1 /*-
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
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 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
35 */
36
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD: releng/9.2/sys/kern/vfs_syscalls.c 250978 2013-05-25 11:05:00Z kib $");
39
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_kdtrace.h"
43 #include "opt_ktrace.h"
44
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capability.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78
79 #include <machine/stdarg.h>
80
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88
89 #include <ufs/ufs/quota.h>
90
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE(vfs, , stat, mode, mode);
95 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
96 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
97 SDT_PROBE_DEFINE(vfs, , stat, reg, reg);
98 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
99 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
100
101 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
102 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
103 static int setfflags(struct thread *td, struct vnode *, int);
104 static int setutimes(struct thread *td, struct vnode *,
105 const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107 struct thread *td);
108
109 /*
110 * The module initialization routine for POSIX asynchronous I/O will
111 * set this to the version of AIO that it implements. (Zero means
112 * that it is not implemented.) This value is used here by pathconf()
113 * and in kern_descrip.c by fpathconf().
114 */
115 int async_io_version;
116
117 #ifdef DEBUG
118 static int syncprt = 0;
119 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
120 #endif
121
122 /*
123 * Sync each mounted filesystem.
124 */
125 #ifndef _SYS_SYSPROTO_H_
126 struct sync_args {
127 int dummy;
128 };
129 #endif
130 /* ARGSUSED */
131 int
132 sys_sync(td, uap)
133 struct thread *td;
134 struct sync_args *uap;
135 {
136 struct mount *mp, *nmp;
137 int save, vfslocked;
138
139 mtx_lock(&mountlist_mtx);
140 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
141 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
142 nmp = TAILQ_NEXT(mp, mnt_list);
143 continue;
144 }
145 vfslocked = VFS_LOCK_GIANT(mp);
146 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
147 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
148 save = curthread_pflags_set(TDP_SYNCIO);
149 vfs_msync(mp, MNT_NOWAIT);
150 VFS_SYNC(mp, MNT_NOWAIT);
151 curthread_pflags_restore(save);
152 vn_finished_write(mp);
153 }
154 VFS_UNLOCK_GIANT(vfslocked);
155 mtx_lock(&mountlist_mtx);
156 nmp = TAILQ_NEXT(mp, mnt_list);
157 vfs_unbusy(mp);
158 }
159 mtx_unlock(&mountlist_mtx);
160 return (0);
161 }
162
163 /*
164 * Change filesystem quotas.
165 */
166 #ifndef _SYS_SYSPROTO_H_
167 struct quotactl_args {
168 char *path;
169 int cmd;
170 int uid;
171 caddr_t arg;
172 };
173 #endif
174 int
175 sys_quotactl(td, uap)
176 struct thread *td;
177 register struct quotactl_args /* {
178 char *path;
179 int cmd;
180 int uid;
181 caddr_t arg;
182 } */ *uap;
183 {
184 struct mount *mp;
185 int vfslocked;
186 int error;
187 struct nameidata nd;
188
189 AUDIT_ARG_CMD(uap->cmd);
190 AUDIT_ARG_UID(uap->uid);
191 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
192 return (EPERM);
193 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
194 UIO_USERSPACE, uap->path, td);
195 if ((error = namei(&nd)) != 0)
196 return (error);
197 vfslocked = NDHASGIANT(&nd);
198 NDFREE(&nd, NDF_ONLY_PNBUF);
199 mp = nd.ni_vp->v_mount;
200 vfs_ref(mp);
201 vput(nd.ni_vp);
202 error = vfs_busy(mp, 0);
203 vfs_rel(mp);
204 if (error) {
205 VFS_UNLOCK_GIANT(vfslocked);
206 return (error);
207 }
208 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
209
210 /*
211 * Since quota on operation typically needs to open quota
212 * file, the Q_QUOTAON handler needs to unbusy the mount point
213 * before calling into namei. Otherwise, unmount might be
214 * started between two vfs_busy() invocations (first is our,
215 * second is from mount point cross-walk code in lookup()),
216 * causing deadlock.
217 *
218 * Require that Q_QUOTAON handles the vfs_busy() reference on
219 * its own, always returning with ubusied mount point.
220 */
221 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
222 vfs_unbusy(mp);
223 VFS_UNLOCK_GIANT(vfslocked);
224 return (error);
225 }
226
227 /*
228 * Used by statfs conversion routines to scale the block size up if
229 * necessary so that all of the block counts are <= 'max_size'. Note
230 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
231 * value of 'n'.
232 */
233 void
234 statfs_scale_blocks(struct statfs *sf, long max_size)
235 {
236 uint64_t count;
237 int shift;
238
239 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
240
241 /*
242 * Attempt to scale the block counts to give a more accurate
243 * overview to userland of the ratio of free space to used
244 * space. To do this, find the largest block count and compute
245 * a divisor that lets it fit into a signed integer <= max_size.
246 */
247 if (sf->f_bavail < 0)
248 count = -sf->f_bavail;
249 else
250 count = sf->f_bavail;
251 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
252 if (count <= max_size)
253 return;
254
255 count >>= flsl(max_size);
256 shift = 0;
257 while (count > 0) {
258 shift++;
259 count >>=1;
260 }
261
262 sf->f_bsize <<= shift;
263 sf->f_blocks >>= shift;
264 sf->f_bfree >>= shift;
265 sf->f_bavail >>= shift;
266 }
267
268 /*
269 * Get filesystem statistics.
270 */
271 #ifndef _SYS_SYSPROTO_H_
272 struct statfs_args {
273 char *path;
274 struct statfs *buf;
275 };
276 #endif
277 int
278 sys_statfs(td, uap)
279 struct thread *td;
280 register struct statfs_args /* {
281 char *path;
282 struct statfs *buf;
283 } */ *uap;
284 {
285 struct statfs sf;
286 int error;
287
288 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
289 if (error == 0)
290 error = copyout(&sf, uap->buf, sizeof(sf));
291 return (error);
292 }
293
294 int
295 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
296 struct statfs *buf)
297 {
298 struct mount *mp;
299 struct statfs *sp, sb;
300 int vfslocked;
301 int error;
302 struct nameidata nd;
303
304 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
305 AUDITVNODE1, pathseg, path, td);
306 error = namei(&nd);
307 if (error)
308 return (error);
309 vfslocked = NDHASGIANT(&nd);
310 mp = nd.ni_vp->v_mount;
311 vfs_ref(mp);
312 NDFREE(&nd, NDF_ONLY_PNBUF);
313 vput(nd.ni_vp);
314 error = vfs_busy(mp, 0);
315 vfs_rel(mp);
316 if (error) {
317 VFS_UNLOCK_GIANT(vfslocked);
318 return (error);
319 }
320 #ifdef MAC
321 error = mac_mount_check_stat(td->td_ucred, mp);
322 if (error)
323 goto out;
324 #endif
325 /*
326 * Set these in case the underlying filesystem fails to do so.
327 */
328 sp = &mp->mnt_stat;
329 sp->f_version = STATFS_VERSION;
330 sp->f_namemax = NAME_MAX;
331 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
332 error = VFS_STATFS(mp, sp);
333 if (error)
334 goto out;
335 if (priv_check(td, PRIV_VFS_GENERATION)) {
336 bcopy(sp, &sb, sizeof(sb));
337 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
338 prison_enforce_statfs(td->td_ucred, mp, &sb);
339 sp = &sb;
340 }
341 *buf = *sp;
342 out:
343 vfs_unbusy(mp);
344 VFS_UNLOCK_GIANT(vfslocked);
345 return (error);
346 }
347
348 /*
349 * Get filesystem statistics.
350 */
351 #ifndef _SYS_SYSPROTO_H_
352 struct fstatfs_args {
353 int fd;
354 struct statfs *buf;
355 };
356 #endif
357 int
358 sys_fstatfs(td, uap)
359 struct thread *td;
360 register struct fstatfs_args /* {
361 int fd;
362 struct statfs *buf;
363 } */ *uap;
364 {
365 struct statfs sf;
366 int error;
367
368 error = kern_fstatfs(td, uap->fd, &sf);
369 if (error == 0)
370 error = copyout(&sf, uap->buf, sizeof(sf));
371 return (error);
372 }
373
374 int
375 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
376 {
377 struct file *fp;
378 struct mount *mp;
379 struct statfs *sp, sb;
380 int vfslocked;
381 struct vnode *vp;
382 int error;
383
384 AUDIT_ARG_FD(fd);
385 error = getvnode(td->td_proc->p_fd, fd, CAP_FSTATFS, &fp);
386 if (error)
387 return (error);
388 vp = fp->f_vnode;
389 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
390 vn_lock(vp, LK_SHARED | LK_RETRY);
391 #ifdef AUDIT
392 AUDIT_ARG_VNODE1(vp);
393 #endif
394 mp = vp->v_mount;
395 if (mp)
396 vfs_ref(mp);
397 VOP_UNLOCK(vp, 0);
398 fdrop(fp, td);
399 if (mp == NULL) {
400 error = EBADF;
401 goto out;
402 }
403 error = vfs_busy(mp, 0);
404 vfs_rel(mp);
405 if (error) {
406 VFS_UNLOCK_GIANT(vfslocked);
407 return (error);
408 }
409 #ifdef MAC
410 error = mac_mount_check_stat(td->td_ucred, mp);
411 if (error)
412 goto out;
413 #endif
414 /*
415 * Set these in case the underlying filesystem fails to do so.
416 */
417 sp = &mp->mnt_stat;
418 sp->f_version = STATFS_VERSION;
419 sp->f_namemax = NAME_MAX;
420 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
421 error = VFS_STATFS(mp, sp);
422 if (error)
423 goto out;
424 if (priv_check(td, PRIV_VFS_GENERATION)) {
425 bcopy(sp, &sb, sizeof(sb));
426 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
427 prison_enforce_statfs(td->td_ucred, mp, &sb);
428 sp = &sb;
429 }
430 *buf = *sp;
431 out:
432 if (mp)
433 vfs_unbusy(mp);
434 VFS_UNLOCK_GIANT(vfslocked);
435 return (error);
436 }
437
438 /*
439 * Get statistics on all filesystems.
440 */
441 #ifndef _SYS_SYSPROTO_H_
442 struct getfsstat_args {
443 struct statfs *buf;
444 long bufsize;
445 int flags;
446 };
447 #endif
448 int
449 sys_getfsstat(td, uap)
450 struct thread *td;
451 register struct getfsstat_args /* {
452 struct statfs *buf;
453 long bufsize;
454 int flags;
455 } */ *uap;
456 {
457
458 return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
459 uap->flags));
460 }
461
462 /*
463 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
464 * The caller is responsible for freeing memory which will be allocated
465 * in '*buf'.
466 */
467 int
468 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
469 enum uio_seg bufseg, int flags)
470 {
471 struct mount *mp, *nmp;
472 struct statfs *sfsp, *sp, sb;
473 size_t count, maxcount;
474 int vfslocked;
475 int error;
476
477 maxcount = bufsize / sizeof(struct statfs);
478 if (bufsize == 0)
479 sfsp = NULL;
480 else if (bufseg == UIO_USERSPACE)
481 sfsp = *buf;
482 else /* if (bufseg == UIO_SYSSPACE) */ {
483 count = 0;
484 mtx_lock(&mountlist_mtx);
485 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
486 count++;
487 }
488 mtx_unlock(&mountlist_mtx);
489 if (maxcount > count)
490 maxcount = count;
491 sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
492 M_WAITOK);
493 }
494 count = 0;
495 mtx_lock(&mountlist_mtx);
496 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
497 if (prison_canseemount(td->td_ucred, mp) != 0) {
498 nmp = TAILQ_NEXT(mp, mnt_list);
499 continue;
500 }
501 #ifdef MAC
502 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
503 nmp = TAILQ_NEXT(mp, mnt_list);
504 continue;
505 }
506 #endif
507 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
508 nmp = TAILQ_NEXT(mp, mnt_list);
509 continue;
510 }
511 vfslocked = VFS_LOCK_GIANT(mp);
512 if (sfsp && count < maxcount) {
513 sp = &mp->mnt_stat;
514 /*
515 * Set these in case the underlying filesystem
516 * fails to do so.
517 */
518 sp->f_version = STATFS_VERSION;
519 sp->f_namemax = NAME_MAX;
520 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
521 /*
522 * If MNT_NOWAIT or MNT_LAZY is specified, do not
523 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
524 * overrides MNT_WAIT.
525 */
526 if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
527 (flags & MNT_WAIT)) &&
528 (error = VFS_STATFS(mp, sp))) {
529 VFS_UNLOCK_GIANT(vfslocked);
530 mtx_lock(&mountlist_mtx);
531 nmp = TAILQ_NEXT(mp, mnt_list);
532 vfs_unbusy(mp);
533 continue;
534 }
535 if (priv_check(td, PRIV_VFS_GENERATION)) {
536 bcopy(sp, &sb, sizeof(sb));
537 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
538 prison_enforce_statfs(td->td_ucred, mp, &sb);
539 sp = &sb;
540 }
541 if (bufseg == UIO_SYSSPACE)
542 bcopy(sp, sfsp, sizeof(*sp));
543 else /* if (bufseg == UIO_USERSPACE) */ {
544 error = copyout(sp, sfsp, sizeof(*sp));
545 if (error) {
546 vfs_unbusy(mp);
547 VFS_UNLOCK_GIANT(vfslocked);
548 return (error);
549 }
550 }
551 sfsp++;
552 }
553 VFS_UNLOCK_GIANT(vfslocked);
554 count++;
555 mtx_lock(&mountlist_mtx);
556 nmp = TAILQ_NEXT(mp, mnt_list);
557 vfs_unbusy(mp);
558 }
559 mtx_unlock(&mountlist_mtx);
560 if (sfsp && count > maxcount)
561 td->td_retval[0] = maxcount;
562 else
563 td->td_retval[0] = count;
564 return (0);
565 }
566
567 #ifdef COMPAT_FREEBSD4
568 /*
569 * Get old format filesystem statistics.
570 */
571 static void cvtstatfs(struct statfs *, struct ostatfs *);
572
573 #ifndef _SYS_SYSPROTO_H_
574 struct freebsd4_statfs_args {
575 char *path;
576 struct ostatfs *buf;
577 };
578 #endif
579 int
580 freebsd4_statfs(td, uap)
581 struct thread *td;
582 struct freebsd4_statfs_args /* {
583 char *path;
584 struct ostatfs *buf;
585 } */ *uap;
586 {
587 struct ostatfs osb;
588 struct statfs sf;
589 int error;
590
591 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
592 if (error)
593 return (error);
594 cvtstatfs(&sf, &osb);
595 return (copyout(&osb, uap->buf, sizeof(osb)));
596 }
597
598 /*
599 * Get filesystem statistics.
600 */
601 #ifndef _SYS_SYSPROTO_H_
602 struct freebsd4_fstatfs_args {
603 int fd;
604 struct ostatfs *buf;
605 };
606 #endif
607 int
608 freebsd4_fstatfs(td, uap)
609 struct thread *td;
610 struct freebsd4_fstatfs_args /* {
611 int fd;
612 struct ostatfs *buf;
613 } */ *uap;
614 {
615 struct ostatfs osb;
616 struct statfs sf;
617 int error;
618
619 error = kern_fstatfs(td, uap->fd, &sf);
620 if (error)
621 return (error);
622 cvtstatfs(&sf, &osb);
623 return (copyout(&osb, uap->buf, sizeof(osb)));
624 }
625
626 /*
627 * Get statistics on all filesystems.
628 */
629 #ifndef _SYS_SYSPROTO_H_
630 struct freebsd4_getfsstat_args {
631 struct ostatfs *buf;
632 long bufsize;
633 int flags;
634 };
635 #endif
636 int
637 freebsd4_getfsstat(td, uap)
638 struct thread *td;
639 register struct freebsd4_getfsstat_args /* {
640 struct ostatfs *buf;
641 long bufsize;
642 int flags;
643 } */ *uap;
644 {
645 struct statfs *buf, *sp;
646 struct ostatfs osb;
647 size_t count, size;
648 int error;
649
650 count = uap->bufsize / sizeof(struct ostatfs);
651 size = count * sizeof(struct statfs);
652 error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
653 if (size > 0) {
654 count = td->td_retval[0];
655 sp = buf;
656 while (count > 0 && error == 0) {
657 cvtstatfs(sp, &osb);
658 error = copyout(&osb, uap->buf, sizeof(osb));
659 sp++;
660 uap->buf++;
661 count--;
662 }
663 free(buf, M_TEMP);
664 }
665 return (error);
666 }
667
668 /*
669 * Implement fstatfs() for (NFS) file handles.
670 */
671 #ifndef _SYS_SYSPROTO_H_
672 struct freebsd4_fhstatfs_args {
673 struct fhandle *u_fhp;
674 struct ostatfs *buf;
675 };
676 #endif
677 int
678 freebsd4_fhstatfs(td, uap)
679 struct thread *td;
680 struct freebsd4_fhstatfs_args /* {
681 struct fhandle *u_fhp;
682 struct ostatfs *buf;
683 } */ *uap;
684 {
685 struct ostatfs osb;
686 struct statfs sf;
687 fhandle_t fh;
688 int error;
689
690 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
691 if (error)
692 return (error);
693 error = kern_fhstatfs(td, fh, &sf);
694 if (error)
695 return (error);
696 cvtstatfs(&sf, &osb);
697 return (copyout(&osb, uap->buf, sizeof(osb)));
698 }
699
700 /*
701 * Convert a new format statfs structure to an old format statfs structure.
702 */
703 static void
704 cvtstatfs(nsp, osp)
705 struct statfs *nsp;
706 struct ostatfs *osp;
707 {
708
709 statfs_scale_blocks(nsp, LONG_MAX);
710 bzero(osp, sizeof(*osp));
711 osp->f_bsize = nsp->f_bsize;
712 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
713 osp->f_blocks = nsp->f_blocks;
714 osp->f_bfree = nsp->f_bfree;
715 osp->f_bavail = nsp->f_bavail;
716 osp->f_files = MIN(nsp->f_files, LONG_MAX);
717 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
718 osp->f_owner = nsp->f_owner;
719 osp->f_type = nsp->f_type;
720 osp->f_flags = nsp->f_flags;
721 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
722 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
723 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
724 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
725 strlcpy(osp->f_fstypename, nsp->f_fstypename,
726 MIN(MFSNAMELEN, OMFSNAMELEN));
727 strlcpy(osp->f_mntonname, nsp->f_mntonname,
728 MIN(MNAMELEN, OMNAMELEN));
729 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
730 MIN(MNAMELEN, OMNAMELEN));
731 osp->f_fsid = nsp->f_fsid;
732 }
733 #endif /* COMPAT_FREEBSD4 */
734
735 /*
736 * Change current working directory to a given file descriptor.
737 */
738 #ifndef _SYS_SYSPROTO_H_
739 struct fchdir_args {
740 int fd;
741 };
742 #endif
743 int
744 sys_fchdir(td, uap)
745 struct thread *td;
746 struct fchdir_args /* {
747 int fd;
748 } */ *uap;
749 {
750 register struct filedesc *fdp = td->td_proc->p_fd;
751 struct vnode *vp, *tdp, *vpold;
752 struct mount *mp;
753 struct file *fp;
754 int vfslocked;
755 int error;
756
757 AUDIT_ARG_FD(uap->fd);
758 if ((error = getvnode(fdp, uap->fd, CAP_FCHDIR, &fp)) != 0)
759 return (error);
760 vp = fp->f_vnode;
761 VREF(vp);
762 fdrop(fp, td);
763 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
764 vn_lock(vp, LK_SHARED | LK_RETRY);
765 AUDIT_ARG_VNODE1(vp);
766 error = change_dir(vp, td);
767 while (!error && (mp = vp->v_mountedhere) != NULL) {
768 int tvfslocked;
769 if (vfs_busy(mp, 0))
770 continue;
771 tvfslocked = VFS_LOCK_GIANT(mp);
772 error = VFS_ROOT(mp, LK_SHARED, &tdp);
773 vfs_unbusy(mp);
774 if (error) {
775 VFS_UNLOCK_GIANT(tvfslocked);
776 break;
777 }
778 vput(vp);
779 VFS_UNLOCK_GIANT(vfslocked);
780 vp = tdp;
781 vfslocked = tvfslocked;
782 }
783 if (error) {
784 vput(vp);
785 VFS_UNLOCK_GIANT(vfslocked);
786 return (error);
787 }
788 VOP_UNLOCK(vp, 0);
789 VFS_UNLOCK_GIANT(vfslocked);
790 FILEDESC_XLOCK(fdp);
791 vpold = fdp->fd_cdir;
792 fdp->fd_cdir = vp;
793 FILEDESC_XUNLOCK(fdp);
794 vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
795 vrele(vpold);
796 VFS_UNLOCK_GIANT(vfslocked);
797 return (0);
798 }
799
800 /*
801 * Change current working directory (``.'').
802 */
803 #ifndef _SYS_SYSPROTO_H_
804 struct chdir_args {
805 char *path;
806 };
807 #endif
808 int
809 sys_chdir(td, uap)
810 struct thread *td;
811 struct chdir_args /* {
812 char *path;
813 } */ *uap;
814 {
815
816 return (kern_chdir(td, uap->path, UIO_USERSPACE));
817 }
818
819 int
820 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
821 {
822 register struct filedesc *fdp = td->td_proc->p_fd;
823 int error;
824 struct nameidata nd;
825 struct vnode *vp;
826 int vfslocked;
827
828 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
829 MPSAFE, pathseg, path, td);
830 if ((error = namei(&nd)) != 0)
831 return (error);
832 vfslocked = NDHASGIANT(&nd);
833 if ((error = change_dir(nd.ni_vp, td)) != 0) {
834 vput(nd.ni_vp);
835 VFS_UNLOCK_GIANT(vfslocked);
836 NDFREE(&nd, NDF_ONLY_PNBUF);
837 return (error);
838 }
839 VOP_UNLOCK(nd.ni_vp, 0);
840 VFS_UNLOCK_GIANT(vfslocked);
841 NDFREE(&nd, NDF_ONLY_PNBUF);
842 FILEDESC_XLOCK(fdp);
843 vp = fdp->fd_cdir;
844 fdp->fd_cdir = nd.ni_vp;
845 FILEDESC_XUNLOCK(fdp);
846 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
847 vrele(vp);
848 VFS_UNLOCK_GIANT(vfslocked);
849 return (0);
850 }
851
852 /*
853 * Helper function for raised chroot(2) security function: Refuse if
854 * any filedescriptors are open directories.
855 */
856 static int
857 chroot_refuse_vdir_fds(fdp)
858 struct filedesc *fdp;
859 {
860 struct vnode *vp;
861 struct file *fp;
862 int fd;
863
864 FILEDESC_LOCK_ASSERT(fdp);
865
866 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
867 fp = fget_locked(fdp, fd);
868 if (fp == NULL)
869 continue;
870 if (fp->f_type == DTYPE_VNODE) {
871 vp = fp->f_vnode;
872 if (vp->v_type == VDIR)
873 return (EPERM);
874 }
875 }
876 return (0);
877 }
878
879 /*
880 * This sysctl determines if we will allow a process to chroot(2) if it
881 * has a directory open:
882 * 0: disallowed for all processes.
883 * 1: allowed for processes that were not already chroot(2)'ed.
884 * 2: allowed for all processes.
885 */
886
887 static int chroot_allow_open_directories = 1;
888
889 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
890 &chroot_allow_open_directories, 0, "");
891
892 /*
893 * Change notion of root (``/'') directory.
894 */
895 #ifndef _SYS_SYSPROTO_H_
896 struct chroot_args {
897 char *path;
898 };
899 #endif
900 int
901 sys_chroot(td, uap)
902 struct thread *td;
903 struct chroot_args /* {
904 char *path;
905 } */ *uap;
906 {
907 int error;
908 struct nameidata nd;
909 int vfslocked;
910
911 error = priv_check(td, PRIV_VFS_CHROOT);
912 if (error)
913 return (error);
914 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
915 AUDITVNODE1, UIO_USERSPACE, uap->path, td);
916 error = namei(&nd);
917 if (error)
918 goto error;
919 vfslocked = NDHASGIANT(&nd);
920 if ((error = change_dir(nd.ni_vp, td)) != 0)
921 goto e_vunlock;
922 #ifdef MAC
923 if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
924 goto e_vunlock;
925 #endif
926 VOP_UNLOCK(nd.ni_vp, 0);
927 error = change_root(nd.ni_vp, td);
928 vrele(nd.ni_vp);
929 VFS_UNLOCK_GIANT(vfslocked);
930 NDFREE(&nd, NDF_ONLY_PNBUF);
931 return (error);
932 e_vunlock:
933 vput(nd.ni_vp);
934 VFS_UNLOCK_GIANT(vfslocked);
935 error:
936 NDFREE(&nd, NDF_ONLY_PNBUF);
937 return (error);
938 }
939
940 /*
941 * Common routine for chroot and chdir. Callers must provide a locked vnode
942 * instance.
943 */
944 int
945 change_dir(vp, td)
946 struct vnode *vp;
947 struct thread *td;
948 {
949 int error;
950
951 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
952 if (vp->v_type != VDIR)
953 return (ENOTDIR);
954 #ifdef MAC
955 error = mac_vnode_check_chdir(td->td_ucred, vp);
956 if (error)
957 return (error);
958 #endif
959 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
960 return (error);
961 }
962
963 /*
964 * Common routine for kern_chroot() and jail_attach(). The caller is
965 * responsible for invoking priv_check() and mac_vnode_check_chroot() to
966 * authorize this operation.
967 */
968 int
969 change_root(vp, td)
970 struct vnode *vp;
971 struct thread *td;
972 {
973 struct filedesc *fdp;
974 struct vnode *oldvp;
975 int vfslocked;
976 int error;
977
978 VFS_ASSERT_GIANT(vp->v_mount);
979 fdp = td->td_proc->p_fd;
980 FILEDESC_XLOCK(fdp);
981 if (chroot_allow_open_directories == 0 ||
982 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
983 error = chroot_refuse_vdir_fds(fdp);
984 if (error) {
985 FILEDESC_XUNLOCK(fdp);
986 return (error);
987 }
988 }
989 oldvp = fdp->fd_rdir;
990 fdp->fd_rdir = vp;
991 VREF(fdp->fd_rdir);
992 if (!fdp->fd_jdir) {
993 fdp->fd_jdir = vp;
994 VREF(fdp->fd_jdir);
995 }
996 FILEDESC_XUNLOCK(fdp);
997 vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
998 vrele(oldvp);
999 VFS_UNLOCK_GIANT(vfslocked);
1000 return (0);
1001 }
1002
1003 static __inline cap_rights_t
1004 flags_to_rights(int flags)
1005 {
1006 cap_rights_t rights = 0;
1007
1008 switch ((flags & O_ACCMODE)) {
1009 case O_RDONLY:
1010 rights |= CAP_READ;
1011 break;
1012
1013 case O_RDWR:
1014 rights |= CAP_READ;
1015 /* fall through */
1016
1017 case O_WRONLY:
1018 rights |= CAP_WRITE;
1019 break;
1020
1021 case O_EXEC:
1022 rights |= CAP_FEXECVE;
1023 break;
1024 }
1025
1026 if (flags & O_CREAT)
1027 rights |= CAP_CREATE;
1028
1029 if (flags & O_TRUNC)
1030 rights |= CAP_FTRUNCATE;
1031
1032 if ((flags & O_EXLOCK) || (flags & O_SHLOCK))
1033 rights |= CAP_FLOCK;
1034
1035 return (rights);
1036 }
1037
1038 /*
1039 * Check permissions, allocate an open file structure, and call the device
1040 * open routine if any.
1041 */
1042 #ifndef _SYS_SYSPROTO_H_
1043 struct open_args {
1044 char *path;
1045 int flags;
1046 int mode;
1047 };
1048 #endif
1049 int
1050 sys_open(td, uap)
1051 struct thread *td;
1052 register struct open_args /* {
1053 char *path;
1054 int flags;
1055 int mode;
1056 } */ *uap;
1057 {
1058
1059 return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1060 }
1061
1062 #ifndef _SYS_SYSPROTO_H_
1063 struct openat_args {
1064 int fd;
1065 char *path;
1066 int flag;
1067 int mode;
1068 };
1069 #endif
1070 int
1071 sys_openat(struct thread *td, struct openat_args *uap)
1072 {
1073
1074 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1075 uap->mode));
1076 }
1077
1078 int
1079 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1080 int mode)
1081 {
1082
1083 return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1084 }
1085
1086 int
1087 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1088 int flags, int mode)
1089 {
1090 struct proc *p = td->td_proc;
1091 struct filedesc *fdp = p->p_fd;
1092 struct file *fp;
1093 struct vnode *vp;
1094 int cmode;
1095 struct file *nfp;
1096 int type, indx = -1, error, error_open;
1097 struct flock lf;
1098 struct nameidata nd;
1099 int vfslocked;
1100 cap_rights_t rights_needed = CAP_LOOKUP;
1101
1102 AUDIT_ARG_FFLAGS(flags);
1103 AUDIT_ARG_MODE(mode);
1104 /* XXX: audit dirfd */
1105 rights_needed |= flags_to_rights(flags);
1106 /*
1107 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1108 * may be specified.
1109 */
1110 if (flags & O_EXEC) {
1111 if (flags & O_ACCMODE)
1112 return (EINVAL);
1113 } else if ((flags & O_ACCMODE) == O_ACCMODE)
1114 return (EINVAL);
1115 else
1116 flags = FFLAGS(flags);
1117
1118 /*
1119 * allocate the file descriptor, but don't install a descriptor yet
1120 */
1121 error = falloc_noinstall(td, &nfp);
1122 if (error)
1123 return (error);
1124 /* An extra reference on `nfp' has been held for us by falloc_noinstall(). */
1125 fp = nfp;
1126 /* Set the flags early so the finit in devfs can pick them up. */
1127 fp->f_flag = flags & FMASK;
1128 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1129 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg,
1130 path, fd, rights_needed, td);
1131 td->td_dupfd = -1; /* XXX check for fdopen */
1132 error = vn_open(&nd, &flags, cmode, fp);
1133 if (error) {
1134 /*
1135 * If the vn_open replaced the method vector, something
1136 * wonderous happened deep below and we just pass it up
1137 * pretending we know what we do.
1138 */
1139 if (error == ENXIO && fp->f_ops != &badfileops)
1140 goto success;
1141
1142 /*
1143 * handle special fdopen() case. bleh. dupfdopen() is
1144 * responsible for dropping the old contents of ofiles[indx]
1145 * if it succeeds.
1146 *
1147 * Don't do this for relative (capability) lookups; we don't
1148 * understand exactly what would happen, and we don't think
1149 * that it ever should.
1150 */
1151 if ((nd.ni_strictrelative == 0) &&
1152 (error == ENODEV || error == ENXIO) &&
1153 (td->td_dupfd >= 0)) {
1154 /* XXX from fdopen */
1155 error_open = error;
1156 if ((error = finstall(td, fp, &indx, flags)) != 0)
1157 goto bad_unlocked;
1158 if ((error = dupfdopen(td, fdp, indx, td->td_dupfd,
1159 flags, error_open)) == 0)
1160 goto success;
1161 }
1162 /*
1163 * Clean up the descriptor, but only if another thread hadn't
1164 * replaced or closed it.
1165 */
1166 if (indx != -1)
1167 fdclose(fdp, fp, indx, td);
1168 fdrop(fp, td);
1169
1170 return (error);
1171 }
1172 td->td_dupfd = 0;
1173 vfslocked = NDHASGIANT(&nd);
1174 NDFREE(&nd, NDF_ONLY_PNBUF);
1175 vp = nd.ni_vp;
1176
1177 /*
1178 * Store the vnode, for any f_type. Typically, the vnode use
1179 * count is decremented by direct call to vn_closefile() for
1180 * files that switched type in the cdevsw fdopen() method.
1181 */
1182 fp->f_vnode = vp;
1183 /*
1184 * If the file wasn't claimed by devfs bind it to the normal
1185 * vnode operations here.
1186 */
1187 if (fp->f_ops == &badfileops) {
1188 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1189 fp->f_seqcount = 1;
1190 finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1191 }
1192
1193 VOP_UNLOCK(vp, 0);
1194 if (fp->f_type == DTYPE_VNODE && (flags & (O_EXLOCK | O_SHLOCK)) != 0) {
1195 lf.l_whence = SEEK_SET;
1196 lf.l_start = 0;
1197 lf.l_len = 0;
1198 if (flags & O_EXLOCK)
1199 lf.l_type = F_WRLCK;
1200 else
1201 lf.l_type = F_RDLCK;
1202 type = F_FLOCK;
1203 if ((flags & FNONBLOCK) == 0)
1204 type |= F_WAIT;
1205 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1206 type)) != 0)
1207 goto bad;
1208 atomic_set_int(&fp->f_flag, FHASLOCK);
1209 }
1210 if (flags & O_TRUNC) {
1211 error = fo_truncate(fp, 0, td->td_ucred, td);
1212 if (error)
1213 goto bad;
1214 }
1215 VFS_UNLOCK_GIANT(vfslocked);
1216 success:
1217 /*
1218 * If we haven't already installed the FD (for dupfdopen), do so now.
1219 */
1220 if (indx == -1) {
1221 #ifdef CAPABILITIES
1222 if (nd.ni_strictrelative == 1) {
1223 /*
1224 * We are doing a strict relative lookup; wrap the
1225 * result in a capability.
1226 */
1227 if ((error = kern_capwrap(td, fp, nd.ni_baserights,
1228 &indx)) != 0)
1229 goto bad_unlocked;
1230 } else
1231 #endif
1232 if ((error = finstall(td, fp, &indx, flags)) != 0)
1233 goto bad_unlocked;
1234
1235 }
1236
1237 /*
1238 * Release our private reference, leaving the one associated with
1239 * the descriptor table intact.
1240 */
1241 fdrop(fp, td);
1242 td->td_retval[0] = indx;
1243 return (0);
1244 bad:
1245 VFS_UNLOCK_GIANT(vfslocked);
1246 bad_unlocked:
1247 if (indx != -1)
1248 fdclose(fdp, fp, indx, td);
1249 fdrop(fp, td);
1250 td->td_retval[0] = -1;
1251 return (error);
1252 }
1253
1254 #ifdef COMPAT_43
1255 /*
1256 * Create a file.
1257 */
1258 #ifndef _SYS_SYSPROTO_H_
1259 struct ocreat_args {
1260 char *path;
1261 int mode;
1262 };
1263 #endif
1264 int
1265 ocreat(td, uap)
1266 struct thread *td;
1267 register struct ocreat_args /* {
1268 char *path;
1269 int mode;
1270 } */ *uap;
1271 {
1272
1273 return (kern_open(td, uap->path, UIO_USERSPACE,
1274 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1275 }
1276 #endif /* COMPAT_43 */
1277
1278 /*
1279 * Create a special file.
1280 */
1281 #ifndef _SYS_SYSPROTO_H_
1282 struct mknod_args {
1283 char *path;
1284 int mode;
1285 int dev;
1286 };
1287 #endif
1288 int
1289 sys_mknod(td, uap)
1290 struct thread *td;
1291 register struct mknod_args /* {
1292 char *path;
1293 int mode;
1294 int dev;
1295 } */ *uap;
1296 {
1297
1298 return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1299 }
1300
1301 #ifndef _SYS_SYSPROTO_H_
1302 struct mknodat_args {
1303 int fd;
1304 char *path;
1305 mode_t mode;
1306 dev_t dev;
1307 };
1308 #endif
1309 int
1310 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1311 {
1312
1313 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1314 uap->dev));
1315 }
1316
1317 int
1318 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1319 int dev)
1320 {
1321
1322 return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1323 }
1324
1325 int
1326 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1327 int mode, int dev)
1328 {
1329 struct vnode *vp;
1330 struct mount *mp;
1331 struct vattr vattr;
1332 int error;
1333 int whiteout = 0;
1334 struct nameidata nd;
1335 int vfslocked;
1336
1337 AUDIT_ARG_MODE(mode);
1338 AUDIT_ARG_DEV(dev);
1339 switch (mode & S_IFMT) {
1340 case S_IFCHR:
1341 case S_IFBLK:
1342 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1343 break;
1344 case S_IFMT:
1345 error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1346 break;
1347 case S_IFWHT:
1348 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1349 break;
1350 case S_IFIFO:
1351 if (dev == 0)
1352 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1353 /* FALLTHROUGH */
1354 default:
1355 error = EINVAL;
1356 break;
1357 }
1358 if (error)
1359 return (error);
1360 restart:
1361 bwillwrite();
1362 NDINIT_ATRIGHTS(&nd, CREATE,
1363 LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1, pathseg, path, fd,
1364 CAP_MKFIFO, td);
1365 if ((error = namei(&nd)) != 0)
1366 return (error);
1367 vfslocked = NDHASGIANT(&nd);
1368 vp = nd.ni_vp;
1369 if (vp != NULL) {
1370 NDFREE(&nd, NDF_ONLY_PNBUF);
1371 if (vp == nd.ni_dvp)
1372 vrele(nd.ni_dvp);
1373 else
1374 vput(nd.ni_dvp);
1375 vrele(vp);
1376 VFS_UNLOCK_GIANT(vfslocked);
1377 return (EEXIST);
1378 } else {
1379 VATTR_NULL(&vattr);
1380 vattr.va_mode = (mode & ALLPERMS) &
1381 ~td->td_proc->p_fd->fd_cmask;
1382 vattr.va_rdev = dev;
1383 whiteout = 0;
1384
1385 switch (mode & S_IFMT) {
1386 case S_IFMT: /* used by badsect to flag bad sectors */
1387 vattr.va_type = VBAD;
1388 break;
1389 case S_IFCHR:
1390 vattr.va_type = VCHR;
1391 break;
1392 case S_IFBLK:
1393 vattr.va_type = VBLK;
1394 break;
1395 case S_IFWHT:
1396 whiteout = 1;
1397 break;
1398 default:
1399 panic("kern_mknod: invalid mode");
1400 }
1401 }
1402 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1403 NDFREE(&nd, NDF_ONLY_PNBUF);
1404 vput(nd.ni_dvp);
1405 VFS_UNLOCK_GIANT(vfslocked);
1406 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1407 return (error);
1408 goto restart;
1409 }
1410 #ifdef MAC
1411 if (error == 0 && !whiteout)
1412 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1413 &nd.ni_cnd, &vattr);
1414 #endif
1415 if (!error) {
1416 if (whiteout)
1417 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1418 else {
1419 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1420 &nd.ni_cnd, &vattr);
1421 if (error == 0)
1422 vput(nd.ni_vp);
1423 }
1424 }
1425 NDFREE(&nd, NDF_ONLY_PNBUF);
1426 vput(nd.ni_dvp);
1427 vn_finished_write(mp);
1428 VFS_UNLOCK_GIANT(vfslocked);
1429 return (error);
1430 }
1431
1432 /*
1433 * Create a named pipe.
1434 */
1435 #ifndef _SYS_SYSPROTO_H_
1436 struct mkfifo_args {
1437 char *path;
1438 int mode;
1439 };
1440 #endif
1441 int
1442 sys_mkfifo(td, uap)
1443 struct thread *td;
1444 register struct mkfifo_args /* {
1445 char *path;
1446 int mode;
1447 } */ *uap;
1448 {
1449
1450 return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1451 }
1452
1453 #ifndef _SYS_SYSPROTO_H_
1454 struct mkfifoat_args {
1455 int fd;
1456 char *path;
1457 mode_t mode;
1458 };
1459 #endif
1460 int
1461 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1462 {
1463
1464 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1465 uap->mode));
1466 }
1467
1468 int
1469 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1470 {
1471
1472 return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1473 }
1474
1475 int
1476 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1477 int mode)
1478 {
1479 struct mount *mp;
1480 struct vattr vattr;
1481 int error;
1482 struct nameidata nd;
1483 int vfslocked;
1484
1485 AUDIT_ARG_MODE(mode);
1486 restart:
1487 bwillwrite();
1488 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1489 pathseg, path, fd, td);
1490 if ((error = namei(&nd)) != 0)
1491 return (error);
1492 vfslocked = NDHASGIANT(&nd);
1493 if (nd.ni_vp != NULL) {
1494 NDFREE(&nd, NDF_ONLY_PNBUF);
1495 if (nd.ni_vp == nd.ni_dvp)
1496 vrele(nd.ni_dvp);
1497 else
1498 vput(nd.ni_dvp);
1499 vrele(nd.ni_vp);
1500 VFS_UNLOCK_GIANT(vfslocked);
1501 return (EEXIST);
1502 }
1503 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1504 NDFREE(&nd, NDF_ONLY_PNBUF);
1505 vput(nd.ni_dvp);
1506 VFS_UNLOCK_GIANT(vfslocked);
1507 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1508 return (error);
1509 goto restart;
1510 }
1511 VATTR_NULL(&vattr);
1512 vattr.va_type = VFIFO;
1513 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1514 #ifdef MAC
1515 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1516 &vattr);
1517 if (error)
1518 goto out;
1519 #endif
1520 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1521 if (error == 0)
1522 vput(nd.ni_vp);
1523 #ifdef MAC
1524 out:
1525 #endif
1526 vput(nd.ni_dvp);
1527 vn_finished_write(mp);
1528 VFS_UNLOCK_GIANT(vfslocked);
1529 NDFREE(&nd, NDF_ONLY_PNBUF);
1530 return (error);
1531 }
1532
1533 /*
1534 * Make a hard file link.
1535 */
1536 #ifndef _SYS_SYSPROTO_H_
1537 struct link_args {
1538 char *path;
1539 char *link;
1540 };
1541 #endif
1542 int
1543 sys_link(td, uap)
1544 struct thread *td;
1545 register struct link_args /* {
1546 char *path;
1547 char *link;
1548 } */ *uap;
1549 {
1550
1551 return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1552 }
1553
1554 #ifndef _SYS_SYSPROTO_H_
1555 struct linkat_args {
1556 int fd1;
1557 char *path1;
1558 int fd2;
1559 char *path2;
1560 int flag;
1561 };
1562 #endif
1563 int
1564 sys_linkat(struct thread *td, struct linkat_args *uap)
1565 {
1566 int flag;
1567
1568 flag = uap->flag;
1569 if (flag & ~AT_SYMLINK_FOLLOW)
1570 return (EINVAL);
1571
1572 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1573 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1574 }
1575
1576 int hardlink_check_uid = 0;
1577 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1578 &hardlink_check_uid, 0,
1579 "Unprivileged processes cannot create hard links to files owned by other "
1580 "users");
1581 static int hardlink_check_gid = 0;
1582 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1583 &hardlink_check_gid, 0,
1584 "Unprivileged processes cannot create hard links to files owned by other "
1585 "groups");
1586
1587 static int
1588 can_hardlink(struct vnode *vp, struct ucred *cred)
1589 {
1590 struct vattr va;
1591 int error;
1592
1593 if (!hardlink_check_uid && !hardlink_check_gid)
1594 return (0);
1595
1596 error = VOP_GETATTR(vp, &va, cred);
1597 if (error != 0)
1598 return (error);
1599
1600 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1601 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1602 if (error)
1603 return (error);
1604 }
1605
1606 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1607 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1608 if (error)
1609 return (error);
1610 }
1611
1612 return (0);
1613 }
1614
1615 int
1616 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1617 {
1618
1619 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1620 }
1621
1622 int
1623 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1624 enum uio_seg segflg, int follow)
1625 {
1626 struct vnode *vp;
1627 struct mount *mp;
1628 struct nameidata nd;
1629 int vfslocked;
1630 int lvfslocked;
1631 int error;
1632
1633 bwillwrite();
1634 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1635 fd1, td);
1636
1637 if ((error = namei(&nd)) != 0)
1638 return (error);
1639 vfslocked = NDHASGIANT(&nd);
1640 NDFREE(&nd, NDF_ONLY_PNBUF);
1641 vp = nd.ni_vp;
1642 if (vp->v_type == VDIR) {
1643 vrele(vp);
1644 VFS_UNLOCK_GIANT(vfslocked);
1645 return (EPERM); /* POSIX */
1646 }
1647 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1648 vrele(vp);
1649 VFS_UNLOCK_GIANT(vfslocked);
1650 return (error);
1651 }
1652 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2,
1653 segflg, path2, fd2, td);
1654 if ((error = namei(&nd)) == 0) {
1655 lvfslocked = NDHASGIANT(&nd);
1656 if (nd.ni_vp != NULL) {
1657 if (nd.ni_dvp == nd.ni_vp)
1658 vrele(nd.ni_dvp);
1659 else
1660 vput(nd.ni_dvp);
1661 vrele(nd.ni_vp);
1662 error = EEXIST;
1663 } else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1664 == 0) {
1665 error = can_hardlink(vp, td->td_ucred);
1666 if (error == 0)
1667 #ifdef MAC
1668 error = mac_vnode_check_link(td->td_ucred,
1669 nd.ni_dvp, vp, &nd.ni_cnd);
1670 if (error == 0)
1671 #endif
1672 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1673 VOP_UNLOCK(vp, 0);
1674 vput(nd.ni_dvp);
1675 }
1676 NDFREE(&nd, NDF_ONLY_PNBUF);
1677 VFS_UNLOCK_GIANT(lvfslocked);
1678 }
1679 vrele(vp);
1680 vn_finished_write(mp);
1681 VFS_UNLOCK_GIANT(vfslocked);
1682 return (error);
1683 }
1684
1685 /*
1686 * Make a symbolic link.
1687 */
1688 #ifndef _SYS_SYSPROTO_H_
1689 struct symlink_args {
1690 char *path;
1691 char *link;
1692 };
1693 #endif
1694 int
1695 sys_symlink(td, uap)
1696 struct thread *td;
1697 register struct symlink_args /* {
1698 char *path;
1699 char *link;
1700 } */ *uap;
1701 {
1702
1703 return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1704 }
1705
1706 #ifndef _SYS_SYSPROTO_H_
1707 struct symlinkat_args {
1708 char *path;
1709 int fd;
1710 char *path2;
1711 };
1712 #endif
1713 int
1714 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1715 {
1716
1717 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1718 UIO_USERSPACE));
1719 }
1720
1721 int
1722 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1723 {
1724
1725 return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1726 }
1727
1728 int
1729 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1730 enum uio_seg segflg)
1731 {
1732 struct mount *mp;
1733 struct vattr vattr;
1734 char *syspath;
1735 int error;
1736 struct nameidata nd;
1737 int vfslocked;
1738
1739 if (segflg == UIO_SYSSPACE) {
1740 syspath = path1;
1741 } else {
1742 syspath = uma_zalloc(namei_zone, M_WAITOK);
1743 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1744 goto out;
1745 }
1746 AUDIT_ARG_TEXT(syspath);
1747 restart:
1748 bwillwrite();
1749 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1750 segflg, path2, fd, td);
1751 if ((error = namei(&nd)) != 0)
1752 goto out;
1753 vfslocked = NDHASGIANT(&nd);
1754 if (nd.ni_vp) {
1755 NDFREE(&nd, NDF_ONLY_PNBUF);
1756 if (nd.ni_vp == nd.ni_dvp)
1757 vrele(nd.ni_dvp);
1758 else
1759 vput(nd.ni_dvp);
1760 vrele(nd.ni_vp);
1761 VFS_UNLOCK_GIANT(vfslocked);
1762 error = EEXIST;
1763 goto out;
1764 }
1765 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1766 NDFREE(&nd, NDF_ONLY_PNBUF);
1767 vput(nd.ni_dvp);
1768 VFS_UNLOCK_GIANT(vfslocked);
1769 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1770 goto out;
1771 goto restart;
1772 }
1773 VATTR_NULL(&vattr);
1774 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1775 #ifdef MAC
1776 vattr.va_type = VLNK;
1777 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1778 &vattr);
1779 if (error)
1780 goto out2;
1781 #endif
1782 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1783 if (error == 0)
1784 vput(nd.ni_vp);
1785 #ifdef MAC
1786 out2:
1787 #endif
1788 NDFREE(&nd, NDF_ONLY_PNBUF);
1789 vput(nd.ni_dvp);
1790 vn_finished_write(mp);
1791 VFS_UNLOCK_GIANT(vfslocked);
1792 out:
1793 if (segflg != UIO_SYSSPACE)
1794 uma_zfree(namei_zone, syspath);
1795 return (error);
1796 }
1797
1798 /*
1799 * Delete a whiteout from the filesystem.
1800 */
1801 int
1802 sys_undelete(td, uap)
1803 struct thread *td;
1804 register struct undelete_args /* {
1805 char *path;
1806 } */ *uap;
1807 {
1808 int error;
1809 struct mount *mp;
1810 struct nameidata nd;
1811 int vfslocked;
1812
1813 restart:
1814 bwillwrite();
1815 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1816 UIO_USERSPACE, uap->path, td);
1817 error = namei(&nd);
1818 if (error)
1819 return (error);
1820 vfslocked = NDHASGIANT(&nd);
1821
1822 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1823 NDFREE(&nd, NDF_ONLY_PNBUF);
1824 if (nd.ni_vp == nd.ni_dvp)
1825 vrele(nd.ni_dvp);
1826 else
1827 vput(nd.ni_dvp);
1828 if (nd.ni_vp)
1829 vrele(nd.ni_vp);
1830 VFS_UNLOCK_GIANT(vfslocked);
1831 return (EEXIST);
1832 }
1833 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1834 NDFREE(&nd, NDF_ONLY_PNBUF);
1835 vput(nd.ni_dvp);
1836 VFS_UNLOCK_GIANT(vfslocked);
1837 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1838 return (error);
1839 goto restart;
1840 }
1841 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1842 NDFREE(&nd, NDF_ONLY_PNBUF);
1843 vput(nd.ni_dvp);
1844 vn_finished_write(mp);
1845 VFS_UNLOCK_GIANT(vfslocked);
1846 return (error);
1847 }
1848
1849 /*
1850 * Delete a name from the filesystem.
1851 */
1852 #ifndef _SYS_SYSPROTO_H_
1853 struct unlink_args {
1854 char *path;
1855 };
1856 #endif
1857 int
1858 sys_unlink(td, uap)
1859 struct thread *td;
1860 struct unlink_args /* {
1861 char *path;
1862 } */ *uap;
1863 {
1864
1865 return (kern_unlink(td, uap->path, UIO_USERSPACE));
1866 }
1867
1868 #ifndef _SYS_SYSPROTO_H_
1869 struct unlinkat_args {
1870 int fd;
1871 char *path;
1872 int flag;
1873 };
1874 #endif
1875 int
1876 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1877 {
1878 int flag = uap->flag;
1879 int fd = uap->fd;
1880 char *path = uap->path;
1881
1882 if (flag & ~AT_REMOVEDIR)
1883 return (EINVAL);
1884
1885 if (flag & AT_REMOVEDIR)
1886 return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1887 else
1888 return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1889 }
1890
1891 int
1892 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1893 {
1894
1895 return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1896 }
1897
1898 int
1899 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1900 ino_t oldinum)
1901 {
1902 struct mount *mp;
1903 struct vnode *vp;
1904 int error;
1905 struct nameidata nd;
1906 struct stat sb;
1907 int vfslocked;
1908
1909 restart:
1910 bwillwrite();
1911 NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1912 pathseg, path, fd, td);
1913 if ((error = namei(&nd)) != 0)
1914 return (error == EINVAL ? EPERM : error);
1915 vfslocked = NDHASGIANT(&nd);
1916 vp = nd.ni_vp;
1917 if (vp->v_type == VDIR && oldinum == 0) {
1918 error = EPERM; /* POSIX */
1919 } else if (oldinum != 0 &&
1920 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1921 sb.st_ino != oldinum) {
1922 error = EIDRM; /* Identifier removed */
1923 } else {
1924 /*
1925 * The root of a mounted filesystem cannot be deleted.
1926 *
1927 * XXX: can this only be a VDIR case?
1928 */
1929 if (vp->v_vflag & VV_ROOT)
1930 error = EBUSY;
1931 }
1932 if (error == 0) {
1933 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1934 NDFREE(&nd, NDF_ONLY_PNBUF);
1935 vput(nd.ni_dvp);
1936 if (vp == nd.ni_dvp)
1937 vrele(vp);
1938 else
1939 vput(vp);
1940 VFS_UNLOCK_GIANT(vfslocked);
1941 if ((error = vn_start_write(NULL, &mp,
1942 V_XSLEEP | PCATCH)) != 0)
1943 return (error);
1944 goto restart;
1945 }
1946 #ifdef MAC
1947 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1948 &nd.ni_cnd);
1949 if (error)
1950 goto out;
1951 #endif
1952 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1953 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1954 #ifdef MAC
1955 out:
1956 #endif
1957 vn_finished_write(mp);
1958 }
1959 NDFREE(&nd, NDF_ONLY_PNBUF);
1960 vput(nd.ni_dvp);
1961 if (vp == nd.ni_dvp)
1962 vrele(vp);
1963 else
1964 vput(vp);
1965 VFS_UNLOCK_GIANT(vfslocked);
1966 return (error);
1967 }
1968
1969 /*
1970 * Reposition read/write file offset.
1971 */
1972 #ifndef _SYS_SYSPROTO_H_
1973 struct lseek_args {
1974 int fd;
1975 int pad;
1976 off_t offset;
1977 int whence;
1978 };
1979 #endif
1980 int
1981 sys_lseek(td, uap)
1982 struct thread *td;
1983 register struct lseek_args /* {
1984 int fd;
1985 int pad;
1986 off_t offset;
1987 int whence;
1988 } */ *uap;
1989 {
1990 struct ucred *cred = td->td_ucred;
1991 struct file *fp;
1992 struct vnode *vp;
1993 struct vattr vattr;
1994 off_t foffset, offset, size;
1995 int error, noneg;
1996 int vfslocked;
1997
1998 AUDIT_ARG_FD(uap->fd);
1999 if ((error = fget(td, uap->fd, CAP_SEEK, &fp)) != 0)
2000 return (error);
2001 if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
2002 fdrop(fp, td);
2003 return (ESPIPE);
2004 }
2005 vp = fp->f_vnode;
2006 foffset = foffset_lock(fp, 0);
2007 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
2008 noneg = (vp->v_type != VCHR);
2009 offset = uap->offset;
2010 switch (uap->whence) {
2011 case L_INCR:
2012 if (noneg &&
2013 (foffset < 0 ||
2014 (offset > 0 && foffset > OFF_MAX - offset))) {
2015 error = EOVERFLOW;
2016 break;
2017 }
2018 offset += foffset;
2019 break;
2020 case L_XTND:
2021 vn_lock(vp, LK_SHARED | LK_RETRY);
2022 error = VOP_GETATTR(vp, &vattr, cred);
2023 VOP_UNLOCK(vp, 0);
2024 if (error)
2025 break;
2026
2027 /*
2028 * If the file references a disk device, then fetch
2029 * the media size and use that to determine the ending
2030 * offset.
2031 */
2032 if (vattr.va_size == 0 && vp->v_type == VCHR &&
2033 fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0)
2034 vattr.va_size = size;
2035 if (noneg &&
2036 (vattr.va_size > OFF_MAX ||
2037 (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
2038 error = EOVERFLOW;
2039 break;
2040 }
2041 offset += vattr.va_size;
2042 break;
2043 case L_SET:
2044 break;
2045 case SEEK_DATA:
2046 error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
2047 break;
2048 case SEEK_HOLE:
2049 error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
2050 break;
2051 default:
2052 error = EINVAL;
2053 }
2054 if (error == 0 && noneg && offset < 0)
2055 error = EINVAL;
2056 if (error != 0)
2057 goto drop;
2058 VFS_KNOTE_UNLOCKED(vp, 0);
2059 *(off_t *)(td->td_retval) = offset;
2060 drop:
2061 fdrop(fp, td);
2062 VFS_UNLOCK_GIANT(vfslocked);
2063 foffset_unlock(fp, offset, error != 0 ? FOF_NOUPDATE : 0);
2064 return (error);
2065 }
2066
2067 #if defined(COMPAT_43)
2068 /*
2069 * Reposition read/write file offset.
2070 */
2071 #ifndef _SYS_SYSPROTO_H_
2072 struct olseek_args {
2073 int fd;
2074 long offset;
2075 int whence;
2076 };
2077 #endif
2078 int
2079 olseek(td, uap)
2080 struct thread *td;
2081 register struct olseek_args /* {
2082 int fd;
2083 long offset;
2084 int whence;
2085 } */ *uap;
2086 {
2087 struct lseek_args /* {
2088 int fd;
2089 int pad;
2090 off_t offset;
2091 int whence;
2092 } */ nuap;
2093
2094 nuap.fd = uap->fd;
2095 nuap.offset = uap->offset;
2096 nuap.whence = uap->whence;
2097 return (sys_lseek(td, &nuap));
2098 }
2099 #endif /* COMPAT_43 */
2100
2101 /* Version with the 'pad' argument */
2102 int
2103 freebsd6_lseek(td, uap)
2104 struct thread *td;
2105 register struct freebsd6_lseek_args *uap;
2106 {
2107 struct lseek_args ouap;
2108
2109 ouap.fd = uap->fd;
2110 ouap.offset = uap->offset;
2111 ouap.whence = uap->whence;
2112 return (sys_lseek(td, &ouap));
2113 }
2114
2115 /*
2116 * Check access permissions using passed credentials.
2117 */
2118 static int
2119 vn_access(vp, user_flags, cred, td)
2120 struct vnode *vp;
2121 int user_flags;
2122 struct ucred *cred;
2123 struct thread *td;
2124 {
2125 int error;
2126 accmode_t accmode;
2127
2128 /* Flags == 0 means only check for existence. */
2129 error = 0;
2130 if (user_flags) {
2131 accmode = 0;
2132 if (user_flags & R_OK)
2133 accmode |= VREAD;
2134 if (user_flags & W_OK)
2135 accmode |= VWRITE;
2136 if (user_flags & X_OK)
2137 accmode |= VEXEC;
2138 #ifdef MAC
2139 error = mac_vnode_check_access(cred, vp, accmode);
2140 if (error)
2141 return (error);
2142 #endif
2143 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2144 error = VOP_ACCESS(vp, accmode, cred, td);
2145 }
2146 return (error);
2147 }
2148
2149 /*
2150 * Check access permissions using "real" credentials.
2151 */
2152 #ifndef _SYS_SYSPROTO_H_
2153 struct access_args {
2154 char *path;
2155 int flags;
2156 };
2157 #endif
2158 int
2159 sys_access(td, uap)
2160 struct thread *td;
2161 register struct access_args /* {
2162 char *path;
2163 int flags;
2164 } */ *uap;
2165 {
2166
2167 return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2168 }
2169
2170 #ifndef _SYS_SYSPROTO_H_
2171 struct faccessat_args {
2172 int dirfd;
2173 char *path;
2174 int mode;
2175 int flag;
2176 }
2177 #endif
2178 int
2179 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2180 {
2181
2182 if (uap->flag & ~AT_EACCESS)
2183 return (EINVAL);
2184 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2185 uap->mode));
2186 }
2187
2188 int
2189 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2190 {
2191
2192 return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2193 }
2194
2195 int
2196 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2197 int flags, int mode)
2198 {
2199 struct ucred *cred, *tmpcred;
2200 struct vnode *vp;
2201 struct nameidata nd;
2202 int vfslocked;
2203 int error;
2204
2205 /*
2206 * Create and modify a temporary credential instead of one that
2207 * is potentially shared.
2208 */
2209 if (!(flags & AT_EACCESS)) {
2210 cred = td->td_ucred;
2211 tmpcred = crdup(cred);
2212 tmpcred->cr_uid = cred->cr_ruid;
2213 tmpcred->cr_groups[0] = cred->cr_rgid;
2214 td->td_ucred = tmpcred;
2215 } else
2216 cred = tmpcred = td->td_ucred;
2217 AUDIT_ARG_VALUE(mode);
2218 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2219 AUDITVNODE1, pathseg, path, fd, CAP_FSTAT, td);
2220 if ((error = namei(&nd)) != 0)
2221 goto out1;
2222 vfslocked = NDHASGIANT(&nd);
2223 vp = nd.ni_vp;
2224
2225 error = vn_access(vp, mode, tmpcred, td);
2226 NDFREE(&nd, NDF_ONLY_PNBUF);
2227 vput(vp);
2228 VFS_UNLOCK_GIANT(vfslocked);
2229 out1:
2230 if (!(flags & AT_EACCESS)) {
2231 td->td_ucred = cred;
2232 crfree(tmpcred);
2233 }
2234 return (error);
2235 }
2236
2237 /*
2238 * Check access permissions using "effective" credentials.
2239 */
2240 #ifndef _SYS_SYSPROTO_H_
2241 struct eaccess_args {
2242 char *path;
2243 int flags;
2244 };
2245 #endif
2246 int
2247 sys_eaccess(td, uap)
2248 struct thread *td;
2249 register struct eaccess_args /* {
2250 char *path;
2251 int flags;
2252 } */ *uap;
2253 {
2254
2255 return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2256 }
2257
2258 int
2259 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2260 {
2261
2262 return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2263 }
2264
2265 #if defined(COMPAT_43)
2266 /*
2267 * Get file status; this version follows links.
2268 */
2269 #ifndef _SYS_SYSPROTO_H_
2270 struct ostat_args {
2271 char *path;
2272 struct ostat *ub;
2273 };
2274 #endif
2275 int
2276 ostat(td, uap)
2277 struct thread *td;
2278 register struct ostat_args /* {
2279 char *path;
2280 struct ostat *ub;
2281 } */ *uap;
2282 {
2283 struct stat sb;
2284 struct ostat osb;
2285 int error;
2286
2287 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2288 if (error)
2289 return (error);
2290 cvtstat(&sb, &osb);
2291 error = copyout(&osb, uap->ub, sizeof (osb));
2292 return (error);
2293 }
2294
2295 /*
2296 * Get file status; this version does not follow links.
2297 */
2298 #ifndef _SYS_SYSPROTO_H_
2299 struct olstat_args {
2300 char *path;
2301 struct ostat *ub;
2302 };
2303 #endif
2304 int
2305 olstat(td, uap)
2306 struct thread *td;
2307 register struct olstat_args /* {
2308 char *path;
2309 struct ostat *ub;
2310 } */ *uap;
2311 {
2312 struct stat sb;
2313 struct ostat osb;
2314 int error;
2315
2316 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2317 if (error)
2318 return (error);
2319 cvtstat(&sb, &osb);
2320 error = copyout(&osb, uap->ub, sizeof (osb));
2321 return (error);
2322 }
2323
2324 /*
2325 * Convert from an old to a new stat structure.
2326 */
2327 void
2328 cvtstat(st, ost)
2329 struct stat *st;
2330 struct ostat *ost;
2331 {
2332
2333 ost->st_dev = st->st_dev;
2334 ost->st_ino = st->st_ino;
2335 ost->st_mode = st->st_mode;
2336 ost->st_nlink = st->st_nlink;
2337 ost->st_uid = st->st_uid;
2338 ost->st_gid = st->st_gid;
2339 ost->st_rdev = st->st_rdev;
2340 if (st->st_size < (quad_t)1 << 32)
2341 ost->st_size = st->st_size;
2342 else
2343 ost->st_size = -2;
2344 ost->st_atim = st->st_atim;
2345 ost->st_mtim = st->st_mtim;
2346 ost->st_ctim = st->st_ctim;
2347 ost->st_blksize = st->st_blksize;
2348 ost->st_blocks = st->st_blocks;
2349 ost->st_flags = st->st_flags;
2350 ost->st_gen = st->st_gen;
2351 }
2352 #endif /* COMPAT_43 */
2353
2354 /*
2355 * Get file status; this version follows links.
2356 */
2357 #ifndef _SYS_SYSPROTO_H_
2358 struct stat_args {
2359 char *path;
2360 struct stat *ub;
2361 };
2362 #endif
2363 int
2364 sys_stat(td, uap)
2365 struct thread *td;
2366 register struct stat_args /* {
2367 char *path;
2368 struct stat *ub;
2369 } */ *uap;
2370 {
2371 struct stat sb;
2372 int error;
2373
2374 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2375 if (error == 0)
2376 error = copyout(&sb, uap->ub, sizeof (sb));
2377 return (error);
2378 }
2379
2380 #ifndef _SYS_SYSPROTO_H_
2381 struct fstatat_args {
2382 int fd;
2383 char *path;
2384 struct stat *buf;
2385 int flag;
2386 }
2387 #endif
2388 int
2389 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2390 {
2391 struct stat sb;
2392 int error;
2393
2394 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2395 UIO_USERSPACE, &sb);
2396 if (error == 0)
2397 error = copyout(&sb, uap->buf, sizeof (sb));
2398 return (error);
2399 }
2400
2401 int
2402 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2403 {
2404
2405 return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2406 }
2407
2408 int
2409 kern_statat(struct thread *td, int flag, int fd, char *path,
2410 enum uio_seg pathseg, struct stat *sbp)
2411 {
2412
2413 return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2414 }
2415
2416 int
2417 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2418 enum uio_seg pathseg, struct stat *sbp,
2419 void (*hook)(struct vnode *vp, struct stat *sbp))
2420 {
2421 struct nameidata nd;
2422 struct stat sb;
2423 int error, vfslocked;
2424
2425 if (flag & ~AT_SYMLINK_NOFOLLOW)
2426 return (EINVAL);
2427
2428 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2429 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2430 path, fd, CAP_FSTAT, td);
2431
2432 if ((error = namei(&nd)) != 0)
2433 return (error);
2434 vfslocked = NDHASGIANT(&nd);
2435 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2436 if (!error) {
2437 SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2438 if (S_ISREG(sb.st_mode))
2439 SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2440 if (__predict_false(hook != NULL))
2441 hook(nd.ni_vp, &sb);
2442 }
2443 NDFREE(&nd, NDF_ONLY_PNBUF);
2444 vput(nd.ni_vp);
2445 VFS_UNLOCK_GIANT(vfslocked);
2446 if (error)
2447 return (error);
2448 *sbp = sb;
2449 #ifdef KTRACE
2450 if (KTRPOINT(td, KTR_STRUCT))
2451 ktrstat(&sb);
2452 #endif
2453 return (0);
2454 }
2455
2456 /*
2457 * Get file status; this version does not follow links.
2458 */
2459 #ifndef _SYS_SYSPROTO_H_
2460 struct lstat_args {
2461 char *path;
2462 struct stat *ub;
2463 };
2464 #endif
2465 int
2466 sys_lstat(td, uap)
2467 struct thread *td;
2468 register struct lstat_args /* {
2469 char *path;
2470 struct stat *ub;
2471 } */ *uap;
2472 {
2473 struct stat sb;
2474 int error;
2475
2476 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2477 if (error == 0)
2478 error = copyout(&sb, uap->ub, sizeof (sb));
2479 return (error);
2480 }
2481
2482 int
2483 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2484 {
2485
2486 return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2487 sbp));
2488 }
2489
2490 /*
2491 * Implementation of the NetBSD [l]stat() functions.
2492 */
2493 void
2494 cvtnstat(sb, nsb)
2495 struct stat *sb;
2496 struct nstat *nsb;
2497 {
2498 bzero(nsb, sizeof *nsb);
2499 nsb->st_dev = sb->st_dev;
2500 nsb->st_ino = sb->st_ino;
2501 nsb->st_mode = sb->st_mode;
2502 nsb->st_nlink = sb->st_nlink;
2503 nsb->st_uid = sb->st_uid;
2504 nsb->st_gid = sb->st_gid;
2505 nsb->st_rdev = sb->st_rdev;
2506 nsb->st_atim = sb->st_atim;
2507 nsb->st_mtim = sb->st_mtim;
2508 nsb->st_ctim = sb->st_ctim;
2509 nsb->st_size = sb->st_size;
2510 nsb->st_blocks = sb->st_blocks;
2511 nsb->st_blksize = sb->st_blksize;
2512 nsb->st_flags = sb->st_flags;
2513 nsb->st_gen = sb->st_gen;
2514 nsb->st_birthtim = sb->st_birthtim;
2515 }
2516
2517 #ifndef _SYS_SYSPROTO_H_
2518 struct nstat_args {
2519 char *path;
2520 struct nstat *ub;
2521 };
2522 #endif
2523 int
2524 sys_nstat(td, uap)
2525 struct thread *td;
2526 register struct nstat_args /* {
2527 char *path;
2528 struct nstat *ub;
2529 } */ *uap;
2530 {
2531 struct stat sb;
2532 struct nstat nsb;
2533 int error;
2534
2535 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2536 if (error)
2537 return (error);
2538 cvtnstat(&sb, &nsb);
2539 error = copyout(&nsb, uap->ub, sizeof (nsb));
2540 return (error);
2541 }
2542
2543 /*
2544 * NetBSD lstat. Get file status; this version does not follow links.
2545 */
2546 #ifndef _SYS_SYSPROTO_H_
2547 struct lstat_args {
2548 char *path;
2549 struct stat *ub;
2550 };
2551 #endif
2552 int
2553 sys_nlstat(td, uap)
2554 struct thread *td;
2555 register struct nlstat_args /* {
2556 char *path;
2557 struct nstat *ub;
2558 } */ *uap;
2559 {
2560 struct stat sb;
2561 struct nstat nsb;
2562 int error;
2563
2564 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2565 if (error)
2566 return (error);
2567 cvtnstat(&sb, &nsb);
2568 error = copyout(&nsb, uap->ub, sizeof (nsb));
2569 return (error);
2570 }
2571
2572 /*
2573 * Get configurable pathname variables.
2574 */
2575 #ifndef _SYS_SYSPROTO_H_
2576 struct pathconf_args {
2577 char *path;
2578 int name;
2579 };
2580 #endif
2581 int
2582 sys_pathconf(td, uap)
2583 struct thread *td;
2584 register struct pathconf_args /* {
2585 char *path;
2586 int name;
2587 } */ *uap;
2588 {
2589
2590 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2591 }
2592
2593 #ifndef _SYS_SYSPROTO_H_
2594 struct lpathconf_args {
2595 char *path;
2596 int name;
2597 };
2598 #endif
2599 int
2600 sys_lpathconf(td, uap)
2601 struct thread *td;
2602 register struct lpathconf_args /* {
2603 char *path;
2604 int name;
2605 } */ *uap;
2606 {
2607
2608 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW));
2609 }
2610
2611 int
2612 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2613 u_long flags)
2614 {
2615 struct nameidata nd;
2616 int error, vfslocked;
2617
2618 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1 |
2619 flags, pathseg, path, td);
2620 if ((error = namei(&nd)) != 0)
2621 return (error);
2622 vfslocked = NDHASGIANT(&nd);
2623 NDFREE(&nd, NDF_ONLY_PNBUF);
2624
2625 /* If asynchronous I/O is available, it works for all files. */
2626 if (name == _PC_ASYNC_IO)
2627 td->td_retval[0] = async_io_version;
2628 else
2629 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2630 vput(nd.ni_vp);
2631 VFS_UNLOCK_GIANT(vfslocked);
2632 return (error);
2633 }
2634
2635 /*
2636 * Return target name of a symbolic link.
2637 */
2638 #ifndef _SYS_SYSPROTO_H_
2639 struct readlink_args {
2640 char *path;
2641 char *buf;
2642 size_t count;
2643 };
2644 #endif
2645 int
2646 sys_readlink(td, uap)
2647 struct thread *td;
2648 register struct readlink_args /* {
2649 char *path;
2650 char *buf;
2651 size_t count;
2652 } */ *uap;
2653 {
2654
2655 return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2656 UIO_USERSPACE, uap->count));
2657 }
2658 #ifndef _SYS_SYSPROTO_H_
2659 struct readlinkat_args {
2660 int fd;
2661 char *path;
2662 char *buf;
2663 size_t bufsize;
2664 };
2665 #endif
2666 int
2667 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2668 {
2669
2670 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2671 uap->buf, UIO_USERSPACE, uap->bufsize));
2672 }
2673
2674 int
2675 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2676 enum uio_seg bufseg, size_t count)
2677 {
2678
2679 return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2680 count));
2681 }
2682
2683 int
2684 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2685 char *buf, enum uio_seg bufseg, size_t count)
2686 {
2687 struct vnode *vp;
2688 struct iovec aiov;
2689 struct uio auio;
2690 int error;
2691 struct nameidata nd;
2692 int vfslocked;
2693
2694 if (count > IOSIZE_MAX)
2695 return (EINVAL);
2696
2697 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2698 AUDITVNODE1, pathseg, path, fd, td);
2699
2700 if ((error = namei(&nd)) != 0)
2701 return (error);
2702 NDFREE(&nd, NDF_ONLY_PNBUF);
2703 vfslocked = NDHASGIANT(&nd);
2704 vp = nd.ni_vp;
2705 #ifdef MAC
2706 error = mac_vnode_check_readlink(td->td_ucred, vp);
2707 if (error) {
2708 vput(vp);
2709 VFS_UNLOCK_GIANT(vfslocked);
2710 return (error);
2711 }
2712 #endif
2713 if (vp->v_type != VLNK)
2714 error = EINVAL;
2715 else {
2716 aiov.iov_base = buf;
2717 aiov.iov_len = count;
2718 auio.uio_iov = &aiov;
2719 auio.uio_iovcnt = 1;
2720 auio.uio_offset = 0;
2721 auio.uio_rw = UIO_READ;
2722 auio.uio_segflg = bufseg;
2723 auio.uio_td = td;
2724 auio.uio_resid = count;
2725 error = VOP_READLINK(vp, &auio, td->td_ucred);
2726 }
2727 vput(vp);
2728 VFS_UNLOCK_GIANT(vfslocked);
2729 td->td_retval[0] = count - auio.uio_resid;
2730 return (error);
2731 }
2732
2733 /*
2734 * Common implementation code for chflags() and fchflags().
2735 */
2736 static int
2737 setfflags(td, vp, flags)
2738 struct thread *td;
2739 struct vnode *vp;
2740 int flags;
2741 {
2742 int error;
2743 struct mount *mp;
2744 struct vattr vattr;
2745
2746 /* We can't support the value matching VNOVAL. */
2747 if (flags == VNOVAL)
2748 return (EOPNOTSUPP);
2749
2750 /*
2751 * Prevent non-root users from setting flags on devices. When
2752 * a device is reused, users can retain ownership of the device
2753 * if they are allowed to set flags and programs assume that
2754 * chown can't fail when done as root.
2755 */
2756 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2757 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2758 if (error)
2759 return (error);
2760 }
2761
2762 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2763 return (error);
2764 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2765 VATTR_NULL(&vattr);
2766 vattr.va_flags = flags;
2767 #ifdef MAC
2768 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2769 if (error == 0)
2770 #endif
2771 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2772 VOP_UNLOCK(vp, 0);
2773 vn_finished_write(mp);
2774 return (error);
2775 }
2776
2777 /*
2778 * Change flags of a file given a path name.
2779 */
2780 #ifndef _SYS_SYSPROTO_H_
2781 struct chflags_args {
2782 char *path;
2783 int flags;
2784 };
2785 #endif
2786 int
2787 sys_chflags(td, uap)
2788 struct thread *td;
2789 register struct chflags_args /* {
2790 char *path;
2791 int flags;
2792 } */ *uap;
2793 {
2794 int error;
2795 struct nameidata nd;
2796 int vfslocked;
2797
2798 AUDIT_ARG_FFLAGS(uap->flags);
2799 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2800 uap->path, td);
2801 if ((error = namei(&nd)) != 0)
2802 return (error);
2803 NDFREE(&nd, NDF_ONLY_PNBUF);
2804 vfslocked = NDHASGIANT(&nd);
2805 error = setfflags(td, nd.ni_vp, uap->flags);
2806 vrele(nd.ni_vp);
2807 VFS_UNLOCK_GIANT(vfslocked);
2808 return (error);
2809 }
2810
2811 /*
2812 * Same as chflags() but doesn't follow symlinks.
2813 */
2814 int
2815 sys_lchflags(td, uap)
2816 struct thread *td;
2817 register struct lchflags_args /* {
2818 char *path;
2819 int flags;
2820 } */ *uap;
2821 {
2822 int error;
2823 struct nameidata nd;
2824 int vfslocked;
2825
2826 AUDIT_ARG_FFLAGS(uap->flags);
2827 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2828 uap->path, td);
2829 if ((error = namei(&nd)) != 0)
2830 return (error);
2831 vfslocked = NDHASGIANT(&nd);
2832 NDFREE(&nd, NDF_ONLY_PNBUF);
2833 error = setfflags(td, nd.ni_vp, uap->flags);
2834 vrele(nd.ni_vp);
2835 VFS_UNLOCK_GIANT(vfslocked);
2836 return (error);
2837 }
2838
2839 /*
2840 * Change flags of a file given a file descriptor.
2841 */
2842 #ifndef _SYS_SYSPROTO_H_
2843 struct fchflags_args {
2844 int fd;
2845 int flags;
2846 };
2847 #endif
2848 int
2849 sys_fchflags(td, uap)
2850 struct thread *td;
2851 register struct fchflags_args /* {
2852 int fd;
2853 int flags;
2854 } */ *uap;
2855 {
2856 struct file *fp;
2857 int vfslocked;
2858 int error;
2859
2860 AUDIT_ARG_FD(uap->fd);
2861 AUDIT_ARG_FFLAGS(uap->flags);
2862 if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FCHFLAGS,
2863 &fp)) != 0)
2864 return (error);
2865 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2866 #ifdef AUDIT
2867 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2868 AUDIT_ARG_VNODE1(fp->f_vnode);
2869 VOP_UNLOCK(fp->f_vnode, 0);
2870 #endif
2871 error = setfflags(td, fp->f_vnode, uap->flags);
2872 VFS_UNLOCK_GIANT(vfslocked);
2873 fdrop(fp, td);
2874 return (error);
2875 }
2876
2877 /*
2878 * Common implementation code for chmod(), lchmod() and fchmod().
2879 */
2880 int
2881 setfmode(td, cred, vp, mode)
2882 struct thread *td;
2883 struct ucred *cred;
2884 struct vnode *vp;
2885 int mode;
2886 {
2887 int error;
2888 struct mount *mp;
2889 struct vattr vattr;
2890
2891 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2892 return (error);
2893 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2894 VATTR_NULL(&vattr);
2895 vattr.va_mode = mode & ALLPERMS;
2896 #ifdef MAC
2897 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2898 if (error == 0)
2899 #endif
2900 error = VOP_SETATTR(vp, &vattr, cred);
2901 VOP_UNLOCK(vp, 0);
2902 vn_finished_write(mp);
2903 return (error);
2904 }
2905
2906 /*
2907 * Change mode of a file given path name.
2908 */
2909 #ifndef _SYS_SYSPROTO_H_
2910 struct chmod_args {
2911 char *path;
2912 int mode;
2913 };
2914 #endif
2915 int
2916 sys_chmod(td, uap)
2917 struct thread *td;
2918 register struct chmod_args /* {
2919 char *path;
2920 int mode;
2921 } */ *uap;
2922 {
2923
2924 return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2925 }
2926
2927 #ifndef _SYS_SYSPROTO_H_
2928 struct fchmodat_args {
2929 int dirfd;
2930 char *path;
2931 mode_t mode;
2932 int flag;
2933 }
2934 #endif
2935 int
2936 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2937 {
2938 int flag = uap->flag;
2939 int fd = uap->fd;
2940 char *path = uap->path;
2941 mode_t mode = uap->mode;
2942
2943 if (flag & ~AT_SYMLINK_NOFOLLOW)
2944 return (EINVAL);
2945
2946 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2947 }
2948
2949 int
2950 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2951 {
2952
2953 return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2954 }
2955
2956 /*
2957 * Change mode of a file given path name (don't follow links.)
2958 */
2959 #ifndef _SYS_SYSPROTO_H_
2960 struct lchmod_args {
2961 char *path;
2962 int mode;
2963 };
2964 #endif
2965 int
2966 sys_lchmod(td, uap)
2967 struct thread *td;
2968 register struct lchmod_args /* {
2969 char *path;
2970 int mode;
2971 } */ *uap;
2972 {
2973
2974 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2975 uap->mode, AT_SYMLINK_NOFOLLOW));
2976 }
2977
2978
2979 int
2980 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2981 mode_t mode, int flag)
2982 {
2983 int error;
2984 struct nameidata nd;
2985 int vfslocked;
2986 int follow;
2987
2988 AUDIT_ARG_MODE(mode);
2989 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2990 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg,
2991 path, fd, CAP_FCHMOD, td);
2992 if ((error = namei(&nd)) != 0)
2993 return (error);
2994 vfslocked = NDHASGIANT(&nd);
2995 NDFREE(&nd, NDF_ONLY_PNBUF);
2996 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2997 vrele(nd.ni_vp);
2998 VFS_UNLOCK_GIANT(vfslocked);
2999 return (error);
3000 }
3001
3002 /*
3003 * Change mode of a file given a file descriptor.
3004 */
3005 #ifndef _SYS_SYSPROTO_H_
3006 struct fchmod_args {
3007 int fd;
3008 int mode;
3009 };
3010 #endif
3011 int
3012 sys_fchmod(struct thread *td, struct fchmod_args *uap)
3013 {
3014 struct file *fp;
3015 int error;
3016
3017 AUDIT_ARG_FD(uap->fd);
3018 AUDIT_ARG_MODE(uap->mode);
3019
3020 error = fget(td, uap->fd, CAP_FCHMOD, &fp);
3021 if (error != 0)
3022 return (error);
3023 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
3024 fdrop(fp, td);
3025 return (error);
3026 }
3027
3028 /*
3029 * Common implementation for chown(), lchown(), and fchown()
3030 */
3031 int
3032 setfown(td, cred, vp, uid, gid)
3033 struct thread *td;
3034 struct ucred *cred;
3035 struct vnode *vp;
3036 uid_t uid;
3037 gid_t gid;
3038 {
3039 int error;
3040 struct mount *mp;
3041 struct vattr vattr;
3042
3043 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3044 return (error);
3045 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3046 VATTR_NULL(&vattr);
3047 vattr.va_uid = uid;
3048 vattr.va_gid = gid;
3049 #ifdef MAC
3050 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3051 vattr.va_gid);
3052 if (error == 0)
3053 #endif
3054 error = VOP_SETATTR(vp, &vattr, cred);
3055 VOP_UNLOCK(vp, 0);
3056 vn_finished_write(mp);
3057 return (error);
3058 }
3059
3060 /*
3061 * Set ownership given a path name.
3062 */
3063 #ifndef _SYS_SYSPROTO_H_
3064 struct chown_args {
3065 char *path;
3066 int uid;
3067 int gid;
3068 };
3069 #endif
3070 int
3071 sys_chown(td, uap)
3072 struct thread *td;
3073 register struct chown_args /* {
3074 char *path;
3075 int uid;
3076 int gid;
3077 } */ *uap;
3078 {
3079
3080 return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3081 }
3082
3083 #ifndef _SYS_SYSPROTO_H_
3084 struct fchownat_args {
3085 int fd;
3086 const char * path;
3087 uid_t uid;
3088 gid_t gid;
3089 int flag;
3090 };
3091 #endif
3092 int
3093 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3094 {
3095 int flag;
3096
3097 flag = uap->flag;
3098 if (flag & ~AT_SYMLINK_NOFOLLOW)
3099 return (EINVAL);
3100
3101 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3102 uap->gid, uap->flag));
3103 }
3104
3105 int
3106 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3107 int gid)
3108 {
3109
3110 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3111 }
3112
3113 int
3114 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3115 int uid, int gid, int flag)
3116 {
3117 struct nameidata nd;
3118 int error, vfslocked, follow;
3119
3120 AUDIT_ARG_OWNER(uid, gid);
3121 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3122 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg,
3123 path, fd, CAP_FCHOWN, td);
3124
3125 if ((error = namei(&nd)) != 0)
3126 return (error);
3127 vfslocked = NDHASGIANT(&nd);
3128 NDFREE(&nd, NDF_ONLY_PNBUF);
3129 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3130 vrele(nd.ni_vp);
3131 VFS_UNLOCK_GIANT(vfslocked);
3132 return (error);
3133 }
3134
3135 /*
3136 * Set ownership given a path name, do not cross symlinks.
3137 */
3138 #ifndef _SYS_SYSPROTO_H_
3139 struct lchown_args {
3140 char *path;
3141 int uid;
3142 int gid;
3143 };
3144 #endif
3145 int
3146 sys_lchown(td, uap)
3147 struct thread *td;
3148 register struct lchown_args /* {
3149 char *path;
3150 int uid;
3151 int gid;
3152 } */ *uap;
3153 {
3154
3155 return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3156 }
3157
3158 int
3159 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3160 int gid)
3161 {
3162
3163 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3164 AT_SYMLINK_NOFOLLOW));
3165 }
3166
3167 /*
3168 * Set ownership given a file descriptor.
3169 */
3170 #ifndef _SYS_SYSPROTO_H_
3171 struct fchown_args {
3172 int fd;
3173 int uid;
3174 int gid;
3175 };
3176 #endif
3177 int
3178 sys_fchown(td, uap)
3179 struct thread *td;
3180 register struct fchown_args /* {
3181 int fd;
3182 int uid;
3183 int gid;
3184 } */ *uap;
3185 {
3186 struct file *fp;
3187 int error;
3188
3189 AUDIT_ARG_FD(uap->fd);
3190 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3191 error = fget(td, uap->fd, CAP_FCHOWN, &fp);
3192 if (error != 0)
3193 return (error);
3194 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3195 fdrop(fp, td);
3196 return (error);
3197 }
3198
3199 /*
3200 * Common implementation code for utimes(), lutimes(), and futimes().
3201 */
3202 static int
3203 getutimes(usrtvp, tvpseg, tsp)
3204 const struct timeval *usrtvp;
3205 enum uio_seg tvpseg;
3206 struct timespec *tsp;
3207 {
3208 struct timeval tv[2];
3209 const struct timeval *tvp;
3210 int error;
3211
3212 if (usrtvp == NULL) {
3213 vfs_timestamp(&tsp[0]);
3214 tsp[1] = tsp[0];
3215 } else {
3216 if (tvpseg == UIO_SYSSPACE) {
3217 tvp = usrtvp;
3218 } else {
3219 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3220 return (error);
3221 tvp = tv;
3222 }
3223
3224 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3225 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3226 return (EINVAL);
3227 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3228 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3229 }
3230 return (0);
3231 }
3232
3233 /*
3234 * Common implementation code for utimes(), lutimes(), and futimes().
3235 */
3236 static int
3237 setutimes(td, vp, ts, numtimes, nullflag)
3238 struct thread *td;
3239 struct vnode *vp;
3240 const struct timespec *ts;
3241 int numtimes;
3242 int nullflag;
3243 {
3244 int error, setbirthtime;
3245 struct mount *mp;
3246 struct vattr vattr;
3247
3248 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3249 return (error);
3250 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3251 setbirthtime = 0;
3252 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3253 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3254 setbirthtime = 1;
3255 VATTR_NULL(&vattr);
3256 vattr.va_atime = ts[0];
3257 vattr.va_mtime = ts[1];
3258 if (setbirthtime)
3259 vattr.va_birthtime = ts[1];
3260 if (numtimes > 2)
3261 vattr.va_birthtime = ts[2];
3262 if (nullflag)
3263 vattr.va_vaflags |= VA_UTIMES_NULL;
3264 #ifdef MAC
3265 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3266 vattr.va_mtime);
3267 #endif
3268 if (error == 0)
3269 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3270 VOP_UNLOCK(vp, 0);
3271 vn_finished_write(mp);
3272 return (error);
3273 }
3274
3275 /*
3276 * Set the access and modification times of a file.
3277 */
3278 #ifndef _SYS_SYSPROTO_H_
3279 struct utimes_args {
3280 char *path;
3281 struct timeval *tptr;
3282 };
3283 #endif
3284 int
3285 sys_utimes(td, uap)
3286 struct thread *td;
3287 register struct utimes_args /* {
3288 char *path;
3289 struct timeval *tptr;
3290 } */ *uap;
3291 {
3292
3293 return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3294 UIO_USERSPACE));
3295 }
3296
3297 #ifndef _SYS_SYSPROTO_H_
3298 struct futimesat_args {
3299 int fd;
3300 const char * path;
3301 const struct timeval * times;
3302 };
3303 #endif
3304 int
3305 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3306 {
3307
3308 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3309 uap->times, UIO_USERSPACE));
3310 }
3311
3312 int
3313 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3314 struct timeval *tptr, enum uio_seg tptrseg)
3315 {
3316
3317 return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3318 }
3319
3320 int
3321 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3322 struct timeval *tptr, enum uio_seg tptrseg)
3323 {
3324 struct nameidata nd;
3325 struct timespec ts[2];
3326 int error, vfslocked;
3327
3328 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3329 return (error);
3330 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg,
3331 path, fd, CAP_FUTIMES, td);
3332
3333 if ((error = namei(&nd)) != 0)
3334 return (error);
3335 vfslocked = NDHASGIANT(&nd);
3336 NDFREE(&nd, NDF_ONLY_PNBUF);
3337 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3338 vrele(nd.ni_vp);
3339 VFS_UNLOCK_GIANT(vfslocked);
3340 return (error);
3341 }
3342
3343 /*
3344 * Set the access and modification times of a file.
3345 */
3346 #ifndef _SYS_SYSPROTO_H_
3347 struct lutimes_args {
3348 char *path;
3349 struct timeval *tptr;
3350 };
3351 #endif
3352 int
3353 sys_lutimes(td, uap)
3354 struct thread *td;
3355 register struct lutimes_args /* {
3356 char *path;
3357 struct timeval *tptr;
3358 } */ *uap;
3359 {
3360
3361 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3362 UIO_USERSPACE));
3363 }
3364
3365 int
3366 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3367 struct timeval *tptr, enum uio_seg tptrseg)
3368 {
3369 struct timespec ts[2];
3370 int error;
3371 struct nameidata nd;
3372 int vfslocked;
3373
3374 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3375 return (error);
3376 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3377 if ((error = namei(&nd)) != 0)
3378 return (error);
3379 vfslocked = NDHASGIANT(&nd);
3380 NDFREE(&nd, NDF_ONLY_PNBUF);
3381 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3382 vrele(nd.ni_vp);
3383 VFS_UNLOCK_GIANT(vfslocked);
3384 return (error);
3385 }
3386
3387 /*
3388 * Set the access and modification times of a file.
3389 */
3390 #ifndef _SYS_SYSPROTO_H_
3391 struct futimes_args {
3392 int fd;
3393 struct timeval *tptr;
3394 };
3395 #endif
3396 int
3397 sys_futimes(td, uap)
3398 struct thread *td;
3399 register struct futimes_args /* {
3400 int fd;
3401 struct timeval *tptr;
3402 } */ *uap;
3403 {
3404
3405 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3406 }
3407
3408 int
3409 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3410 enum uio_seg tptrseg)
3411 {
3412 struct timespec ts[2];
3413 struct file *fp;
3414 int vfslocked;
3415 int error;
3416
3417 AUDIT_ARG_FD(fd);
3418 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3419 return (error);
3420 if ((error = getvnode(td->td_proc->p_fd, fd, CAP_FUTIMES, &fp))
3421 != 0)
3422 return (error);
3423 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3424 #ifdef AUDIT
3425 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3426 AUDIT_ARG_VNODE1(fp->f_vnode);
3427 VOP_UNLOCK(fp->f_vnode, 0);
3428 #endif
3429 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3430 VFS_UNLOCK_GIANT(vfslocked);
3431 fdrop(fp, td);
3432 return (error);
3433 }
3434
3435 /*
3436 * Truncate a file given its path name.
3437 */
3438 #ifndef _SYS_SYSPROTO_H_
3439 struct truncate_args {
3440 char *path;
3441 int pad;
3442 off_t length;
3443 };
3444 #endif
3445 int
3446 sys_truncate(td, uap)
3447 struct thread *td;
3448 register struct truncate_args /* {
3449 char *path;
3450 int pad;
3451 off_t length;
3452 } */ *uap;
3453 {
3454
3455 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3456 }
3457
3458 int
3459 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3460 {
3461 struct mount *mp;
3462 struct vnode *vp;
3463 void *rl_cookie;
3464 struct vattr vattr;
3465 struct nameidata nd;
3466 int error, vfslocked;
3467
3468 if (length < 0)
3469 return(EINVAL);
3470 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3471 if ((error = namei(&nd)) != 0)
3472 return (error);
3473 vfslocked = NDHASGIANT(&nd);
3474 vp = nd.ni_vp;
3475 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3476 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3477 vn_rangelock_unlock(vp, rl_cookie);
3478 vrele(vp);
3479 VFS_UNLOCK_GIANT(vfslocked);
3480 return (error);
3481 }
3482 NDFREE(&nd, NDF_ONLY_PNBUF);
3483 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3484 if (vp->v_type == VDIR)
3485 error = EISDIR;
3486 #ifdef MAC
3487 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3488 }
3489 #endif
3490 else if ((error = vn_writechk(vp)) == 0 &&
3491 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3492 VATTR_NULL(&vattr);
3493 vattr.va_size = length;
3494 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3495 }
3496 VOP_UNLOCK(vp, 0);
3497 vn_finished_write(mp);
3498 vn_rangelock_unlock(vp, rl_cookie);
3499 vrele(vp);
3500 VFS_UNLOCK_GIANT(vfslocked);
3501 return (error);
3502 }
3503
3504 #if defined(COMPAT_43)
3505 /*
3506 * Truncate a file given its path name.
3507 */
3508 #ifndef _SYS_SYSPROTO_H_
3509 struct otruncate_args {
3510 char *path;
3511 long length;
3512 };
3513 #endif
3514 int
3515 otruncate(td, uap)
3516 struct thread *td;
3517 register struct otruncate_args /* {
3518 char *path;
3519 long length;
3520 } */ *uap;
3521 {
3522 struct truncate_args /* {
3523 char *path;
3524 int pad;
3525 off_t length;
3526 } */ nuap;
3527
3528 nuap.path = uap->path;
3529 nuap.length = uap->length;
3530 return (sys_truncate(td, &nuap));
3531 }
3532 #endif /* COMPAT_43 */
3533
3534 /* Versions with the pad argument */
3535 int
3536 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3537 {
3538 struct truncate_args ouap;
3539
3540 ouap.path = uap->path;
3541 ouap.length = uap->length;
3542 return (sys_truncate(td, &ouap));
3543 }
3544
3545 int
3546 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3547 {
3548 struct ftruncate_args ouap;
3549
3550 ouap.fd = uap->fd;
3551 ouap.length = uap->length;
3552 return (sys_ftruncate(td, &ouap));
3553 }
3554
3555 /*
3556 * Sync an open file.
3557 */
3558 #ifndef _SYS_SYSPROTO_H_
3559 struct fsync_args {
3560 int fd;
3561 };
3562 #endif
3563 int
3564 sys_fsync(td, uap)
3565 struct thread *td;
3566 struct fsync_args /* {
3567 int fd;
3568 } */ *uap;
3569 {
3570 struct vnode *vp;
3571 struct mount *mp;
3572 struct file *fp;
3573 int vfslocked;
3574 int error, lock_flags;
3575
3576 AUDIT_ARG_FD(uap->fd);
3577 if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FSYNC,
3578 &fp)) != 0)
3579 return (error);
3580 vp = fp->f_vnode;
3581 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3582 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3583 goto drop;
3584 if (MNT_SHARED_WRITES(mp) ||
3585 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3586 lock_flags = LK_SHARED;
3587 } else {
3588 lock_flags = LK_EXCLUSIVE;
3589 }
3590 vn_lock(vp, lock_flags | LK_RETRY);
3591 AUDIT_ARG_VNODE1(vp);
3592 if (vp->v_object != NULL) {
3593 VM_OBJECT_LOCK(vp->v_object);
3594 vm_object_page_clean(vp->v_object, 0, 0, 0);
3595 VM_OBJECT_UNLOCK(vp->v_object);
3596 }
3597 error = VOP_FSYNC(vp, MNT_WAIT, td);
3598
3599 VOP_UNLOCK(vp, 0);
3600 vn_finished_write(mp);
3601 drop:
3602 VFS_UNLOCK_GIANT(vfslocked);
3603 fdrop(fp, td);
3604 return (error);
3605 }
3606
3607 /*
3608 * Rename files. Source and destination must either both be directories, or
3609 * both not be directories. If target is a directory, it must be empty.
3610 */
3611 #ifndef _SYS_SYSPROTO_H_
3612 struct rename_args {
3613 char *from;
3614 char *to;
3615 };
3616 #endif
3617 int
3618 sys_rename(td, uap)
3619 struct thread *td;
3620 register struct rename_args /* {
3621 char *from;
3622 char *to;
3623 } */ *uap;
3624 {
3625
3626 return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3627 }
3628
3629 #ifndef _SYS_SYSPROTO_H_
3630 struct renameat_args {
3631 int oldfd;
3632 char *old;
3633 int newfd;
3634 char *new;
3635 };
3636 #endif
3637 int
3638 sys_renameat(struct thread *td, struct renameat_args *uap)
3639 {
3640
3641 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3642 UIO_USERSPACE));
3643 }
3644
3645 int
3646 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3647 {
3648
3649 return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3650 }
3651
3652 int
3653 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3654 enum uio_seg pathseg)
3655 {
3656 struct mount *mp = NULL;
3657 struct vnode *tvp, *fvp, *tdvp;
3658 struct nameidata fromnd, tond;
3659 int tvfslocked;
3660 int fvfslocked;
3661 int error;
3662
3663 bwillwrite();
3664 #ifdef MAC
3665 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3666 MPSAFE | AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3667 #else
3668 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3669 AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3670 #endif
3671
3672 if ((error = namei(&fromnd)) != 0)
3673 return (error);
3674 fvfslocked = NDHASGIANT(&fromnd);
3675 tvfslocked = 0;
3676 #ifdef MAC
3677 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3678 fromnd.ni_vp, &fromnd.ni_cnd);
3679 VOP_UNLOCK(fromnd.ni_dvp, 0);
3680 if (fromnd.ni_dvp != fromnd.ni_vp)
3681 VOP_UNLOCK(fromnd.ni_vp, 0);
3682 #endif
3683 fvp = fromnd.ni_vp;
3684 if (error == 0)
3685 error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3686 if (error != 0) {
3687 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3688 vrele(fromnd.ni_dvp);
3689 vrele(fvp);
3690 goto out1;
3691 }
3692 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3693 SAVESTART | MPSAFE | AUDITVNODE2, pathseg, new, newfd, CAP_CREATE,
3694 td);
3695 if (fromnd.ni_vp->v_type == VDIR)
3696 tond.ni_cnd.cn_flags |= WILLBEDIR;
3697 if ((error = namei(&tond)) != 0) {
3698 /* Translate error code for rename("dir1", "dir2/."). */
3699 if (error == EISDIR && fvp->v_type == VDIR)
3700 error = EINVAL;
3701 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3702 vrele(fromnd.ni_dvp);
3703 vrele(fvp);
3704 vn_finished_write(mp);
3705 goto out1;
3706 }
3707 tvfslocked = NDHASGIANT(&tond);
3708 tdvp = tond.ni_dvp;
3709 tvp = tond.ni_vp;
3710 if (tvp != NULL) {
3711 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3712 error = ENOTDIR;
3713 goto out;
3714 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3715 error = EISDIR;
3716 goto out;
3717 }
3718 }
3719 if (fvp == tdvp) {
3720 error = EINVAL;
3721 goto out;
3722 }
3723 /*
3724 * If the source is the same as the destination (that is, if they
3725 * are links to the same vnode), then there is nothing to do.
3726 */
3727 if (fvp == tvp)
3728 error = -1;
3729 #ifdef MAC
3730 else
3731 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3732 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3733 #endif
3734 out:
3735 if (!error) {
3736 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3737 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3738 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3739 NDFREE(&tond, NDF_ONLY_PNBUF);
3740 } else {
3741 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3742 NDFREE(&tond, NDF_ONLY_PNBUF);
3743 if (tvp)
3744 vput(tvp);
3745 if (tdvp == tvp)
3746 vrele(tdvp);
3747 else
3748 vput(tdvp);
3749 vrele(fromnd.ni_dvp);
3750 vrele(fvp);
3751 }
3752 vrele(tond.ni_startdir);
3753 vn_finished_write(mp);
3754 out1:
3755 if (fromnd.ni_startdir)
3756 vrele(fromnd.ni_startdir);
3757 VFS_UNLOCK_GIANT(fvfslocked);
3758 VFS_UNLOCK_GIANT(tvfslocked);
3759 if (error == -1)
3760 return (0);
3761 return (error);
3762 }
3763
3764 /*
3765 * Make a directory file.
3766 */
3767 #ifndef _SYS_SYSPROTO_H_
3768 struct mkdir_args {
3769 char *path;
3770 int mode;
3771 };
3772 #endif
3773 int
3774 sys_mkdir(td, uap)
3775 struct thread *td;
3776 register struct mkdir_args /* {
3777 char *path;
3778 int mode;
3779 } */ *uap;
3780 {
3781
3782 return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3783 }
3784
3785 #ifndef _SYS_SYSPROTO_H_
3786 struct mkdirat_args {
3787 int fd;
3788 char *path;
3789 mode_t mode;
3790 };
3791 #endif
3792 int
3793 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3794 {
3795
3796 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3797 }
3798
3799 int
3800 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3801 {
3802
3803 return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3804 }
3805
3806 int
3807 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3808 int mode)
3809 {
3810 struct mount *mp;
3811 struct vnode *vp;
3812 struct vattr vattr;
3813 int error;
3814 struct nameidata nd;
3815 int vfslocked;
3816
3817 AUDIT_ARG_MODE(mode);
3818 restart:
3819 bwillwrite();
3820 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE |
3821 AUDITVNODE1, segflg, path, fd, CAP_MKDIR, td);
3822 nd.ni_cnd.cn_flags |= WILLBEDIR;
3823 if ((error = namei(&nd)) != 0)
3824 return (error);
3825 vfslocked = NDHASGIANT(&nd);
3826 vp = nd.ni_vp;
3827 if (vp != NULL) {
3828 NDFREE(&nd, NDF_ONLY_PNBUF);
3829 /*
3830 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3831 * the strange behaviour of leaving the vnode unlocked
3832 * if the target is the same vnode as the parent.
3833 */
3834 if (vp == nd.ni_dvp)
3835 vrele(nd.ni_dvp);
3836 else
3837 vput(nd.ni_dvp);
3838 vrele(vp);
3839 VFS_UNLOCK_GIANT(vfslocked);
3840 return (EEXIST);
3841 }
3842 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3843 NDFREE(&nd, NDF_ONLY_PNBUF);
3844 vput(nd.ni_dvp);
3845 VFS_UNLOCK_GIANT(vfslocked);
3846 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3847 return (error);
3848 goto restart;
3849 }
3850 VATTR_NULL(&vattr);
3851 vattr.va_type = VDIR;
3852 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3853 #ifdef MAC
3854 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3855 &vattr);
3856 if (error)
3857 goto out;
3858 #endif
3859 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3860 #ifdef MAC
3861 out:
3862 #endif
3863 NDFREE(&nd, NDF_ONLY_PNBUF);
3864 vput(nd.ni_dvp);
3865 if (!error)
3866 vput(nd.ni_vp);
3867 vn_finished_write(mp);
3868 VFS_UNLOCK_GIANT(vfslocked);
3869 return (error);
3870 }
3871
3872 /*
3873 * Remove a directory file.
3874 */
3875 #ifndef _SYS_SYSPROTO_H_
3876 struct rmdir_args {
3877 char *path;
3878 };
3879 #endif
3880 int
3881 sys_rmdir(td, uap)
3882 struct thread *td;
3883 struct rmdir_args /* {
3884 char *path;
3885 } */ *uap;
3886 {
3887
3888 return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3889 }
3890
3891 int
3892 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3893 {
3894
3895 return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3896 }
3897
3898 int
3899 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3900 {
3901 struct mount *mp;
3902 struct vnode *vp;
3903 int error;
3904 struct nameidata nd;
3905 int vfslocked;
3906
3907 restart:
3908 bwillwrite();
3909 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE |
3910 AUDITVNODE1, pathseg, path, fd, CAP_RMDIR, td);
3911 if ((error = namei(&nd)) != 0)
3912 return (error);
3913 vfslocked = NDHASGIANT(&nd);
3914 vp = nd.ni_vp;
3915 if (vp->v_type != VDIR) {
3916 error = ENOTDIR;
3917 goto out;
3918 }
3919 /*
3920 * No rmdir "." please.
3921 */
3922 if (nd.ni_dvp == vp) {
3923 error = EINVAL;
3924 goto out;
3925 }
3926 /*
3927 * The root of a mounted filesystem cannot be deleted.
3928 */
3929 if (vp->v_vflag & VV_ROOT) {
3930 error = EBUSY;
3931 goto out;
3932 }
3933 #ifdef MAC
3934 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3935 &nd.ni_cnd);
3936 if (error)
3937 goto out;
3938 #endif
3939 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3940 NDFREE(&nd, NDF_ONLY_PNBUF);
3941 vput(vp);
3942 if (nd.ni_dvp == vp)
3943 vrele(nd.ni_dvp);
3944 else
3945 vput(nd.ni_dvp);
3946 VFS_UNLOCK_GIANT(vfslocked);
3947 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3948 return (error);
3949 goto restart;
3950 }
3951 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3952 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3953 vn_finished_write(mp);
3954 out:
3955 NDFREE(&nd, NDF_ONLY_PNBUF);
3956 vput(vp);
3957 if (nd.ni_dvp == vp)
3958 vrele(nd.ni_dvp);
3959 else
3960 vput(nd.ni_dvp);
3961 VFS_UNLOCK_GIANT(vfslocked);
3962 return (error);
3963 }
3964
3965 #ifdef COMPAT_43
3966 /*
3967 * Read a block of directory entries in a filesystem independent format.
3968 */
3969 #ifndef _SYS_SYSPROTO_H_
3970 struct ogetdirentries_args {
3971 int fd;
3972 char *buf;
3973 u_int count;
3974 long *basep;
3975 };
3976 #endif
3977 int
3978 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3979 {
3980 long loff;
3981 int error;
3982
3983 error = kern_ogetdirentries(td, uap, &loff);
3984 if (error == 0)
3985 error = copyout(&loff, uap->basep, sizeof(long));
3986 return (error);
3987 }
3988
3989 int
3990 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3991 long *ploff)
3992 {
3993 struct vnode *vp;
3994 struct file *fp;
3995 struct uio auio, kuio;
3996 struct iovec aiov, kiov;
3997 struct dirent *dp, *edp;
3998 caddr_t dirbuf;
3999 int error, eofflag, readcnt, vfslocked;
4000 long loff;
4001 off_t foffset;
4002
4003 /* XXX arbitrary sanity limit on `count'. */
4004 if (uap->count > 64 * 1024)
4005 return (EINVAL);
4006 if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_READ,
4007 &fp)) != 0)
4008 return (error);
4009 if ((fp->f_flag & FREAD) == 0) {
4010 fdrop(fp, td);
4011 return (EBADF);
4012 }
4013 vp = fp->f_vnode;
4014 foffset = foffset_lock(fp, 0);
4015 unionread:
4016 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4017 if (vp->v_type != VDIR) {
4018 VFS_UNLOCK_GIANT(vfslocked);
4019 foffset_unlock(fp, foffset, 0);
4020 fdrop(fp, td);
4021 return (EINVAL);
4022 }
4023 aiov.iov_base = uap->buf;
4024 aiov.iov_len = uap->count;
4025 auio.uio_iov = &aiov;
4026 auio.uio_iovcnt = 1;
4027 auio.uio_rw = UIO_READ;
4028 auio.uio_segflg = UIO_USERSPACE;
4029 auio.uio_td = td;
4030 auio.uio_resid = uap->count;
4031 vn_lock(vp, LK_SHARED | LK_RETRY);
4032 loff = auio.uio_offset = foffset;
4033 #ifdef MAC
4034 error = mac_vnode_check_readdir(td->td_ucred, vp);
4035 if (error) {
4036 VOP_UNLOCK(vp, 0);
4037 VFS_UNLOCK_GIANT(vfslocked);
4038 foffset_unlock(fp, foffset, FOF_NOUPDATE);
4039 fdrop(fp, td);
4040 return (error);
4041 }
4042 #endif
4043 # if (BYTE_ORDER != LITTLE_ENDIAN)
4044 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
4045 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
4046 NULL, NULL);
4047 foffset = auio.uio_offset;
4048 } else
4049 # endif
4050 {
4051 kuio = auio;
4052 kuio.uio_iov = &kiov;
4053 kuio.uio_segflg = UIO_SYSSPACE;
4054 kiov.iov_len = uap->count;
4055 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
4056 kiov.iov_base = dirbuf;
4057 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
4058 NULL, NULL);
4059 foffset = kuio.uio_offset;
4060 if (error == 0) {
4061 readcnt = uap->count - kuio.uio_resid;
4062 edp = (struct dirent *)&dirbuf[readcnt];
4063 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
4064 # if (BYTE_ORDER == LITTLE_ENDIAN)
4065 /*
4066 * The expected low byte of
4067 * dp->d_namlen is our dp->d_type.
4068 * The high MBZ byte of dp->d_namlen
4069 * is our dp->d_namlen.
4070 */
4071 dp->d_type = dp->d_namlen;
4072 dp->d_namlen = 0;
4073 # else
4074 /*
4075 * The dp->d_type is the high byte
4076 * of the expected dp->d_namlen,
4077 * so must be zero'ed.
4078 */
4079 dp->d_type = 0;
4080 # endif
4081 if (dp->d_reclen > 0) {
4082 dp = (struct dirent *)
4083 ((char *)dp + dp->d_reclen);
4084 } else {
4085 error = EIO;
4086 break;
4087 }
4088 }
4089 if (dp >= edp)
4090 error = uiomove(dirbuf, readcnt, &auio);
4091 }
4092 free(dirbuf, M_TEMP);
4093 }
4094 if (error) {
4095 VOP_UNLOCK(vp, 0);
4096 VFS_UNLOCK_GIANT(vfslocked);
4097 foffset_unlock(fp, foffset, 0);
4098 fdrop(fp, td);
4099 return (error);
4100 }
4101 if (uap->count == auio.uio_resid &&
4102 (vp->v_vflag & VV_ROOT) &&
4103 (vp->v_mount->mnt_flag & MNT_UNION)) {
4104 struct vnode *tvp = vp;
4105 vp = vp->v_mount->mnt_vnodecovered;
4106 VREF(vp);
4107 fp->f_vnode = vp;
4108 fp->f_data = vp;
4109 foffset = 0;
4110 vput(tvp);
4111 VFS_UNLOCK_GIANT(vfslocked);
4112 goto unionread;
4113 }
4114 VOP_UNLOCK(vp, 0);
4115 VFS_UNLOCK_GIANT(vfslocked);
4116 foffset_unlock(fp, foffset, 0);
4117 fdrop(fp, td);
4118 td->td_retval[0] = uap->count - auio.uio_resid;
4119 if (error == 0)
4120 *ploff = loff;
4121 return (error);
4122 }
4123 #endif /* COMPAT_43 */
4124
4125 /*
4126 * Read a block of directory entries in a filesystem independent format.
4127 */
4128 #ifndef _SYS_SYSPROTO_H_
4129 struct getdirentries_args {
4130 int fd;
4131 char *buf;
4132 u_int count;
4133 long *basep;
4134 };
4135 #endif
4136 int
4137 sys_getdirentries(td, uap)
4138 struct thread *td;
4139 register struct getdirentries_args /* {
4140 int fd;
4141 char *buf;
4142 u_int count;
4143 long *basep;
4144 } */ *uap;
4145 {
4146 long base;
4147 int error;
4148
4149 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4150 if (error)
4151 return (error);
4152 if (uap->basep != NULL)
4153 error = copyout(&base, uap->basep, sizeof(long));
4154 return (error);
4155 }
4156
4157 int
4158 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4159 long *basep)
4160 {
4161 struct vnode *vp;
4162 struct file *fp;
4163 struct uio auio;
4164 struct iovec aiov;
4165 int vfslocked;
4166 long loff;
4167 int error, eofflag;
4168 off_t foffset;
4169
4170 AUDIT_ARG_FD(fd);
4171 auio.uio_resid = count;
4172 if (auio.uio_resid > IOSIZE_MAX)
4173 return (EINVAL);
4174 if ((error = getvnode(td->td_proc->p_fd, fd, CAP_READ | CAP_SEEK,
4175 &fp)) != 0)
4176 return (error);
4177 if ((fp->f_flag & FREAD) == 0) {
4178 fdrop(fp, td);
4179 return (EBADF);
4180 }
4181 vp = fp->f_vnode;
4182 foffset = foffset_lock(fp, 0);
4183 unionread:
4184 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4185 if (vp->v_type != VDIR) {
4186 VFS_UNLOCK_GIANT(vfslocked);
4187 error = EINVAL;
4188 goto fail;
4189 }
4190 aiov.iov_base = buf;
4191 aiov.iov_len = count;
4192 auio.uio_iov = &aiov;
4193 auio.uio_iovcnt = 1;
4194 auio.uio_rw = UIO_READ;
4195 auio.uio_segflg = UIO_USERSPACE;
4196 auio.uio_td = td;
4197 vn_lock(vp, LK_SHARED | LK_RETRY);
4198 AUDIT_ARG_VNODE1(vp);
4199 loff = auio.uio_offset = foffset;
4200 #ifdef MAC
4201 error = mac_vnode_check_readdir(td->td_ucred, vp);
4202 if (error == 0)
4203 #endif
4204 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4205 NULL);
4206 foffset = auio.uio_offset;
4207 if (error) {
4208 VOP_UNLOCK(vp, 0);
4209 VFS_UNLOCK_GIANT(vfslocked);
4210 goto fail;
4211 }
4212 if (count == auio.uio_resid &&
4213 (vp->v_vflag & VV_ROOT) &&
4214 (vp->v_mount->mnt_flag & MNT_UNION)) {
4215 struct vnode *tvp = vp;
4216 vp = vp->v_mount->mnt_vnodecovered;
4217 VREF(vp);
4218 fp->f_vnode = vp;
4219 fp->f_data = vp;
4220 foffset = 0;
4221 vput(tvp);
4222 VFS_UNLOCK_GIANT(vfslocked);
4223 goto unionread;
4224 }
4225 VOP_UNLOCK(vp, 0);
4226 VFS_UNLOCK_GIANT(vfslocked);
4227 *basep = loff;
4228 td->td_retval[0] = count - auio.uio_resid;
4229 fail:
4230 foffset_unlock(fp, foffset, 0);
4231 fdrop(fp, td);
4232 return (error);
4233 }
4234
4235 #ifndef _SYS_SYSPROTO_H_
4236 struct getdents_args {
4237 int fd;
4238 char *buf;
4239 size_t count;
4240 };
4241 #endif
4242 int
4243 sys_getdents(td, uap)
4244 struct thread *td;
4245 register struct getdents_args /* {
4246 int fd;
4247 char *buf;
4248 u_int count;
4249 } */ *uap;
4250 {
4251 struct getdirentries_args ap;
4252 ap.fd = uap->fd;
4253 ap.buf = uap->buf;
4254 ap.count = uap->count;
4255 ap.basep = NULL;
4256 return (sys_getdirentries(td, &ap));
4257 }
4258
4259 /*
4260 * Set the mode mask for creation of filesystem nodes.
4261 */
4262 #ifndef _SYS_SYSPROTO_H_
4263 struct umask_args {
4264 int newmask;
4265 };
4266 #endif
4267 int
4268 sys_umask(td, uap)
4269 struct thread *td;
4270 struct umask_args /* {
4271 int newmask;
4272 } */ *uap;
4273 {
4274 register struct filedesc *fdp;
4275
4276 FILEDESC_XLOCK(td->td_proc->p_fd);
4277 fdp = td->td_proc->p_fd;
4278 td->td_retval[0] = fdp->fd_cmask;
4279 fdp->fd_cmask = uap->newmask & ALLPERMS;
4280 FILEDESC_XUNLOCK(td->td_proc->p_fd);
4281 return (0);
4282 }
4283
4284 /*
4285 * Void all references to file by ripping underlying filesystem away from
4286 * vnode.
4287 */
4288 #ifndef _SYS_SYSPROTO_H_
4289 struct revoke_args {
4290 char *path;
4291 };
4292 #endif
4293 int
4294 sys_revoke(td, uap)
4295 struct thread *td;
4296 register struct revoke_args /* {
4297 char *path;
4298 } */ *uap;
4299 {
4300 struct vnode *vp;
4301 struct vattr vattr;
4302 int error;
4303 struct nameidata nd;
4304 int vfslocked;
4305
4306 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4307 UIO_USERSPACE, uap->path, td);
4308 if ((error = namei(&nd)) != 0)
4309 return (error);
4310 vfslocked = NDHASGIANT(&nd);
4311 vp = nd.ni_vp;
4312 NDFREE(&nd, NDF_ONLY_PNBUF);
4313 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4314 error = EINVAL;
4315 goto out;
4316 }
4317 #ifdef MAC
4318 error = mac_vnode_check_revoke(td->td_ucred, vp);
4319 if (error)
4320 goto out;
4321 #endif
4322 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4323 if (error)
4324 goto out;
4325 if (td->td_ucred->cr_uid != vattr.va_uid) {
4326 error = priv_check(td, PRIV_VFS_ADMIN);
4327 if (error)
4328 goto out;
4329 }
4330 if (vcount(vp) > 1)
4331 VOP_REVOKE(vp, REVOKEALL);
4332 out:
4333 vput(vp);
4334 VFS_UNLOCK_GIANT(vfslocked);
4335 return (error);
4336 }
4337
4338 /*
4339 * Convert a user file descriptor to a kernel file entry and check that, if it
4340 * is a capability, the correct rights are present. A reference on the file
4341 * entry is held upon returning.
4342 */
4343 int
4344 getvnode(struct filedesc *fdp, int fd, cap_rights_t rights,
4345 struct file **fpp)
4346 {
4347 struct file *fp;
4348 #ifdef CAPABILITIES
4349 struct file *fp_fromcap;
4350 #endif
4351 int error;
4352
4353 error = 0;
4354 fp = NULL;
4355 if ((fdp == NULL) || (fp = fget_unlocked(fdp, fd)) == NULL)
4356 return (EBADF);
4357 #ifdef CAPABILITIES
4358 /*
4359 * If the file descriptor is for a capability, test rights and use the
4360 * file descriptor referenced by the capability.
4361 */
4362 error = cap_funwrap(fp, rights, &fp_fromcap);
4363 if (error) {
4364 fdrop(fp, curthread);
4365 return (error);
4366 }
4367 if (fp != fp_fromcap) {
4368 fhold(fp_fromcap);
4369 fdrop(fp, curthread);
4370 fp = fp_fromcap;
4371 }
4372 #endif /* CAPABILITIES */
4373
4374 /*
4375 * The file could be not of the vnode type, or it may be not
4376 * yet fully initialized, in which case the f_vnode pointer
4377 * may be set, but f_ops is still badfileops. E.g.,
4378 * devfs_open() transiently create such situation to
4379 * facilitate csw d_fdopen().
4380 *
4381 * Dupfdopen() handling in kern_openat() installs the
4382 * half-baked file into the process descriptor table, allowing
4383 * other thread to dereference it. Guard against the race by
4384 * checking f_ops.
4385 */
4386 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4387 fdrop(fp, curthread);
4388 return (EINVAL);
4389 }
4390 *fpp = fp;
4391 return (0);
4392 }
4393
4394
4395 /*
4396 * Get an (NFS) file handle.
4397 */
4398 #ifndef _SYS_SYSPROTO_H_
4399 struct lgetfh_args {
4400 char *fname;
4401 fhandle_t *fhp;
4402 };
4403 #endif
4404 int
4405 sys_lgetfh(td, uap)
4406 struct thread *td;
4407 register struct lgetfh_args *uap;
4408 {
4409 struct nameidata nd;
4410 fhandle_t fh;
4411 register struct vnode *vp;
4412 int vfslocked;
4413 int error;
4414
4415 error = priv_check(td, PRIV_VFS_GETFH);
4416 if (error)
4417 return (error);
4418 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4419 UIO_USERSPACE, uap->fname, td);
4420 error = namei(&nd);
4421 if (error)
4422 return (error);
4423 vfslocked = NDHASGIANT(&nd);
4424 NDFREE(&nd, NDF_ONLY_PNBUF);
4425 vp = nd.ni_vp;
4426 bzero(&fh, sizeof(fh));
4427 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4428 error = VOP_VPTOFH(vp, &fh.fh_fid);
4429 vput(vp);
4430 VFS_UNLOCK_GIANT(vfslocked);
4431 if (error)
4432 return (error);
4433 error = copyout(&fh, uap->fhp, sizeof (fh));
4434 return (error);
4435 }
4436
4437 #ifndef _SYS_SYSPROTO_H_
4438 struct getfh_args {
4439 char *fname;
4440 fhandle_t *fhp;
4441 };
4442 #endif
4443 int
4444 sys_getfh(td, uap)
4445 struct thread *td;
4446 register struct getfh_args *uap;
4447 {
4448 struct nameidata nd;
4449 fhandle_t fh;
4450 register struct vnode *vp;
4451 int vfslocked;
4452 int error;
4453
4454 error = priv_check(td, PRIV_VFS_GETFH);
4455 if (error)
4456 return (error);
4457 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4458 UIO_USERSPACE, uap->fname, td);
4459 error = namei(&nd);
4460 if (error)
4461 return (error);
4462 vfslocked = NDHASGIANT(&nd);
4463 NDFREE(&nd, NDF_ONLY_PNBUF);
4464 vp = nd.ni_vp;
4465 bzero(&fh, sizeof(fh));
4466 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4467 error = VOP_VPTOFH(vp, &fh.fh_fid);
4468 vput(vp);
4469 VFS_UNLOCK_GIANT(vfslocked);
4470 if (error)
4471 return (error);
4472 error = copyout(&fh, uap->fhp, sizeof (fh));
4473 return (error);
4474 }
4475
4476 /*
4477 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4478 * open descriptor.
4479 *
4480 * warning: do not remove the priv_check() call or this becomes one giant
4481 * security hole.
4482 */
4483 #ifndef _SYS_SYSPROTO_H_
4484 struct fhopen_args {
4485 const struct fhandle *u_fhp;
4486 int flags;
4487 };
4488 #endif
4489 int
4490 sys_fhopen(td, uap)
4491 struct thread *td;
4492 struct fhopen_args /* {
4493 const struct fhandle *u_fhp;
4494 int flags;
4495 } */ *uap;
4496 {
4497 struct proc *p = td->td_proc;
4498 struct mount *mp;
4499 struct vnode *vp;
4500 struct fhandle fhp;
4501 struct vattr vat;
4502 struct vattr *vap = &vat;
4503 struct flock lf;
4504 struct file *fp;
4505 register struct filedesc *fdp = p->p_fd;
4506 int fmode, error, type;
4507 accmode_t accmode;
4508 struct file *nfp;
4509 int vfslocked;
4510 int indx;
4511
4512 error = priv_check(td, PRIV_VFS_FHOPEN);
4513 if (error)
4514 return (error);
4515 fmode = FFLAGS(uap->flags);
4516 /* why not allow a non-read/write open for our lockd? */
4517 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4518 return (EINVAL);
4519 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4520 if (error)
4521 return(error);
4522 /* find the mount point */
4523 mp = vfs_busyfs(&fhp.fh_fsid);
4524 if (mp == NULL)
4525 return (ESTALE);
4526 vfslocked = VFS_LOCK_GIANT(mp);
4527 /* now give me my vnode, it gets returned to me locked */
4528 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4529 vfs_unbusy(mp);
4530 if (error)
4531 goto out;
4532 /*
4533 * from now on we have to make sure not
4534 * to forget about the vnode
4535 * any error that causes an abort must vput(vp)
4536 * just set error = err and 'goto bad;'.
4537 */
4538
4539 /*
4540 * from vn_open
4541 */
4542 if (vp->v_type == VLNK) {
4543 error = EMLINK;
4544 goto bad;
4545 }
4546 if (vp->v_type == VSOCK) {
4547 error = EOPNOTSUPP;
4548 goto bad;
4549 }
4550 if (vp->v_type != VDIR && fmode & O_DIRECTORY) {
4551 error = ENOTDIR;
4552 goto bad;
4553 }
4554 accmode = 0;
4555 if (fmode & (FWRITE | O_TRUNC)) {
4556 if (vp->v_type == VDIR) {
4557 error = EISDIR;
4558 goto bad;
4559 }
4560 error = vn_writechk(vp);
4561 if (error)
4562 goto bad;
4563 accmode |= VWRITE;
4564 }
4565 if (fmode & FREAD)
4566 accmode |= VREAD;
4567 if ((fmode & O_APPEND) && (fmode & FWRITE))
4568 accmode |= VAPPEND;
4569 #ifdef MAC
4570 error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4571 if (error)
4572 goto bad;
4573 #endif
4574 if (accmode) {
4575 error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4576 if (error)
4577 goto bad;
4578 }
4579 if (fmode & O_TRUNC) {
4580 vfs_ref(mp);
4581 VOP_UNLOCK(vp, 0); /* XXX */
4582 if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4583 vrele(vp);
4584 vfs_rel(mp);
4585 goto out;
4586 }
4587 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
4588 vfs_rel(mp);
4589 #ifdef MAC
4590 /*
4591 * We don't yet have fp->f_cred, so use td->td_ucred, which
4592 * should be right.
4593 */
4594 error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4595 if (error == 0) {
4596 #endif
4597 VATTR_NULL(vap);
4598 vap->va_size = 0;
4599 error = VOP_SETATTR(vp, vap, td->td_ucred);
4600 #ifdef MAC
4601 }
4602 #endif
4603 vn_finished_write(mp);
4604 if (error)
4605 goto bad;
4606 }
4607 error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4608 if (error)
4609 goto bad;
4610
4611 if (fmode & FWRITE)
4612 vp->v_writecount++;
4613
4614 /*
4615 * end of vn_open code
4616 */
4617
4618 if ((error = falloc(td, &nfp, &indx, fmode)) != 0) {
4619 if (fmode & FWRITE)
4620 vp->v_writecount--;
4621 goto bad;
4622 }
4623 /* An extra reference on `nfp' has been held for us by falloc(). */
4624 fp = nfp;
4625 nfp->f_vnode = vp;
4626 finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4627 if (fmode & (O_EXLOCK | O_SHLOCK)) {
4628 lf.l_whence = SEEK_SET;
4629 lf.l_start = 0;
4630 lf.l_len = 0;
4631 if (fmode & O_EXLOCK)
4632 lf.l_type = F_WRLCK;
4633 else
4634 lf.l_type = F_RDLCK;
4635 type = F_FLOCK;
4636 if ((fmode & FNONBLOCK) == 0)
4637 type |= F_WAIT;
4638 VOP_UNLOCK(vp, 0);
4639 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4640 type)) != 0) {
4641 /*
4642 * The lock request failed. Normally close the
4643 * descriptor but handle the case where someone might
4644 * have dup()d or close()d it when we weren't looking.
4645 */
4646 fdclose(fdp, fp, indx, td);
4647
4648 /*
4649 * release our private reference
4650 */
4651 fdrop(fp, td);
4652 goto out;
4653 }
4654 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4655 atomic_set_int(&fp->f_flag, FHASLOCK);
4656 }
4657
4658 VOP_UNLOCK(vp, 0);
4659 fdrop(fp, td);
4660 VFS_UNLOCK_GIANT(vfslocked);
4661 td->td_retval[0] = indx;
4662 return (0);
4663
4664 bad:
4665 vput(vp);
4666 out:
4667 VFS_UNLOCK_GIANT(vfslocked);
4668 return (error);
4669 }
4670
4671 /*
4672 * Stat an (NFS) file handle.
4673 */
4674 #ifndef _SYS_SYSPROTO_H_
4675 struct fhstat_args {
4676 struct fhandle *u_fhp;
4677 struct stat *sb;
4678 };
4679 #endif
4680 int
4681 sys_fhstat(td, uap)
4682 struct thread *td;
4683 register struct fhstat_args /* {
4684 struct fhandle *u_fhp;
4685 struct stat *sb;
4686 } */ *uap;
4687 {
4688 struct stat sb;
4689 fhandle_t fh;
4690 struct mount *mp;
4691 struct vnode *vp;
4692 int vfslocked;
4693 int error;
4694
4695 error = priv_check(td, PRIV_VFS_FHSTAT);
4696 if (error)
4697 return (error);
4698 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4699 if (error)
4700 return (error);
4701 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4702 return (ESTALE);
4703 vfslocked = VFS_LOCK_GIANT(mp);
4704 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4705 vfs_unbusy(mp);
4706 if (error) {
4707 VFS_UNLOCK_GIANT(vfslocked);
4708 return (error);
4709 }
4710 error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4711 vput(vp);
4712 VFS_UNLOCK_GIANT(vfslocked);
4713 if (error)
4714 return (error);
4715 error = copyout(&sb, uap->sb, sizeof(sb));
4716 return (error);
4717 }
4718
4719 /*
4720 * Implement fstatfs() for (NFS) file handles.
4721 */
4722 #ifndef _SYS_SYSPROTO_H_
4723 struct fhstatfs_args {
4724 struct fhandle *u_fhp;
4725 struct statfs *buf;
4726 };
4727 #endif
4728 int
4729 sys_fhstatfs(td, uap)
4730 struct thread *td;
4731 struct fhstatfs_args /* {
4732 struct fhandle *u_fhp;
4733 struct statfs *buf;
4734 } */ *uap;
4735 {
4736 struct statfs sf;
4737 fhandle_t fh;
4738 int error;
4739
4740 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4741 if (error)
4742 return (error);
4743 error = kern_fhstatfs(td, fh, &sf);
4744 if (error)
4745 return (error);
4746 return (copyout(&sf, uap->buf, sizeof(sf)));
4747 }
4748
4749 int
4750 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4751 {
4752 struct statfs *sp;
4753 struct mount *mp;
4754 struct vnode *vp;
4755 int vfslocked;
4756 int error;
4757
4758 error = priv_check(td, PRIV_VFS_FHSTATFS);
4759 if (error)
4760 return (error);
4761 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4762 return (ESTALE);
4763 vfslocked = VFS_LOCK_GIANT(mp);
4764 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4765 if (error) {
4766 vfs_unbusy(mp);
4767 VFS_UNLOCK_GIANT(vfslocked);
4768 return (error);
4769 }
4770 vput(vp);
4771 error = prison_canseemount(td->td_ucred, mp);
4772 if (error)
4773 goto out;
4774 #ifdef MAC
4775 error = mac_mount_check_stat(td->td_ucred, mp);
4776 if (error)
4777 goto out;
4778 #endif
4779 /*
4780 * Set these in case the underlying filesystem fails to do so.
4781 */
4782 sp = &mp->mnt_stat;
4783 sp->f_version = STATFS_VERSION;
4784 sp->f_namemax = NAME_MAX;
4785 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4786 error = VFS_STATFS(mp, sp);
4787 if (error == 0)
4788 *buf = *sp;
4789 out:
4790 vfs_unbusy(mp);
4791 VFS_UNLOCK_GIANT(vfslocked);
4792 return (error);
4793 }
4794
4795 int
4796 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4797 {
4798 struct file *fp;
4799 struct mount *mp;
4800 struct vnode *vp;
4801 off_t olen, ooffset;
4802 int error, vfslocked;
4803
4804 fp = NULL;
4805 vfslocked = 0;
4806 error = fget(td, fd, CAP_WRITE, &fp);
4807 if (error != 0)
4808 goto out;
4809
4810 switch (fp->f_type) {
4811 case DTYPE_VNODE:
4812 break;
4813 case DTYPE_PIPE:
4814 case DTYPE_FIFO:
4815 error = ESPIPE;
4816 goto out;
4817 default:
4818 error = ENODEV;
4819 goto out;
4820 }
4821 if ((fp->f_flag & FWRITE) == 0) {
4822 error = EBADF;
4823 goto out;
4824 }
4825 vp = fp->f_vnode;
4826 if (vp->v_type != VREG) {
4827 error = ENODEV;
4828 goto out;
4829 }
4830 if (offset < 0 || len <= 0) {
4831 error = EINVAL;
4832 goto out;
4833 }
4834 /* Check for wrap. */
4835 if (offset > OFF_MAX - len) {
4836 error = EFBIG;
4837 goto out;
4838 }
4839
4840 /* Allocating blocks may take a long time, so iterate. */
4841 for (;;) {
4842 olen = len;
4843 ooffset = offset;
4844
4845 bwillwrite();
4846 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4847 mp = NULL;
4848 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4849 if (error != 0) {
4850 VFS_UNLOCK_GIANT(vfslocked);
4851 break;
4852 }
4853 error = vn_lock(vp, LK_EXCLUSIVE);
4854 if (error != 0) {
4855 vn_finished_write(mp);
4856 VFS_UNLOCK_GIANT(vfslocked);
4857 break;
4858 }
4859 #ifdef MAC
4860 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4861 if (error == 0)
4862 #endif
4863 error = VOP_ALLOCATE(vp, &offset, &len);
4864 VOP_UNLOCK(vp, 0);
4865 vn_finished_write(mp);
4866 VFS_UNLOCK_GIANT(vfslocked);
4867
4868 if (olen + ooffset != offset + len) {
4869 panic("offset + len changed from %jx/%jx to %jx/%jx",
4870 ooffset, olen, offset, len);
4871 }
4872 if (error != 0 || len == 0)
4873 break;
4874 KASSERT(olen > len, ("Iteration did not make progress?"));
4875 maybe_yield();
4876 }
4877 out:
4878 if (fp != NULL)
4879 fdrop(fp, td);
4880 return (error);
4881 }
4882
4883 int
4884 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4885 {
4886
4887 return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4888 }
4889
4890 /*
4891 * Unlike madvise(2), we do not make a best effort to remember every
4892 * possible caching hint. Instead, we remember the last setting with
4893 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4894 * region of any current setting.
4895 */
4896 int
4897 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4898 int advice)
4899 {
4900 struct fadvise_info *fa, *new;
4901 struct file *fp;
4902 struct vnode *vp;
4903 off_t end;
4904 int error;
4905
4906 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4907 return (EINVAL);
4908 switch (advice) {
4909 case POSIX_FADV_SEQUENTIAL:
4910 case POSIX_FADV_RANDOM:
4911 case POSIX_FADV_NOREUSE:
4912 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4913 break;
4914 case POSIX_FADV_NORMAL:
4915 case POSIX_FADV_WILLNEED:
4916 case POSIX_FADV_DONTNEED:
4917 new = NULL;
4918 break;
4919 default:
4920 return (EINVAL);
4921 }
4922 /* XXX: CAP_POSIX_FADVISE? */
4923 error = fget(td, fd, 0, &fp);
4924 if (error != 0)
4925 goto out;
4926
4927 switch (fp->f_type) {
4928 case DTYPE_VNODE:
4929 break;
4930 case DTYPE_PIPE:
4931 case DTYPE_FIFO:
4932 error = ESPIPE;
4933 goto out;
4934 default:
4935 error = ENODEV;
4936 goto out;
4937 }
4938 vp = fp->f_vnode;
4939 if (vp->v_type != VREG) {
4940 error = ENODEV;
4941 goto out;
4942 }
4943 if (len == 0)
4944 end = OFF_MAX;
4945 else
4946 end = offset + len - 1;
4947 switch (advice) {
4948 case POSIX_FADV_SEQUENTIAL:
4949 case POSIX_FADV_RANDOM:
4950 case POSIX_FADV_NOREUSE:
4951 /*
4952 * Try to merge any existing non-standard region with
4953 * this new region if possible, otherwise create a new
4954 * non-standard region for this request.
4955 */
4956 mtx_pool_lock(mtxpool_sleep, fp);
4957 fa = fp->f_advice;
4958 if (fa != NULL && fa->fa_advice == advice &&
4959 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4960 (end != OFF_MAX && fa->fa_start == end + 1) ||
4961 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4962 if (offset < fa->fa_start)
4963 fa->fa_start = offset;
4964 if (end > fa->fa_end)
4965 fa->fa_end = end;
4966 } else {
4967 new->fa_advice = advice;
4968 new->fa_start = offset;
4969 new->fa_end = end;
4970 new->fa_prevstart = 0;
4971 new->fa_prevend = 0;
4972 fp->f_advice = new;
4973 new = fa;
4974 }
4975 mtx_pool_unlock(mtxpool_sleep, fp);
4976 break;
4977 case POSIX_FADV_NORMAL:
4978 /*
4979 * If a the "normal" region overlaps with an existing
4980 * non-standard region, trim or remove the
4981 * non-standard region.
4982 */
4983 mtx_pool_lock(mtxpool_sleep, fp);
4984 fa = fp->f_advice;
4985 if (fa != NULL) {
4986 if (offset <= fa->fa_start && end >= fa->fa_end) {
4987 new = fa;
4988 fp->f_advice = NULL;
4989 } else if (offset <= fa->fa_start &&
4990 end >= fa->fa_start)
4991 fa->fa_start = end + 1;
4992 else if (offset <= fa->fa_end && end >= fa->fa_end)
4993 fa->fa_end = offset - 1;
4994 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4995 /*
4996 * If the "normal" region is a middle
4997 * portion of the existing
4998 * non-standard region, just remove
4999 * the whole thing rather than picking
5000 * one side or the other to
5001 * preserve.
5002 */
5003 new = fa;
5004 fp->f_advice = NULL;
5005 }
5006 }
5007 mtx_pool_unlock(mtxpool_sleep, fp);
5008 break;
5009 case POSIX_FADV_WILLNEED:
5010 case POSIX_FADV_DONTNEED:
5011 error = VOP_ADVISE(vp, offset, end, advice);
5012 break;
5013 }
5014 out:
5015 if (fp != NULL)
5016 fdrop(fp, td);
5017 free(new, M_FADVISE);
5018 return (error);
5019 }
5020
5021 int
5022 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
5023 {
5024
5025 return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
5026 uap->advice));
5027 }
Cache object: 50f8ffa130e76e381b0835a9ed4c7d4d
|