1/* $NetBSD: vfs_vnops.c,v 1.193 2015/02/04 07:09:37 msaitoh 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 1.193 2015/02/04 07:09:37 msaitoh 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#include <sys/mman.h>
93
94#include <miscfs/specfs/specdev.h>
95#include <miscfs/fifofs/fifo.h>
96
97#include <uvm/uvm_extern.h>
98#include <uvm/uvm_readahead.h>
99#include <uvm/uvm_device.h>
100
101#ifdef UNION
102#include <fs/union/union.h>
103#endif
104
105#ifndef COMPAT_ZERODEV
106#define COMPAT_ZERODEV(dev) (0)
107#endif
108
109int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct lwp *);
110
111#include <sys/verified_exec.h>
112
113static int vn_read(file_t *fp, off_t *offset, struct uio *uio,
114 kauth_cred_t cred, int flags);
115static int vn_write(file_t *fp, off_t *offset, struct uio *uio,
116 kauth_cred_t cred, int flags);
117static int vn_closefile(file_t *fp);
118static int vn_poll(file_t *fp, int events);
119static int vn_fcntl(file_t *fp, u_int com, void *data);
120static int vn_statfile(file_t *fp, struct stat *sb);
121static int vn_ioctl(file_t *fp, u_long com, void *data);
122static int vn_mmap(struct file *, off_t *, size_t, int, int *, int *,
123 struct uvm_object **, int *);
124
125const struct fileops vnops = {
126 .fo_read = vn_read,
127 .fo_write = vn_write,
128 .fo_ioctl = vn_ioctl,
129 .fo_fcntl = vn_fcntl,
130 .fo_poll = vn_poll,
131 .fo_stat = vn_statfile,
132 .fo_close = vn_closefile,
133 .fo_kqfilter = vn_kqfilter,
134 .fo_restart = fnullop_restart,
135 .fo_mmap = vn_mmap,
136};
137
138/*
139 * Common code for vnode open operations.
140 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
141 */
142int
143vn_open(struct nameidata *ndp, int fmode, int cmode)
144{
145 struct vnode *vp;
146 struct lwp *l = curlwp;
147 kauth_cred_t cred = l->l_cred;
148 struct vattr va;
149 int error;
150 const char *pathstring;
151
152 if ((fmode & (O_CREAT | O_DIRECTORY)) == (O_CREAT | O_DIRECTORY))
153 return EINVAL;
154
155 ndp->ni_cnd.cn_flags &= TRYEMULROOT | NOCHROOT;
156
157 if (fmode & O_CREAT) {
158 ndp->ni_cnd.cn_nameiop = CREATE;
159 ndp->ni_cnd.cn_flags |= LOCKPARENT | LOCKLEAF;
160 if ((fmode & O_EXCL) == 0 &&
161 ((fmode & O_NOFOLLOW) == 0))
162 ndp->ni_cnd.cn_flags |= FOLLOW;
163 } else {
164 ndp->ni_cnd.cn_nameiop = LOOKUP;
165 ndp->ni_cnd.cn_flags |= LOCKLEAF;
166 if ((fmode & O_NOFOLLOW) == 0)
167 ndp->ni_cnd.cn_flags |= FOLLOW;
168 }
169
170 pathstring = pathbuf_stringcopy_get(ndp->ni_pathbuf);
171 if (pathstring == NULL) {
172 return ENOMEM;
173 }
174
175 error = namei(ndp);
176 if (error)
177 goto out;
178
179 vp = ndp->ni_vp;
180
181#if NVERIEXEC > 0
182 error = veriexec_openchk(l, ndp->ni_vp, pathstring, fmode);
183 if (error) {
184 /* We have to release the locks ourselves */
185 if (fmode & O_CREAT) {
186 if (vp == NULL) {
187 vput(ndp->ni_dvp);
188 } else {
189 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
190 if (ndp->ni_dvp == ndp->ni_vp)
191 vrele(ndp->ni_dvp);
192 else
193 vput(ndp->ni_dvp);
194 ndp->ni_dvp = NULL;
195 vput(vp);
196 }
197 } else {
198 vput(vp);
199 }
200 goto out;
201 }
202#endif /* NVERIEXEC > 0 */
203
204 if (fmode & O_CREAT) {
205 if (ndp->ni_vp == NULL) {
206 vattr_null(&va);
207 va.va_type = VREG;
208 va.va_mode = cmode;
209 if (fmode & O_EXCL)
210 va.va_vaflags |= VA_EXCLUSIVE;
211 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
212 &ndp->ni_cnd, &va);
213 vput(ndp->ni_dvp);
214 if (error)
215 goto out;
216 fmode &= ~O_TRUNC;
217 vp = ndp->ni_vp;
218 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
219 } else {
220 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
221 if (ndp->ni_dvp == ndp->ni_vp)
222 vrele(ndp->ni_dvp);
223 else
224 vput(ndp->ni_dvp);
225 ndp->ni_dvp = NULL;
226 vp = ndp->ni_vp;
227 if (fmode & O_EXCL) {
228 error = EEXIST;
229 goto bad;
230 }
231 fmode &= ~O_CREAT;
232 }
233 } else {
234 vp = ndp->ni_vp;
235 }
236 if (vp->v_type == VSOCK) {
237 error = EOPNOTSUPP;
238 goto bad;
239 }
240 if (ndp->ni_vp->v_type == VLNK) {
241 error = EFTYPE;
242 goto bad;
243 }
244
245 if ((fmode & O_CREAT) == 0) {
246 error = vn_openchk(vp, cred, fmode);
247 if (error != 0)
248 goto bad;
249 }
250
251 if (fmode & O_TRUNC) {
252 vattr_null(&va);
253 va.va_size = 0;
254 error = VOP_SETATTR(vp, &va, cred);
255 if (error != 0)
256 goto bad;
257 }
258 if ((error = VOP_OPEN(vp, fmode, cred)) != 0)
259 goto bad;
260 if (fmode & FWRITE) {
261 mutex_enter(vp->v_interlock);
262 vp->v_writecount++;
263 mutex_exit(vp->v_interlock);
264 }
265
266bad:
267 if (error)
268 vput(vp);
269out:
270 pathbuf_stringcopy_put(ndp->ni_pathbuf, pathstring);
271 return (error);
272}
273
274/*
275 * Check for write permissions on the specified vnode.
276 * Prototype text segments cannot be written.
277 */
278int
279vn_writechk(struct vnode *vp)
280{
281
282 /*
283 * If the vnode is in use as a process's text,
284 * we can't allow writing.
285 */
286 if (vp->v_iflag & VI_TEXT)
287 return (ETXTBSY);
288 return (0);
289}
290
291int
292vn_openchk(struct vnode *vp, kauth_cred_t cred, int fflags)
293{
294 int permbits = 0;
295 int error;
296
297 if ((fflags & O_DIRECTORY) != 0 && vp->v_type != VDIR)
298 return ENOTDIR;
299
300 if ((fflags & FREAD) != 0) {
301 permbits = VREAD;
302 }
303 if ((fflags & (FWRITE | O_TRUNC)) != 0) {
304 permbits |= VWRITE;
305 if (vp->v_type == VDIR) {
306 error = EISDIR;
307 goto bad;
308 }
309 error = vn_writechk(vp);
310 if (error != 0)
311 goto bad;
312 }
313 error = VOP_ACCESS(vp, permbits, cred);
314bad:
315 return error;
316}
317
318/*
319 * Mark a vnode as having executable mappings.
320 */
321void
322vn_markexec(struct vnode *vp)
323{
324
325 if ((vp->v_iflag & VI_EXECMAP) != 0) {
326 /* Safe unlocked, as long as caller holds a reference. */
327 return;
328 }
329
330 mutex_enter(vp->v_interlock);
331 if ((vp->v_iflag & VI_EXECMAP) == 0) {
332 atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
333 atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
334 vp->v_iflag |= VI_EXECMAP;
335 }
336 mutex_exit(vp->v_interlock);
337}
338
339/*
340 * Mark a vnode as being the text of a process.
341 * Fail if the vnode is currently writable.
342 */
343int
344vn_marktext(struct vnode *vp)
345{
346
347 if ((vp->v_iflag & (VI_TEXT|VI_EXECMAP)) == (VI_TEXT|VI_EXECMAP)) {
348 /* Safe unlocked, as long as caller holds a reference. */
349 return (0);
350 }
351
352 mutex_enter(vp->v_interlock);
353 if (vp->v_writecount != 0) {
354 KASSERT((vp->v_iflag & VI_TEXT) == 0);
355 mutex_exit(vp->v_interlock);
356 return (ETXTBSY);
357 }
358 if ((vp->v_iflag & VI_EXECMAP) == 0) {
359 atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
360 atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
361 }
362 vp->v_iflag |= (VI_TEXT | VI_EXECMAP);
363 mutex_exit(vp->v_interlock);
364 return (0);
365}
366
367/*
368 * Vnode close call
369 *
370 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
371 */
372int
373vn_close(struct vnode *vp, int flags, kauth_cred_t cred)
374{
375 int error;
376
377 if (flags & FWRITE) {
378 mutex_enter(vp->v_interlock);
379 KASSERT(vp->v_writecount > 0);
380 vp->v_writecount--;
381 mutex_exit(vp->v_interlock);
382 }
383 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
384 error = VOP_CLOSE(vp, flags, cred);
385 vput(vp);
386 return (error);
387}
388
389static int
390enforce_rlimit_fsize(struct vnode *vp, struct uio *uio, int ioflag)
391{
392 struct lwp *l = curlwp;
393 off_t testoff;
394
395 if (uio->uio_rw != UIO_WRITE || vp->v_type != VREG)
396 return 0;
397
398 KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
399 if (ioflag & IO_APPEND)
400 testoff = vp->v_size;
401 else
402 testoff = uio->uio_offset;
403
404 if (testoff + uio->uio_resid >
405 l->l_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
406 mutex_enter(proc_lock);
407 psignal(l->l_proc, SIGXFSZ);
408 mutex_exit(proc_lock);
409 return EFBIG;
410 }
411
412 return 0;
413}
414
415/*
416 * Package up an I/O request on a vnode into a uio and do it.
417 */
418int
419vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset,
420 enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid,
421 struct lwp *l)
422{
423 struct uio auio;
424 struct iovec aiov;
425 int error;
426
427 if ((ioflg & IO_NODELOCKED) == 0) {
428 if (rw == UIO_READ) {
429 vn_lock(vp, LK_SHARED | LK_RETRY);
430 } else /* UIO_WRITE */ {
431 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
432 }
433 }
434 auio.uio_iov = &aiov;
435 auio.uio_iovcnt = 1;
436 aiov.iov_base = base;
437 aiov.iov_len = len;
438 auio.uio_resid = len;
439 auio.uio_offset = offset;
440 auio.uio_rw = rw;
441 if (segflg == UIO_SYSSPACE) {
442 UIO_SETUP_SYSSPACE(&auio);
443 } else {
444 auio.uio_vmspace = l->l_proc->p_vmspace;
445 }
446
447 if ((error = enforce_rlimit_fsize(vp, &auio, ioflg)) != 0)
448 goto out;
449
450 if (rw == UIO_READ) {
451 error = VOP_READ(vp, &auio, ioflg, cred);
452 } else {
453 error = VOP_WRITE(vp, &auio, ioflg, cred);
454 }
455
456 if (aresid)
457 *aresid = auio.uio_resid;
458 else
459 if (auio.uio_resid && error == 0)
460 error = EIO;
461
462 out:
463 if ((ioflg & IO_NODELOCKED) == 0) {
464 VOP_UNLOCK(vp);
465 }
466 return (error);
467}
468
469int
470vn_readdir(file_t *fp, char *bf, int segflg, u_int count, int *done,
471 struct lwp *l, off_t **cookies, int *ncookies)
472{
473 struct vnode *vp = fp->f_vnode;
474 struct iovec aiov;
475 struct uio auio;
476 int error, eofflag;
477
478 /* Limit the size on any kernel buffers used by VOP_READDIR */
479 count = min(MAXBSIZE, count);
480
481unionread:
482 if (vp->v_type != VDIR)
483 return (EINVAL);
484 aiov.iov_base = bf;
485 aiov.iov_len = count;
486 auio.uio_iov = &aiov;
487 auio.uio_iovcnt = 1;
488 auio.uio_rw = UIO_READ;
489 if (segflg == UIO_SYSSPACE) {
490 UIO_SETUP_SYSSPACE(&auio);
491 } else {
492 KASSERT(l == curlwp);
493 auio.uio_vmspace = l->l_proc->p_vmspace;
494 }
495 auio.uio_resid = count;
496 vn_lock(vp, LK_SHARED | LK_RETRY);
497 auio.uio_offset = fp->f_offset;
498 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
499 ncookies);
500 mutex_enter(&fp->f_lock);
501 fp->f_offset = auio.uio_offset;
502 mutex_exit(&fp->f_lock);
503 VOP_UNLOCK(vp);
504 if (error)
505 return (error);
506
507 if (count == auio.uio_resid && vn_union_readdir_hook) {
508 struct vnode *ovp = vp;
509
510 error = (*vn_union_readdir_hook)(&vp, fp, l);
511 if (error)
512 return (error);
513 if (vp != ovp)
514 goto unionread;
515 }
516
517 if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) &&
518 (vp->v_mount->mnt_flag & MNT_UNION)) {
519 struct vnode *tvp = vp;
520 vp = vp->v_mount->mnt_vnodecovered;
521 vref(vp);
522 mutex_enter(&fp->f_lock);
523 fp->f_vnode = vp;
524 fp->f_offset = 0;
525 mutex_exit(&fp->f_lock);
526 vrele(tvp);
527 goto unionread;
528 }
529 *done = count - auio.uio_resid;
530 return error;
531}
532
533/*
534 * File table vnode read routine.
535 */
536static int
537vn_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
538 int flags)
539{
540 struct vnode *vp = fp->f_vnode;
541 int error, ioflag, fflag;
542 size_t count;
543
544 ioflag = IO_ADV_ENCODE(fp->f_advice);
545 fflag = fp->f_flag;
546 if (fflag & FNONBLOCK)
547 ioflag |= IO_NDELAY;
548 if ((fflag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
549 ioflag |= IO_SYNC;
550 if (fflag & FALTIO)
551 ioflag |= IO_ALTSEMANTICS;
552 if (fflag & FDIRECT)
553 ioflag |= IO_DIRECT;
554 vn_lock(vp, LK_SHARED | LK_RETRY);
555 uio->uio_offset = *offset;
556 count = uio->uio_resid;
557 error = VOP_READ(vp, uio, ioflag, cred);
558 if (flags & FOF_UPDATE_OFFSET)
559 *offset += count - uio->uio_resid;
560 VOP_UNLOCK(vp);
561 return (error);
562}
563
564/*
565 * File table vnode write routine.
566 */
567static int
568vn_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
569 int flags)
570{
571 struct vnode *vp = fp->f_vnode;
572 int error, ioflag, fflag;
573 size_t count;
574
575 ioflag = IO_ADV_ENCODE(fp->f_advice) | IO_UNIT;
576 fflag = fp->f_flag;
577 if (vp->v_type == VREG && (fflag & O_APPEND))
578 ioflag |= IO_APPEND;
579 if (fflag & FNONBLOCK)
580 ioflag |= IO_NDELAY;
581 if (fflag & FFSYNC ||
582 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
583 ioflag |= IO_SYNC;
584 else if (fflag & FDSYNC)
585 ioflag |= IO_DSYNC;
586 if (fflag & FALTIO)
587 ioflag |= IO_ALTSEMANTICS;
588 if (fflag & FDIRECT)
589 ioflag |= IO_DIRECT;
590 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
591 uio->uio_offset = *offset;
592 count = uio->uio_resid;
593
594 if ((error = enforce_rlimit_fsize(vp, uio, ioflag)) != 0)
595 goto out;
596
597 error = VOP_WRITE(vp, uio, ioflag, cred);
598
599 if (flags & FOF_UPDATE_OFFSET) {
600 if (ioflag & IO_APPEND) {
601 /*
602 * SUSv3 describes behaviour for count = 0 as following:
603 * "Before any action ... is taken, and if nbyte is zero
604 * and the file is a regular file, the write() function
605 * ... in the absence of errors ... shall return zero
606 * and have no other results."
607 */
608 if (count)
609 *offset = uio->uio_offset;
610 } else
611 *offset += count - uio->uio_resid;
612 }
613
614 out:
615 VOP_UNLOCK(vp);
616 return (error);
617}
618
619/*
620 * File table vnode stat routine.
621 */
622static int
623vn_statfile(file_t *fp, struct stat *sb)
624{
625 struct vnode *vp = fp->f_vnode;
626 int error;
627
628 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
629 error = vn_stat(vp, sb);
630 VOP_UNLOCK(vp);
631 return error;
632}
633
634int
635vn_stat(struct vnode *vp, struct stat *sb)
636{
637 struct vattr va;
638 int error;
639 mode_t mode;
640
641 memset(&va, 0, sizeof(va));
642 error = VOP_GETATTR(vp, &va, kauth_cred_get());
643 if (error)
644 return (error);
645 /*
646 * Copy from vattr table
647 */
648 memset(sb, 0, sizeof(*sb));
649 sb->st_dev = va.va_fsid;
650 sb->st_ino = va.va_fileid;
651 mode = va.va_mode;
652 switch (vp->v_type) {
653 case VREG:
654 mode |= S_IFREG;
655 break;
656 case VDIR:
657 mode |= S_IFDIR;
658 break;
659 case VBLK:
660 mode |= S_IFBLK;
661 break;
662 case VCHR:
663 mode |= S_IFCHR;
664 break;
665 case VLNK:
666 mode |= S_IFLNK;
667 break;
668 case VSOCK:
669 mode |= S_IFSOCK;
670 break;
671 case VFIFO:
672 mode |= S_IFIFO;
673 break;
674 default:
675 return (EBADF);
676 }
677 sb->st_mode = mode;
678 sb->st_nlink = va.va_nlink;
679 sb->st_uid = va.va_uid;
680 sb->st_gid = va.va_gid;
681 sb->st_rdev = va.va_rdev;
682 sb->st_size = va.va_size;
683 sb->st_atimespec = va.va_atime;
684 sb->st_mtimespec = va.va_mtime;
685 sb->st_ctimespec = va.va_ctime;
686 sb->st_birthtimespec = va.va_birthtime;
687 sb->st_blksize = va.va_blocksize;
688 sb->st_flags = va.va_flags;
689 sb->st_gen = 0;
690 sb->st_blocks = va.va_bytes / S_BLKSIZE;
691 return (0);
692}
693
694/*
695 * File table vnode fcntl routine.
696 */
697static int
698vn_fcntl(file_t *fp, u_int com, void *data)
699{
700 struct vnode *vp = fp->f_vnode;
701 int error;
702
703 error = VOP_FCNTL(vp, com, data, fp->f_flag, kauth_cred_get());
704 return (error);
705}
706
707/*
708 * File table vnode ioctl routine.
709 */
710static int
711vn_ioctl(file_t *fp, u_long com, void *data)
712{
713 struct vnode *vp = fp->f_vnode, *ovp;
714 struct vattr vattr;
715 int error;
716
717 switch (vp->v_type) {
718
719 case VREG:
720 case VDIR:
721 if (com == FIONREAD) {
722 vn_lock(vp, LK_SHARED | LK_RETRY);
723 error = VOP_GETATTR(vp, &vattr, kauth_cred_get());
724 VOP_UNLOCK(vp);
725 if (error)
726 return (error);
727 *(int *)data = vattr.va_size - fp->f_offset;
728 return (0);
729 }
730 if ((com == FIONWRITE) || (com == FIONSPACE)) {
731 /*
732 * Files don't have send queues, so there never
733 * are any bytes in them, nor is there any
734 * open space in them.
735 */
736 *(int *)data = 0;
737 return (0);
738 }
739 if (com == FIOGETBMAP) {
740 daddr_t *block;
741
742 if (*(daddr_t *)data < 0)
743 return (EINVAL);
744 block = (daddr_t *)data;
745 return (VOP_BMAP(vp, *block, NULL, block, NULL));
746 }
747 if (com == OFIOGETBMAP) {
748 daddr_t ibn, obn;
749
750 if (*(int32_t *)data < 0)
751 return (EINVAL);
752 ibn = (daddr_t)*(int32_t *)data;
753 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
754 *(int32_t *)data = (int32_t)obn;
755 return error;
756 }
757 if (com == FIONBIO || com == FIOASYNC) /* XXX */
758 return (0); /* XXX */
759 /* fall into ... */
760 case VFIFO:
761 case VCHR:
762 case VBLK:
763 error = VOP_IOCTL(vp, com, data, fp->f_flag,
764 kauth_cred_get());
765 if (error == 0 && com == TIOCSCTTY) {
766 vref(vp);
767 mutex_enter(proc_lock);
768 ovp = curproc->p_session->s_ttyvp;
769 curproc->p_session->s_ttyvp = vp;
770 mutex_exit(proc_lock);
771 if (ovp != NULL)
772 vrele(ovp);
773 }
774 return (error);
775
776 default:
777 return (EPASSTHROUGH);
778 }
779}
780
781/*
782 * File table vnode poll routine.
783 */
784static int
785vn_poll(file_t *fp, int events)
786{
787
788 return (VOP_POLL(fp->f_vnode, events));
789}
790
791/*
792 * File table vnode kqfilter routine.
793 */
794int
795vn_kqfilter(file_t *fp, struct knote *kn)
796{
797
798 return (VOP_KQFILTER(fp->f_vnode, kn));
799}
800
801static int
802vn_mmap(struct file *fp, off_t *offp, size_t size, int prot, int *flagsp,
803 int *advicep, struct uvm_object **uobjp, int *maxprotp)
804{
805 struct uvm_object *uobj;
806 struct vnode *vp;
807 struct vattr va;
808 struct lwp *l;
809 vm_prot_t maxprot;
810 off_t off;
811 int error, flags;
812 bool needwritemap;
813
814 l = curlwp;
815
816 off = *offp;
817 flags = *flagsp;
818 maxprot = VM_PROT_EXECUTE;
819
820 vp = fp->f_vnode;
821 if (vp->v_type != VREG && vp->v_type != VCHR &&
822 vp->v_type != VBLK) {
823 /* only REG/CHR/BLK support mmap */
824 return ENODEV;
825 }
826 if (vp->v_type != VCHR && off < 0) {
827 return EINVAL;
828 }
829 if (vp->v_type != VCHR && (off_t)(off + size) < off) {
830 /* no offset wrapping */
831 return EOVERFLOW;
832 }
833
834 /* special case: catch SunOS style /dev/zero */
835 if (vp->v_type == VCHR &&
836 (vp->v_rdev == zerodev || COMPAT_ZERODEV(vp->v_rdev))) {
837 *uobjp = NULL;
838 *maxprotp = VM_PROT_ALL;
839 return 0;
840 }
841
842 /*
843 * Old programs may not select a specific sharing type, so
844 * default to an appropriate one.
845 *
846 * XXX: how does MAP_ANON fit in the picture?
847 */
848 if ((flags & (MAP_SHARED|MAP_PRIVATE)) == 0) {
849#if defined(DEBUG)
850 struct proc *p = l->l_proc;
851 printf("WARNING: defaulted mmap() share type to "
852 "%s (pid %d command %s)\n", vp->v_type == VCHR ?
853 "MAP_SHARED" : "MAP_PRIVATE", p->p_pid,
854 p->p_comm);
855#endif
856 if (vp->v_type == VCHR)
857 flags |= MAP_SHARED; /* for a device */
858 else
859 flags |= MAP_PRIVATE; /* for a file */
860 }
861
862 /*
863 * MAP_PRIVATE device mappings don't make sense (and aren't
864 * supported anyway). However, some programs rely on this,
865 * so just change it to MAP_SHARED.
866 */
867 if (vp->v_type == VCHR && (flags & MAP_PRIVATE) != 0) {
868 flags = (flags & ~MAP_PRIVATE) | MAP_SHARED;
869 }
870
871 /*
872 * now check protection
873 */
874
875 /* check read access */
876 if (fp->f_flag & FREAD)
877 maxprot |= VM_PROT_READ;
878 else if (prot & PROT_READ) {
879 return EACCES;
880 }
881
882 /* check write access, shared case first */
883 if (flags & MAP_SHARED) {
884 /*
885 * if the file is writable, only add PROT_WRITE to
886 * maxprot if the file is not immutable, append-only.
887 * otherwise, if we have asked for PROT_WRITE, return
888 * EPERM.
889 */
890 if (fp->f_flag & FWRITE) {
891 vn_lock(vp, LK_SHARED | LK_RETRY);
892 error = VOP_GETATTR(vp, &va, l->l_cred);
893 VOP_UNLOCK(vp);
894 if (error) {
895 return error;
896 }
897 if ((va.va_flags &
898 (SF_SNAPSHOT|IMMUTABLE|APPEND)) == 0)
899 maxprot |= VM_PROT_WRITE;
900 else if (prot & PROT_WRITE) {
901 return EPERM;
902 }
903 } else if (prot & PROT_WRITE) {
904 return EACCES;
905 }
906 } else {
907 /* MAP_PRIVATE mappings can always write to */
908 maxprot |= VM_PROT_WRITE;
909 }
910
911 /*
912 * Don't allow mmap for EXEC if the file system
913 * is mounted NOEXEC.
914 */
915 if ((prot & PROT_EXEC) != 0 &&
916 (vp->v_mount->mnt_flag & MNT_NOEXEC) != 0) {
917 return EACCES;
918 }
919
920 if (vp->v_type != VCHR) {
921 error = VOP_MMAP(vp, prot, curlwp->l_cred);
922 if (error) {
923 return error;
924 }
925 vref(vp);
926 uobj = &vp->v_uobj;
927
928 /*
929 * If the vnode is being mapped with PROT_EXEC,
930 * then mark it as text.
931 */
932 if (prot & PROT_EXEC) {
933 vn_markexec(vp);
934 }
935 } else {
936 int i = maxprot;
937
938 /*
939 * XXX Some devices don't like to be mapped with
940 * XXX PROT_EXEC or PROT_WRITE, but we don't really
941 * XXX have a better way of handling this, right now
942 */
943 do {
944 uobj = udv_attach(vp->v_rdev,
945 (flags & MAP_SHARED) ? i :
946 (i & ~VM_PROT_WRITE), off, size);
947 i--;
948 } while ((uobj == NULL) && (i > 0));
949 if (uobj == NULL) {
950 return EINVAL;
951 }
952 *advicep = UVM_ADV_RANDOM;
953 }
954
955 /*
956 * Set vnode flags to indicate the new kinds of mapping.
957 * We take the vnode lock in exclusive mode here to serialize
958 * with direct I/O.
959 *
960 * Safe to check for these flag values without a lock, as
961 * long as a reference to the vnode is held.
962 */
963 needwritemap = (vp->v_iflag & VI_WRMAP) == 0 &&
964 (flags & MAP_SHARED) != 0 &&
965 (maxprot & VM_PROT_WRITE) != 0;
966 if ((vp->v_vflag & VV_MAPPED) == 0 || needwritemap) {
967 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
968 vp->v_vflag |= VV_MAPPED;
969 if (needwritemap) {
970 mutex_enter(vp->v_interlock);
971 vp->v_iflag |= VI_WRMAP;
972 mutex_exit(vp->v_interlock);
973 }
974 VOP_UNLOCK(vp);
975 }
976
977#if NVERIEXEC > 0
978
979 /*
980 * Check if the file can be executed indirectly.
981 *
982 * XXX: This gives false warnings about "Incorrect access type"
983 * XXX: if the mapping is not executable. Harmless, but will be
984 * XXX: fixed as part of other changes.
985 */
986 if (veriexec_verify(l, vp, "(mmap)", VERIEXEC_INDIRECT,
987 NULL)) {
988
989 /*
990 * Don't allow executable mappings if we can't
991 * indirectly execute the file.
992 */
993 if (prot & VM_PROT_EXECUTE) {
994 return EPERM;
995 }
996
997 /*
998 * Strip the executable bit from 'maxprot' to make sure
999 * it can't be made executable later.
1000 */
1001 maxprot &= ~VM_PROT_EXECUTE;
1002 }
1003#endif /* NVERIEXEC > 0 */
1004
1005 *uobjp = uobj;
1006 *maxprotp = maxprot;
1007 *flagsp = flags;
1008
1009 return 0;
1010}
1011
1012
1013
1014/*
1015 * Check that the vnode is still valid, and if so
1016 * acquire requested lock.
1017 */
1018int
1019vn_lock(struct vnode *vp, int flags)
1020{
1021 int error;
1022
1023#if 0
1024 KASSERT(vp->v_usecount > 0 || (vp->v_iflag & VI_ONWORKLST) != 0);
1025#endif
1026 KASSERT((flags & ~(LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT|LK_RETRY)) == 0);
1027 KASSERT(!mutex_owned(vp->v_interlock));
1028
1029#ifdef DIAGNOSTIC
1030 if (wapbl_vphaswapbl(vp))
1031 WAPBL_JUNLOCK_ASSERT(wapbl_vptomp(vp));
1032#endif
1033
1034 error = VOP_LOCK(vp, flags);
1035 if ((flags & LK_RETRY) != 0 && error == ENOENT)
1036 error = VOP_LOCK(vp, flags);
1037
1038 KASSERT((flags & LK_RETRY) == 0 || (flags & LK_NOWAIT) != 0 ||
1039 error == 0);
1040
1041 return error;
1042}
1043
1044/*
1045 * File table vnode close routine.
1046 */
1047static int
1048vn_closefile(file_t *fp)
1049{
1050
1051 return vn_close(fp->f_vnode, fp->f_flag, fp->f_cred);
1052}
1053
1054/*
1055 * Simplified in-kernel wrapper calls for extended attribute access.
1056 * Both calls pass in a NULL credential, authorizing a "kernel" access.
1057 * Set IO_NODELOCKED in ioflg if the vnode is already locked.
1058 */
1059int
1060vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
1061 const char *attrname, size_t *buflen, void *bf, struct lwp *l)
1062{
1063 struct uio auio;
1064 struct iovec aiov;
1065 int error;
1066
1067 aiov.iov_len = *buflen;
1068 aiov.iov_base = bf;
1069
1070 auio.uio_iov = &aiov;
1071 auio.uio_iovcnt = 1;
1072 auio.uio_rw = UIO_READ;
1073 auio.uio_offset = 0;
1074 auio.uio_resid = *buflen;
1075 UIO_SETUP_SYSSPACE(&auio);
1076
1077 if ((ioflg & IO_NODELOCKED) == 0)
1078 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1079
1080 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL);
1081
1082 if ((ioflg & IO_NODELOCKED) == 0)
1083 VOP_UNLOCK(vp);
1084
1085 if (error == 0)
1086 *buflen = *buflen - auio.uio_resid;
1087
1088 return (error);
1089}
1090
1091/*
1092 * XXX Failure mode if partially written?
1093 */
1094int
1095vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
1096 const char *attrname, size_t buflen, const void *bf, struct lwp *l)
1097{
1098 struct uio auio;
1099 struct iovec aiov;
1100 int error;
1101
1102 aiov.iov_len = buflen;
1103 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */
1104
1105 auio.uio_iov = &aiov;
1106 auio.uio_iovcnt = 1;
1107 auio.uio_rw = UIO_WRITE;
1108 auio.uio_offset = 0;
1109 auio.uio_resid = buflen;
1110 UIO_SETUP_SYSSPACE(&auio);
1111
1112 if ((ioflg & IO_NODELOCKED) == 0) {
1113 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1114 }
1115
1116 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL);
1117
1118 if ((ioflg & IO_NODELOCKED) == 0) {
1119 VOP_UNLOCK(vp);
1120 }
1121
1122 return (error);
1123}
1124
1125int
1126vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
1127 const char *attrname, struct lwp *l)
1128{
1129 int error;
1130
1131 if ((ioflg & IO_NODELOCKED) == 0) {
1132 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1133 }
1134
1135 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL);
1136 if (error == EOPNOTSUPP)
1137 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL);
1138
1139 if ((ioflg & IO_NODELOCKED) == 0) {
1140 VOP_UNLOCK(vp);
1141 }
1142
1143 return (error);
1144}
1145
1146void
1147vn_ra_allocctx(struct vnode *vp)
1148{
1149 struct uvm_ractx *ra = NULL;
1150
1151 KASSERT(mutex_owned(vp->v_interlock));
1152
1153 if (vp->v_type != VREG) {
1154 return;
1155 }
1156 if (vp->v_ractx != NULL) {
1157 return;
1158 }
1159 if (vp->v_ractx == NULL) {
1160 mutex_exit(vp->v_interlock);
1161 ra = uvm_ra_allocctx();
1162 mutex_enter(vp->v_interlock);
1163 if (ra != NULL && vp->v_ractx == NULL) {
1164 vp->v_ractx = ra;
1165 ra = NULL;
1166 }
1167 }
1168 if (ra != NULL) {
1169 uvm_ra_freectx(ra);
1170 }
1171}
1172
1173int
1174vn_fifo_bypass(void *v)
1175{
1176 struct vop_generic_args *ap = v;
1177
1178 return VOCALL(fifo_vnodeop_p, ap->a_desc->vdesc_offset, v);
1179}
1180