xref: /netbsd-src/sys/kern/vfs_vnops.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: vfs_vnops.c,v 1.190 2014/06/22 18:32:27 maxv 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 1.190 2014/06/22 18:32:27 maxv Exp $");
70 
71 #include "veriexec.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/kernel.h>
76 #include <sys/file.h>
77 #include <sys/stat.h>
78 #include <sys/buf.h>
79 #include <sys/proc.h>
80 #include <sys/mount.h>
81 #include <sys/namei.h>
82 #include <sys/vnode.h>
83 #include <sys/ioctl.h>
84 #include <sys/tty.h>
85 #include <sys/poll.h>
86 #include <sys/kauth.h>
87 #include <sys/syslog.h>
88 #include <sys/fstrans.h>
89 #include <sys/atomic.h>
90 #include <sys/filedesc.h>
91 #include <sys/wapbl.h>
92 
93 #include <miscfs/specfs/specdev.h>
94 #include <miscfs/fifofs/fifo.h>
95 
96 #include <uvm/uvm_extern.h>
97 #include <uvm/uvm_readahead.h>
98 
99 #ifdef UNION
100 #include <fs/union/union.h>
101 #endif
102 
103 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct lwp *);
104 
105 #include <sys/verified_exec.h>
106 
107 static int vn_read(file_t *fp, off_t *offset, struct uio *uio,
108 	    kauth_cred_t cred, int flags);
109 static int vn_write(file_t *fp, off_t *offset, struct uio *uio,
110 	    kauth_cred_t cred, int flags);
111 static int vn_closefile(file_t *fp);
112 static int vn_poll(file_t *fp, int events);
113 static int vn_fcntl(file_t *fp, u_int com, void *data);
114 static int vn_statfile(file_t *fp, struct stat *sb);
115 static int vn_ioctl(file_t *fp, u_long com, void *data);
116 
117 const struct fileops vnops = {
118 	.fo_read = vn_read,
119 	.fo_write = vn_write,
120 	.fo_ioctl = vn_ioctl,
121 	.fo_fcntl = vn_fcntl,
122 	.fo_poll = vn_poll,
123 	.fo_stat = vn_statfile,
124 	.fo_close = vn_closefile,
125 	.fo_kqfilter = vn_kqfilter,
126 	.fo_restart = fnullop_restart,
127 };
128 
129 /*
130  * Common code for vnode open operations.
131  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
132  */
133 int
134 vn_open(struct nameidata *ndp, int fmode, int cmode)
135 {
136 	struct vnode *vp;
137 	struct lwp *l = curlwp;
138 	kauth_cred_t cred = l->l_cred;
139 	struct vattr va;
140 	int error;
141 	const char *pathstring;
142 
143 	if ((fmode & (O_CREAT | O_DIRECTORY)) == (O_CREAT | O_DIRECTORY))
144 		return EINVAL;
145 
146 	ndp->ni_cnd.cn_flags &= TRYEMULROOT | NOCHROOT;
147 
148 	if (fmode & O_CREAT) {
149 		ndp->ni_cnd.cn_nameiop = CREATE;
150 		ndp->ni_cnd.cn_flags |= LOCKPARENT | LOCKLEAF;
151 		if ((fmode & O_EXCL) == 0 &&
152 		    ((fmode & O_NOFOLLOW) == 0))
153 			ndp->ni_cnd.cn_flags |= FOLLOW;
154 	} else {
155 		ndp->ni_cnd.cn_nameiop = LOOKUP;
156 		ndp->ni_cnd.cn_flags |= LOCKLEAF;
157 		if ((fmode & O_NOFOLLOW) == 0)
158 			ndp->ni_cnd.cn_flags |= FOLLOW;
159 	}
160 
161 	pathstring = pathbuf_stringcopy_get(ndp->ni_pathbuf);
162 	if (pathstring == NULL) {
163 		return ENOMEM;
164 	}
165 
166 	error = namei(ndp);
167 	if (error)
168 		goto out;
169 
170 	vp = ndp->ni_vp;
171 
172 #if NVERIEXEC > 0
173 	error = veriexec_openchk(l, ndp->ni_vp, pathstring, fmode);
174 	if (error) {
175 		/* We have to release the locks ourselves */
176 		if (fmode & O_CREAT) {
177 			if (vp == NULL) {
178 				vput(ndp->ni_dvp);
179 			} else {
180 				VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
181 				if (ndp->ni_dvp == ndp->ni_vp)
182 					vrele(ndp->ni_dvp);
183 				else
184 					vput(ndp->ni_dvp);
185 				ndp->ni_dvp = NULL;
186 				vput(vp);
187 			}
188 		} else {
189 			vput(vp);
190 		}
191 		goto out;
192 	}
193 #endif /* NVERIEXEC > 0 */
194 
195 	if (fmode & O_CREAT) {
196 		if (ndp->ni_vp == NULL) {
197 			vattr_null(&va);
198 			va.va_type = VREG;
199 			va.va_mode = cmode;
200 			if (fmode & O_EXCL)
201 				 va.va_vaflags |= VA_EXCLUSIVE;
202 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
203 					   &ndp->ni_cnd, &va);
204 			vput(ndp->ni_dvp);
205 			if (error)
206 				goto out;
207 			fmode &= ~O_TRUNC;
208 			vp = ndp->ni_vp;
209 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
210 		} else {
211 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
212 			if (ndp->ni_dvp == ndp->ni_vp)
213 				vrele(ndp->ni_dvp);
214 			else
215 				vput(ndp->ni_dvp);
216 			ndp->ni_dvp = NULL;
217 			vp = ndp->ni_vp;
218 			if (fmode & O_EXCL) {
219 				error = EEXIST;
220 				goto bad;
221 			}
222 			fmode &= ~O_CREAT;
223 		}
224 	} else {
225 		vp = ndp->ni_vp;
226 	}
227 	if (vp->v_type == VSOCK) {
228 		error = EOPNOTSUPP;
229 		goto bad;
230 	}
231 	if (ndp->ni_vp->v_type == VLNK) {
232 		error = EFTYPE;
233 		goto bad;
234 	}
235 
236 	if ((fmode & O_CREAT) == 0) {
237 		error = vn_openchk(vp, cred, fmode);
238 		if (error != 0)
239 			goto bad;
240 	}
241 
242 	if (fmode & O_TRUNC) {
243 		vattr_null(&va);
244 		va.va_size = 0;
245 		error = VOP_SETATTR(vp, &va, cred);
246 		if (error != 0)
247 			goto bad;
248 	}
249 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0)
250 		goto bad;
251 	if (fmode & FWRITE) {
252 		mutex_enter(vp->v_interlock);
253 		vp->v_writecount++;
254 		mutex_exit(vp->v_interlock);
255 	}
256 
257 bad:
258 	if (error)
259 		vput(vp);
260 out:
261 	pathbuf_stringcopy_put(ndp->ni_pathbuf, pathstring);
262 	return (error);
263 }
264 
265 /*
266  * Check for write permissions on the specified vnode.
267  * Prototype text segments cannot be written.
268  */
269 int
270 vn_writechk(struct vnode *vp)
271 {
272 
273 	/*
274 	 * If the vnode is in use as a process's text,
275 	 * we can't allow writing.
276 	 */
277 	if (vp->v_iflag & VI_TEXT)
278 		return (ETXTBSY);
279 	return (0);
280 }
281 
282 int
283 vn_openchk(struct vnode *vp, kauth_cred_t cred, int fflags)
284 {
285 	int permbits = 0;
286 	int error;
287 
288 	if ((fflags & O_DIRECTORY) != 0 && vp->v_type != VDIR)
289 		return ENOTDIR;
290 
291 	if ((fflags & FREAD) != 0) {
292 		permbits = VREAD;
293 	}
294 	if ((fflags & (FWRITE | O_TRUNC)) != 0) {
295 		permbits |= VWRITE;
296 		if (vp->v_type == VDIR) {
297 			error = EISDIR;
298 			goto bad;
299 		}
300 		error = vn_writechk(vp);
301 		if (error != 0)
302 			goto bad;
303 	}
304 	error = VOP_ACCESS(vp, permbits, cred);
305 bad:
306 	return error;
307 }
308 
309 /*
310  * Mark a vnode as having executable mappings.
311  */
312 void
313 vn_markexec(struct vnode *vp)
314 {
315 
316 	if ((vp->v_iflag & VI_EXECMAP) != 0) {
317 		/* Safe unlocked, as long as caller holds a reference. */
318 		return;
319 	}
320 
321 	mutex_enter(vp->v_interlock);
322 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
323 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
324 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
325 		vp->v_iflag |= VI_EXECMAP;
326 	}
327 	mutex_exit(vp->v_interlock);
328 }
329 
330 /*
331  * Mark a vnode as being the text of a process.
332  * Fail if the vnode is currently writable.
333  */
334 int
335 vn_marktext(struct vnode *vp)
336 {
337 
338 	if ((vp->v_iflag & (VI_TEXT|VI_EXECMAP)) == (VI_TEXT|VI_EXECMAP)) {
339 		/* Safe unlocked, as long as caller holds a reference. */
340 		return (0);
341 	}
342 
343 	mutex_enter(vp->v_interlock);
344 	if (vp->v_writecount != 0) {
345 		KASSERT((vp->v_iflag & VI_TEXT) == 0);
346 		mutex_exit(vp->v_interlock);
347 		return (ETXTBSY);
348 	}
349 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
350 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
351 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
352 	}
353 	vp->v_iflag |= (VI_TEXT | VI_EXECMAP);
354 	mutex_exit(vp->v_interlock);
355 	return (0);
356 }
357 
358 /*
359  * Vnode close call
360  *
361  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
362  */
363 int
364 vn_close(struct vnode *vp, int flags, kauth_cred_t cred)
365 {
366 	int error;
367 
368 	if (flags & FWRITE) {
369 		mutex_enter(vp->v_interlock);
370 		KASSERT(vp->v_writecount > 0);
371 		vp->v_writecount--;
372 		mutex_exit(vp->v_interlock);
373 	}
374 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
375 	error = VOP_CLOSE(vp, flags, cred);
376 	vput(vp);
377 	return (error);
378 }
379 
380 static int
381 enforce_rlimit_fsize(struct vnode *vp, struct uio *uio, int ioflag)
382 {
383 	struct lwp *l = curlwp;
384 	off_t testoff;
385 
386 	if (uio->uio_rw != UIO_WRITE || vp->v_type != VREG)
387 		return 0;
388 
389 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
390 	if (ioflag & IO_APPEND)
391 		testoff = vp->v_size;
392 	else
393 		testoff = uio->uio_offset;
394 
395 	if (testoff + uio->uio_resid >
396 	    l->l_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
397 		mutex_enter(proc_lock);
398 		psignal(l->l_proc, SIGXFSZ);
399 		mutex_exit(proc_lock);
400 		return EFBIG;
401 	}
402 
403 	return 0;
404 }
405 
406 /*
407  * Package up an I/O request on a vnode into a uio and do it.
408  */
409 int
410 vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset,
411     enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid,
412     struct lwp *l)
413 {
414 	struct uio auio;
415 	struct iovec aiov;
416 	int error;
417 
418 	if ((ioflg & IO_NODELOCKED) == 0) {
419 		if (rw == UIO_READ) {
420 			vn_lock(vp, LK_SHARED | LK_RETRY);
421 		} else /* UIO_WRITE */ {
422 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
423 		}
424 	}
425 	auio.uio_iov = &aiov;
426 	auio.uio_iovcnt = 1;
427 	aiov.iov_base = base;
428 	aiov.iov_len = len;
429 	auio.uio_resid = len;
430 	auio.uio_offset = offset;
431 	auio.uio_rw = rw;
432 	if (segflg == UIO_SYSSPACE) {
433 		UIO_SETUP_SYSSPACE(&auio);
434 	} else {
435 		auio.uio_vmspace = l->l_proc->p_vmspace;
436 	}
437 
438 	if ((error = enforce_rlimit_fsize(vp, &auio, ioflg)) != 0)
439 		goto out;
440 
441 	if (rw == UIO_READ) {
442 		error = VOP_READ(vp, &auio, ioflg, cred);
443 	} else {
444 		error = VOP_WRITE(vp, &auio, ioflg, cred);
445 	}
446 
447 	if (aresid)
448 		*aresid = auio.uio_resid;
449 	else
450 		if (auio.uio_resid && error == 0)
451 			error = EIO;
452 
453  out:
454 	if ((ioflg & IO_NODELOCKED) == 0) {
455 		VOP_UNLOCK(vp);
456 	}
457 	return (error);
458 }
459 
460 int
461 vn_readdir(file_t *fp, char *bf, int segflg, u_int count, int *done,
462     struct lwp *l, off_t **cookies, int *ncookies)
463 {
464 	struct vnode *vp = (struct vnode *)fp->f_data;
465 	struct iovec aiov;
466 	struct uio auio;
467 	int error, eofflag;
468 
469 	/* Limit the size on any kernel buffers used by VOP_READDIR */
470 	count = min(MAXBSIZE, count);
471 
472 unionread:
473 	if (vp->v_type != VDIR)
474 		return (EINVAL);
475 	aiov.iov_base = bf;
476 	aiov.iov_len = count;
477 	auio.uio_iov = &aiov;
478 	auio.uio_iovcnt = 1;
479 	auio.uio_rw = UIO_READ;
480 	if (segflg == UIO_SYSSPACE) {
481 		UIO_SETUP_SYSSPACE(&auio);
482 	} else {
483 		KASSERT(l == curlwp);
484 		auio.uio_vmspace = l->l_proc->p_vmspace;
485 	}
486 	auio.uio_resid = count;
487 	vn_lock(vp, LK_SHARED | LK_RETRY);
488 	auio.uio_offset = fp->f_offset;
489 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
490 		    ncookies);
491 	mutex_enter(&fp->f_lock);
492 	fp->f_offset = auio.uio_offset;
493 	mutex_exit(&fp->f_lock);
494 	VOP_UNLOCK(vp);
495 	if (error)
496 		return (error);
497 
498 	if (count == auio.uio_resid && vn_union_readdir_hook) {
499 		struct vnode *ovp = vp;
500 
501 		error = (*vn_union_readdir_hook)(&vp, fp, l);
502 		if (error)
503 			return (error);
504 		if (vp != ovp)
505 			goto unionread;
506 	}
507 
508 	if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) &&
509 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
510 		struct vnode *tvp = vp;
511 		vp = vp->v_mount->mnt_vnodecovered;
512 		vref(vp);
513 		mutex_enter(&fp->f_lock);
514 		fp->f_data = vp;
515 		fp->f_offset = 0;
516 		mutex_exit(&fp->f_lock);
517 		vrele(tvp);
518 		goto unionread;
519 	}
520 	*done = count - auio.uio_resid;
521 	return error;
522 }
523 
524 /*
525  * File table vnode read routine.
526  */
527 static int
528 vn_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
529     int flags)
530 {
531 	struct vnode *vp = (struct vnode *)fp->f_data;
532 	int error, ioflag, fflag;
533 	size_t count;
534 
535 	ioflag = IO_ADV_ENCODE(fp->f_advice);
536 	fflag = fp->f_flag;
537 	if (fflag & FNONBLOCK)
538 		ioflag |= IO_NDELAY;
539 	if ((fflag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
540 		ioflag |= IO_SYNC;
541 	if (fflag & FALTIO)
542 		ioflag |= IO_ALTSEMANTICS;
543 	if (fflag & FDIRECT)
544 		ioflag |= IO_DIRECT;
545 	vn_lock(vp, LK_SHARED | LK_RETRY);
546 	uio->uio_offset = *offset;
547 	count = uio->uio_resid;
548 	error = VOP_READ(vp, uio, ioflag, cred);
549 	if (flags & FOF_UPDATE_OFFSET)
550 		*offset += count - uio->uio_resid;
551 	VOP_UNLOCK(vp);
552 	return (error);
553 }
554 
555 /*
556  * File table vnode write routine.
557  */
558 static int
559 vn_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
560     int flags)
561 {
562 	struct vnode *vp = (struct vnode *)fp->f_data;
563 	int error, ioflag, fflag;
564 	size_t count;
565 
566 	ioflag = IO_ADV_ENCODE(fp->f_advice) | IO_UNIT;
567 	fflag = fp->f_flag;
568 	if (vp->v_type == VREG && (fflag & O_APPEND))
569 		ioflag |= IO_APPEND;
570 	if (fflag & FNONBLOCK)
571 		ioflag |= IO_NDELAY;
572 	if (fflag & FFSYNC ||
573 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
574 		ioflag |= IO_SYNC;
575 	else if (fflag & FDSYNC)
576 		ioflag |= IO_DSYNC;
577 	if (fflag & FALTIO)
578 		ioflag |= IO_ALTSEMANTICS;
579 	if (fflag & FDIRECT)
580 		ioflag |= IO_DIRECT;
581 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
582 	uio->uio_offset = *offset;
583 	count = uio->uio_resid;
584 
585 	if ((error = enforce_rlimit_fsize(vp, uio, ioflag)) != 0)
586 		goto out;
587 
588 	error = VOP_WRITE(vp, uio, ioflag, cred);
589 
590 	if (flags & FOF_UPDATE_OFFSET) {
591 		if (ioflag & IO_APPEND) {
592 			/*
593 			 * SUSv3 describes behaviour for count = 0 as following:
594 			 * "Before any action ... is taken, and if nbyte is zero
595 			 * and the file is a regular file, the write() function
596 			 * ... in the absence of errors ... shall return zero
597 			 * and have no other results."
598 			 */
599 			if (count)
600 				*offset = uio->uio_offset;
601 		} else
602 			*offset += count - uio->uio_resid;
603 	}
604 
605  out:
606 	VOP_UNLOCK(vp);
607 	return (error);
608 }
609 
610 /*
611  * File table vnode stat routine.
612  */
613 static int
614 vn_statfile(file_t *fp, struct stat *sb)
615 {
616 	struct vnode *vp = fp->f_data;
617 	int error;
618 
619 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
620 	error = vn_stat(vp, sb);
621 	VOP_UNLOCK(vp);
622 	return error;
623 }
624 
625 int
626 vn_stat(struct vnode *vp, struct stat *sb)
627 {
628 	struct vattr va;
629 	int error;
630 	mode_t mode;
631 
632 	memset(&va, 0, sizeof(va));
633 	error = VOP_GETATTR(vp, &va, kauth_cred_get());
634 	if (error)
635 		return (error);
636 	/*
637 	 * Copy from vattr table
638 	 */
639 	memset(sb, 0, sizeof(*sb));
640 	sb->st_dev = va.va_fsid;
641 	sb->st_ino = va.va_fileid;
642 	mode = va.va_mode;
643 	switch (vp->v_type) {
644 	case VREG:
645 		mode |= S_IFREG;
646 		break;
647 	case VDIR:
648 		mode |= S_IFDIR;
649 		break;
650 	case VBLK:
651 		mode |= S_IFBLK;
652 		break;
653 	case VCHR:
654 		mode |= S_IFCHR;
655 		break;
656 	case VLNK:
657 		mode |= S_IFLNK;
658 		break;
659 	case VSOCK:
660 		mode |= S_IFSOCK;
661 		break;
662 	case VFIFO:
663 		mode |= S_IFIFO;
664 		break;
665 	default:
666 		return (EBADF);
667 	};
668 	sb->st_mode = mode;
669 	sb->st_nlink = va.va_nlink;
670 	sb->st_uid = va.va_uid;
671 	sb->st_gid = va.va_gid;
672 	sb->st_rdev = va.va_rdev;
673 	sb->st_size = va.va_size;
674 	sb->st_atimespec = va.va_atime;
675 	sb->st_mtimespec = va.va_mtime;
676 	sb->st_ctimespec = va.va_ctime;
677 	sb->st_birthtimespec = va.va_birthtime;
678 	sb->st_blksize = va.va_blocksize;
679 	sb->st_flags = va.va_flags;
680 	sb->st_gen = 0;
681 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
682 	return (0);
683 }
684 
685 /*
686  * File table vnode fcntl routine.
687  */
688 static int
689 vn_fcntl(file_t *fp, u_int com, void *data)
690 {
691 	struct vnode *vp = fp->f_data;
692 	int error;
693 
694 	error = VOP_FCNTL(vp, com, data, fp->f_flag, kauth_cred_get());
695 	return (error);
696 }
697 
698 /*
699  * File table vnode ioctl routine.
700  */
701 static int
702 vn_ioctl(file_t *fp, u_long com, void *data)
703 {
704 	struct vnode *vp = fp->f_data, *ovp;
705 	struct vattr vattr;
706 	int error;
707 
708 	switch (vp->v_type) {
709 
710 	case VREG:
711 	case VDIR:
712 		if (com == FIONREAD) {
713 			vn_lock(vp, LK_SHARED | LK_RETRY);
714 			error = VOP_GETATTR(vp, &vattr, kauth_cred_get());
715 			VOP_UNLOCK(vp);
716 			if (error)
717 				return (error);
718 			*(int *)data = vattr.va_size - fp->f_offset;
719 			return (0);
720 		}
721 		if ((com == FIONWRITE) || (com == FIONSPACE)) {
722 			/*
723 			 * Files don't have send queues, so there never
724 			 * are any bytes in them, nor is there any
725 			 * open space in them.
726 			 */
727 			*(int *)data = 0;
728 			return (0);
729 		}
730 		if (com == FIOGETBMAP) {
731 			daddr_t *block;
732 
733 			if (*(daddr_t *)data < 0)
734 				return (EINVAL);
735 			block = (daddr_t *)data;
736 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
737 		}
738 		if (com == OFIOGETBMAP) {
739 			daddr_t ibn, obn;
740 
741 			if (*(int32_t *)data < 0)
742 				return (EINVAL);
743 			ibn = (daddr_t)*(int32_t *)data;
744 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
745 			*(int32_t *)data = (int32_t)obn;
746 			return error;
747 		}
748 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
749 			return (0);			/* XXX */
750 		/* fall into ... */
751 	case VFIFO:
752 	case VCHR:
753 	case VBLK:
754 		error = VOP_IOCTL(vp, com, data, fp->f_flag,
755 		    kauth_cred_get());
756 		if (error == 0 && com == TIOCSCTTY) {
757 			vref(vp);
758 			mutex_enter(proc_lock);
759 			ovp = curproc->p_session->s_ttyvp;
760 			curproc->p_session->s_ttyvp = vp;
761 			mutex_exit(proc_lock);
762 			if (ovp != NULL)
763 				vrele(ovp);
764 		}
765 		return (error);
766 
767 	default:
768 		return (EPASSTHROUGH);
769 	}
770 }
771 
772 /*
773  * File table vnode poll routine.
774  */
775 static int
776 vn_poll(file_t *fp, int events)
777 {
778 
779 	return (VOP_POLL(fp->f_data, events));
780 }
781 
782 /*
783  * File table vnode kqfilter routine.
784  */
785 int
786 vn_kqfilter(file_t *fp, struct knote *kn)
787 {
788 
789 	return (VOP_KQFILTER(fp->f_data, kn));
790 }
791 
792 /*
793  * Check that the vnode is still valid, and if so
794  * acquire requested lock.
795  */
796 int
797 vn_lock(struct vnode *vp, int flags)
798 {
799 	int error;
800 
801 #if 0
802 	KASSERT(vp->v_usecount > 0 || (vp->v_iflag & VI_ONWORKLST) != 0);
803 #endif
804 	KASSERT((flags & ~(LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT|LK_RETRY)) == 0);
805 	KASSERT(!mutex_owned(vp->v_interlock));
806 
807 #ifdef DIAGNOSTIC
808 	if (wapbl_vphaswapbl(vp))
809 		WAPBL_JUNLOCK_ASSERT(wapbl_vptomp(vp));
810 #endif
811 
812 	error = VOP_LOCK(vp, flags);
813 	if ((flags & LK_RETRY) != 0 && error == ENOENT)
814 		error = VOP_LOCK(vp, flags);
815 
816 	KASSERT((flags & LK_RETRY) == 0 || (flags & LK_NOWAIT) != 0 ||
817 	    error == 0);
818 
819 	return error;
820 }
821 
822 /*
823  * File table vnode close routine.
824  */
825 static int
826 vn_closefile(file_t *fp)
827 {
828 
829 	return vn_close(fp->f_data, fp->f_flag, fp->f_cred);
830 }
831 
832 /*
833  * Simplified in-kernel wrapper calls for extended attribute access.
834  * Both calls pass in a NULL credential, authorizing a "kernel" access.
835  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
836  */
837 int
838 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
839     const char *attrname, size_t *buflen, void *bf, struct lwp *l)
840 {
841 	struct uio auio;
842 	struct iovec aiov;
843 	int error;
844 
845 	aiov.iov_len = *buflen;
846 	aiov.iov_base = bf;
847 
848 	auio.uio_iov = &aiov;
849 	auio.uio_iovcnt = 1;
850 	auio.uio_rw = UIO_READ;
851 	auio.uio_offset = 0;
852 	auio.uio_resid = *buflen;
853 	UIO_SETUP_SYSSPACE(&auio);
854 
855 	if ((ioflg & IO_NODELOCKED) == 0)
856 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
857 
858 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL);
859 
860 	if ((ioflg & IO_NODELOCKED) == 0)
861 		VOP_UNLOCK(vp);
862 
863 	if (error == 0)
864 		*buflen = *buflen - auio.uio_resid;
865 
866 	return (error);
867 }
868 
869 /*
870  * XXX Failure mode if partially written?
871  */
872 int
873 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
874     const char *attrname, size_t buflen, const void *bf, struct lwp *l)
875 {
876 	struct uio auio;
877 	struct iovec aiov;
878 	int error;
879 
880 	aiov.iov_len = buflen;
881 	aiov.iov_base = __UNCONST(bf);		/* XXXUNCONST kills const */
882 
883 	auio.uio_iov = &aiov;
884 	auio.uio_iovcnt = 1;
885 	auio.uio_rw = UIO_WRITE;
886 	auio.uio_offset = 0;
887 	auio.uio_resid = buflen;
888 	UIO_SETUP_SYSSPACE(&auio);
889 
890 	if ((ioflg & IO_NODELOCKED) == 0) {
891 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
892 	}
893 
894 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL);
895 
896 	if ((ioflg & IO_NODELOCKED) == 0) {
897 		VOP_UNLOCK(vp);
898 	}
899 
900 	return (error);
901 }
902 
903 int
904 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
905     const char *attrname, struct lwp *l)
906 {
907 	int error;
908 
909 	if ((ioflg & IO_NODELOCKED) == 0) {
910 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
911 	}
912 
913 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL);
914 	if (error == EOPNOTSUPP)
915 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL);
916 
917 	if ((ioflg & IO_NODELOCKED) == 0) {
918 		VOP_UNLOCK(vp);
919 	}
920 
921 	return (error);
922 }
923 
924 void
925 vn_ra_allocctx(struct vnode *vp)
926 {
927 	struct uvm_ractx *ra = NULL;
928 
929 	KASSERT(mutex_owned(vp->v_interlock));
930 
931 	if (vp->v_type != VREG) {
932 		return;
933 	}
934 	if (vp->v_ractx != NULL) {
935 		return;
936 	}
937 	if (vp->v_ractx == NULL) {
938 		mutex_exit(vp->v_interlock);
939 		ra = uvm_ra_allocctx();
940 		mutex_enter(vp->v_interlock);
941 		if (ra != NULL && vp->v_ractx == NULL) {
942 			vp->v_ractx = ra;
943 			ra = NULL;
944 		}
945 	}
946 	if (ra != NULL) {
947 		uvm_ra_freectx(ra);
948 	}
949 }
950 
951 int
952 vn_fifo_bypass(void *v)
953 {
954 	struct vop_generic_args *ap = v;
955 
956 	return VOCALL(fifo_vnodeop_p, ap->a_desc->vdesc_offset, v);
957 }
958