xref: /netbsd-src/sys/kern/vfs_vnops.c (revision 2980e352a13e8f0b545a366830c411e7a542ada8)
1 /*	$NetBSD: vfs_vnops.c,v 1.158 2008/06/02 16:08:41 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 1.158 2008/06/02 16:08:41 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 	if ((vp->v_iflag & VI_EXECMAP) != 0) {
258 		/* Safe unlocked, as long as caller holds a reference. */
259 		return;
260 	}
261 
262 	mutex_enter(&vp->v_interlock);
263 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
264 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
265 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
266 		vp->v_iflag |= VI_EXECMAP;
267 	}
268 	mutex_exit(&vp->v_interlock);
269 }
270 
271 /*
272  * Mark a vnode as being the text of a process.
273  * Fail if the vnode is currently writable.
274  */
275 int
276 vn_marktext(struct vnode *vp)
277 {
278 
279 	if ((vp->v_iflag & (VI_TEXT|VI_EXECMAP)) == (VI_TEXT|VI_EXECMAP)) {
280 		/* Safe unlocked, as long as caller holds a reference. */
281 		return (0);
282 	}
283 
284 	mutex_enter(&vp->v_interlock);
285 	if (vp->v_writecount != 0) {
286 		KASSERT((vp->v_iflag & VI_TEXT) == 0);
287 		mutex_exit(&vp->v_interlock);
288 		return (ETXTBSY);
289 	}
290 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
291 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
292 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
293 	}
294 	vp->v_iflag |= (VI_TEXT | VI_EXECMAP);
295 	mutex_exit(&vp->v_interlock);
296 	return (0);
297 }
298 
299 /*
300  * Vnode close call
301  *
302  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
303  */
304 int
305 vn_close(struct vnode *vp, int flags, kauth_cred_t cred)
306 {
307 	int error;
308 
309 	if (flags & FWRITE) {
310 		mutex_enter(&vp->v_interlock);
311 		vp->v_writecount--;
312 		mutex_exit(&vp->v_interlock);
313 	}
314 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
315 	error = VOP_CLOSE(vp, flags, cred);
316 	vput(vp);
317 	return (error);
318 }
319 
320 /*
321  * Package up an I/O request on a vnode into a uio and do it.
322  */
323 int
324 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset,
325     enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid,
326     struct lwp *l)
327 {
328 	struct uio auio;
329 	struct iovec aiov;
330 	int error;
331 
332 	if ((ioflg & IO_NODELOCKED) == 0) {
333 		if (rw == UIO_READ) {
334 			vn_lock(vp, LK_SHARED | LK_RETRY);
335 		} else /* UIO_WRITE */ {
336 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
337 		}
338 	}
339 	auio.uio_iov = &aiov;
340 	auio.uio_iovcnt = 1;
341 	aiov.iov_base = base;
342 	aiov.iov_len = len;
343 	auio.uio_resid = len;
344 	auio.uio_offset = offset;
345 	auio.uio_rw = rw;
346 	if (segflg == UIO_SYSSPACE) {
347 		UIO_SETUP_SYSSPACE(&auio);
348 	} else {
349 		auio.uio_vmspace = l->l_proc->p_vmspace;
350 	}
351 	if (rw == UIO_READ) {
352 		error = VOP_READ(vp, &auio, ioflg, cred);
353 	} else {
354 		error = VOP_WRITE(vp, &auio, ioflg, cred);
355 	}
356 	if (aresid)
357 		*aresid = auio.uio_resid;
358 	else
359 		if (auio.uio_resid && error == 0)
360 			error = EIO;
361 	if ((ioflg & IO_NODELOCKED) == 0) {
362 		VOP_UNLOCK(vp, 0);
363 	}
364 	return (error);
365 }
366 
367 int
368 vn_readdir(file_t *fp, char *bf, int segflg, u_int count, int *done,
369     struct lwp *l, off_t **cookies, int *ncookies)
370 {
371 	struct vnode *vp = (struct vnode *)fp->f_data;
372 	struct iovec aiov;
373 	struct uio auio;
374 	int error, eofflag;
375 
376 	/* Limit the size on any kernel buffers used by VOP_READDIR */
377 	count = min(MAXBSIZE, count);
378 
379 unionread:
380 	if (vp->v_type != VDIR)
381 		return (EINVAL);
382 	aiov.iov_base = bf;
383 	aiov.iov_len = count;
384 	auio.uio_iov = &aiov;
385 	auio.uio_iovcnt = 1;
386 	auio.uio_rw = UIO_READ;
387 	if (segflg == UIO_SYSSPACE) {
388 		UIO_SETUP_SYSSPACE(&auio);
389 	} else {
390 		KASSERT(l == curlwp);
391 		auio.uio_vmspace = l->l_proc->p_vmspace;
392 	}
393 	auio.uio_resid = count;
394 	vn_lock(vp, LK_SHARED | LK_RETRY);
395 	auio.uio_offset = fp->f_offset;
396 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
397 		    ncookies);
398 	FILE_LOCK(fp);
399 	fp->f_offset = auio.uio_offset;
400 	FILE_UNLOCK(fp);
401 	VOP_UNLOCK(vp, 0);
402 	if (error)
403 		return (error);
404 
405 #if defined(UNION) || defined(LKM)
406 	if (count == auio.uio_resid && vn_union_readdir_hook) {
407 		struct vnode *ovp = vp;
408 
409 		error = (*vn_union_readdir_hook)(&vp, fp, l);
410 		if (error)
411 			return (error);
412 		if (vp != ovp)
413 			goto unionread;
414 	}
415 #endif /* UNION || LKM */
416 
417 	if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) &&
418 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
419 		struct vnode *tvp = vp;
420 		vp = vp->v_mount->mnt_vnodecovered;
421 		VREF(vp);
422 		FILE_LOCK(fp);
423 		fp->f_data = vp;
424 		fp->f_offset = 0;
425 		FILE_UNLOCK(fp);
426 		vrele(tvp);
427 		goto unionread;
428 	}
429 	*done = count - auio.uio_resid;
430 	return error;
431 }
432 
433 /*
434  * File table vnode read routine.
435  */
436 static int
437 vn_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
438     int flags)
439 {
440 	struct vnode *vp = (struct vnode *)fp->f_data;
441 	int count, error, ioflag, fflag;
442 
443 	ioflag = IO_ADV_ENCODE(fp->f_advice);
444 	fflag = fp->f_flag;
445 	if (fflag & FNONBLOCK)
446 		ioflag |= IO_NDELAY;
447 	if ((fflag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
448 		ioflag |= IO_SYNC;
449 	if (fflag & FALTIO)
450 		ioflag |= IO_ALTSEMANTICS;
451 	if (fflag & FDIRECT)
452 		ioflag |= IO_DIRECT;
453 	vn_lock(vp, LK_SHARED | LK_RETRY);
454 	uio->uio_offset = *offset;
455 	count = uio->uio_resid;
456 	error = VOP_READ(vp, uio, ioflag, cred);
457 	if (flags & FOF_UPDATE_OFFSET)
458 		*offset += count - uio->uio_resid;
459 	VOP_UNLOCK(vp, 0);
460 	return (error);
461 }
462 
463 /*
464  * File table vnode write routine.
465  */
466 static int
467 vn_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
468     int flags)
469 {
470 	struct vnode *vp = (struct vnode *)fp->f_data;
471 	int count, error, ioflag, fflag;
472 
473 	ioflag = IO_ADV_ENCODE(fp->f_advice) | IO_UNIT;
474 	fflag = fp->f_flag;
475 	if (vp->v_type == VREG && (fflag & O_APPEND))
476 		ioflag |= IO_APPEND;
477 	if (fflag & FNONBLOCK)
478 		ioflag |= IO_NDELAY;
479 	if (fflag & FFSYNC ||
480 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
481 		ioflag |= IO_SYNC;
482 	else if (fflag & FDSYNC)
483 		ioflag |= IO_DSYNC;
484 	if (fflag & FALTIO)
485 		ioflag |= IO_ALTSEMANTICS;
486 	if (fflag & FDIRECT)
487 		ioflag |= IO_DIRECT;
488 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
489 	uio->uio_offset = *offset;
490 	count = uio->uio_resid;
491 	error = VOP_WRITE(vp, uio, ioflag, cred);
492 	if (flags & FOF_UPDATE_OFFSET) {
493 		if (ioflag & IO_APPEND)
494 			*offset = uio->uio_offset;
495 		else
496 			*offset += count - uio->uio_resid;
497 	}
498 	VOP_UNLOCK(vp, 0);
499 	return (error);
500 }
501 
502 /*
503  * File table vnode stat routine.
504  */
505 static int
506 vn_statfile(file_t *fp, struct stat *sb)
507 {
508 	struct vnode *vp = (struct vnode *)fp->f_data;
509 
510 	return vn_stat(vp, sb);
511 }
512 
513 int
514 vn_stat(struct vnode *vp, struct stat *sb)
515 {
516 	struct vattr va;
517 	int error;
518 	mode_t mode;
519 
520 	error = VOP_GETATTR(vp, &va, kauth_cred_get());
521 	if (error)
522 		return (error);
523 	/*
524 	 * Copy from vattr table
525 	 */
526 	sb->st_dev = va.va_fsid;
527 	sb->st_ino = va.va_fileid;
528 	mode = va.va_mode;
529 	switch (vp->v_type) {
530 	case VREG:
531 		mode |= S_IFREG;
532 		break;
533 	case VDIR:
534 		mode |= S_IFDIR;
535 		break;
536 	case VBLK:
537 		mode |= S_IFBLK;
538 		break;
539 	case VCHR:
540 		mode |= S_IFCHR;
541 		break;
542 	case VLNK:
543 		mode |= S_IFLNK;
544 		break;
545 	case VSOCK:
546 		mode |= S_IFSOCK;
547 		break;
548 	case VFIFO:
549 		mode |= S_IFIFO;
550 		break;
551 	default:
552 		return (EBADF);
553 	};
554 	sb->st_mode = mode;
555 	sb->st_nlink = va.va_nlink;
556 	sb->st_uid = va.va_uid;
557 	sb->st_gid = va.va_gid;
558 	sb->st_rdev = va.va_rdev;
559 	sb->st_size = va.va_size;
560 	sb->st_atimespec = va.va_atime;
561 	sb->st_mtimespec = va.va_mtime;
562 	sb->st_ctimespec = va.va_ctime;
563 	sb->st_birthtimespec = va.va_birthtime;
564 	sb->st_blksize = va.va_blocksize;
565 	sb->st_flags = va.va_flags;
566 	sb->st_gen = 0;
567 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
568 	return (0);
569 }
570 
571 /*
572  * File table vnode fcntl routine.
573  */
574 static int
575 vn_fcntl(file_t *fp, u_int com, void *data)
576 {
577 	struct vnode *vp = fp->f_data;
578 	int error;
579 
580 	error = VOP_FCNTL(vp, com, data, fp->f_flag, kauth_cred_get());
581 	return (error);
582 }
583 
584 /*
585  * File table vnode ioctl routine.
586  */
587 static int
588 vn_ioctl(file_t *fp, u_long com, void *data)
589 {
590 	struct vnode *vp = fp->f_data, *ovp;
591 	struct vattr vattr;
592 	int error;
593 
594 	switch (vp->v_type) {
595 
596 	case VREG:
597 	case VDIR:
598 		if (com == FIONREAD) {
599 			error = VOP_GETATTR(vp, &vattr,
600 			    kauth_cred_get());
601 			if (error)
602 				return (error);
603 			*(int *)data = vattr.va_size - fp->f_offset;
604 			return (0);
605 		}
606 		if ((com == FIONWRITE) || (com == FIONSPACE)) {
607 			/*
608 			 * Files don't have send queues, so there never
609 			 * are any bytes in them, nor is there any
610 			 * open space in them.
611 			 */
612 			*(int *)data = 0;
613 			return (0);
614 		}
615 		if (com == FIOGETBMAP) {
616 			daddr_t *block;
617 
618 			if (*(daddr_t *)data < 0)
619 				return (EINVAL);
620 			block = (daddr_t *)data;
621 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
622 		}
623 		if (com == OFIOGETBMAP) {
624 			daddr_t ibn, obn;
625 
626 			if (*(int32_t *)data < 0)
627 				return (EINVAL);
628 			ibn = (daddr_t)*(int32_t *)data;
629 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
630 			*(int32_t *)data = (int32_t)obn;
631 			return error;
632 		}
633 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
634 			return (0);			/* XXX */
635 		/* fall into ... */
636 	case VFIFO:
637 	case VCHR:
638 	case VBLK:
639 		error = VOP_IOCTL(vp, com, data, fp->f_flag,
640 		    kauth_cred_get());
641 		if (error == 0 && com == TIOCSCTTY) {
642 			VREF(vp);
643 			mutex_enter(proc_lock);
644 			ovp = curproc->p_session->s_ttyvp;
645 			curproc->p_session->s_ttyvp = vp;
646 			mutex_exit(proc_lock);
647 			if (ovp != NULL)
648 				vrele(ovp);
649 		}
650 		return (error);
651 
652 	default:
653 		return (EPASSTHROUGH);
654 	}
655 }
656 
657 /*
658  * File table vnode poll routine.
659  */
660 static int
661 vn_poll(file_t *fp, int events)
662 {
663 
664 	return (VOP_POLL(fp->f_data, events));
665 }
666 
667 /*
668  * File table vnode kqfilter routine.
669  */
670 int
671 vn_kqfilter(file_t *fp, struct knote *kn)
672 {
673 
674 	return (VOP_KQFILTER(fp->f_data, kn));
675 }
676 
677 /*
678  * Check that the vnode is still valid, and if so
679  * acquire requested lock.
680  */
681 int
682 vn_lock(struct vnode *vp, int flags)
683 {
684 	int error;
685 
686 #if 0
687 	KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0
688 	    || (vp->v_iflag & VI_ONWORKLST) != 0);
689 #endif
690 	KASSERT((flags &
691 	    ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT|LK_RETRY|
692 	    LK_CANRECURSE))
693 	    == 0);
694 
695 	do {
696 		/*
697 		 * XXX PR 37706 forced unmount of file systems is unsafe.
698 		 * Race between vclean() and this the remaining problem.
699 		 */
700 		if (vp->v_iflag & VI_XLOCK) {
701 			if ((flags & LK_INTERLOCK) == 0) {
702 				mutex_enter(&vp->v_interlock);
703 			}
704 			flags &= ~LK_INTERLOCK;
705 			if (flags & LK_NOWAIT) {
706 				mutex_exit(&vp->v_interlock);
707 				return EBUSY;
708 			}
709 			vwait(vp, VI_XLOCK);
710 			mutex_exit(&vp->v_interlock);
711 			error = ENOENT;
712 		} else {
713 			if ((flags & LK_INTERLOCK) != 0) {
714 				mutex_exit(&vp->v_interlock);
715 			}
716 			flags &= ~LK_INTERLOCK;
717 			error = VOP_LOCK(vp, (flags & ~LK_RETRY));
718 			if (error == 0 || error == EDEADLK || error == EBUSY)
719 				return (error);
720 		}
721 	} while (flags & LK_RETRY);
722 	return (error);
723 }
724 
725 /*
726  * File table vnode close routine.
727  */
728 static int
729 vn_closefile(file_t *fp)
730 {
731 
732 	return vn_close(fp->f_data, fp->f_flag, fp->f_cred);
733 }
734 
735 /*
736  * Enable LK_CANRECURSE on lock. Return prior status.
737  */
738 u_int
739 vn_setrecurse(struct vnode *vp)
740 {
741 	struct vnlock *lkp;
742 
743 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
744 	atomic_inc_uint(&lkp->vl_canrecurse);
745 
746 	return 0;
747 }
748 
749 /*
750  * Called when done with locksetrecurse.
751  */
752 void
753 vn_restorerecurse(struct vnode *vp, u_int flags)
754 {
755 	struct vnlock *lkp;
756 
757 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
758 	atomic_dec_uint(&lkp->vl_canrecurse);
759 }
760 
761 /*
762  * Simplified in-kernel wrapper calls for extended attribute access.
763  * Both calls pass in a NULL credential, authorizing a "kernel" access.
764  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
765  */
766 int
767 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
768     const char *attrname, size_t *buflen, void *bf, struct lwp *l)
769 {
770 	struct uio auio;
771 	struct iovec aiov;
772 	int error;
773 
774 	aiov.iov_len = *buflen;
775 	aiov.iov_base = bf;
776 
777 	auio.uio_iov = &aiov;
778 	auio.uio_iovcnt = 1;
779 	auio.uio_rw = UIO_READ;
780 	auio.uio_offset = 0;
781 	auio.uio_resid = *buflen;
782 	UIO_SETUP_SYSSPACE(&auio);
783 
784 	if ((ioflg & IO_NODELOCKED) == 0)
785 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
786 
787 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL);
788 
789 	if ((ioflg & IO_NODELOCKED) == 0)
790 		VOP_UNLOCK(vp, 0);
791 
792 	if (error == 0)
793 		*buflen = *buflen - auio.uio_resid;
794 
795 	return (error);
796 }
797 
798 /*
799  * XXX Failure mode if partially written?
800  */
801 int
802 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
803     const char *attrname, size_t buflen, const void *bf, struct lwp *l)
804 {
805 	struct uio auio;
806 	struct iovec aiov;
807 	int error;
808 
809 	aiov.iov_len = buflen;
810 	aiov.iov_base = __UNCONST(bf);		/* XXXUNCONST kills const */
811 
812 	auio.uio_iov = &aiov;
813 	auio.uio_iovcnt = 1;
814 	auio.uio_rw = UIO_WRITE;
815 	auio.uio_offset = 0;
816 	auio.uio_resid = buflen;
817 	UIO_SETUP_SYSSPACE(&auio);
818 
819 	if ((ioflg & IO_NODELOCKED) == 0) {
820 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
821 	}
822 
823 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL);
824 
825 	if ((ioflg & IO_NODELOCKED) == 0) {
826 		VOP_UNLOCK(vp, 0);
827 	}
828 
829 	return (error);
830 }
831 
832 int
833 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
834     const char *attrname, struct lwp *l)
835 {
836 	int error;
837 
838 	if ((ioflg & IO_NODELOCKED) == 0) {
839 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
840 	}
841 
842 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL);
843 	if (error == EOPNOTSUPP)
844 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL);
845 
846 	if ((ioflg & IO_NODELOCKED) == 0) {
847 		VOP_UNLOCK(vp, 0);
848 	}
849 
850 	return (error);
851 }
852 
853 void
854 vn_ra_allocctx(struct vnode *vp)
855 {
856 	struct uvm_ractx *ra = NULL;
857 
858 	KASSERT(mutex_owned(&vp->v_interlock));
859 
860 	if (vp->v_type != VREG) {
861 		return;
862 	}
863 	if (vp->v_ractx != NULL) {
864 		return;
865 	}
866 	if (vp->v_ractx == NULL) {
867 		mutex_exit(&vp->v_interlock);
868 		ra = uvm_ra_allocctx();
869 		mutex_enter(&vp->v_interlock);
870 		if (ra != NULL && vp->v_ractx == NULL) {
871 			vp->v_ractx = ra;
872 			ra = NULL;
873 		}
874 	}
875 	if (ra != NULL) {
876 		uvm_ra_freectx(ra);
877 	}
878 }
879