xref: /netbsd-src/sys/kern/vfs_vnops.c (revision cac8e449158efc7261bebc8657cbb0125a2cfdde)
1 /*	$NetBSD: vfs_vnops.c,v 1.159 2008/07/31 05:38:05 simonb 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 1.159 2008/07/31 05:38:05 simonb 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 #include <sys/wapbl.h>
65 
66 #include <miscfs/specfs/specdev.h>
67 
68 #include <uvm/uvm_extern.h>
69 #include <uvm/uvm_readahead.h>
70 
71 #ifdef UNION
72 #include <fs/union/union.h>
73 #endif
74 
75 #if defined(LKM) || defined(UNION)
76 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct lwp *);
77 #endif
78 
79 #include <sys/verified_exec.h>
80 
81 static int vn_read(file_t *fp, off_t *offset, struct uio *uio,
82 	    kauth_cred_t cred, int flags);
83 static int vn_write(file_t *fp, off_t *offset, struct uio *uio,
84 	    kauth_cred_t cred, int flags);
85 static int vn_closefile(file_t *fp);
86 static int vn_poll(file_t *fp, int events);
87 static int vn_fcntl(file_t *fp, u_int com, void *data);
88 static int vn_statfile(file_t *fp, struct stat *sb);
89 static int vn_ioctl(file_t *fp, u_long com, void *data);
90 
91 const struct fileops vnops = {
92 	vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll,
93 	vn_statfile, vn_closefile, vn_kqfilter
94 };
95 
96 /*
97  * Common code for vnode open operations.
98  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
99  */
100 int
101 vn_open(struct nameidata *ndp, int fmode, int cmode)
102 {
103 	struct vnode *vp;
104 	struct lwp *l = curlwp;
105 	kauth_cred_t cred = l->l_cred;
106 	struct vattr va;
107 	int error;
108 	char *path;
109 
110 	ndp->ni_cnd.cn_flags &= TRYEMULROOT;
111 
112 	if (fmode & O_CREAT) {
113 		ndp->ni_cnd.cn_nameiop = CREATE;
114 		ndp->ni_cnd.cn_flags |= LOCKPARENT | LOCKLEAF;
115 		if ((fmode & O_EXCL) == 0 &&
116 		    ((fmode & O_NOFOLLOW) == 0))
117 			ndp->ni_cnd.cn_flags |= FOLLOW;
118 	} else {
119 		ndp->ni_cnd.cn_nameiop = LOOKUP;
120 		ndp->ni_cnd.cn_flags |= LOCKLEAF;
121 		if ((fmode & O_NOFOLLOW) == 0)
122 			ndp->ni_cnd.cn_flags |= FOLLOW;
123 	}
124 
125 	VERIEXEC_PATH_GET(ndp->ni_dirp, ndp->ni_segflg, ndp->ni_dirp, path);
126 
127 	error = namei(ndp);
128 	if (error)
129 		goto out;
130 
131 	vp = ndp->ni_vp;
132 
133 #if NVERIEXEC > 0
134 	error = veriexec_openchk(l, ndp->ni_vp, ndp->ni_dirp, fmode);
135 	if (error)
136 		goto bad;
137 #endif /* NVERIEXEC > 0 */
138 
139 	if (fmode & O_CREAT) {
140 		if (ndp->ni_vp == NULL) {
141 			VATTR_NULL(&va);
142 			va.va_type = VREG;
143 			va.va_mode = cmode;
144 			if (fmode & O_EXCL)
145 				 va.va_vaflags |= VA_EXCLUSIVE;
146 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
147 					   &ndp->ni_cnd, &va);
148 			if (error)
149 				goto out;
150 			fmode &= ~O_TRUNC;
151 			vp = ndp->ni_vp;
152 		} else {
153 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
154 			if (ndp->ni_dvp == ndp->ni_vp)
155 				vrele(ndp->ni_dvp);
156 			else
157 				vput(ndp->ni_dvp);
158 			ndp->ni_dvp = NULL;
159 			vp = ndp->ni_vp;
160 			if (fmode & O_EXCL) {
161 				error = EEXIST;
162 				goto bad;
163 			}
164 			fmode &= ~O_CREAT;
165 		}
166 	} else {
167 		vp = ndp->ni_vp;
168 	}
169 	if (vp->v_type == VSOCK) {
170 		error = EOPNOTSUPP;
171 		goto bad;
172 	}
173 	if (ndp->ni_vp->v_type == VLNK) {
174 		error = EFTYPE;
175 		goto bad;
176 	}
177 
178 	if ((fmode & O_CREAT) == 0) {
179 		error = vn_openchk(vp, cred, fmode);
180 		if (error != 0)
181 			goto bad;
182 	}
183 
184 	if (fmode & O_TRUNC) {
185 		VOP_UNLOCK(vp, 0);			/* XXX */
186 
187 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
188 		VATTR_NULL(&va);
189 		va.va_size = 0;
190 		error = VOP_SETATTR(vp, &va, cred);
191 		if (error != 0)
192 			goto bad;
193 	}
194 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0)
195 		goto bad;
196 	if (fmode & FWRITE) {
197 		mutex_enter(&vp->v_interlock);
198 		vp->v_writecount++;
199 		mutex_exit(&vp->v_interlock);
200 	}
201 
202 bad:
203 	if (error)
204 		vput(vp);
205 out:
206 	VERIEXEC_PATH_PUT(path);
207 	return (error);
208 }
209 
210 /*
211  * Check for write permissions on the specified vnode.
212  * Prototype text segments cannot be written.
213  */
214 int
215 vn_writechk(struct vnode *vp)
216 {
217 
218 	/*
219 	 * If the vnode is in use as a process's text,
220 	 * we can't allow writing.
221 	 */
222 	if (vp->v_iflag & VI_TEXT)
223 		return (ETXTBSY);
224 	return (0);
225 }
226 
227 int
228 vn_openchk(struct vnode *vp, kauth_cred_t cred, int fflags)
229 {
230 	int permbits = 0;
231 	int error;
232 
233 	if ((fflags & FREAD) != 0) {
234 		permbits = VREAD;
235 	}
236 	if ((fflags & (FWRITE | O_TRUNC)) != 0) {
237 		permbits |= VWRITE;
238 		if (vp->v_type == VDIR) {
239 			error = EISDIR;
240 			goto bad;
241 		}
242 		error = vn_writechk(vp);
243 		if (error != 0)
244 			goto bad;
245 	}
246 	error = VOP_ACCESS(vp, permbits, cred);
247 bad:
248 	return error;
249 }
250 
251 /*
252  * Mark a vnode as having executable mappings.
253  */
254 void
255 vn_markexec(struct vnode *vp)
256 {
257 
258 	if ((vp->v_iflag & VI_EXECMAP) != 0) {
259 		/* Safe unlocked, as long as caller holds a reference. */
260 		return;
261 	}
262 
263 	mutex_enter(&vp->v_interlock);
264 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
265 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
266 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
267 		vp->v_iflag |= VI_EXECMAP;
268 	}
269 	mutex_exit(&vp->v_interlock);
270 }
271 
272 /*
273  * Mark a vnode as being the text of a process.
274  * Fail if the vnode is currently writable.
275  */
276 int
277 vn_marktext(struct vnode *vp)
278 {
279 
280 	if ((vp->v_iflag & (VI_TEXT|VI_EXECMAP)) == (VI_TEXT|VI_EXECMAP)) {
281 		/* Safe unlocked, as long as caller holds a reference. */
282 		return (0);
283 	}
284 
285 	mutex_enter(&vp->v_interlock);
286 	if (vp->v_writecount != 0) {
287 		KASSERT((vp->v_iflag & VI_TEXT) == 0);
288 		mutex_exit(&vp->v_interlock);
289 		return (ETXTBSY);
290 	}
291 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
292 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
293 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
294 	}
295 	vp->v_iflag |= (VI_TEXT | VI_EXECMAP);
296 	mutex_exit(&vp->v_interlock);
297 	return (0);
298 }
299 
300 /*
301  * Vnode close call
302  *
303  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
304  */
305 int
306 vn_close(struct vnode *vp, int flags, kauth_cred_t cred)
307 {
308 	int error;
309 
310 	if (flags & FWRITE) {
311 		mutex_enter(&vp->v_interlock);
312 		vp->v_writecount--;
313 		mutex_exit(&vp->v_interlock);
314 	}
315 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
316 	error = VOP_CLOSE(vp, flags, cred);
317 	vput(vp);
318 	return (error);
319 }
320 
321 /*
322  * Package up an I/O request on a vnode into a uio and do it.
323  */
324 int
325 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset,
326     enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid,
327     struct lwp *l)
328 {
329 	struct uio auio;
330 	struct iovec aiov;
331 	int error;
332 
333 	if ((ioflg & IO_NODELOCKED) == 0) {
334 		if (rw == UIO_READ) {
335 			vn_lock(vp, LK_SHARED | LK_RETRY);
336 		} else /* UIO_WRITE */ {
337 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
338 		}
339 	}
340 	auio.uio_iov = &aiov;
341 	auio.uio_iovcnt = 1;
342 	aiov.iov_base = base;
343 	aiov.iov_len = len;
344 	auio.uio_resid = len;
345 	auio.uio_offset = offset;
346 	auio.uio_rw = rw;
347 	if (segflg == UIO_SYSSPACE) {
348 		UIO_SETUP_SYSSPACE(&auio);
349 	} else {
350 		auio.uio_vmspace = l->l_proc->p_vmspace;
351 	}
352 	if (rw == UIO_READ) {
353 		error = VOP_READ(vp, &auio, ioflg, cred);
354 	} else {
355 		error = VOP_WRITE(vp, &auio, ioflg, cred);
356 	}
357 	if (aresid)
358 		*aresid = auio.uio_resid;
359 	else
360 		if (auio.uio_resid && error == 0)
361 			error = EIO;
362 	if ((ioflg & IO_NODELOCKED) == 0) {
363 		VOP_UNLOCK(vp, 0);
364 	}
365 	return (error);
366 }
367 
368 int
369 vn_readdir(file_t *fp, char *bf, int segflg, u_int count, int *done,
370     struct lwp *l, off_t **cookies, int *ncookies)
371 {
372 	struct vnode *vp = (struct vnode *)fp->f_data;
373 	struct iovec aiov;
374 	struct uio auio;
375 	int error, eofflag;
376 
377 	/* Limit the size on any kernel buffers used by VOP_READDIR */
378 	count = min(MAXBSIZE, count);
379 
380 unionread:
381 	if (vp->v_type != VDIR)
382 		return (EINVAL);
383 	aiov.iov_base = bf;
384 	aiov.iov_len = count;
385 	auio.uio_iov = &aiov;
386 	auio.uio_iovcnt = 1;
387 	auio.uio_rw = UIO_READ;
388 	if (segflg == UIO_SYSSPACE) {
389 		UIO_SETUP_SYSSPACE(&auio);
390 	} else {
391 		KASSERT(l == curlwp);
392 		auio.uio_vmspace = l->l_proc->p_vmspace;
393 	}
394 	auio.uio_resid = count;
395 	vn_lock(vp, LK_SHARED | LK_RETRY);
396 	auio.uio_offset = fp->f_offset;
397 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
398 		    ncookies);
399 	FILE_LOCK(fp);
400 	fp->f_offset = auio.uio_offset;
401 	FILE_UNLOCK(fp);
402 	VOP_UNLOCK(vp, 0);
403 	if (error)
404 		return (error);
405 
406 #if defined(UNION) || defined(LKM)
407 	if (count == auio.uio_resid && vn_union_readdir_hook) {
408 		struct vnode *ovp = vp;
409 
410 		error = (*vn_union_readdir_hook)(&vp, fp, l);
411 		if (error)
412 			return (error);
413 		if (vp != ovp)
414 			goto unionread;
415 	}
416 #endif /* UNION || LKM */
417 
418 	if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) &&
419 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
420 		struct vnode *tvp = vp;
421 		vp = vp->v_mount->mnt_vnodecovered;
422 		VREF(vp);
423 		FILE_LOCK(fp);
424 		fp->f_data = vp;
425 		fp->f_offset = 0;
426 		FILE_UNLOCK(fp);
427 		vrele(tvp);
428 		goto unionread;
429 	}
430 	*done = count - auio.uio_resid;
431 	return error;
432 }
433 
434 /*
435  * File table vnode read routine.
436  */
437 static int
438 vn_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
439     int flags)
440 {
441 	struct vnode *vp = (struct vnode *)fp->f_data;
442 	int count, error, ioflag, fflag;
443 
444 	ioflag = IO_ADV_ENCODE(fp->f_advice);
445 	fflag = fp->f_flag;
446 	if (fflag & FNONBLOCK)
447 		ioflag |= IO_NDELAY;
448 	if ((fflag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
449 		ioflag |= IO_SYNC;
450 	if (fflag & FALTIO)
451 		ioflag |= IO_ALTSEMANTICS;
452 	if (fflag & FDIRECT)
453 		ioflag |= IO_DIRECT;
454 	vn_lock(vp, LK_SHARED | LK_RETRY);
455 	uio->uio_offset = *offset;
456 	count = uio->uio_resid;
457 	error = VOP_READ(vp, uio, ioflag, cred);
458 	if (flags & FOF_UPDATE_OFFSET)
459 		*offset += count - uio->uio_resid;
460 	VOP_UNLOCK(vp, 0);
461 	return (error);
462 }
463 
464 /*
465  * File table vnode write routine.
466  */
467 static int
468 vn_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
469     int flags)
470 {
471 	struct vnode *vp = (struct vnode *)fp->f_data;
472 	int count, error, ioflag, fflag;
473 
474 	ioflag = IO_ADV_ENCODE(fp->f_advice) | IO_UNIT;
475 	fflag = fp->f_flag;
476 	if (vp->v_type == VREG && (fflag & O_APPEND))
477 		ioflag |= IO_APPEND;
478 	if (fflag & FNONBLOCK)
479 		ioflag |= IO_NDELAY;
480 	if (fflag & FFSYNC ||
481 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
482 		ioflag |= IO_SYNC;
483 	else if (fflag & FDSYNC)
484 		ioflag |= IO_DSYNC;
485 	if (fflag & FALTIO)
486 		ioflag |= IO_ALTSEMANTICS;
487 	if (fflag & FDIRECT)
488 		ioflag |= IO_DIRECT;
489 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
490 	uio->uio_offset = *offset;
491 	count = uio->uio_resid;
492 	error = VOP_WRITE(vp, uio, ioflag, cred);
493 	if (flags & FOF_UPDATE_OFFSET) {
494 		if (ioflag & IO_APPEND)
495 			*offset = uio->uio_offset;
496 		else
497 			*offset += count - uio->uio_resid;
498 	}
499 	VOP_UNLOCK(vp, 0);
500 	return (error);
501 }
502 
503 /*
504  * File table vnode stat routine.
505  */
506 static int
507 vn_statfile(file_t *fp, struct stat *sb)
508 {
509 	struct vnode *vp = (struct vnode *)fp->f_data;
510 
511 	return vn_stat(vp, sb);
512 }
513 
514 int
515 vn_stat(struct vnode *vp, struct stat *sb)
516 {
517 	struct vattr va;
518 	int error;
519 	mode_t mode;
520 
521 	error = VOP_GETATTR(vp, &va, kauth_cred_get());
522 	if (error)
523 		return (error);
524 	/*
525 	 * Copy from vattr table
526 	 */
527 	sb->st_dev = va.va_fsid;
528 	sb->st_ino = va.va_fileid;
529 	mode = va.va_mode;
530 	switch (vp->v_type) {
531 	case VREG:
532 		mode |= S_IFREG;
533 		break;
534 	case VDIR:
535 		mode |= S_IFDIR;
536 		break;
537 	case VBLK:
538 		mode |= S_IFBLK;
539 		break;
540 	case VCHR:
541 		mode |= S_IFCHR;
542 		break;
543 	case VLNK:
544 		mode |= S_IFLNK;
545 		break;
546 	case VSOCK:
547 		mode |= S_IFSOCK;
548 		break;
549 	case VFIFO:
550 		mode |= S_IFIFO;
551 		break;
552 	default:
553 		return (EBADF);
554 	};
555 	sb->st_mode = mode;
556 	sb->st_nlink = va.va_nlink;
557 	sb->st_uid = va.va_uid;
558 	sb->st_gid = va.va_gid;
559 	sb->st_rdev = va.va_rdev;
560 	sb->st_size = va.va_size;
561 	sb->st_atimespec = va.va_atime;
562 	sb->st_mtimespec = va.va_mtime;
563 	sb->st_ctimespec = va.va_ctime;
564 	sb->st_birthtimespec = va.va_birthtime;
565 	sb->st_blksize = va.va_blocksize;
566 	sb->st_flags = va.va_flags;
567 	sb->st_gen = 0;
568 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
569 	return (0);
570 }
571 
572 /*
573  * File table vnode fcntl routine.
574  */
575 static int
576 vn_fcntl(file_t *fp, u_int com, void *data)
577 {
578 	struct vnode *vp = fp->f_data;
579 	int error;
580 
581 	error = VOP_FCNTL(vp, com, data, fp->f_flag, kauth_cred_get());
582 	return (error);
583 }
584 
585 /*
586  * File table vnode ioctl routine.
587  */
588 static int
589 vn_ioctl(file_t *fp, u_long com, void *data)
590 {
591 	struct vnode *vp = fp->f_data, *ovp;
592 	struct vattr vattr;
593 	int error;
594 
595 	switch (vp->v_type) {
596 
597 	case VREG:
598 	case VDIR:
599 		if (com == FIONREAD) {
600 			error = VOP_GETATTR(vp, &vattr,
601 			    kauth_cred_get());
602 			if (error)
603 				return (error);
604 			*(int *)data = vattr.va_size - fp->f_offset;
605 			return (0);
606 		}
607 		if ((com == FIONWRITE) || (com == FIONSPACE)) {
608 			/*
609 			 * Files don't have send queues, so there never
610 			 * are any bytes in them, nor is there any
611 			 * open space in them.
612 			 */
613 			*(int *)data = 0;
614 			return (0);
615 		}
616 		if (com == FIOGETBMAP) {
617 			daddr_t *block;
618 
619 			if (*(daddr_t *)data < 0)
620 				return (EINVAL);
621 			block = (daddr_t *)data;
622 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
623 		}
624 		if (com == OFIOGETBMAP) {
625 			daddr_t ibn, obn;
626 
627 			if (*(int32_t *)data < 0)
628 				return (EINVAL);
629 			ibn = (daddr_t)*(int32_t *)data;
630 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
631 			*(int32_t *)data = (int32_t)obn;
632 			return error;
633 		}
634 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
635 			return (0);			/* XXX */
636 		/* fall into ... */
637 	case VFIFO:
638 	case VCHR:
639 	case VBLK:
640 		error = VOP_IOCTL(vp, com, data, fp->f_flag,
641 		    kauth_cred_get());
642 		if (error == 0 && com == TIOCSCTTY) {
643 			VREF(vp);
644 			mutex_enter(proc_lock);
645 			ovp = curproc->p_session->s_ttyvp;
646 			curproc->p_session->s_ttyvp = vp;
647 			mutex_exit(proc_lock);
648 			if (ovp != NULL)
649 				vrele(ovp);
650 		}
651 		return (error);
652 
653 	default:
654 		return (EPASSTHROUGH);
655 	}
656 }
657 
658 /*
659  * File table vnode poll routine.
660  */
661 static int
662 vn_poll(file_t *fp, int events)
663 {
664 
665 	return (VOP_POLL(fp->f_data, events));
666 }
667 
668 /*
669  * File table vnode kqfilter routine.
670  */
671 int
672 vn_kqfilter(file_t *fp, struct knote *kn)
673 {
674 
675 	return (VOP_KQFILTER(fp->f_data, kn));
676 }
677 
678 /*
679  * Check that the vnode is still valid, and if so
680  * acquire requested lock.
681  */
682 int
683 vn_lock(struct vnode *vp, int flags)
684 {
685 	int error;
686 
687 #if 0
688 	KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0
689 	    || (vp->v_iflag & VI_ONWORKLST) != 0);
690 #endif
691 	KASSERT((flags &
692 	    ~(LK_INTERLOCK|LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT|LK_RETRY|
693 	    LK_CANRECURSE))
694 	    == 0);
695 
696 #ifdef DIAGNOSTIC
697 	if (wapbl_vphaswapbl(vp))
698 		WAPBL_JUNLOCK_ASSERT(wapbl_vptomp(vp));
699 #endif
700 
701 	do {
702 		/*
703 		 * XXX PR 37706 forced unmount of file systems is unsafe.
704 		 * Race between vclean() and this the remaining problem.
705 		 */
706 		if (vp->v_iflag & VI_XLOCK) {
707 			if ((flags & LK_INTERLOCK) == 0) {
708 				mutex_enter(&vp->v_interlock);
709 			}
710 			flags &= ~LK_INTERLOCK;
711 			if (flags & LK_NOWAIT) {
712 				mutex_exit(&vp->v_interlock);
713 				return EBUSY;
714 			}
715 			vwait(vp, VI_XLOCK);
716 			mutex_exit(&vp->v_interlock);
717 			error = ENOENT;
718 		} else {
719 			if ((flags & LK_INTERLOCK) != 0) {
720 				mutex_exit(&vp->v_interlock);
721 			}
722 			flags &= ~LK_INTERLOCK;
723 			error = VOP_LOCK(vp, (flags & ~LK_RETRY));
724 			if (error == 0 || error == EDEADLK || error == EBUSY)
725 				return (error);
726 		}
727 	} while (flags & LK_RETRY);
728 	return (error);
729 }
730 
731 /*
732  * File table vnode close routine.
733  */
734 static int
735 vn_closefile(file_t *fp)
736 {
737 
738 	return vn_close(fp->f_data, fp->f_flag, fp->f_cred);
739 }
740 
741 /*
742  * Enable LK_CANRECURSE on lock. Return prior status.
743  */
744 u_int
745 vn_setrecurse(struct vnode *vp)
746 {
747 	struct vnlock *lkp;
748 
749 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
750 	atomic_inc_uint(&lkp->vl_canrecurse);
751 
752 	return 0;
753 }
754 
755 /*
756  * Called when done with locksetrecurse.
757  */
758 void
759 vn_restorerecurse(struct vnode *vp, u_int flags)
760 {
761 	struct vnlock *lkp;
762 
763 	lkp = (vp->v_vnlock != NULL ? vp->v_vnlock : &vp->v_lock);
764 	atomic_dec_uint(&lkp->vl_canrecurse);
765 }
766 
767 /*
768  * Simplified in-kernel wrapper calls for extended attribute access.
769  * Both calls pass in a NULL credential, authorizing a "kernel" access.
770  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
771  */
772 int
773 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
774     const char *attrname, size_t *buflen, void *bf, struct lwp *l)
775 {
776 	struct uio auio;
777 	struct iovec aiov;
778 	int error;
779 
780 	aiov.iov_len = *buflen;
781 	aiov.iov_base = bf;
782 
783 	auio.uio_iov = &aiov;
784 	auio.uio_iovcnt = 1;
785 	auio.uio_rw = UIO_READ;
786 	auio.uio_offset = 0;
787 	auio.uio_resid = *buflen;
788 	UIO_SETUP_SYSSPACE(&auio);
789 
790 	if ((ioflg & IO_NODELOCKED) == 0)
791 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
792 
793 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL);
794 
795 	if ((ioflg & IO_NODELOCKED) == 0)
796 		VOP_UNLOCK(vp, 0);
797 
798 	if (error == 0)
799 		*buflen = *buflen - auio.uio_resid;
800 
801 	return (error);
802 }
803 
804 /*
805  * XXX Failure mode if partially written?
806  */
807 int
808 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
809     const char *attrname, size_t buflen, const void *bf, struct lwp *l)
810 {
811 	struct uio auio;
812 	struct iovec aiov;
813 	int error;
814 
815 	aiov.iov_len = buflen;
816 	aiov.iov_base = __UNCONST(bf);		/* XXXUNCONST kills const */
817 
818 	auio.uio_iov = &aiov;
819 	auio.uio_iovcnt = 1;
820 	auio.uio_rw = UIO_WRITE;
821 	auio.uio_offset = 0;
822 	auio.uio_resid = buflen;
823 	UIO_SETUP_SYSSPACE(&auio);
824 
825 	if ((ioflg & IO_NODELOCKED) == 0) {
826 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
827 	}
828 
829 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL);
830 
831 	if ((ioflg & IO_NODELOCKED) == 0) {
832 		VOP_UNLOCK(vp, 0);
833 	}
834 
835 	return (error);
836 }
837 
838 int
839 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
840     const char *attrname, struct lwp *l)
841 {
842 	int error;
843 
844 	if ((ioflg & IO_NODELOCKED) == 0) {
845 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
846 	}
847 
848 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL);
849 	if (error == EOPNOTSUPP)
850 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL);
851 
852 	if ((ioflg & IO_NODELOCKED) == 0) {
853 		VOP_UNLOCK(vp, 0);
854 	}
855 
856 	return (error);
857 }
858 
859 void
860 vn_ra_allocctx(struct vnode *vp)
861 {
862 	struct uvm_ractx *ra = NULL;
863 
864 	KASSERT(mutex_owned(&vp->v_interlock));
865 
866 	if (vp->v_type != VREG) {
867 		return;
868 	}
869 	if (vp->v_ractx != NULL) {
870 		return;
871 	}
872 	if (vp->v_ractx == NULL) {
873 		mutex_exit(&vp->v_interlock);
874 		ra = uvm_ra_allocctx();
875 		mutex_enter(&vp->v_interlock);
876 		if (ra != NULL && vp->v_ractx == NULL) {
877 			vp->v_ractx = ra;
878 			ra = NULL;
879 		}
880 	}
881 	if (ra != NULL) {
882 		uvm_ra_freectx(ra);
883 	}
884 }
885