239 			goto bad;
240 		}
241 		error = vn_writechk(vp);
242 		if (error != 0)
243 			goto bad;
244 	}
245 	error = VOP_ACCESS(vp, permbits, cred);
246 bad:
247 	return error;
248 }
249 
250 /*
251  * Mark a vnode as having executable mappings.
252  */
253 void
254 vn_markexec(struct vnode *vp)
255 {
256 
257 	KASSERT(mutex_owned(&vp->v_interlock));
258 
259 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
260 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
261 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
262 	}
263 	vp->v_iflag |= VI_EXECMAP;
264 }
265 
266 /*
267  * Mark a vnode as being the text of a process.
268  * Fail if the vnode is currently writable.
269  */
270 int
271 vn_marktext(struct vnode *vp)
272 {
273 
274 	mutex_enter(&vp->v_interlock);
275 	if (vp->v_writecount != 0) {
276 		KASSERT((vp->v_iflag & VI_TEXT) == 0);
277 		mutex_exit(&vp->v_interlock);
278 		return (ETXTBSY);
279 	}
280 	vp->v_iflag |= VI_TEXT;
281 	vn_markexec(vp);
282 	mutex_exit(&vp->v_interlock);
283 	return (0);
284 }
285 
286 /*
287  * Vnode close call
288  *
289  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
290  */
291 int
292 vn_close(struct vnode *vp, int flags, kauth_cred_t cred)
293 {
294 	int error;
295 
296 	mutex_enter(&vp->v_interlock);
297 	if (flags & FWRITE)
298 		vp->v_writecount--;
299 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_INTERLOCK);
300 	error = VOP_CLOSE(vp, flags, cred);
301 	vput(vp);
302 	return (error);
303 }
304 
305 /*
306  * Package up an I/O request on a vnode into a uio and do it.
307  */
308 int
309 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset,
310     enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid,
311     struct lwp *l)
312 {
313 	struct uio auio;
314 	struct iovec aiov;
315 	int error;
316 
317 	if ((ioflg & IO_NODELOCKED) == 0) {
318 		if (rw == UIO_READ) {
319 			vn_lock(vp, LK_SHARED | LK_RETRY);
320 		} else /* UIO_WRITE */ {
321 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
322 		}
323 	}
324 	auio.uio_iov = &aiov;
325 	auio.uio_iovcnt = 1;
326 	aiov.iov_base = base;
327 	aiov.iov_len = len;
328 	auio.uio_resid = len;
329 	auio.uio_offset = offset;
330 	auio.uio_rw = rw;
331 	if (segflg == UIO_SYSSPACE) {
332 		UIO_SETUP_SYSSPACE(&auio);
333 	} else {
334 		auio.uio_vmspace = l->l_proc->p_vmspace;
335 	}
336 	if (rw == UIO_READ) {
337 		error = VOP_READ(vp, &auio, ioflg, cred);
338 	} else {
339 		error = VOP_WRITE(vp, &auio, ioflg, cred);
340 	}
341 	if (aresid)
342 		*aresid = auio.uio_resid;
343 	else
344 		if (auio.uio_resid && error == 0)
345 			error = EIO;
346 	if ((ioflg & IO_NODELOCKED) == 0) {
347 		VOP_UNLOCK(vp, 0);
348 	}
349 	return (error);
350 }
351 
352 int
353 vn_readdir(file_t *fp, char *bf, int segflg, u_int count, int *done,
354     struct lwp *l, off_t **cookies, int *ncookies)
355 {
356 	struct vnode *vp = (struct vnode *)fp->f_data;
357 	struct iovec aiov;
358 	struct uio auio;
359 	int error, eofflag;
360 
361 	/* Limit the size on any kernel buffers used by VOP_READDIR */
362 	count = min(MAXBSIZE, count);
363 
364 unionread:
365 	if (vp->v_type != VDIR)
366 		return (EINVAL);
367 	aiov.iov_base = bf;
368 	aiov.iov_len = count;
369 	auio.uio_iov = &aiov;
370 	auio.uio_iovcnt = 1;
371 	auio.uio_rw = UIO_READ;
372 	if (segflg == UIO_SYSSPACE) {
373 		UIO_SETUP_SYSSPACE(&auio);
374 	} else {
375 		KASSERT(l == curlwp);
376 		auio.uio_vmspace = l->l_proc->p_vmspace;
377 	}
378 	auio.uio_resid = count;
379 	vn_lock(vp, LK_SHARED | LK_RETRY);
380 	auio.uio_offset = fp->f_offset;
381 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
382 		    ncookies);
383 	FILE_LOCK(fp);
384 	fp->f_offset = auio.uio_offset;
385 	FILE_UNLOCK(fp);
386 	VOP_UNLOCK(vp, 0);
387 	if (error)
388 		return (error);
389 
390 #if defined(UNION) || defined(LKM)
391 	if (count == auio.uio_resid && vn_union_readdir_hook) {
392 		struct vnode *ovp = vp;
393 
394 		error = (*vn_union_readdir_hook)(&vp, fp, l);
395 		if (error)
396 			return (error);
397 		if (vp != ovp)
398 			goto unionread;
399 	}
400 #endif /* UNION || LKM */
401 
402 	if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) &&
403 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
404 		struct vnode *tvp = vp;
405 		vp = vp->v_mount->mnt_vnodecovered;
406 		VREF(vp);
407 		FILE_LOCK(fp);
408 		fp->f_data = vp;
409 		fp->f_offset = 0;
410 		FILE_UNLOCK(fp);
411 		vrele(tvp);
412 		goto unionread;
413 	}
414 	*done = count - auio.uio_resid;
415 	return error;
416 }
417 
418 /*
419  * File table vnode read routine.
420  */
421 static int
422 vn_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
423     int flags)
424 {
425 	struct vnode *vp = (struct vnode *)fp->f_data;
426 	int count, error, ioflag, fflag;
427 
428 	ioflag = IO_ADV_ENCODE(fp->f_advice);
429 	fflag = fp->f_flag;
430 	if (fflag & FNONBLOCK)
431 		ioflag |= IO_NDELAY;
432 	if ((fflag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
433 		ioflag |= IO_SYNC;
434 	if (fflag & FALTIO)
435 		ioflag |= IO_ALTSEMANTICS;
436 	if (fflag & FDIRECT)
437 		ioflag |= IO_DIRECT;
438 	vn_lock(vp, LK_SHARED | LK_RETRY);
439 	uio->uio_offset = *offset;
440 	count = uio->uio_resid;
441 	error = VOP_READ(vp, uio, ioflag, cred);
442 	if (flags & FOF_UPDATE_OFFSET)
443 		*offset += count - uio->uio_resid;
444 	VOP_UNLOCK(vp, 0);
445 	return (error);
446 }
447 
448 /*
449  * File table vnode write routine.
450  */
451 static int
452 vn_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
453     int flags)
454 {
455 	struct vnode *vp = (struct vnode *)fp->f_data;
456 	int count, error, ioflag, fflag;
457 
458 	ioflag = IO_ADV_ENCODE(fp->f_advice) | IO_UNIT;
459 	fflag = fp->f_flag;
460 	if (vp->v_type == VREG && (fflag & O_APPEND))
461 		ioflag |= IO_APPEND;
462 	if (fflag & FNONBLOCK)
463 		ioflag |= IO_NDELAY;
464 	if (fflag & FFSYNC ||
465 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
466 		ioflag |= IO_SYNC;
467 	else if (fflag & FDSYNC)
468 		ioflag |= IO_DSYNC;
469 	if (fflag & FALTIO)
470 		ioflag |= IO_ALTSEMANTICS;
471 	if (fflag & FDIRECT)
472 		ioflag |= IO_DIRECT;
473 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
474 	uio->uio_offset = *offset;
475 	count = uio->uio_resid;
476 	error = VOP_WRITE(vp, uio, ioflag, cred);
477 	if (flags & FOF_UPDATE_OFFSET) {
478 		if (ioflag & IO_APPEND)
479 			*offset = uio->uio_offset;
480 		else
481 			*offset += count - uio->uio_resid;
482 	}
483 	VOP_UNLOCK(vp, 0);
484 	return (error);
485 }
486 
487 /*
488  * File table vnode stat routine.
489  */
490 static int
491 vn_statfile(file_t *fp, struct stat *sb)
492 {
493 	struct vnode *vp = (struct vnode *)fp->f_data;
494 
495 	return vn_stat(vp, sb);
496 }
497 
498 int
499 vn_stat(struct vnode *vp, struct stat *sb)
500 {
501 	struct vattr va;
502 	int error;
503 	mode_t mode;
504 
505 	error = VOP_GETATTR(vp, &va, kauth_cred_get());
506 	if (error)
507 		return (error);
508 	/*
509 	 * Copy from vattr table
510 	 */
511 	sb->st_dev = va.va_fsid;
512 	sb->st_ino = va.va_fileid;
513 	mode = va.va_mode;
514 	switch (vp->v_type) {
515 	case VREG:
516 		mode |= S_IFREG;
517 		break;
518 	case VDIR:
519 		mode |= S_IFDIR;
520 		break;
521 	case VBLK:
522 		mode |= S_IFBLK;
523 		break;
524 	case VCHR:
525 		mode |= S_IFCHR;
526 		break;
527 	case VLNK:
528 		mode |= S_IFLNK;
529 		break;
530 	case VSOCK:
531 		mode |= S_IFSOCK;
532 		break;
533 	case VFIFO:
534 		mode |= S_IFIFO;
535 		break;
536 	default:
537 		return (EBADF);
538 	};
539 	sb->st_mode = mode;
540 	sb->st_nlink = va.va_nlink;
541 	sb->st_uid = va.va_uid;
542 	sb->st_gid = va.va_gid;
543 	sb->st_rdev = va.va_rdev;
544 	sb->st_size = va.va_size;
545 	sb->st_atimespec = va.va_atime;
546 	sb->st_mtimespec = va.va_mtime;
547 	sb->st_ctimespec = va.va_ctime;
548 	sb->st_birthtimespec = va.va_birthtime;
549 	sb->st_blksize = va.va_blocksize;
550 	sb->st_flags = va.va_flags;
551 	sb->st_gen = 0;
552 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
553 	return (0);
554 }
555 
556 /*
557  * File table vnode fcntl routine.
558  */
559 static int
560 vn_fcntl(file_t *fp, u_int com, void *data)
561 {
562 	struct vnode *vp = fp->f_data;
563 	int error;
564 
565 	error = VOP_FCNTL(vp, com, data, fp->f_flag, kauth_cred_get());
566 	return (error);
567 }
568 
569 /*
570  * File table vnode ioctl routine.
571  */
572 static int
573 vn_ioctl(file_t *fp, u_long com, void *data)
574 {
575 	struct vnode *vp = fp->f_data, *ovp;
576 	struct vattr vattr;
577 	int error;
578 
579 	switch (vp->v_type) {
580 
581 	case VREG:
582 	case VDIR:
583 		if (com == FIONREAD) {
584 			error = VOP_GETATTR(vp, &vattr,
585 			    kauth_cred_get());
586 			if (error)
587 				return (error);
588 			*(int *)data = vattr.va_size - fp->f_offset;
589 			return (0);
590 		}
591 		if ((com == FIONWRITE) || (com == FIONSPACE)) {
592 			/*
593 			 * Files don't have send queues, so there never
594 			 * are any bytes in them, nor is there any
595 			 * open space in them.
596 			 */
597 			*(int *)data = 0;
598 			return (0);
599 		}
600 		if (com == FIOGETBMAP) {
601 			daddr_t *block;
602 
603 			if (*(daddr_t *)data < 0)
604 				return (EINVAL);
605 			block = (daddr_t *)data;
606 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
607 		}
608 		if (com == OFIOGETBMAP) {
609 			daddr_t ibn, obn;
610 
611 			if (*(int32_t *)data < 0)
612 				return (EINVAL);
613 			ibn = (daddr_t)*(int32_t *)data;
614 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
615 			*(int32_t *)data = (int32_t)obn;
616 			return error;
617 		}
618 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
619 			return (0);			/* XXX */
620 		/* fall into ... */
621 	case VFIFO:
622 	case VCHR:
623 	case VBLK:
624 		error = VOP_IOCTL(vp, com, data, fp->f_flag,
625 		    kauth_cred_get());
626 		if (error == 0 && com == TIOCSCTTY) {
627 			VREF(vp);
628 			mutex_enter(proc_lock);
629 			ovp = curproc->p_session->s_ttyvp;
630 			curproc->p_session->s_ttyvp = vp;
631 			mutex_exit(proc_lock);
632 			if (ovp != NULL)
633 				vrele(ovp);
634 		}
635 		return (error);
636 
637 	default:
638 		return (EPASSTHROUGH);
639 	}
640 }
641 
642 /*
643  * File table vnode poll routine.
644  */
645 static int
646 vn_poll(file_t *fp, int events)
647 {
648 
649 	return (VOP_POLL(fp->f_data, events));
650 }
651 
652 /*
653  * File table vnode kqfilter routine.
654  */
655 int
656 vn_kqfilter(file_t *fp, struct knote *kn)
657 {
658 
659 	return (VOP_KQFILTER(fp->f_data, kn));
660 }
661 
662 /*
663  * Check that the vnode is still valid, and if so
664  * acquire requested lock.
665  */
666 int
667 vn_lock(struct vnode *vp, int flags)
668 {
669 	int error;
670 
671 #if 0
672 	KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0
673 	    || (vp->v_iflag & VI_ONWORKLST) != 0);
674 #endif
675 	KASSERT((flags &
676 	    ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT|LK_RETRY|
677 	    LK_CANRECURSE))
678 	    == 0);
679 
680 	do {
681 		if ((flags & LK_INTERLOCK) == 0)
682 			mutex_enter(&vp->v_interlock);
683 		if (vp->v_iflag & VI_XLOCK) {
684 			if (flags & LK_NOWAIT) {
685 				mutex_exit(&vp->v_interlock);
686 				return EBUSY;
687 			}
688 			vwait(vp, VI_XLOCK);
689 			mutex_exit(&vp->v_interlock);
690 			error = ENOENT;
691 		} else {
692 			error = VOP_LOCK(vp,
693 			    (flags & ~LK_RETRY) | LK_INTERLOCK);
694 			if (error == 0 || error == EDEADLK || error == EBUSY)
695 				return (error);
696 		}
697 		flags &= ~LK_INTERLOCK;
698 	} while (flags & LK_RETRY);
699 	return (error);
700 }
701 
702 /*
703  * File table vnode close routine.
704  */
705 static int
706 vn_closefile(file_t *fp)
707 {
708 
709 	return vn_close(fp->f_data, fp->f_flag, fp->f_cred);
710 }
711 
712 /*
713  * Enable LK_CANRECURSE on lock. Return prior status.
714  */
715 u_int
716 vn_setrecurse(struct vnode *vp)
717 {
718 	struct vnlock *lkp;
719 
720 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
721 	atomic_inc_uint(&lkp->vl_canrecurse);
722 
723 	return 0;
724 }
725 
726 /*
727  * Called when done with locksetrecurse.
728  */
729 void
730 vn_restorerecurse(struct vnode *vp, u_int flags)
731 {
732 	struct vnlock *lkp;
733 
734 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
735 	atomic_dec_uint(&lkp->vl_canrecurse);
736 }
737 
738 /*
739  * Simplified in-kernel wrapper calls for extended attribute access.
740  * Both calls pass in a NULL credential, authorizing a "kernel" access.
741  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
742  */
743 int
744 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
745     const char *attrname, size_t *buflen, void *bf, struct lwp *l)
746 {
747 	struct uio auio;
748 	struct iovec aiov;
749 	int error;
750 
751 	aiov.iov_len = *buflen;
752 	aiov.iov_base = bf;
753 
754 	auio.uio_iov = &aiov;
755 	auio.uio_iovcnt = 1;
756 	auio.uio_rw = UIO_READ;
757 	auio.uio_offset = 0;
758 	auio.uio_resid = *buflen;
759 	UIO_SETUP_SYSSPACE(&auio);
760 
761 	if ((ioflg & IO_NODELOCKED) == 0)
762 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
763 
764 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL);
765 
766 	if ((ioflg & IO_NODELOCKED) == 0)
767 		VOP_UNLOCK(vp, 0);
768 
769 	if (error == 0)
770 		*buflen = *buflen - auio.uio_resid;
771 
772 	return (error);
773 }
774 
775 /*
776  * XXX Failure mode if partially written?
777  */
778 int
779 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
780     const char *attrname, size_t buflen, const void *bf, struct lwp *l)
781 {
782 	struct uio auio;
783 	struct iovec aiov;
784 	int error;
785 
786 	aiov.iov_len = buflen;
787 	aiov.iov_base = __UNCONST(bf);		/* XXXUNCONST kills const */
788 
789 	auio.uio_iov = &aiov;
790 	auio.uio_iovcnt = 1;
791 	auio.uio_rw = UIO_WRITE;
792 	auio.uio_offset = 0;
793 	auio.uio_resid = buflen;
794 	UIO_SETUP_SYSSPACE(&auio);
795 
796 	if ((ioflg & IO_NODELOCKED) == 0) {
797 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
798 	}
799 
800 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL);
801 
802 	if ((ioflg & IO_NODELOCKED) == 0) {
803 		VOP_UNLOCK(vp, 0);
804 	}
805 
806 	return (error);
807 }
808 
809 int
810 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
811     const char *attrname, struct lwp *l)
812 {
813 	int error;
814 
815 	if ((ioflg & IO_NODELOCKED) == 0) {
816 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
817 	}
818 
819 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL);
820 	if (error == EOPNOTSUPP)
821 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL);
822 
823 	if ((ioflg & IO_NODELOCKED) == 0) {
824 		VOP_UNLOCK(vp, 0);
825 	}
826 
827 	return (error);
828 }
829 
830 void
831 vn_ra_allocctx(struct vnode *vp)
832 {
833 	struct uvm_ractx *ra = NULL;
834 
835 	KASSERT(mutex_owned(&vp->v_interlock));
836 
837 	if (vp->v_type != VREG) {
838 		return;
839 	}
840 	if (vp->v_ractx != NULL) {
841 		return;
842 	}
843 	if (vp->v_ractx == NULL) {
844 		mutex_exit(&vp->v_interlock);
845 		ra = uvm_ra_allocctx();
846 		mutex_enter(&vp->v_interlock);
847 		if (ra != NULL && vp->v_ractx == NULL) {
848 			vp->v_ractx = ra;
849 			ra = NULL;
850 		}
851 	}
852 	if (ra != NULL) {
853 		uvm_ra_freectx(ra);
854 	}
855 }
856