1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
37 */
38
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41
42 #include "opt_capsicum.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/capsicum.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/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79
80 #include <machine/stdarg.h>
81
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89
90 #include <fs/devfs/devfs.h>
91
92 #include <ufs/ufs/quota.h>
93
94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
95
96 static int kern_chflagsat(struct thread *td, int fd, const char *path,
97 enum uio_seg pathseg, u_long flags, int atflag);
98 static int setfflags(struct thread *td, struct vnode *, u_long);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int getutimens(const struct timespec *, enum uio_seg,
101 struct timespec *, int *);
102 static int setutimes(struct thread *td, struct vnode *,
103 const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105 struct thread *td);
106 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
107 enum uio_seg pathseg, fhandle_t *fhp);
108 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
109 size_t count, struct thread *td);
110 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
111 const char *path, enum uio_seg segflag);
112
113 static uint64_t
114 at2cnpflags(u_int at_flags, u_int mask)
115 {
116 u_int64_t res;
117
118 MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
119 (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
120
121 res = 0;
122 at_flags &= mask;
123 if ((at_flags & AT_RESOLVE_BENEATH) != 0)
124 res |= RBENEATH;
125 if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
126 res |= FOLLOW;
127 /* NOFOLLOW is pseudo flag */
128 if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
129 res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
130 FOLLOW;
131 }
132 if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
133 res |= EMPTYPATH;
134 return (res);
135 }
136
137 int
138 kern_sync(struct thread *td)
139 {
140 struct mount *mp, *nmp;
141 int save;
142
143 mtx_lock(&mountlist_mtx);
144 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
145 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
146 nmp = TAILQ_NEXT(mp, mnt_list);
147 continue;
148 }
149 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
150 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
151 save = curthread_pflags_set(TDP_SYNCIO);
152 vfs_periodic(mp, MNT_NOWAIT);
153 VFS_SYNC(mp, MNT_NOWAIT);
154 curthread_pflags_restore(save);
155 vn_finished_write(mp);
156 }
157 mtx_lock(&mountlist_mtx);
158 nmp = TAILQ_NEXT(mp, mnt_list);
159 vfs_unbusy(mp);
160 }
161 mtx_unlock(&mountlist_mtx);
162 return (0);
163 }
164
165 /*
166 * Sync each mounted filesystem.
167 */
168 #ifndef _SYS_SYSPROTO_H_
169 struct sync_args {
170 int dummy;
171 };
172 #endif
173 /* ARGSUSED */
174 int
175 sys_sync(struct thread *td, struct sync_args *uap)
176 {
177
178 return (kern_sync(td));
179 }
180
181 /*
182 * Change filesystem quotas.
183 */
184 #ifndef _SYS_SYSPROTO_H_
185 struct quotactl_args {
186 char *path;
187 int cmd;
188 int uid;
189 caddr_t arg;
190 };
191 #endif
192 int
193 sys_quotactl(struct thread *td, struct quotactl_args *uap)
194 {
195 struct mount *mp;
196 struct nameidata nd;
197 int error;
198
199 AUDIT_ARG_CMD(uap->cmd);
200 AUDIT_ARG_UID(uap->uid);
201 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
202 return (EPERM);
203 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
204 uap->path, td);
205 if ((error = namei(&nd)) != 0)
206 return (error);
207 NDFREE(&nd, NDF_ONLY_PNBUF);
208 mp = nd.ni_vp->v_mount;
209 vfs_ref(mp);
210 vput(nd.ni_vp);
211 error = vfs_busy(mp, 0);
212 if (error != 0) {
213 vfs_rel(mp);
214 return (error);
215 }
216 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
217
218 /*
219 * Since quota on operation typically needs to open quota
220 * file, the Q_QUOTAON handler needs to unbusy the mount point
221 * before calling into namei. Otherwise, unmount might be
222 * started between two vfs_busy() invocations (first is our,
223 * second is from mount point cross-walk code in lookup()),
224 * causing deadlock.
225 *
226 * Require that Q_QUOTAON handles the vfs_busy() reference on
227 * its own, always returning with ubusied mount point.
228 */
229 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
230 (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
231 vfs_unbusy(mp);
232 vfs_rel(mp);
233 return (error);
234 }
235
236 /*
237 * Used by statfs conversion routines to scale the block size up if
238 * necessary so that all of the block counts are <= 'max_size'. Note
239 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
240 * value of 'n'.
241 */
242 void
243 statfs_scale_blocks(struct statfs *sf, long max_size)
244 {
245 uint64_t count;
246 int shift;
247
248 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
249
250 /*
251 * Attempt to scale the block counts to give a more accurate
252 * overview to userland of the ratio of free space to used
253 * space. To do this, find the largest block count and compute
254 * a divisor that lets it fit into a signed integer <= max_size.
255 */
256 if (sf->f_bavail < 0)
257 count = -sf->f_bavail;
258 else
259 count = sf->f_bavail;
260 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
261 if (count <= max_size)
262 return;
263
264 count >>= flsl(max_size);
265 shift = 0;
266 while (count > 0) {
267 shift++;
268 count >>=1;
269 }
270
271 sf->f_bsize <<= shift;
272 sf->f_blocks >>= shift;
273 sf->f_bfree >>= shift;
274 sf->f_bavail >>= shift;
275 }
276
277 static int
278 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
279 {
280 int error;
281
282 if (mp == NULL)
283 return (EBADF);
284 error = vfs_busy(mp, 0);
285 vfs_rel(mp);
286 if (error != 0)
287 return (error);
288 #ifdef MAC
289 error = mac_mount_check_stat(td->td_ucred, mp);
290 if (error != 0)
291 goto out;
292 #endif
293 error = VFS_STATFS(mp, buf);
294 if (error != 0)
295 goto out;
296 if (priv_check_cred_vfs_generation(td->td_ucred)) {
297 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
298 prison_enforce_statfs(td->td_ucred, mp, buf);
299 }
300 out:
301 vfs_unbusy(mp);
302 return (error);
303 }
304
305 /*
306 * Get filesystem statistics.
307 */
308 #ifndef _SYS_SYSPROTO_H_
309 struct statfs_args {
310 char *path;
311 struct statfs *buf;
312 };
313 #endif
314 int
315 sys_statfs(struct thread *td, struct statfs_args *uap)
316 {
317 struct statfs *sfp;
318 int error;
319
320 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
321 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
322 if (error == 0)
323 error = copyout(sfp, uap->buf, sizeof(struct statfs));
324 free(sfp, M_STATFS);
325 return (error);
326 }
327
328 int
329 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
330 struct statfs *buf)
331 {
332 struct mount *mp;
333 struct nameidata nd;
334 int error;
335
336 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
337 error = namei(&nd);
338 if (error != 0)
339 return (error);
340 mp = vfs_ref_from_vp(nd.ni_vp);
341 NDFREE_NOTHING(&nd);
342 vrele(nd.ni_vp);
343 return (kern_do_statfs(td, mp, buf));
344 }
345
346 /*
347 * Get filesystem statistics.
348 */
349 #ifndef _SYS_SYSPROTO_H_
350 struct fstatfs_args {
351 int fd;
352 struct statfs *buf;
353 };
354 #endif
355 int
356 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
357 {
358 struct statfs *sfp;
359 int error;
360
361 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
362 error = kern_fstatfs(td, uap->fd, sfp);
363 if (error == 0)
364 error = copyout(sfp, uap->buf, sizeof(struct statfs));
365 free(sfp, M_STATFS);
366 return (error);
367 }
368
369 int
370 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
371 {
372 struct file *fp;
373 struct mount *mp;
374 struct vnode *vp;
375 int error;
376
377 AUDIT_ARG_FD(fd);
378 error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
379 if (error != 0)
380 return (error);
381 vp = fp->f_vnode;
382 #ifdef AUDIT
383 if (AUDITING_TD(td)) {
384 vn_lock(vp, LK_SHARED | LK_RETRY);
385 AUDIT_ARG_VNODE1(vp);
386 VOP_UNLOCK(vp);
387 }
388 #endif
389 mp = vfs_ref_from_vp(vp);
390 fdrop(fp, td);
391 return (kern_do_statfs(td, mp, buf));
392 }
393
394 /*
395 * Get statistics on all filesystems.
396 */
397 #ifndef _SYS_SYSPROTO_H_
398 struct getfsstat_args {
399 struct statfs *buf;
400 long bufsize;
401 int mode;
402 };
403 #endif
404 int
405 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
406 {
407 size_t count;
408 int error;
409
410 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
411 return (EINVAL);
412 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
413 UIO_USERSPACE, uap->mode);
414 if (error == 0)
415 td->td_retval[0] = count;
416 return (error);
417 }
418
419 /*
420 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
421 * The caller is responsible for freeing memory which will be allocated
422 * in '*buf'.
423 */
424 int
425 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
426 size_t *countp, enum uio_seg bufseg, int mode)
427 {
428 struct mount *mp, *nmp;
429 struct statfs *sfsp, *sp, *sptmp, *tofree;
430 size_t count, maxcount;
431 int error;
432
433 switch (mode) {
434 case MNT_WAIT:
435 case MNT_NOWAIT:
436 break;
437 default:
438 if (bufseg == UIO_SYSSPACE)
439 *buf = NULL;
440 return (EINVAL);
441 }
442 restart:
443 maxcount = bufsize / sizeof(struct statfs);
444 if (bufsize == 0) {
445 sfsp = NULL;
446 tofree = NULL;
447 } else if (bufseg == UIO_USERSPACE) {
448 sfsp = *buf;
449 tofree = NULL;
450 } else /* if (bufseg == UIO_SYSSPACE) */ {
451 count = 0;
452 mtx_lock(&mountlist_mtx);
453 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
454 count++;
455 }
456 mtx_unlock(&mountlist_mtx);
457 if (maxcount > count)
458 maxcount = count;
459 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
460 M_STATFS, M_WAITOK);
461 }
462
463 count = 0;
464
465 /*
466 * If there is no target buffer they only want the count.
467 *
468 * This could be TAILQ_FOREACH but it is open-coded to match the original
469 * code below.
470 */
471 if (sfsp == NULL) {
472 mtx_lock(&mountlist_mtx);
473 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
474 if (prison_canseemount(td->td_ucred, mp) != 0) {
475 nmp = TAILQ_NEXT(mp, mnt_list);
476 continue;
477 }
478 #ifdef MAC
479 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
480 nmp = TAILQ_NEXT(mp, mnt_list);
481 continue;
482 }
483 #endif
484 count++;
485 nmp = TAILQ_NEXT(mp, mnt_list);
486 }
487 mtx_unlock(&mountlist_mtx);
488 *countp = count;
489 return (0);
490 }
491
492 /*
493 * They want the entire thing.
494 *
495 * Short-circuit the corner case of no room for anything, avoids
496 * relocking below.
497 */
498 if (maxcount < 1) {
499 goto out;
500 }
501
502 mtx_lock(&mountlist_mtx);
503 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
504 if (prison_canseemount(td->td_ucred, mp) != 0) {
505 nmp = TAILQ_NEXT(mp, mnt_list);
506 continue;
507 }
508 #ifdef MAC
509 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
510 nmp = TAILQ_NEXT(mp, mnt_list);
511 continue;
512 }
513 #endif
514 if (mode == MNT_WAIT) {
515 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
516 /*
517 * If vfs_busy() failed, and MBF_NOWAIT
518 * wasn't passed, then the mp is gone.
519 * Furthermore, because of MBF_MNTLSTLOCK,
520 * the mountlist_mtx was dropped. We have
521 * no other choice than to start over.
522 */
523 mtx_unlock(&mountlist_mtx);
524 free(tofree, M_STATFS);
525 goto restart;
526 }
527 } else {
528 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
529 nmp = TAILQ_NEXT(mp, mnt_list);
530 continue;
531 }
532 }
533 sp = &mp->mnt_stat;
534 /*
535 * If MNT_NOWAIT is specified, do not refresh
536 * the fsstat cache.
537 */
538 if (mode != MNT_NOWAIT) {
539 error = VFS_STATFS(mp, sp);
540 if (error != 0) {
541 mtx_lock(&mountlist_mtx);
542 nmp = TAILQ_NEXT(mp, mnt_list);
543 vfs_unbusy(mp);
544 continue;
545 }
546 }
547 if (priv_check_cred_vfs_generation(td->td_ucred)) {
548 sptmp = malloc(sizeof(struct statfs), M_STATFS,
549 M_WAITOK);
550 *sptmp = *sp;
551 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
552 prison_enforce_statfs(td->td_ucred, mp, sptmp);
553 sp = sptmp;
554 } else
555 sptmp = NULL;
556 if (bufseg == UIO_SYSSPACE) {
557 bcopy(sp, sfsp, sizeof(*sp));
558 free(sptmp, M_STATFS);
559 } else /* if (bufseg == UIO_USERSPACE) */ {
560 error = copyout(sp, sfsp, sizeof(*sp));
561 free(sptmp, M_STATFS);
562 if (error != 0) {
563 vfs_unbusy(mp);
564 return (error);
565 }
566 }
567 sfsp++;
568 count++;
569
570 if (count == maxcount) {
571 vfs_unbusy(mp);
572 goto out;
573 }
574
575 mtx_lock(&mountlist_mtx);
576 nmp = TAILQ_NEXT(mp, mnt_list);
577 vfs_unbusy(mp);
578 }
579 mtx_unlock(&mountlist_mtx);
580 out:
581 *countp = count;
582 return (0);
583 }
584
585 #ifdef COMPAT_FREEBSD4
586 /*
587 * Get old format filesystem statistics.
588 */
589 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
590
591 #ifndef _SYS_SYSPROTO_H_
592 struct freebsd4_statfs_args {
593 char *path;
594 struct ostatfs *buf;
595 };
596 #endif
597 int
598 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
599 {
600 struct ostatfs osb;
601 struct statfs *sfp;
602 int error;
603
604 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
605 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
606 if (error == 0) {
607 freebsd4_cvtstatfs(sfp, &osb);
608 error = copyout(&osb, uap->buf, sizeof(osb));
609 }
610 free(sfp, M_STATFS);
611 return (error);
612 }
613
614 /*
615 * Get filesystem statistics.
616 */
617 #ifndef _SYS_SYSPROTO_H_
618 struct freebsd4_fstatfs_args {
619 int fd;
620 struct ostatfs *buf;
621 };
622 #endif
623 int
624 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
625 {
626 struct ostatfs osb;
627 struct statfs *sfp;
628 int error;
629
630 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
631 error = kern_fstatfs(td, uap->fd, sfp);
632 if (error == 0) {
633 freebsd4_cvtstatfs(sfp, &osb);
634 error = copyout(&osb, uap->buf, sizeof(osb));
635 }
636 free(sfp, M_STATFS);
637 return (error);
638 }
639
640 /*
641 * Get statistics on all filesystems.
642 */
643 #ifndef _SYS_SYSPROTO_H_
644 struct freebsd4_getfsstat_args {
645 struct ostatfs *buf;
646 long bufsize;
647 int mode;
648 };
649 #endif
650 int
651 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
652 {
653 struct statfs *buf, *sp;
654 struct ostatfs osb;
655 size_t count, size;
656 int error;
657
658 if (uap->bufsize < 0)
659 return (EINVAL);
660 count = uap->bufsize / sizeof(struct ostatfs);
661 if (count > SIZE_MAX / sizeof(struct statfs))
662 return (EINVAL);
663 size = count * sizeof(struct statfs);
664 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
665 uap->mode);
666 if (error == 0)
667 td->td_retval[0] = count;
668 if (size != 0) {
669 sp = buf;
670 while (count != 0 && error == 0) {
671 freebsd4_cvtstatfs(sp, &osb);
672 error = copyout(&osb, uap->buf, sizeof(osb));
673 sp++;
674 uap->buf++;
675 count--;
676 }
677 free(buf, M_STATFS);
678 }
679 return (error);
680 }
681
682 /*
683 * Implement fstatfs() for (NFS) file handles.
684 */
685 #ifndef _SYS_SYSPROTO_H_
686 struct freebsd4_fhstatfs_args {
687 struct fhandle *u_fhp;
688 struct ostatfs *buf;
689 };
690 #endif
691 int
692 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
693 {
694 struct ostatfs osb;
695 struct statfs *sfp;
696 fhandle_t fh;
697 int error;
698
699 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
700 if (error != 0)
701 return (error);
702 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
703 error = kern_fhstatfs(td, fh, sfp);
704 if (error == 0) {
705 freebsd4_cvtstatfs(sfp, &osb);
706 error = copyout(&osb, uap->buf, sizeof(osb));
707 }
708 free(sfp, M_STATFS);
709 return (error);
710 }
711
712 /*
713 * Convert a new format statfs structure to an old format statfs structure.
714 */
715 static void
716 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
717 {
718
719 statfs_scale_blocks(nsp, LONG_MAX);
720 bzero(osp, sizeof(*osp));
721 osp->f_bsize = nsp->f_bsize;
722 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
723 osp->f_blocks = nsp->f_blocks;
724 osp->f_bfree = nsp->f_bfree;
725 osp->f_bavail = nsp->f_bavail;
726 osp->f_files = MIN(nsp->f_files, LONG_MAX);
727 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
728 osp->f_owner = nsp->f_owner;
729 osp->f_type = nsp->f_type;
730 osp->f_flags = nsp->f_flags;
731 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
732 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
733 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
734 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
735 strlcpy(osp->f_fstypename, nsp->f_fstypename,
736 MIN(MFSNAMELEN, OMFSNAMELEN));
737 strlcpy(osp->f_mntonname, nsp->f_mntonname,
738 MIN(MNAMELEN, OMNAMELEN));
739 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
740 MIN(MNAMELEN, OMNAMELEN));
741 osp->f_fsid = nsp->f_fsid;
742 }
743 #endif /* COMPAT_FREEBSD4 */
744
745 #if defined(COMPAT_FREEBSD11)
746 /*
747 * Get old format filesystem statistics.
748 */
749 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
750
751 int
752 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
753 {
754 struct freebsd11_statfs osb;
755 struct statfs *sfp;
756 int error;
757
758 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
759 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
760 if (error == 0) {
761 freebsd11_cvtstatfs(sfp, &osb);
762 error = copyout(&osb, uap->buf, sizeof(osb));
763 }
764 free(sfp, M_STATFS);
765 return (error);
766 }
767
768 /*
769 * Get filesystem statistics.
770 */
771 int
772 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
773 {
774 struct freebsd11_statfs osb;
775 struct statfs *sfp;
776 int error;
777
778 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
779 error = kern_fstatfs(td, uap->fd, sfp);
780 if (error == 0) {
781 freebsd11_cvtstatfs(sfp, &osb);
782 error = copyout(&osb, uap->buf, sizeof(osb));
783 }
784 free(sfp, M_STATFS);
785 return (error);
786 }
787
788 /*
789 * Get statistics on all filesystems.
790 */
791 int
792 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
793 {
794 struct freebsd11_statfs osb;
795 struct statfs *buf, *sp;
796 size_t count, size;
797 int error;
798
799 count = uap->bufsize / sizeof(struct ostatfs);
800 size = count * sizeof(struct statfs);
801 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
802 uap->mode);
803 if (error == 0)
804 td->td_retval[0] = count;
805 if (size > 0) {
806 sp = buf;
807 while (count > 0 && error == 0) {
808 freebsd11_cvtstatfs(sp, &osb);
809 error = copyout(&osb, uap->buf, sizeof(osb));
810 sp++;
811 uap->buf++;
812 count--;
813 }
814 free(buf, M_STATFS);
815 }
816 return (error);
817 }
818
819 /*
820 * Implement fstatfs() for (NFS) file handles.
821 */
822 int
823 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
824 {
825 struct freebsd11_statfs osb;
826 struct statfs *sfp;
827 fhandle_t fh;
828 int error;
829
830 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
831 if (error)
832 return (error);
833 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
834 error = kern_fhstatfs(td, fh, sfp);
835 if (error == 0) {
836 freebsd11_cvtstatfs(sfp, &osb);
837 error = copyout(&osb, uap->buf, sizeof(osb));
838 }
839 free(sfp, M_STATFS);
840 return (error);
841 }
842
843 /*
844 * Convert a new format statfs structure to an old format statfs structure.
845 */
846 static void
847 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
848 {
849
850 bzero(osp, sizeof(*osp));
851 osp->f_version = FREEBSD11_STATFS_VERSION;
852 osp->f_type = nsp->f_type;
853 osp->f_flags = nsp->f_flags;
854 osp->f_bsize = nsp->f_bsize;
855 osp->f_iosize = nsp->f_iosize;
856 osp->f_blocks = nsp->f_blocks;
857 osp->f_bfree = nsp->f_bfree;
858 osp->f_bavail = nsp->f_bavail;
859 osp->f_files = nsp->f_files;
860 osp->f_ffree = nsp->f_ffree;
861 osp->f_syncwrites = nsp->f_syncwrites;
862 osp->f_asyncwrites = nsp->f_asyncwrites;
863 osp->f_syncreads = nsp->f_syncreads;
864 osp->f_asyncreads = nsp->f_asyncreads;
865 osp->f_namemax = nsp->f_namemax;
866 osp->f_owner = nsp->f_owner;
867 osp->f_fsid = nsp->f_fsid;
868 strlcpy(osp->f_fstypename, nsp->f_fstypename,
869 MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
870 strlcpy(osp->f_mntonname, nsp->f_mntonname,
871 MIN(MNAMELEN, sizeof(osp->f_mntonname)));
872 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
873 MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
874 }
875 #endif /* COMPAT_FREEBSD11 */
876
877 /*
878 * Change current working directory to a given file descriptor.
879 */
880 #ifndef _SYS_SYSPROTO_H_
881 struct fchdir_args {
882 int fd;
883 };
884 #endif
885 int
886 sys_fchdir(struct thread *td, struct fchdir_args *uap)
887 {
888 struct vnode *vp, *tdp;
889 struct mount *mp;
890 struct file *fp;
891 int error;
892
893 AUDIT_ARG_FD(uap->fd);
894 error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
895 &fp);
896 if (error != 0)
897 return (error);
898 vp = fp->f_vnode;
899 vref(vp);
900 fdrop(fp, td);
901 vn_lock(vp, LK_SHARED | LK_RETRY);
902 AUDIT_ARG_VNODE1(vp);
903 error = change_dir(vp, td);
904 while (!error && (mp = vp->v_mountedhere) != NULL) {
905 if (vfs_busy(mp, 0))
906 continue;
907 error = VFS_ROOT(mp, LK_SHARED, &tdp);
908 vfs_unbusy(mp);
909 if (error != 0)
910 break;
911 vput(vp);
912 vp = tdp;
913 }
914 if (error != 0) {
915 vput(vp);
916 return (error);
917 }
918 VOP_UNLOCK(vp);
919 pwd_chdir(td, vp);
920 return (0);
921 }
922
923 /*
924 * Change current working directory (``.'').
925 */
926 #ifndef _SYS_SYSPROTO_H_
927 struct chdir_args {
928 char *path;
929 };
930 #endif
931 int
932 sys_chdir(struct thread *td, struct chdir_args *uap)
933 {
934
935 return (kern_chdir(td, uap->path, UIO_USERSPACE));
936 }
937
938 int
939 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
940 {
941 struct nameidata nd;
942 int error;
943
944 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
945 pathseg, path, td);
946 if ((error = namei(&nd)) != 0)
947 return (error);
948 if ((error = change_dir(nd.ni_vp, td)) != 0) {
949 vput(nd.ni_vp);
950 NDFREE_NOTHING(&nd);
951 return (error);
952 }
953 VOP_UNLOCK(nd.ni_vp);
954 NDFREE_NOTHING(&nd);
955 pwd_chdir(td, nd.ni_vp);
956 return (0);
957 }
958
959 /*
960 * Change notion of root (``/'') directory.
961 */
962 #ifndef _SYS_SYSPROTO_H_
963 struct chroot_args {
964 char *path;
965 };
966 #endif
967 int
968 sys_chroot(struct thread *td, struct chroot_args *uap)
969 {
970 struct nameidata nd;
971 int error;
972
973 error = priv_check(td, PRIV_VFS_CHROOT);
974 if (error != 0)
975 return (error);
976 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
977 UIO_USERSPACE, uap->path, td);
978 error = namei(&nd);
979 if (error != 0)
980 goto error;
981 error = change_dir(nd.ni_vp, td);
982 if (error != 0)
983 goto e_vunlock;
984 #ifdef MAC
985 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
986 if (error != 0)
987 goto e_vunlock;
988 #endif
989 VOP_UNLOCK(nd.ni_vp);
990 error = pwd_chroot(td, nd.ni_vp);
991 vrele(nd.ni_vp);
992 NDFREE_NOTHING(&nd);
993 return (error);
994 e_vunlock:
995 vput(nd.ni_vp);
996 error:
997 NDFREE_NOTHING(&nd);
998 return (error);
999 }
1000
1001 /*
1002 * Common routine for chroot and chdir. Callers must provide a locked vnode
1003 * instance.
1004 */
1005 int
1006 change_dir(struct vnode *vp, struct thread *td)
1007 {
1008 #ifdef MAC
1009 int error;
1010 #endif
1011
1012 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1013 if (vp->v_type != VDIR)
1014 return (ENOTDIR);
1015 #ifdef MAC
1016 error = mac_vnode_check_chdir(td->td_ucred, vp);
1017 if (error != 0)
1018 return (error);
1019 #endif
1020 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1021 }
1022
1023 static __inline void
1024 flags_to_rights(int flags, cap_rights_t *rightsp)
1025 {
1026 if (flags & O_EXEC) {
1027 cap_rights_set_one(rightsp, CAP_FEXECVE);
1028 if (flags & O_PATH)
1029 return;
1030 } else {
1031 switch ((flags & O_ACCMODE)) {
1032 case O_RDONLY:
1033 cap_rights_set_one(rightsp, CAP_READ);
1034 break;
1035 case O_RDWR:
1036 cap_rights_set_one(rightsp, CAP_READ);
1037 /* FALLTHROUGH */
1038 case O_WRONLY:
1039 cap_rights_set_one(rightsp, CAP_WRITE);
1040 if (!(flags & (O_APPEND | O_TRUNC)))
1041 cap_rights_set_one(rightsp, CAP_SEEK);
1042 break;
1043 }
1044 }
1045
1046 if (flags & O_CREAT)
1047 cap_rights_set_one(rightsp, CAP_CREATE);
1048
1049 if (flags & O_TRUNC)
1050 cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1051
1052 if (flags & (O_SYNC | O_FSYNC))
1053 cap_rights_set_one(rightsp, CAP_FSYNC);
1054
1055 if (flags & (O_EXLOCK | O_SHLOCK))
1056 cap_rights_set_one(rightsp, CAP_FLOCK);
1057 }
1058
1059 /*
1060 * Check permissions, allocate an open file structure, and call the device
1061 * open routine if any.
1062 */
1063 #ifndef _SYS_SYSPROTO_H_
1064 struct open_args {
1065 char *path;
1066 int flags;
1067 int mode;
1068 };
1069 #endif
1070 int
1071 sys_open(struct thread *td, struct open_args *uap)
1072 {
1073
1074 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1075 uap->flags, uap->mode));
1076 }
1077
1078 #ifndef _SYS_SYSPROTO_H_
1079 struct openat_args {
1080 int fd;
1081 char *path;
1082 int flag;
1083 int mode;
1084 };
1085 #endif
1086 int
1087 sys_openat(struct thread *td, struct openat_args *uap)
1088 {
1089
1090 AUDIT_ARG_FD(uap->fd);
1091 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1092 uap->mode));
1093 }
1094
1095 int
1096 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1097 int flags, int mode)
1098 {
1099 struct proc *p = td->td_proc;
1100 struct filedesc *fdp;
1101 struct pwddesc *pdp;
1102 struct file *fp;
1103 struct vnode *vp;
1104 struct nameidata nd;
1105 cap_rights_t rights;
1106 int cmode, error, indx;
1107
1108 indx = -1;
1109 fdp = p->p_fd;
1110 pdp = p->p_pd;
1111
1112 AUDIT_ARG_FFLAGS(flags);
1113 AUDIT_ARG_MODE(mode);
1114 cap_rights_init_one(&rights, CAP_LOOKUP);
1115 flags_to_rights(flags, &rights);
1116
1117 /*
1118 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1119 * may be specified. On the other hand, for O_PATH any mode
1120 * except O_EXEC is ignored.
1121 */
1122 if ((flags & O_PATH) != 0) {
1123 flags &= ~(O_CREAT | O_ACCMODE);
1124 } else if ((flags & O_EXEC) != 0) {
1125 if (flags & O_ACCMODE)
1126 return (EINVAL);
1127 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1128 return (EINVAL);
1129 } else {
1130 flags = FFLAGS(flags);
1131 }
1132
1133 /*
1134 * Allocate a file structure. The descriptor to reference it
1135 * is allocated and used by finstall_refed() below.
1136 */
1137 error = falloc_noinstall(td, &fp);
1138 if (error != 0)
1139 return (error);
1140 /* Set the flags early so the finit in devfs can pick them up. */
1141 fp->f_flag = flags & FMASK;
1142 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1143 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1144 &rights, td);
1145 td->td_dupfd = -1; /* XXX check for fdopen */
1146 error = vn_open(&nd, &flags, cmode, fp);
1147 if (error != 0) {
1148 /*
1149 * If the vn_open replaced the method vector, something
1150 * wonderous happened deep below and we just pass it up
1151 * pretending we know what we do.
1152 */
1153 if (error == ENXIO && fp->f_ops != &badfileops) {
1154 MPASS((flags & O_PATH) == 0);
1155 goto success;
1156 }
1157
1158 /*
1159 * Handle special fdopen() case. bleh.
1160 *
1161 * Don't do this for relative (capability) lookups; we don't
1162 * understand exactly what would happen, and we don't think
1163 * that it ever should.
1164 */
1165 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1166 (error == ENODEV || error == ENXIO) &&
1167 td->td_dupfd >= 0) {
1168 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1169 &indx);
1170 if (error == 0)
1171 goto success;
1172 }
1173
1174 goto bad;
1175 }
1176 td->td_dupfd = 0;
1177 NDFREE(&nd, NDF_ONLY_PNBUF);
1178 vp = nd.ni_vp;
1179
1180 /*
1181 * Store the vnode, for any f_type. Typically, the vnode use
1182 * count is decremented by direct call to vn_closefile() for
1183 * files that switched type in the cdevsw fdopen() method.
1184 */
1185 fp->f_vnode = vp;
1186
1187 /*
1188 * If the file wasn't claimed by devfs bind it to the normal
1189 * vnode operations here.
1190 */
1191 if (fp->f_ops == &badfileops) {
1192 KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1193 ("Unexpected fifo fp %p vp %p", fp, vp));
1194 if ((flags & O_PATH) != 0) {
1195 finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1196 DTYPE_VNODE, NULL, &path_fileops);
1197 vhold(vp);
1198 vunref(vp);
1199 } else {
1200 finit_vnode(fp, flags, NULL, &vnops);
1201 }
1202 }
1203
1204 VOP_UNLOCK(vp);
1205 if (flags & O_TRUNC) {
1206 error = fo_truncate(fp, 0, td->td_ucred, td);
1207 if (error != 0)
1208 goto bad;
1209 }
1210 success:
1211 /*
1212 * If we haven't already installed the FD (for dupfdopen), do so now.
1213 */
1214 if (indx == -1) {
1215 struct filecaps *fcaps;
1216
1217 #ifdef CAPABILITIES
1218 if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1219 fcaps = &nd.ni_filecaps;
1220 else
1221 #endif
1222 fcaps = NULL;
1223 error = finstall_refed(td, fp, &indx, flags, fcaps);
1224 /* On success finstall_refed() consumes fcaps. */
1225 if (error != 0) {
1226 filecaps_free(&nd.ni_filecaps);
1227 goto bad;
1228 }
1229 } else {
1230 filecaps_free(&nd.ni_filecaps);
1231 falloc_abort(td, fp);
1232 }
1233
1234 td->td_retval[0] = indx;
1235 return (0);
1236 bad:
1237 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1238 falloc_abort(td, fp);
1239 return (error);
1240 }
1241
1242 #ifdef COMPAT_43
1243 /*
1244 * Create a file.
1245 */
1246 #ifndef _SYS_SYSPROTO_H_
1247 struct ocreat_args {
1248 char *path;
1249 int mode;
1250 };
1251 #endif
1252 int
1253 ocreat(struct thread *td, struct ocreat_args *uap)
1254 {
1255
1256 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1257 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1258 }
1259 #endif /* COMPAT_43 */
1260
1261 /*
1262 * Create a special file.
1263 */
1264 #ifndef _SYS_SYSPROTO_H_
1265 struct mknodat_args {
1266 int fd;
1267 char *path;
1268 mode_t mode;
1269 dev_t dev;
1270 };
1271 #endif
1272 int
1273 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1274 {
1275
1276 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1277 uap->dev));
1278 }
1279
1280 #if defined(COMPAT_FREEBSD11)
1281 int
1282 freebsd11_mknod(struct thread *td,
1283 struct freebsd11_mknod_args *uap)
1284 {
1285
1286 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1287 uap->mode, uap->dev));
1288 }
1289
1290 int
1291 freebsd11_mknodat(struct thread *td,
1292 struct freebsd11_mknodat_args *uap)
1293 {
1294
1295 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1296 uap->dev));
1297 }
1298 #endif /* COMPAT_FREEBSD11 */
1299
1300 int
1301 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1302 int mode, dev_t dev)
1303 {
1304 struct vnode *vp;
1305 struct mount *mp;
1306 struct vattr vattr;
1307 struct nameidata nd;
1308 int error, whiteout = 0;
1309
1310 AUDIT_ARG_MODE(mode);
1311 AUDIT_ARG_DEV(dev);
1312 switch (mode & S_IFMT) {
1313 case S_IFCHR:
1314 case S_IFBLK:
1315 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1316 if (error == 0 && dev == VNOVAL)
1317 error = EINVAL;
1318 break;
1319 case S_IFWHT:
1320 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1321 break;
1322 case S_IFIFO:
1323 if (dev == 0)
1324 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1325 /* FALLTHROUGH */
1326 default:
1327 error = EINVAL;
1328 break;
1329 }
1330 if (error != 0)
1331 return (error);
1332 NDPREINIT(&nd);
1333 restart:
1334 bwillwrite();
1335 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1336 NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1337 td);
1338 if ((error = namei(&nd)) != 0)
1339 return (error);
1340 vp = nd.ni_vp;
1341 if (vp != NULL) {
1342 NDFREE(&nd, NDF_ONLY_PNBUF);
1343 if (vp == nd.ni_dvp)
1344 vrele(nd.ni_dvp);
1345 else
1346 vput(nd.ni_dvp);
1347 vrele(vp);
1348 return (EEXIST);
1349 } else {
1350 VATTR_NULL(&vattr);
1351 vattr.va_mode = (mode & ALLPERMS) &
1352 ~td->td_proc->p_pd->pd_cmask;
1353 vattr.va_rdev = dev;
1354 whiteout = 0;
1355
1356 switch (mode & S_IFMT) {
1357 case S_IFCHR:
1358 vattr.va_type = VCHR;
1359 break;
1360 case S_IFBLK:
1361 vattr.va_type = VBLK;
1362 break;
1363 case S_IFWHT:
1364 whiteout = 1;
1365 break;
1366 default:
1367 panic("kern_mknod: invalid mode");
1368 }
1369 }
1370 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1371 NDFREE(&nd, NDF_ONLY_PNBUF);
1372 vput(nd.ni_dvp);
1373 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1374 return (error);
1375 goto restart;
1376 }
1377 #ifdef MAC
1378 if (error == 0 && !whiteout)
1379 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1380 &nd.ni_cnd, &vattr);
1381 #endif
1382 if (error == 0) {
1383 if (whiteout)
1384 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1385 else {
1386 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1387 &nd.ni_cnd, &vattr);
1388 }
1389 }
1390 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1391 true);
1392 vn_finished_write(mp);
1393 NDFREE(&nd, NDF_ONLY_PNBUF);
1394 if (error == ERELOOKUP)
1395 goto restart;
1396 return (error);
1397 }
1398
1399 /*
1400 * Create a named pipe.
1401 */
1402 #ifndef _SYS_SYSPROTO_H_
1403 struct mkfifo_args {
1404 char *path;
1405 int mode;
1406 };
1407 #endif
1408 int
1409 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1410 {
1411
1412 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1413 uap->mode));
1414 }
1415
1416 #ifndef _SYS_SYSPROTO_H_
1417 struct mkfifoat_args {
1418 int fd;
1419 char *path;
1420 mode_t mode;
1421 };
1422 #endif
1423 int
1424 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1425 {
1426
1427 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1428 uap->mode));
1429 }
1430
1431 int
1432 kern_mkfifoat(struct thread *td, int fd, const char *path,
1433 enum uio_seg pathseg, int mode)
1434 {
1435 struct mount *mp;
1436 struct vattr vattr;
1437 struct nameidata nd;
1438 int error;
1439
1440 AUDIT_ARG_MODE(mode);
1441 NDPREINIT(&nd);
1442 restart:
1443 bwillwrite();
1444 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1445 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1446 td);
1447 if ((error = namei(&nd)) != 0)
1448 return (error);
1449 if (nd.ni_vp != NULL) {
1450 NDFREE(&nd, NDF_ONLY_PNBUF);
1451 if (nd.ni_vp == nd.ni_dvp)
1452 vrele(nd.ni_dvp);
1453 else
1454 vput(nd.ni_dvp);
1455 vrele(nd.ni_vp);
1456 return (EEXIST);
1457 }
1458 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1459 NDFREE(&nd, NDF_ONLY_PNBUF);
1460 vput(nd.ni_dvp);
1461 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1462 return (error);
1463 goto restart;
1464 }
1465 VATTR_NULL(&vattr);
1466 vattr.va_type = VFIFO;
1467 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1468 #ifdef MAC
1469 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1470 &vattr);
1471 if (error != 0)
1472 goto out;
1473 #endif
1474 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1475 #ifdef MAC
1476 out:
1477 #endif
1478 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1479 vn_finished_write(mp);
1480 NDFREE(&nd, NDF_ONLY_PNBUF);
1481 if (error == ERELOOKUP)
1482 goto restart;
1483 return (error);
1484 }
1485
1486 /*
1487 * Make a hard file link.
1488 */
1489 #ifndef _SYS_SYSPROTO_H_
1490 struct link_args {
1491 char *path;
1492 char *link;
1493 };
1494 #endif
1495 int
1496 sys_link(struct thread *td, struct link_args *uap)
1497 {
1498
1499 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1500 UIO_USERSPACE, FOLLOW));
1501 }
1502
1503 #ifndef _SYS_SYSPROTO_H_
1504 struct linkat_args {
1505 int fd1;
1506 char *path1;
1507 int fd2;
1508 char *path2;
1509 int flag;
1510 };
1511 #endif
1512 int
1513 sys_linkat(struct thread *td, struct linkat_args *uap)
1514 {
1515 int flag;
1516
1517 flag = uap->flag;
1518 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1519 AT_EMPTY_PATH)) != 0)
1520 return (EINVAL);
1521
1522 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1523 UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW |
1524 AT_RESOLVE_BENEATH | AT_EMPTY_PATH)));
1525 }
1526
1527 int hardlink_check_uid = 0;
1528 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1529 &hardlink_check_uid, 0,
1530 "Unprivileged processes cannot create hard links to files owned by other "
1531 "users");
1532 static int hardlink_check_gid = 0;
1533 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1534 &hardlink_check_gid, 0,
1535 "Unprivileged processes cannot create hard links to files owned by other "
1536 "groups");
1537
1538 static int
1539 can_hardlink(struct vnode *vp, struct ucred *cred)
1540 {
1541 struct vattr va;
1542 int error;
1543
1544 if (!hardlink_check_uid && !hardlink_check_gid)
1545 return (0);
1546
1547 error = VOP_GETATTR(vp, &va, cred);
1548 if (error != 0)
1549 return (error);
1550
1551 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1552 error = priv_check_cred(cred, PRIV_VFS_LINK);
1553 if (error != 0)
1554 return (error);
1555 }
1556
1557 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1558 error = priv_check_cred(cred, PRIV_VFS_LINK);
1559 if (error != 0)
1560 return (error);
1561 }
1562
1563 return (0);
1564 }
1565
1566 int
1567 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1568 const char *path2, enum uio_seg segflag, int follow)
1569 {
1570 struct nameidata nd;
1571 int error;
1572
1573 NDPREINIT(&nd);
1574 do {
1575 bwillwrite();
1576 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1577 path1, fd1, &cap_linkat_source_rights, td);
1578 if ((error = namei(&nd)) != 0)
1579 return (error);
1580 NDFREE(&nd, NDF_ONLY_PNBUF);
1581 if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1582 error = priv_check(td, PRIV_VFS_FHOPEN);
1583 if (error != 0) {
1584 vrele(nd.ni_vp);
1585 return (error);
1586 }
1587 }
1588 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1589 } while (error == EAGAIN || error == ERELOOKUP);
1590 return (error);
1591 }
1592
1593 static int
1594 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1595 enum uio_seg segflag)
1596 {
1597 struct nameidata nd;
1598 struct mount *mp;
1599 int error;
1600
1601 if (vp->v_type == VDIR) {
1602 vrele(vp);
1603 return (EPERM); /* POSIX */
1604 }
1605 NDINIT_ATRIGHTS(&nd, CREATE,
1606 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1607 &cap_linkat_target_rights, td);
1608 if ((error = namei(&nd)) == 0) {
1609 if (nd.ni_vp != NULL) {
1610 NDFREE(&nd, NDF_ONLY_PNBUF);
1611 if (nd.ni_dvp == nd.ni_vp)
1612 vrele(nd.ni_dvp);
1613 else
1614 vput(nd.ni_dvp);
1615 vrele(nd.ni_vp);
1616 vrele(vp);
1617 return (EEXIST);
1618 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1619 /*
1620 * Cross-device link. No need to recheck
1621 * vp->v_type, since it cannot change, except
1622 * to VBAD.
1623 */
1624 NDFREE(&nd, NDF_ONLY_PNBUF);
1625 vput(nd.ni_dvp);
1626 vrele(vp);
1627 return (EXDEV);
1628 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1629 error = can_hardlink(vp, td->td_ucred);
1630 #ifdef MAC
1631 if (error == 0)
1632 error = mac_vnode_check_link(td->td_ucred,
1633 nd.ni_dvp, vp, &nd.ni_cnd);
1634 #endif
1635 if (error != 0) {
1636 vput(vp);
1637 vput(nd.ni_dvp);
1638 NDFREE(&nd, NDF_ONLY_PNBUF);
1639 return (error);
1640 }
1641 error = vn_start_write(vp, &mp, V_NOWAIT);
1642 if (error != 0) {
1643 vput(vp);
1644 vput(nd.ni_dvp);
1645 NDFREE(&nd, NDF_ONLY_PNBUF);
1646 error = vn_start_write(NULL, &mp,
1647 V_XSLEEP | PCATCH);
1648 if (error != 0)
1649 return (error);
1650 return (EAGAIN);
1651 }
1652 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1653 VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1654 vn_finished_write(mp);
1655 NDFREE(&nd, NDF_ONLY_PNBUF);
1656 vp = NULL;
1657 } else {
1658 vput(nd.ni_dvp);
1659 NDFREE(&nd, NDF_ONLY_PNBUF);
1660 vrele(vp);
1661 return (EAGAIN);
1662 }
1663 }
1664 if (vp != NULL)
1665 vrele(vp);
1666 return (error);
1667 }
1668
1669 /*
1670 * Make a symbolic link.
1671 */
1672 #ifndef _SYS_SYSPROTO_H_
1673 struct symlink_args {
1674 char *path;
1675 char *link;
1676 };
1677 #endif
1678 int
1679 sys_symlink(struct thread *td, struct symlink_args *uap)
1680 {
1681
1682 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1683 UIO_USERSPACE));
1684 }
1685
1686 #ifndef _SYS_SYSPROTO_H_
1687 struct symlinkat_args {
1688 char *path;
1689 int fd;
1690 char *path2;
1691 };
1692 #endif
1693 int
1694 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1695 {
1696
1697 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1698 UIO_USERSPACE));
1699 }
1700
1701 int
1702 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1703 enum uio_seg segflg)
1704 {
1705 struct mount *mp;
1706 struct vattr vattr;
1707 const char *syspath;
1708 char *tmppath;
1709 struct nameidata nd;
1710 int error;
1711
1712 if (segflg == UIO_SYSSPACE) {
1713 syspath = path1;
1714 } else {
1715 tmppath = uma_zalloc(namei_zone, M_WAITOK);
1716 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1717 goto out;
1718 syspath = tmppath;
1719 }
1720 AUDIT_ARG_TEXT(syspath);
1721 NDPREINIT(&nd);
1722 restart:
1723 bwillwrite();
1724 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1725 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1726 td);
1727 if ((error = namei(&nd)) != 0)
1728 goto out;
1729 if (nd.ni_vp) {
1730 NDFREE(&nd, NDF_ONLY_PNBUF);
1731 if (nd.ni_vp == nd.ni_dvp)
1732 vrele(nd.ni_dvp);
1733 else
1734 vput(nd.ni_dvp);
1735 vrele(nd.ni_vp);
1736 nd.ni_vp = NULL;
1737 error = EEXIST;
1738 goto out;
1739 }
1740 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1741 NDFREE(&nd, NDF_ONLY_PNBUF);
1742 vput(nd.ni_dvp);
1743 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1744 goto out;
1745 goto restart;
1746 }
1747 VATTR_NULL(&vattr);
1748 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1749 #ifdef MAC
1750 vattr.va_type = VLNK;
1751 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1752 &vattr);
1753 if (error != 0)
1754 goto out2;
1755 #endif
1756 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1757 #ifdef MAC
1758 out2:
1759 #endif
1760 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1761 vn_finished_write(mp);
1762 NDFREE(&nd, NDF_ONLY_PNBUF);
1763 if (error == ERELOOKUP)
1764 goto restart;
1765 out:
1766 if (segflg != UIO_SYSSPACE)
1767 uma_zfree(namei_zone, tmppath);
1768 return (error);
1769 }
1770
1771 /*
1772 * Delete a whiteout from the filesystem.
1773 */
1774 #ifndef _SYS_SYSPROTO_H_
1775 struct undelete_args {
1776 char *path;
1777 };
1778 #endif
1779 int
1780 sys_undelete(struct thread *td, struct undelete_args *uap)
1781 {
1782 struct mount *mp;
1783 struct nameidata nd;
1784 int error;
1785
1786 NDPREINIT(&nd);
1787 restart:
1788 bwillwrite();
1789 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1790 UIO_USERSPACE, uap->path, td);
1791 error = namei(&nd);
1792 if (error != 0)
1793 return (error);
1794
1795 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1796 NDFREE(&nd, NDF_ONLY_PNBUF);
1797 if (nd.ni_vp == nd.ni_dvp)
1798 vrele(nd.ni_dvp);
1799 else
1800 vput(nd.ni_dvp);
1801 if (nd.ni_vp)
1802 vrele(nd.ni_vp);
1803 return (EEXIST);
1804 }
1805 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1806 NDFREE(&nd, NDF_ONLY_PNBUF);
1807 vput(nd.ni_dvp);
1808 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1809 return (error);
1810 goto restart;
1811 }
1812 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1813 NDFREE(&nd, NDF_ONLY_PNBUF);
1814 vput(nd.ni_dvp);
1815 vn_finished_write(mp);
1816 if (error == ERELOOKUP)
1817 goto restart;
1818 return (error);
1819 }
1820
1821 /*
1822 * Delete a name from the filesystem.
1823 */
1824 #ifndef _SYS_SYSPROTO_H_
1825 struct unlink_args {
1826 char *path;
1827 };
1828 #endif
1829 int
1830 sys_unlink(struct thread *td, struct unlink_args *uap)
1831 {
1832
1833 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1834 0, 0));
1835 }
1836
1837 static int
1838 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1839 int flag, enum uio_seg pathseg, ino_t oldinum)
1840 {
1841
1842 if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1843 return (EINVAL);
1844
1845 if ((flag & AT_REMOVEDIR) != 0)
1846 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1847
1848 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1849 }
1850
1851 #ifndef _SYS_SYSPROTO_H_
1852 struct unlinkat_args {
1853 int fd;
1854 char *path;
1855 int flag;
1856 };
1857 #endif
1858 int
1859 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1860 {
1861
1862 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1863 UIO_USERSPACE, 0));
1864 }
1865
1866 #ifndef _SYS_SYSPROTO_H_
1867 struct funlinkat_args {
1868 int dfd;
1869 const char *path;
1870 int fd;
1871 int flag;
1872 };
1873 #endif
1874 int
1875 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1876 {
1877
1878 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1879 UIO_USERSPACE, 0));
1880 }
1881
1882 int
1883 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1884 enum uio_seg pathseg, int flag, ino_t oldinum)
1885 {
1886 struct mount *mp;
1887 struct file *fp;
1888 struct vnode *vp;
1889 struct nameidata nd;
1890 struct stat sb;
1891 int error;
1892
1893 fp = NULL;
1894 if (fd != FD_NONE) {
1895 error = getvnode_path(td, fd, &cap_no_rights, &fp);
1896 if (error != 0)
1897 return (error);
1898 }
1899
1900 NDPREINIT(&nd);
1901 restart:
1902 bwillwrite();
1903 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1904 at2cnpflags(flag, AT_RESOLVE_BENEATH),
1905 pathseg, path, dfd, &cap_unlinkat_rights, td);
1906 if ((error = namei(&nd)) != 0) {
1907 if (error == EINVAL)
1908 error = EPERM;
1909 goto fdout;
1910 }
1911 vp = nd.ni_vp;
1912 if (vp->v_type == VDIR && oldinum == 0) {
1913 error = EPERM; /* POSIX */
1914 } else if (oldinum != 0 &&
1915 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1916 sb.st_ino != oldinum) {
1917 error = EIDRM; /* Identifier removed */
1918 } else if (fp != NULL && fp->f_vnode != vp) {
1919 if (VN_IS_DOOMED(fp->f_vnode))
1920 error = EBADF;
1921 else
1922 error = EDEADLK;
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 if ((error = vn_start_write(NULL, &mp,
1941 V_XSLEEP | PCATCH)) != 0) {
1942 goto fdout;
1943 }
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 != 0)
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 if (error == ERELOOKUP)
1966 goto restart;
1967 fdout:
1968 if (fp != NULL)
1969 fdrop(fp, td);
1970 return (error);
1971 }
1972
1973 /*
1974 * Reposition read/write file offset.
1975 */
1976 #ifndef _SYS_SYSPROTO_H_
1977 struct lseek_args {
1978 int fd;
1979 int pad;
1980 off_t offset;
1981 int whence;
1982 };
1983 #endif
1984 int
1985 sys_lseek(struct thread *td, struct lseek_args *uap)
1986 {
1987
1988 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1989 }
1990
1991 int
1992 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1993 {
1994 struct file *fp;
1995 int error;
1996
1997 AUDIT_ARG_FD(fd);
1998 error = fget(td, fd, &cap_seek_rights, &fp);
1999 if (error != 0)
2000 return (error);
2001 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2002 fo_seek(fp, offset, whence, td) : ESPIPE;
2003 fdrop(fp, td);
2004 return (error);
2005 }
2006
2007 #if defined(COMPAT_43)
2008 /*
2009 * Reposition read/write file offset.
2010 */
2011 #ifndef _SYS_SYSPROTO_H_
2012 struct olseek_args {
2013 int fd;
2014 long offset;
2015 int whence;
2016 };
2017 #endif
2018 int
2019 olseek(struct thread *td, struct olseek_args *uap)
2020 {
2021
2022 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2023 }
2024 #endif /* COMPAT_43 */
2025
2026 #if defined(COMPAT_FREEBSD6)
2027 /* Version with the 'pad' argument */
2028 int
2029 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2030 {
2031
2032 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2033 }
2034 #endif
2035
2036 /*
2037 * Check access permissions using passed credentials.
2038 */
2039 static int
2040 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2041 struct thread *td)
2042 {
2043 accmode_t accmode;
2044 int error;
2045
2046 /* Flags == 0 means only check for existence. */
2047 if (user_flags == 0)
2048 return (0);
2049
2050 accmode = 0;
2051 if (user_flags & R_OK)
2052 accmode |= VREAD;
2053 if (user_flags & W_OK)
2054 accmode |= VWRITE;
2055 if (user_flags & X_OK)
2056 accmode |= VEXEC;
2057 #ifdef MAC
2058 error = mac_vnode_check_access(cred, vp, accmode);
2059 if (error != 0)
2060 return (error);
2061 #endif
2062 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2063 error = VOP_ACCESS(vp, accmode, cred, td);
2064 return (error);
2065 }
2066
2067 /*
2068 * Check access permissions using "real" credentials.
2069 */
2070 #ifndef _SYS_SYSPROTO_H_
2071 struct access_args {
2072 char *path;
2073 int amode;
2074 };
2075 #endif
2076 int
2077 sys_access(struct thread *td, struct access_args *uap)
2078 {
2079
2080 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2081 0, uap->amode));
2082 }
2083
2084 #ifndef _SYS_SYSPROTO_H_
2085 struct faccessat_args {
2086 int dirfd;
2087 char *path;
2088 int amode;
2089 int flag;
2090 }
2091 #endif
2092 int
2093 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2094 {
2095
2096 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2097 uap->amode));
2098 }
2099
2100 int
2101 kern_accessat(struct thread *td, int fd, const char *path,
2102 enum uio_seg pathseg, int flag, int amode)
2103 {
2104 struct ucred *cred, *usecred;
2105 struct vnode *vp;
2106 struct nameidata nd;
2107 int error;
2108
2109 if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2110 return (EINVAL);
2111 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2112 return (EINVAL);
2113
2114 /*
2115 * Create and modify a temporary credential instead of one that
2116 * is potentially shared (if we need one).
2117 */
2118 cred = td->td_ucred;
2119 if ((flag & AT_EACCESS) == 0 &&
2120 ((cred->cr_uid != cred->cr_ruid ||
2121 cred->cr_rgid != cred->cr_groups[0]))) {
2122 usecred = crdup(cred);
2123 usecred->cr_uid = cred->cr_ruid;
2124 usecred->cr_groups[0] = cred->cr_rgid;
2125 td->td_ucred = usecred;
2126 } else
2127 usecred = cred;
2128 AUDIT_ARG_VALUE(amode);
2129 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2130 AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2131 AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights, td);
2132 if ((error = namei(&nd)) != 0)
2133 goto out;
2134 vp = nd.ni_vp;
2135
2136 error = vn_access(vp, amode, usecred, td);
2137 NDFREE_NOTHING(&nd);
2138 vput(vp);
2139 out:
2140 if (usecred != cred) {
2141 td->td_ucred = cred;
2142 crfree(usecred);
2143 }
2144 return (error);
2145 }
2146
2147 /*
2148 * Check access permissions using "effective" credentials.
2149 */
2150 #ifndef _SYS_SYSPROTO_H_
2151 struct eaccess_args {
2152 char *path;
2153 int amode;
2154 };
2155 #endif
2156 int
2157 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2158 {
2159
2160 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2161 AT_EACCESS, uap->amode));
2162 }
2163
2164 #if defined(COMPAT_43)
2165 /*
2166 * Get file status; this version follows links.
2167 */
2168 #ifndef _SYS_SYSPROTO_H_
2169 struct ostat_args {
2170 char *path;
2171 struct ostat *ub;
2172 };
2173 #endif
2174 int
2175 ostat(struct thread *td, struct ostat_args *uap)
2176 {
2177 struct stat sb;
2178 struct ostat osb;
2179 int error;
2180
2181 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2182 &sb, NULL);
2183 if (error != 0)
2184 return (error);
2185 cvtstat(&sb, &osb);
2186 return (copyout(&osb, uap->ub, sizeof (osb)));
2187 }
2188
2189 /*
2190 * Get file status; this version does not follow links.
2191 */
2192 #ifndef _SYS_SYSPROTO_H_
2193 struct olstat_args {
2194 char *path;
2195 struct ostat *ub;
2196 };
2197 #endif
2198 int
2199 olstat(struct thread *td, struct olstat_args *uap)
2200 {
2201 struct stat sb;
2202 struct ostat osb;
2203 int error;
2204
2205 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2206 UIO_USERSPACE, &sb, NULL);
2207 if (error != 0)
2208 return (error);
2209 cvtstat(&sb, &osb);
2210 return (copyout(&osb, uap->ub, sizeof (osb)));
2211 }
2212
2213 /*
2214 * Convert from an old to a new stat structure.
2215 * XXX: many values are blindly truncated.
2216 */
2217 void
2218 cvtstat(struct stat *st, struct ostat *ost)
2219 {
2220
2221 bzero(ost, sizeof(*ost));
2222 ost->st_dev = st->st_dev;
2223 ost->st_ino = st->st_ino;
2224 ost->st_mode = st->st_mode;
2225 ost->st_nlink = st->st_nlink;
2226 ost->st_uid = st->st_uid;
2227 ost->st_gid = st->st_gid;
2228 ost->st_rdev = st->st_rdev;
2229 ost->st_size = MIN(st->st_size, INT32_MAX);
2230 ost->st_atim = st->st_atim;
2231 ost->st_mtim = st->st_mtim;
2232 ost->st_ctim = st->st_ctim;
2233 ost->st_blksize = st->st_blksize;
2234 ost->st_blocks = st->st_blocks;
2235 ost->st_flags = st->st_flags;
2236 ost->st_gen = st->st_gen;
2237 }
2238 #endif /* COMPAT_43 */
2239
2240 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2241 int ino64_trunc_error;
2242 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2243 &ino64_trunc_error, 0,
2244 "Error on truncation of device, file or inode number, or link count");
2245
2246 int
2247 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2248 {
2249
2250 ost->st_dev = st->st_dev;
2251 if (ost->st_dev != st->st_dev) {
2252 switch (ino64_trunc_error) {
2253 default:
2254 /*
2255 * Since dev_t is almost raw, don't clamp to the
2256 * maximum for case 2, but ignore the error.
2257 */
2258 break;
2259 case 1:
2260 return (EOVERFLOW);
2261 }
2262 }
2263 ost->st_ino = st->st_ino;
2264 if (ost->st_ino != st->st_ino) {
2265 switch (ino64_trunc_error) {
2266 default:
2267 case 0:
2268 break;
2269 case 1:
2270 return (EOVERFLOW);
2271 case 2:
2272 ost->st_ino = UINT32_MAX;
2273 break;
2274 }
2275 }
2276 ost->st_mode = st->st_mode;
2277 ost->st_nlink = st->st_nlink;
2278 if (ost->st_nlink != st->st_nlink) {
2279 switch (ino64_trunc_error) {
2280 default:
2281 case 0:
2282 break;
2283 case 1:
2284 return (EOVERFLOW);
2285 case 2:
2286 ost->st_nlink = UINT16_MAX;
2287 break;
2288 }
2289 }
2290 ost->st_uid = st->st_uid;
2291 ost->st_gid = st->st_gid;
2292 ost->st_rdev = st->st_rdev;
2293 if (ost->st_rdev != st->st_rdev) {
2294 switch (ino64_trunc_error) {
2295 default:
2296 break;
2297 case 1:
2298 return (EOVERFLOW);
2299 }
2300 }
2301 ost->st_atim = st->st_atim;
2302 ost->st_mtim = st->st_mtim;
2303 ost->st_ctim = st->st_ctim;
2304 ost->st_size = st->st_size;
2305 ost->st_blocks = st->st_blocks;
2306 ost->st_blksize = st->st_blksize;
2307 ost->st_flags = st->st_flags;
2308 ost->st_gen = st->st_gen;
2309 ost->st_lspare = 0;
2310 ost->st_birthtim = st->st_birthtim;
2311 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2312 sizeof(*ost) - offsetof(struct freebsd11_stat,
2313 st_birthtim) - sizeof(ost->st_birthtim));
2314 return (0);
2315 }
2316
2317 int
2318 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2319 {
2320 struct stat sb;
2321 struct freebsd11_stat osb;
2322 int error;
2323
2324 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2325 &sb, NULL);
2326 if (error != 0)
2327 return (error);
2328 error = freebsd11_cvtstat(&sb, &osb);
2329 if (error == 0)
2330 error = copyout(&osb, uap->ub, sizeof(osb));
2331 return (error);
2332 }
2333
2334 int
2335 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2336 {
2337 struct stat sb;
2338 struct freebsd11_stat osb;
2339 int error;
2340
2341 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2342 UIO_USERSPACE, &sb, NULL);
2343 if (error != 0)
2344 return (error);
2345 error = freebsd11_cvtstat(&sb, &osb);
2346 if (error == 0)
2347 error = copyout(&osb, uap->ub, sizeof(osb));
2348 return (error);
2349 }
2350
2351 int
2352 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2353 {
2354 struct fhandle fh;
2355 struct stat sb;
2356 struct freebsd11_stat osb;
2357 int error;
2358
2359 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2360 if (error != 0)
2361 return (error);
2362 error = kern_fhstat(td, fh, &sb);
2363 if (error != 0)
2364 return (error);
2365 error = freebsd11_cvtstat(&sb, &osb);
2366 if (error == 0)
2367 error = copyout(&osb, uap->sb, sizeof(osb));
2368 return (error);
2369 }
2370
2371 int
2372 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2373 {
2374 struct stat sb;
2375 struct freebsd11_stat osb;
2376 int error;
2377
2378 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2379 UIO_USERSPACE, &sb, NULL);
2380 if (error != 0)
2381 return (error);
2382 error = freebsd11_cvtstat(&sb, &osb);
2383 if (error == 0)
2384 error = copyout(&osb, uap->buf, sizeof(osb));
2385 return (error);
2386 }
2387 #endif /* COMPAT_FREEBSD11 */
2388
2389 /*
2390 * Get file status
2391 */
2392 #ifndef _SYS_SYSPROTO_H_
2393 struct fstatat_args {
2394 int fd;
2395 char *path;
2396 struct stat *buf;
2397 int flag;
2398 }
2399 #endif
2400 int
2401 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2402 {
2403 struct stat sb;
2404 int error;
2405
2406 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2407 UIO_USERSPACE, &sb, NULL);
2408 if (error == 0)
2409 error = copyout(&sb, uap->buf, sizeof (sb));
2410 return (error);
2411 }
2412
2413 int
2414 kern_statat(struct thread *td, int flag, int fd, const char *path,
2415 enum uio_seg pathseg, struct stat *sbp,
2416 void (*hook)(struct vnode *vp, struct stat *sbp))
2417 {
2418 struct nameidata nd;
2419 int error;
2420
2421 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2422 AT_EMPTY_PATH)) != 0)
2423 return (EINVAL);
2424
2425 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2426 AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2427 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2428
2429 if ((error = namei(&nd)) != 0) {
2430 if (error == ENOTDIR &&
2431 (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2432 error = kern_fstat(td, fd, sbp);
2433 return (error);
2434 }
2435 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2436 if (error == 0) {
2437 if (__predict_false(hook != NULL))
2438 hook(nd.ni_vp, sbp);
2439 }
2440 NDFREE_NOTHING(&nd);
2441 vput(nd.ni_vp);
2442 #ifdef __STAT_TIME_T_EXT
2443 sbp->st_atim_ext = 0;
2444 sbp->st_mtim_ext = 0;
2445 sbp->st_ctim_ext = 0;
2446 sbp->st_btim_ext = 0;
2447 #endif
2448 #ifdef KTRACE
2449 if (KTRPOINT(td, KTR_STRUCT))
2450 ktrstat_error(sbp, error);
2451 #endif
2452 return (error);
2453 }
2454
2455 #if defined(COMPAT_FREEBSD11)
2456 /*
2457 * Implementation of the NetBSD [l]stat() functions.
2458 */
2459 void
2460 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2461 {
2462
2463 bzero(nsb, sizeof(*nsb));
2464 nsb->st_dev = sb->st_dev;
2465 nsb->st_ino = sb->st_ino;
2466 nsb->st_mode = sb->st_mode;
2467 nsb->st_nlink = sb->st_nlink;
2468 nsb->st_uid = sb->st_uid;
2469 nsb->st_gid = sb->st_gid;
2470 nsb->st_rdev = sb->st_rdev;
2471 nsb->st_atim = sb->st_atim;
2472 nsb->st_mtim = sb->st_mtim;
2473 nsb->st_ctim = sb->st_ctim;
2474 nsb->st_size = sb->st_size;
2475 nsb->st_blocks = sb->st_blocks;
2476 nsb->st_blksize = sb->st_blksize;
2477 nsb->st_flags = sb->st_flags;
2478 nsb->st_gen = sb->st_gen;
2479 nsb->st_birthtim = sb->st_birthtim;
2480 }
2481
2482 #ifndef _SYS_SYSPROTO_H_
2483 struct freebsd11_nstat_args {
2484 char *path;
2485 struct nstat *ub;
2486 };
2487 #endif
2488 int
2489 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2490 {
2491 struct stat sb;
2492 struct nstat nsb;
2493 int error;
2494
2495 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2496 &sb, NULL);
2497 if (error != 0)
2498 return (error);
2499 freebsd11_cvtnstat(&sb, &nsb);
2500 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2501 }
2502
2503 /*
2504 * NetBSD lstat. Get file status; this version does not follow links.
2505 */
2506 #ifndef _SYS_SYSPROTO_H_
2507 struct freebsd11_nlstat_args {
2508 char *path;
2509 struct nstat *ub;
2510 };
2511 #endif
2512 int
2513 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2514 {
2515 struct stat sb;
2516 struct nstat nsb;
2517 int error;
2518
2519 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2520 UIO_USERSPACE, &sb, NULL);
2521 if (error != 0)
2522 return (error);
2523 freebsd11_cvtnstat(&sb, &nsb);
2524 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2525 }
2526 #endif /* COMPAT_FREEBSD11 */
2527
2528 /*
2529 * Get configurable pathname variables.
2530 */
2531 #ifndef _SYS_SYSPROTO_H_
2532 struct pathconf_args {
2533 char *path;
2534 int name;
2535 };
2536 #endif
2537 int
2538 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2539 {
2540 long value;
2541 int error;
2542
2543 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2544 &value);
2545 if (error == 0)
2546 td->td_retval[0] = value;
2547 return (error);
2548 }
2549
2550 #ifndef _SYS_SYSPROTO_H_
2551 struct lpathconf_args {
2552 char *path;
2553 int name;
2554 };
2555 #endif
2556 int
2557 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2558 {
2559 long value;
2560 int error;
2561
2562 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2563 NOFOLLOW, &value);
2564 if (error == 0)
2565 td->td_retval[0] = value;
2566 return (error);
2567 }
2568
2569 int
2570 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2571 int name, u_long flags, long *valuep)
2572 {
2573 struct nameidata nd;
2574 int error;
2575
2576 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2577 pathseg, path, td);
2578 if ((error = namei(&nd)) != 0)
2579 return (error);
2580 NDFREE_NOTHING(&nd);
2581
2582 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2583 vput(nd.ni_vp);
2584 return (error);
2585 }
2586
2587 /*
2588 * Return target name of a symbolic link.
2589 */
2590 #ifndef _SYS_SYSPROTO_H_
2591 struct readlink_args {
2592 char *path;
2593 char *buf;
2594 size_t count;
2595 };
2596 #endif
2597 int
2598 sys_readlink(struct thread *td, struct readlink_args *uap)
2599 {
2600
2601 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2602 uap->buf, UIO_USERSPACE, uap->count));
2603 }
2604 #ifndef _SYS_SYSPROTO_H_
2605 struct readlinkat_args {
2606 int fd;
2607 char *path;
2608 char *buf;
2609 size_t bufsize;
2610 };
2611 #endif
2612 int
2613 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2614 {
2615
2616 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2617 uap->buf, UIO_USERSPACE, uap->bufsize));
2618 }
2619
2620 int
2621 kern_readlinkat(struct thread *td, int fd, const char *path,
2622 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2623 {
2624 struct vnode *vp;
2625 struct nameidata nd;
2626 int error;
2627
2628 if (count > IOSIZE_MAX)
2629 return (EINVAL);
2630
2631 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2632 EMPTYPATH, pathseg, path, fd, td);
2633
2634 if ((error = namei(&nd)) != 0)
2635 return (error);
2636 NDFREE_NOTHING(&nd);
2637 vp = nd.ni_vp;
2638
2639 error = kern_readlink_vp(vp, buf, bufseg, count, td);
2640 vput(vp);
2641
2642 return (error);
2643 }
2644
2645 /*
2646 * Helper function to readlink from a vnode
2647 */
2648 static int
2649 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2650 struct thread *td)
2651 {
2652 struct iovec aiov;
2653 struct uio auio;
2654 int error;
2655
2656 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2657 #ifdef MAC
2658 error = mac_vnode_check_readlink(td->td_ucred, vp);
2659 if (error != 0)
2660 return (error);
2661 #endif
2662 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2663 return (EINVAL);
2664
2665 aiov.iov_base = buf;
2666 aiov.iov_len = count;
2667 auio.uio_iov = &aiov;
2668 auio.uio_iovcnt = 1;
2669 auio.uio_offset = 0;
2670 auio.uio_rw = UIO_READ;
2671 auio.uio_segflg = bufseg;
2672 auio.uio_td = td;
2673 auio.uio_resid = count;
2674 error = VOP_READLINK(vp, &auio, td->td_ucred);
2675 td->td_retval[0] = count - auio.uio_resid;
2676 return (error);
2677 }
2678
2679 /*
2680 * Common implementation code for chflags() and fchflags().
2681 */
2682 static int
2683 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2684 {
2685 struct mount *mp;
2686 struct vattr vattr;
2687 int error;
2688
2689 /* We can't support the value matching VNOVAL. */
2690 if (flags == VNOVAL)
2691 return (EOPNOTSUPP);
2692
2693 /*
2694 * Prevent non-root users from setting flags on devices. When
2695 * a device is reused, users can retain ownership of the device
2696 * if they are allowed to set flags and programs assume that
2697 * chown can't fail when done as root.
2698 */
2699 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2700 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2701 if (error != 0)
2702 return (error);
2703 }
2704
2705 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2706 return (error);
2707 VATTR_NULL(&vattr);
2708 vattr.va_flags = flags;
2709 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2710 #ifdef MAC
2711 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2712 if (error == 0)
2713 #endif
2714 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2715 VOP_UNLOCK(vp);
2716 vn_finished_write(mp);
2717 return (error);
2718 }
2719
2720 /*
2721 * Change flags of a file given a path name.
2722 */
2723 #ifndef _SYS_SYSPROTO_H_
2724 struct chflags_args {
2725 const char *path;
2726 u_long flags;
2727 };
2728 #endif
2729 int
2730 sys_chflags(struct thread *td, struct chflags_args *uap)
2731 {
2732
2733 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2734 uap->flags, 0));
2735 }
2736
2737 #ifndef _SYS_SYSPROTO_H_
2738 struct chflagsat_args {
2739 int fd;
2740 const char *path;
2741 u_long flags;
2742 int atflag;
2743 }
2744 #endif
2745 int
2746 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2747 {
2748
2749 if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2750 AT_EMPTY_PATH)) != 0)
2751 return (EINVAL);
2752
2753 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2754 uap->flags, uap->atflag));
2755 }
2756
2757 /*
2758 * Same as chflags() but doesn't follow symlinks.
2759 */
2760 #ifndef _SYS_SYSPROTO_H_
2761 struct lchflags_args {
2762 const char *path;
2763 u_long flags;
2764 };
2765 #endif
2766 int
2767 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2768 {
2769
2770 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2771 uap->flags, AT_SYMLINK_NOFOLLOW));
2772 }
2773
2774 static int
2775 kern_chflagsat(struct thread *td, int fd, const char *path,
2776 enum uio_seg pathseg, u_long flags, int atflag)
2777 {
2778 struct nameidata nd;
2779 int error;
2780
2781 AUDIT_ARG_FFLAGS(flags);
2782 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2783 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2784 fd, &cap_fchflags_rights, td);
2785 if ((error = namei(&nd)) != 0)
2786 return (error);
2787 NDFREE_NOTHING(&nd);
2788 error = setfflags(td, nd.ni_vp, flags);
2789 vrele(nd.ni_vp);
2790 return (error);
2791 }
2792
2793 /*
2794 * Change flags of a file given a file descriptor.
2795 */
2796 #ifndef _SYS_SYSPROTO_H_
2797 struct fchflags_args {
2798 int fd;
2799 u_long flags;
2800 };
2801 #endif
2802 int
2803 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2804 {
2805 struct file *fp;
2806 int error;
2807
2808 AUDIT_ARG_FD(uap->fd);
2809 AUDIT_ARG_FFLAGS(uap->flags);
2810 error = getvnode(td, uap->fd, &cap_fchflags_rights,
2811 &fp);
2812 if (error != 0)
2813 return (error);
2814 #ifdef AUDIT
2815 if (AUDITING_TD(td)) {
2816 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2817 AUDIT_ARG_VNODE1(fp->f_vnode);
2818 VOP_UNLOCK(fp->f_vnode);
2819 }
2820 #endif
2821 error = setfflags(td, fp->f_vnode, uap->flags);
2822 fdrop(fp, td);
2823 return (error);
2824 }
2825
2826 /*
2827 * Common implementation code for chmod(), lchmod() and fchmod().
2828 */
2829 int
2830 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2831 {
2832 struct mount *mp;
2833 struct vattr vattr;
2834 int error;
2835
2836 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2837 return (error);
2838 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2839 VATTR_NULL(&vattr);
2840 vattr.va_mode = mode & ALLPERMS;
2841 #ifdef MAC
2842 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2843 if (error == 0)
2844 #endif
2845 error = VOP_SETATTR(vp, &vattr, cred);
2846 VOP_UNLOCK(vp);
2847 vn_finished_write(mp);
2848 return (error);
2849 }
2850
2851 /*
2852 * Change mode of a file given path name.
2853 */
2854 #ifndef _SYS_SYSPROTO_H_
2855 struct chmod_args {
2856 char *path;
2857 int mode;
2858 };
2859 #endif
2860 int
2861 sys_chmod(struct thread *td, struct chmod_args *uap)
2862 {
2863
2864 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2865 uap->mode, 0));
2866 }
2867
2868 #ifndef _SYS_SYSPROTO_H_
2869 struct fchmodat_args {
2870 int dirfd;
2871 char *path;
2872 mode_t mode;
2873 int flag;
2874 }
2875 #endif
2876 int
2877 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2878 {
2879
2880 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2881 AT_EMPTY_PATH)) != 0)
2882 return (EINVAL);
2883
2884 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2885 uap->mode, uap->flag));
2886 }
2887
2888 /*
2889 * Change mode of a file given path name (don't follow links.)
2890 */
2891 #ifndef _SYS_SYSPROTO_H_
2892 struct lchmod_args {
2893 char *path;
2894 int mode;
2895 };
2896 #endif
2897 int
2898 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2899 {
2900
2901 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2902 uap->mode, AT_SYMLINK_NOFOLLOW));
2903 }
2904
2905 int
2906 kern_fchmodat(struct thread *td, int fd, const char *path,
2907 enum uio_seg pathseg, mode_t mode, int flag)
2908 {
2909 struct nameidata nd;
2910 int error;
2911
2912 AUDIT_ARG_MODE(mode);
2913 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2914 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2915 fd, &cap_fchmod_rights, td);
2916 if ((error = namei(&nd)) != 0)
2917 return (error);
2918 NDFREE_NOTHING(&nd);
2919 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2920 vrele(nd.ni_vp);
2921 return (error);
2922 }
2923
2924 /*
2925 * Change mode of a file given a file descriptor.
2926 */
2927 #ifndef _SYS_SYSPROTO_H_
2928 struct fchmod_args {
2929 int fd;
2930 int mode;
2931 };
2932 #endif
2933 int
2934 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2935 {
2936 struct file *fp;
2937 int error;
2938
2939 AUDIT_ARG_FD(uap->fd);
2940 AUDIT_ARG_MODE(uap->mode);
2941
2942 error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2943 if (error != 0)
2944 return (error);
2945 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2946 fdrop(fp, td);
2947 return (error);
2948 }
2949
2950 /*
2951 * Common implementation for chown(), lchown(), and fchown()
2952 */
2953 int
2954 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2955 gid_t gid)
2956 {
2957 struct mount *mp;
2958 struct vattr vattr;
2959 int error;
2960
2961 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2962 return (error);
2963 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2964 VATTR_NULL(&vattr);
2965 vattr.va_uid = uid;
2966 vattr.va_gid = gid;
2967 #ifdef MAC
2968 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2969 vattr.va_gid);
2970 if (error == 0)
2971 #endif
2972 error = VOP_SETATTR(vp, &vattr, cred);
2973 VOP_UNLOCK(vp);
2974 vn_finished_write(mp);
2975 return (error);
2976 }
2977
2978 /*
2979 * Set ownership given a path name.
2980 */
2981 #ifndef _SYS_SYSPROTO_H_
2982 struct chown_args {
2983 char *path;
2984 int uid;
2985 int gid;
2986 };
2987 #endif
2988 int
2989 sys_chown(struct thread *td, struct chown_args *uap)
2990 {
2991
2992 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2993 uap->gid, 0));
2994 }
2995
2996 #ifndef _SYS_SYSPROTO_H_
2997 struct fchownat_args {
2998 int fd;
2999 const char * path;
3000 uid_t uid;
3001 gid_t gid;
3002 int flag;
3003 };
3004 #endif
3005 int
3006 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3007 {
3008
3009 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3010 AT_EMPTY_PATH)) != 0)
3011 return (EINVAL);
3012
3013 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3014 uap->gid, uap->flag));
3015 }
3016
3017 int
3018 kern_fchownat(struct thread *td, int fd, const char *path,
3019 enum uio_seg pathseg, int uid, int gid, int flag)
3020 {
3021 struct nameidata nd;
3022 int error;
3023
3024 AUDIT_ARG_OWNER(uid, gid);
3025 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3026 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3027 fd, &cap_fchown_rights, td);
3028
3029 if ((error = namei(&nd)) != 0)
3030 return (error);
3031 NDFREE_NOTHING(&nd);
3032 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3033 vrele(nd.ni_vp);
3034 return (error);
3035 }
3036
3037 /*
3038 * Set ownership given a path name, do not cross symlinks.
3039 */
3040 #ifndef _SYS_SYSPROTO_H_
3041 struct lchown_args {
3042 char *path;
3043 int uid;
3044 int gid;
3045 };
3046 #endif
3047 int
3048 sys_lchown(struct thread *td, struct lchown_args *uap)
3049 {
3050
3051 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3052 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3053 }
3054
3055 /*
3056 * Set ownership given a file descriptor.
3057 */
3058 #ifndef _SYS_SYSPROTO_H_
3059 struct fchown_args {
3060 int fd;
3061 int uid;
3062 int gid;
3063 };
3064 #endif
3065 int
3066 sys_fchown(struct thread *td, struct fchown_args *uap)
3067 {
3068 struct file *fp;
3069 int error;
3070
3071 AUDIT_ARG_FD(uap->fd);
3072 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3073 error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3074 if (error != 0)
3075 return (error);
3076 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3077 fdrop(fp, td);
3078 return (error);
3079 }
3080
3081 /*
3082 * Common implementation code for utimes(), lutimes(), and futimes().
3083 */
3084 static int
3085 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3086 struct timespec *tsp)
3087 {
3088 struct timeval tv[2];
3089 const struct timeval *tvp;
3090 int error;
3091
3092 if (usrtvp == NULL) {
3093 vfs_timestamp(&tsp[0]);
3094 tsp[1] = tsp[0];
3095 } else {
3096 if (tvpseg == UIO_SYSSPACE) {
3097 tvp = usrtvp;
3098 } else {
3099 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3100 return (error);
3101 tvp = tv;
3102 }
3103
3104 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3105 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3106 return (EINVAL);
3107 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3108 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3109 }
3110 return (0);
3111 }
3112
3113 /*
3114 * Common implementation code for futimens(), utimensat().
3115 */
3116 #define UTIMENS_NULL 0x1
3117 #define UTIMENS_EXIT 0x2
3118 static int
3119 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3120 struct timespec *tsp, int *retflags)
3121 {
3122 struct timespec tsnow;
3123 int error;
3124
3125 vfs_timestamp(&tsnow);
3126 *retflags = 0;
3127 if (usrtsp == NULL) {
3128 tsp[0] = tsnow;
3129 tsp[1] = tsnow;
3130 *retflags |= UTIMENS_NULL;
3131 return (0);
3132 }
3133 if (tspseg == UIO_SYSSPACE) {
3134 tsp[0] = usrtsp[0];
3135 tsp[1] = usrtsp[1];
3136 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3137 return (error);
3138 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3139 *retflags |= UTIMENS_EXIT;
3140 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3141 *retflags |= UTIMENS_NULL;
3142 if (tsp[0].tv_nsec == UTIME_OMIT)
3143 tsp[0].tv_sec = VNOVAL;
3144 else if (tsp[0].tv_nsec == UTIME_NOW)
3145 tsp[0] = tsnow;
3146 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3147 return (EINVAL);
3148 if (tsp[1].tv_nsec == UTIME_OMIT)
3149 tsp[1].tv_sec = VNOVAL;
3150 else if (tsp[1].tv_nsec == UTIME_NOW)
3151 tsp[1] = tsnow;
3152 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3153 return (EINVAL);
3154
3155 return (0);
3156 }
3157
3158 /*
3159 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3160 * and utimensat().
3161 */
3162 static int
3163 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3164 int numtimes, int nullflag)
3165 {
3166 struct mount *mp;
3167 struct vattr vattr;
3168 int error;
3169 bool setbirthtime;
3170
3171 setbirthtime = false;
3172 vattr.va_birthtime.tv_sec = VNOVAL;
3173 vattr.va_birthtime.tv_nsec = 0;
3174
3175 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3176 return (error);
3177 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3178 if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3179 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3180 setbirthtime = true;
3181 VATTR_NULL(&vattr);
3182 vattr.va_atime = ts[0];
3183 vattr.va_mtime = ts[1];
3184 if (setbirthtime)
3185 vattr.va_birthtime = ts[1];
3186 if (numtimes > 2)
3187 vattr.va_birthtime = ts[2];
3188 if (nullflag)
3189 vattr.va_vaflags |= VA_UTIMES_NULL;
3190 #ifdef MAC
3191 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3192 vattr.va_mtime);
3193 #endif
3194 if (error == 0)
3195 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3196 VOP_UNLOCK(vp);
3197 vn_finished_write(mp);
3198 return (error);
3199 }
3200
3201 /*
3202 * Set the access and modification times of a file.
3203 */
3204 #ifndef _SYS_SYSPROTO_H_
3205 struct utimes_args {
3206 char *path;
3207 struct timeval *tptr;
3208 };
3209 #endif
3210 int
3211 sys_utimes(struct thread *td, struct utimes_args *uap)
3212 {
3213
3214 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3215 uap->tptr, UIO_USERSPACE));
3216 }
3217
3218 #ifndef _SYS_SYSPROTO_H_
3219 struct futimesat_args {
3220 int fd;
3221 const char * path;
3222 const struct timeval * times;
3223 };
3224 #endif
3225 int
3226 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3227 {
3228
3229 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3230 uap->times, UIO_USERSPACE));
3231 }
3232
3233 int
3234 kern_utimesat(struct thread *td, int fd, const char *path,
3235 enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3236 {
3237 struct nameidata nd;
3238 struct timespec ts[2];
3239 int error;
3240
3241 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3242 return (error);
3243 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3244 &cap_futimes_rights, td);
3245
3246 if ((error = namei(&nd)) != 0)
3247 return (error);
3248 NDFREE_NOTHING(&nd);
3249 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3250 vrele(nd.ni_vp);
3251 return (error);
3252 }
3253
3254 /*
3255 * Set the access and modification times of a file.
3256 */
3257 #ifndef _SYS_SYSPROTO_H_
3258 struct lutimes_args {
3259 char *path;
3260 struct timeval *tptr;
3261 };
3262 #endif
3263 int
3264 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3265 {
3266
3267 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3268 UIO_USERSPACE));
3269 }
3270
3271 int
3272 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3273 struct timeval *tptr, enum uio_seg tptrseg)
3274 {
3275 struct timespec ts[2];
3276 struct nameidata nd;
3277 int error;
3278
3279 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3280 return (error);
3281 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3282 if ((error = namei(&nd)) != 0)
3283 return (error);
3284 NDFREE_NOTHING(&nd);
3285 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3286 vrele(nd.ni_vp);
3287 return (error);
3288 }
3289
3290 /*
3291 * Set the access and modification times of a file.
3292 */
3293 #ifndef _SYS_SYSPROTO_H_
3294 struct futimes_args {
3295 int fd;
3296 struct timeval *tptr;
3297 };
3298 #endif
3299 int
3300 sys_futimes(struct thread *td, struct futimes_args *uap)
3301 {
3302
3303 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3304 }
3305
3306 int
3307 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3308 enum uio_seg tptrseg)
3309 {
3310 struct timespec ts[2];
3311 struct file *fp;
3312 int error;
3313
3314 AUDIT_ARG_FD(fd);
3315 error = getutimes(tptr, tptrseg, ts);
3316 if (error != 0)
3317 return (error);
3318 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3319 if (error != 0)
3320 return (error);
3321 #ifdef AUDIT
3322 if (AUDITING_TD(td)) {
3323 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3324 AUDIT_ARG_VNODE1(fp->f_vnode);
3325 VOP_UNLOCK(fp->f_vnode);
3326 }
3327 #endif
3328 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3329 fdrop(fp, td);
3330 return (error);
3331 }
3332
3333 int
3334 sys_futimens(struct thread *td, struct futimens_args *uap)
3335 {
3336
3337 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3338 }
3339
3340 int
3341 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3342 enum uio_seg tptrseg)
3343 {
3344 struct timespec ts[2];
3345 struct file *fp;
3346 int error, flags;
3347
3348 AUDIT_ARG_FD(fd);
3349 error = getutimens(tptr, tptrseg, ts, &flags);
3350 if (error != 0)
3351 return (error);
3352 if (flags & UTIMENS_EXIT)
3353 return (0);
3354 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3355 if (error != 0)
3356 return (error);
3357 #ifdef AUDIT
3358 if (AUDITING_TD(td)) {
3359 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3360 AUDIT_ARG_VNODE1(fp->f_vnode);
3361 VOP_UNLOCK(fp->f_vnode);
3362 }
3363 #endif
3364 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3365 fdrop(fp, td);
3366 return (error);
3367 }
3368
3369 int
3370 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3371 {
3372
3373 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3374 uap->times, UIO_USERSPACE, uap->flag));
3375 }
3376
3377 int
3378 kern_utimensat(struct thread *td, int fd, const char *path,
3379 enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3380 int flag)
3381 {
3382 struct nameidata nd;
3383 struct timespec ts[2];
3384 int error, flags;
3385
3386 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3387 AT_EMPTY_PATH)) != 0)
3388 return (EINVAL);
3389
3390 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3391 return (error);
3392 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3393 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3394 pathseg, path, fd, &cap_futimes_rights, td);
3395 if ((error = namei(&nd)) != 0)
3396 return (error);
3397 /*
3398 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3399 * POSIX states:
3400 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3401 * "Search permission is denied by a component of the path prefix."
3402 */
3403 NDFREE_NOTHING(&nd);
3404 if ((flags & UTIMENS_EXIT) == 0)
3405 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3406 vrele(nd.ni_vp);
3407 return (error);
3408 }
3409
3410 /*
3411 * Truncate a file given its path name.
3412 */
3413 #ifndef _SYS_SYSPROTO_H_
3414 struct truncate_args {
3415 char *path;
3416 int pad;
3417 off_t length;
3418 };
3419 #endif
3420 int
3421 sys_truncate(struct thread *td, struct truncate_args *uap)
3422 {
3423
3424 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3425 }
3426
3427 int
3428 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3429 off_t length)
3430 {
3431 struct mount *mp;
3432 struct vnode *vp;
3433 void *rl_cookie;
3434 struct vattr vattr;
3435 struct nameidata nd;
3436 int error;
3437
3438 if (length < 0)
3439 return (EINVAL);
3440 NDPREINIT(&nd);
3441 retry:
3442 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3443 if ((error = namei(&nd)) != 0)
3444 return (error);
3445 vp = nd.ni_vp;
3446 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3447 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3448 vn_rangelock_unlock(vp, rl_cookie);
3449 vrele(vp);
3450 return (error);
3451 }
3452 NDFREE(&nd, NDF_ONLY_PNBUF);
3453 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3454 if (vp->v_type == VDIR)
3455 error = EISDIR;
3456 #ifdef MAC
3457 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3458 }
3459 #endif
3460 else if ((error = vn_writechk(vp)) == 0 &&
3461 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3462 VATTR_NULL(&vattr);
3463 vattr.va_size = length;
3464 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3465 }
3466 VOP_UNLOCK(vp);
3467 vn_finished_write(mp);
3468 vn_rangelock_unlock(vp, rl_cookie);
3469 vrele(vp);
3470 if (error == ERELOOKUP)
3471 goto retry;
3472 return (error);
3473 }
3474
3475 #if defined(COMPAT_43)
3476 /*
3477 * Truncate a file given its path name.
3478 */
3479 #ifndef _SYS_SYSPROTO_H_
3480 struct otruncate_args {
3481 char *path;
3482 long length;
3483 };
3484 #endif
3485 int
3486 otruncate(struct thread *td, struct otruncate_args *uap)
3487 {
3488
3489 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3490 }
3491 #endif /* COMPAT_43 */
3492
3493 #if defined(COMPAT_FREEBSD6)
3494 /* Versions with the pad argument */
3495 int
3496 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3497 {
3498
3499 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3500 }
3501
3502 int
3503 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3504 {
3505
3506 return (kern_ftruncate(td, uap->fd, uap->length));
3507 }
3508 #endif
3509
3510 int
3511 kern_fsync(struct thread *td, int fd, bool fullsync)
3512 {
3513 struct vnode *vp;
3514 struct mount *mp;
3515 struct file *fp;
3516 int error;
3517
3518 AUDIT_ARG_FD(fd);
3519 error = getvnode(td, fd, &cap_fsync_rights, &fp);
3520 if (error != 0)
3521 return (error);
3522 vp = fp->f_vnode;
3523 #if 0
3524 if (!fullsync)
3525 /* XXXKIB: compete outstanding aio writes */;
3526 #endif
3527 retry:
3528 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3529 if (error != 0)
3530 goto drop;
3531 vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3532 AUDIT_ARG_VNODE1(vp);
3533 if (vp->v_object != NULL) {
3534 VM_OBJECT_WLOCK(vp->v_object);
3535 vm_object_page_clean(vp->v_object, 0, 0, 0);
3536 VM_OBJECT_WUNLOCK(vp->v_object);
3537 }
3538 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3539 VOP_UNLOCK(vp);
3540 vn_finished_write(mp);
3541 if (error == ERELOOKUP)
3542 goto retry;
3543 drop:
3544 fdrop(fp, td);
3545 return (error);
3546 }
3547
3548 /*
3549 * Sync an open file.
3550 */
3551 #ifndef _SYS_SYSPROTO_H_
3552 struct fsync_args {
3553 int fd;
3554 };
3555 #endif
3556 int
3557 sys_fsync(struct thread *td, struct fsync_args *uap)
3558 {
3559
3560 return (kern_fsync(td, uap->fd, true));
3561 }
3562
3563 int
3564 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3565 {
3566
3567 return (kern_fsync(td, uap->fd, false));
3568 }
3569
3570 /*
3571 * Rename files. Source and destination must either both be directories, or
3572 * both not be directories. If target is a directory, it must be empty.
3573 */
3574 #ifndef _SYS_SYSPROTO_H_
3575 struct rename_args {
3576 char *from;
3577 char *to;
3578 };
3579 #endif
3580 int
3581 sys_rename(struct thread *td, struct rename_args *uap)
3582 {
3583
3584 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3585 uap->to, UIO_USERSPACE));
3586 }
3587
3588 #ifndef _SYS_SYSPROTO_H_
3589 struct renameat_args {
3590 int oldfd;
3591 char *old;
3592 int newfd;
3593 char *new;
3594 };
3595 #endif
3596 int
3597 sys_renameat(struct thread *td, struct renameat_args *uap)
3598 {
3599
3600 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3601 UIO_USERSPACE));
3602 }
3603
3604 #ifdef MAC
3605 static int
3606 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3607 const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3608 {
3609 int error;
3610
3611 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3612 AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3613 if ((error = namei(fromnd)) != 0)
3614 return (error);
3615 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3616 fromnd->ni_vp, &fromnd->ni_cnd);
3617 VOP_UNLOCK(fromnd->ni_dvp);
3618 if (fromnd->ni_dvp != fromnd->ni_vp)
3619 VOP_UNLOCK(fromnd->ni_vp);
3620 if (error != 0) {
3621 NDFREE(fromnd, NDF_ONLY_PNBUF);
3622 vrele(fromnd->ni_dvp);
3623 vrele(fromnd->ni_vp);
3624 if (fromnd->ni_startdir)
3625 vrele(fromnd->ni_startdir);
3626 }
3627 return (error);
3628 }
3629 #endif
3630
3631 int
3632 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3633 const char *new, enum uio_seg pathseg)
3634 {
3635 struct mount *mp = NULL;
3636 struct vnode *tvp, *fvp, *tdvp;
3637 struct nameidata fromnd, tond;
3638 u_int64_t tondflags;
3639 int error;
3640
3641 again:
3642 bwillwrite();
3643 #ifdef MAC
3644 if (mac_vnode_check_rename_from_enabled()) {
3645 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3646 &fromnd);
3647 if (error != 0)
3648 return (error);
3649 } else {
3650 #endif
3651 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3652 pathseg, old, oldfd, &cap_renameat_source_rights, td);
3653 if ((error = namei(&fromnd)) != 0)
3654 return (error);
3655 #ifdef MAC
3656 }
3657 #endif
3658 fvp = fromnd.ni_vp;
3659 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3660 if (fromnd.ni_vp->v_type == VDIR)
3661 tondflags |= WILLBEDIR;
3662 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3663 &cap_renameat_target_rights, td);
3664 if ((error = namei(&tond)) != 0) {
3665 /* Translate error code for rename("dir1", "dir2/."). */
3666 if (error == EISDIR && fvp->v_type == VDIR)
3667 error = EINVAL;
3668 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3669 vrele(fromnd.ni_dvp);
3670 vrele(fvp);
3671 goto out1;
3672 }
3673 tdvp = tond.ni_dvp;
3674 tvp = tond.ni_vp;
3675 error = vn_start_write(fvp, &mp, V_NOWAIT);
3676 if (error != 0) {
3677 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3678 NDFREE(&tond, NDF_ONLY_PNBUF);
3679 if (tvp != NULL)
3680 vput(tvp);
3681 if (tdvp == tvp)
3682 vrele(tdvp);
3683 else
3684 vput(tdvp);
3685 vrele(fromnd.ni_dvp);
3686 vrele(fvp);
3687 vrele(tond.ni_startdir);
3688 if (fromnd.ni_startdir != NULL)
3689 vrele(fromnd.ni_startdir);
3690 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3691 if (error != 0)
3692 return (error);
3693 goto again;
3694 }
3695 if (tvp != NULL) {
3696 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3697 error = ENOTDIR;
3698 goto out;
3699 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3700 error = EISDIR;
3701 goto out;
3702 }
3703 #ifdef CAPABILITIES
3704 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3705 /*
3706 * If the target already exists we require CAP_UNLINKAT
3707 * from 'newfd', when newfd was used for the lookup.
3708 */
3709 error = cap_check(&tond.ni_filecaps.fc_rights,
3710 &cap_unlinkat_rights);
3711 if (error != 0)
3712 goto out;
3713 }
3714 #endif
3715 }
3716 if (fvp == tdvp) {
3717 error = EINVAL;
3718 goto out;
3719 }
3720 /*
3721 * If the source is the same as the destination (that is, if they
3722 * are links to the same vnode), then there is nothing to do.
3723 */
3724 if (fvp == tvp)
3725 error = ERESTART;
3726 #ifdef MAC
3727 else
3728 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3729 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3730 #endif
3731 out:
3732 if (error == 0) {
3733 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3734 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3735 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3736 NDFREE(&tond, NDF_ONLY_PNBUF);
3737 } else {
3738 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3739 NDFREE(&tond, NDF_ONLY_PNBUF);
3740 if (tvp != NULL)
3741 vput(tvp);
3742 if (tdvp == tvp)
3743 vrele(tdvp);
3744 else
3745 vput(tdvp);
3746 vrele(fromnd.ni_dvp);
3747 vrele(fvp);
3748 }
3749 vrele(tond.ni_startdir);
3750 vn_finished_write(mp);
3751 out1:
3752 if (fromnd.ni_startdir)
3753 vrele(fromnd.ni_startdir);
3754 if (error == ERESTART)
3755 return (0);
3756 if (error == ERELOOKUP)
3757 goto again;
3758 return (error);
3759 }
3760
3761 /*
3762 * Make a directory file.
3763 */
3764 #ifndef _SYS_SYSPROTO_H_
3765 struct mkdir_args {
3766 char *path;
3767 int mode;
3768 };
3769 #endif
3770 int
3771 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3772 {
3773
3774 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3775 uap->mode));
3776 }
3777
3778 #ifndef _SYS_SYSPROTO_H_
3779 struct mkdirat_args {
3780 int fd;
3781 char *path;
3782 mode_t mode;
3783 };
3784 #endif
3785 int
3786 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3787 {
3788
3789 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3790 }
3791
3792 int
3793 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3794 int mode)
3795 {
3796 struct mount *mp;
3797 struct vattr vattr;
3798 struct nameidata nd;
3799 int error;
3800
3801 AUDIT_ARG_MODE(mode);
3802 NDPREINIT(&nd);
3803 restart:
3804 bwillwrite();
3805 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3806 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3807 segflg, path, fd, &cap_mkdirat_rights, td);
3808 if ((error = namei(&nd)) != 0)
3809 return (error);
3810 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3811 NDFREE(&nd, NDF_ONLY_PNBUF);
3812 vput(nd.ni_dvp);
3813 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3814 return (error);
3815 goto restart;
3816 }
3817 VATTR_NULL(&vattr);
3818 vattr.va_type = VDIR;
3819 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3820 #ifdef MAC
3821 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3822 &vattr);
3823 if (error != 0)
3824 goto out;
3825 #endif
3826 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3827 #ifdef MAC
3828 out:
3829 #endif
3830 NDFREE(&nd, NDF_ONLY_PNBUF);
3831 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3832 vn_finished_write(mp);
3833 if (error == ERELOOKUP)
3834 goto restart;
3835 return (error);
3836 }
3837
3838 /*
3839 * Remove a directory file.
3840 */
3841 #ifndef _SYS_SYSPROTO_H_
3842 struct rmdir_args {
3843 char *path;
3844 };
3845 #endif
3846 int
3847 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3848 {
3849
3850 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3851 0));
3852 }
3853
3854 int
3855 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3856 enum uio_seg pathseg, int flag)
3857 {
3858 struct mount *mp;
3859 struct vnode *vp;
3860 struct file *fp;
3861 struct nameidata nd;
3862 cap_rights_t rights;
3863 int error;
3864
3865 fp = NULL;
3866 if (fd != FD_NONE) {
3867 error = getvnode(td, fd, cap_rights_init_one(&rights,
3868 CAP_LOOKUP), &fp);
3869 if (error != 0)
3870 return (error);
3871 }
3872
3873 NDPREINIT(&nd);
3874 restart:
3875 bwillwrite();
3876 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3877 at2cnpflags(flag, AT_RESOLVE_BENEATH),
3878 pathseg, path, dfd, &cap_unlinkat_rights, td);
3879 if ((error = namei(&nd)) != 0)
3880 goto fdout;
3881 vp = nd.ni_vp;
3882 if (vp->v_type != VDIR) {
3883 error = ENOTDIR;
3884 goto out;
3885 }
3886 /*
3887 * No rmdir "." please.
3888 */
3889 if (nd.ni_dvp == vp) {
3890 error = EINVAL;
3891 goto out;
3892 }
3893 /*
3894 * The root of a mounted filesystem cannot be deleted.
3895 */
3896 if (vp->v_vflag & VV_ROOT) {
3897 error = EBUSY;
3898 goto out;
3899 }
3900
3901 if (fp != NULL && fp->f_vnode != vp) {
3902 if (VN_IS_DOOMED(fp->f_vnode))
3903 error = EBADF;
3904 else
3905 error = EDEADLK;
3906 goto out;
3907 }
3908
3909 #ifdef MAC
3910 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3911 &nd.ni_cnd);
3912 if (error != 0)
3913 goto out;
3914 #endif
3915 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3916 NDFREE(&nd, NDF_ONLY_PNBUF);
3917 vput(vp);
3918 if (nd.ni_dvp == vp)
3919 vrele(nd.ni_dvp);
3920 else
3921 vput(nd.ni_dvp);
3922 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3923 goto fdout;
3924 goto restart;
3925 }
3926 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3927 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3928 vn_finished_write(mp);
3929 out:
3930 NDFREE(&nd, NDF_ONLY_PNBUF);
3931 vput(vp);
3932 if (nd.ni_dvp == vp)
3933 vrele(nd.ni_dvp);
3934 else
3935 vput(nd.ni_dvp);
3936 if (error == ERELOOKUP)
3937 goto restart;
3938 fdout:
3939 if (fp != NULL)
3940 fdrop(fp, td);
3941 return (error);
3942 }
3943
3944 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3945 int
3946 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3947 long *basep, void (*func)(struct freebsd11_dirent *))
3948 {
3949 struct freebsd11_dirent dstdp;
3950 struct dirent *dp, *edp;
3951 char *dirbuf;
3952 off_t base;
3953 ssize_t resid, ucount;
3954 int error;
3955
3956 /* XXX arbitrary sanity limit on `count'. */
3957 count = min(count, 64 * 1024);
3958
3959 dirbuf = malloc(count, M_TEMP, M_WAITOK);
3960
3961 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3962 UIO_SYSSPACE);
3963 if (error != 0)
3964 goto done;
3965 if (basep != NULL)
3966 *basep = base;
3967
3968 ucount = 0;
3969 for (dp = (struct dirent *)dirbuf,
3970 edp = (struct dirent *)&dirbuf[count - resid];
3971 ucount < count && dp < edp; ) {
3972 if (dp->d_reclen == 0)
3973 break;
3974 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3975 if (dp->d_namlen >= sizeof(dstdp.d_name))
3976 continue;
3977 dstdp.d_type = dp->d_type;
3978 dstdp.d_namlen = dp->d_namlen;
3979 dstdp.d_fileno = dp->d_fileno; /* truncate */
3980 if (dstdp.d_fileno != dp->d_fileno) {
3981 switch (ino64_trunc_error) {
3982 default:
3983 case 0:
3984 break;
3985 case 1:
3986 error = EOVERFLOW;
3987 goto done;
3988 case 2:
3989 dstdp.d_fileno = UINT32_MAX;
3990 break;
3991 }
3992 }
3993 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3994 ((dp->d_namlen + 1 + 3) &~ 3);
3995 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3996 bzero(dstdp.d_name + dstdp.d_namlen,
3997 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3998 dstdp.d_namlen);
3999 MPASS(dstdp.d_reclen <= dp->d_reclen);
4000 MPASS(ucount + dstdp.d_reclen <= count);
4001 if (func != NULL)
4002 func(&dstdp);
4003 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4004 if (error != 0)
4005 break;
4006 dp = (struct dirent *)((char *)dp + dp->d_reclen);
4007 ucount += dstdp.d_reclen;
4008 }
4009
4010 done:
4011 free(dirbuf, M_TEMP);
4012 if (error == 0)
4013 td->td_retval[0] = ucount;
4014 return (error);
4015 }
4016 #endif /* COMPAT */
4017
4018 #ifdef COMPAT_43
4019 static void
4020 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4021 {
4022 #if (BYTE_ORDER == LITTLE_ENDIAN)
4023 /*
4024 * The expected low byte of dp->d_namlen is our dp->d_type.
4025 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4026 */
4027 dp->d_type = dp->d_namlen;
4028 dp->d_namlen = 0;
4029 #else
4030 /*
4031 * The dp->d_type is the high byte of the expected dp->d_namlen,
4032 * so must be zero'ed.
4033 */
4034 dp->d_type = 0;
4035 #endif
4036 }
4037
4038 /*
4039 * Read a block of directory entries in a filesystem independent format.
4040 */
4041 #ifndef _SYS_SYSPROTO_H_
4042 struct ogetdirentries_args {
4043 int fd;
4044 char *buf;
4045 u_int count;
4046 long *basep;
4047 };
4048 #endif
4049 int
4050 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4051 {
4052 long loff;
4053 int error;
4054
4055 error = kern_ogetdirentries(td, uap, &loff);
4056 if (error == 0)
4057 error = copyout(&loff, uap->basep, sizeof(long));
4058 return (error);
4059 }
4060
4061 int
4062 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4063 long *ploff)
4064 {
4065 long base;
4066 int error;
4067
4068 /* XXX arbitrary sanity limit on `count'. */
4069 if (uap->count > 64 * 1024)
4070 return (EINVAL);
4071
4072 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4073 &base, ogetdirentries_cvt);
4074
4075 if (error == 0 && uap->basep != NULL)
4076 error = copyout(&base, uap->basep, sizeof(long));
4077
4078 return (error);
4079 }
4080 #endif /* COMPAT_43 */
4081
4082 #if defined(COMPAT_FREEBSD11)
4083 #ifndef _SYS_SYSPROTO_H_
4084 struct freebsd11_getdirentries_args {
4085 int fd;
4086 char *buf;
4087 u_int count;
4088 long *basep;
4089 };
4090 #endif
4091 int
4092 freebsd11_getdirentries(struct thread *td,
4093 struct freebsd11_getdirentries_args *uap)
4094 {
4095 long base;
4096 int error;
4097
4098 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4099 &base, NULL);
4100
4101 if (error == 0 && uap->basep != NULL)
4102 error = copyout(&base, uap->basep, sizeof(long));
4103 return (error);
4104 }
4105
4106 int
4107 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4108 {
4109 struct freebsd11_getdirentries_args ap;
4110
4111 ap.fd = uap->fd;
4112 ap.buf = uap->buf;
4113 ap.count = uap->count;
4114 ap.basep = NULL;
4115 return (freebsd11_getdirentries(td, &ap));
4116 }
4117 #endif /* COMPAT_FREEBSD11 */
4118
4119 /*
4120 * Read a block of directory entries in a filesystem independent format.
4121 */
4122 int
4123 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4124 {
4125 off_t base;
4126 int error;
4127
4128 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4129 NULL, UIO_USERSPACE);
4130 if (error != 0)
4131 return (error);
4132 if (uap->basep != NULL)
4133 error = copyout(&base, uap->basep, sizeof(off_t));
4134 return (error);
4135 }
4136
4137 int
4138 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4139 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4140 {
4141 struct vnode *vp;
4142 struct file *fp;
4143 struct uio auio;
4144 struct iovec aiov;
4145 off_t loff;
4146 int error, eofflag;
4147 off_t foffset;
4148
4149 AUDIT_ARG_FD(fd);
4150 if (count > IOSIZE_MAX)
4151 return (EINVAL);
4152 auio.uio_resid = count;
4153 error = getvnode(td, fd, &cap_read_rights, &fp);
4154 if (error != 0)
4155 return (error);
4156 if ((fp->f_flag & FREAD) == 0) {
4157 fdrop(fp, td);
4158 return (EBADF);
4159 }
4160 vp = fp->f_vnode;
4161 foffset = foffset_lock(fp, 0);
4162 unionread:
4163 if (vp->v_type != VDIR) {
4164 error = EINVAL;
4165 goto fail;
4166 }
4167 aiov.iov_base = buf;
4168 aiov.iov_len = count;
4169 auio.uio_iov = &aiov;
4170 auio.uio_iovcnt = 1;
4171 auio.uio_rw = UIO_READ;
4172 auio.uio_segflg = bufseg;
4173 auio.uio_td = td;
4174 vn_lock(vp, LK_SHARED | LK_RETRY);
4175 AUDIT_ARG_VNODE1(vp);
4176 loff = auio.uio_offset = foffset;
4177 #ifdef MAC
4178 error = mac_vnode_check_readdir(td->td_ucred, vp);
4179 if (error == 0)
4180 #endif
4181 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4182 NULL);
4183 foffset = auio.uio_offset;
4184 if (error != 0) {
4185 VOP_UNLOCK(vp);
4186 goto fail;
4187 }
4188 if (count == auio.uio_resid &&
4189 (vp->v_vflag & VV_ROOT) &&
4190 (vp->v_mount->mnt_flag & MNT_UNION)) {
4191 struct vnode *tvp = vp;
4192
4193 vp = vp->v_mount->mnt_vnodecovered;
4194 VREF(vp);
4195 fp->f_vnode = vp;
4196 foffset = 0;
4197 vput(tvp);
4198 goto unionread;
4199 }
4200 VOP_UNLOCK(vp);
4201 *basep = loff;
4202 if (residp != NULL)
4203 *residp = auio.uio_resid;
4204 td->td_retval[0] = count - auio.uio_resid;
4205 fail:
4206 foffset_unlock(fp, foffset, 0);
4207 fdrop(fp, td);
4208 return (error);
4209 }
4210
4211 /*
4212 * Set the mode mask for creation of filesystem nodes.
4213 */
4214 #ifndef _SYS_SYSPROTO_H_
4215 struct umask_args {
4216 int newmask;
4217 };
4218 #endif
4219 int
4220 sys_umask(struct thread *td, struct umask_args *uap)
4221 {
4222 struct pwddesc *pdp;
4223
4224 pdp = td->td_proc->p_pd;
4225 PWDDESC_XLOCK(pdp);
4226 td->td_retval[0] = pdp->pd_cmask;
4227 pdp->pd_cmask = uap->newmask & ALLPERMS;
4228 PWDDESC_XUNLOCK(pdp);
4229 return (0);
4230 }
4231
4232 /*
4233 * Void all references to file by ripping underlying filesystem away from
4234 * vnode.
4235 */
4236 #ifndef _SYS_SYSPROTO_H_
4237 struct revoke_args {
4238 char *path;
4239 };
4240 #endif
4241 int
4242 sys_revoke(struct thread *td, struct revoke_args *uap)
4243 {
4244 struct vnode *vp;
4245 struct vattr vattr;
4246 struct nameidata nd;
4247 int error;
4248
4249 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4250 uap->path, td);
4251 if ((error = namei(&nd)) != 0)
4252 return (error);
4253 vp = nd.ni_vp;
4254 NDFREE_NOTHING(&nd);
4255 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4256 error = EINVAL;
4257 goto out;
4258 }
4259 #ifdef MAC
4260 error = mac_vnode_check_revoke(td->td_ucred, vp);
4261 if (error != 0)
4262 goto out;
4263 #endif
4264 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4265 if (error != 0)
4266 goto out;
4267 if (td->td_ucred->cr_uid != vattr.va_uid) {
4268 error = priv_check(td, PRIV_VFS_ADMIN);
4269 if (error != 0)
4270 goto out;
4271 }
4272 if (devfs_usecount(vp) > 0)
4273 VOP_REVOKE(vp, REVOKEALL);
4274 out:
4275 vput(vp);
4276 return (error);
4277 }
4278
4279 /*
4280 * This variant of getvnode() allows O_PATH files. Caller should
4281 * ensure that returned file and vnode are only used for compatible
4282 * semantics.
4283 */
4284 int
4285 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4286 struct file **fpp)
4287 {
4288 struct file *fp;
4289 int error;
4290
4291 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4292 if (error != 0)
4293 return (error);
4294
4295 /*
4296 * The file could be not of the vnode type, or it may be not
4297 * yet fully initialized, in which case the f_vnode pointer
4298 * may be set, but f_ops is still badfileops. E.g.,
4299 * devfs_open() transiently create such situation to
4300 * facilitate csw d_fdopen().
4301 *
4302 * Dupfdopen() handling in kern_openat() installs the
4303 * half-baked file into the process descriptor table, allowing
4304 * other thread to dereference it. Guard against the race by
4305 * checking f_ops.
4306 */
4307 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4308 fdrop(fp, td);
4309 return (EINVAL);
4310 }
4311
4312 *fpp = fp;
4313 return (0);
4314 }
4315
4316 /*
4317 * Convert a user file descriptor to a kernel file entry and check
4318 * that, if it is a capability, the correct rights are present.
4319 * A reference on the file entry is held upon returning.
4320 */
4321 int
4322 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4323 {
4324 int error;
4325
4326 error = getvnode_path(td, fd, rightsp, fpp);
4327
4328 /*
4329 * Filter out O_PATH file descriptors, most getvnode() callers
4330 * do not call fo_ methods.
4331 */
4332 if (error == 0 && (*fpp)->f_ops == &path_fileops) {
4333 fdrop(*fpp, td);
4334 error = EBADF;
4335 }
4336
4337 return (error);
4338 }
4339
4340 /*
4341 * Get an (NFS) file handle.
4342 */
4343 #ifndef _SYS_SYSPROTO_H_
4344 struct lgetfh_args {
4345 char *fname;
4346 fhandle_t *fhp;
4347 };
4348 #endif
4349 int
4350 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4351 {
4352
4353 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4354 UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4355 }
4356
4357 #ifndef _SYS_SYSPROTO_H_
4358 struct getfh_args {
4359 char *fname;
4360 fhandle_t *fhp;
4361 };
4362 #endif
4363 int
4364 sys_getfh(struct thread *td, struct getfh_args *uap)
4365 {
4366
4367 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4368 uap->fhp, UIO_USERSPACE));
4369 }
4370
4371 /*
4372 * syscall for the rpc.lockd to use to translate an open descriptor into
4373 * a NFS file handle.
4374 *
4375 * warning: do not remove the priv_check() call or this becomes one giant
4376 * security hole.
4377 */
4378 #ifndef _SYS_SYSPROTO_H_
4379 struct getfhat_args {
4380 int fd;
4381 char *path;
4382 fhandle_t *fhp;
4383 int flags;
4384 };
4385 #endif
4386 int
4387 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4388 {
4389
4390 if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4391 return (EINVAL);
4392 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4393 uap->fhp, UIO_USERSPACE));
4394 }
4395
4396 int
4397 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4398 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4399 {
4400 struct nameidata nd;
4401 fhandle_t fh;
4402 struct vnode *vp;
4403 int error;
4404
4405 error = priv_check(td, PRIV_VFS_GETFH);
4406 if (error != 0)
4407 return (error);
4408 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4409 AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4410 fd, td);
4411 error = namei(&nd);
4412 if (error != 0)
4413 return (error);
4414 NDFREE_NOTHING(&nd);
4415 vp = nd.ni_vp;
4416 bzero(&fh, sizeof(fh));
4417 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4418 error = VOP_VPTOFH(vp, &fh.fh_fid);
4419 vput(vp);
4420 if (error == 0) {
4421 if (fhseg == UIO_USERSPACE)
4422 error = copyout(&fh, fhp, sizeof (fh));
4423 else
4424 memcpy(fhp, &fh, sizeof(fh));
4425 }
4426 return (error);
4427 }
4428
4429 #ifndef _SYS_SYSPROTO_H_
4430 struct fhlink_args {
4431 fhandle_t *fhp;
4432 const char *to;
4433 };
4434 #endif
4435 int
4436 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4437 {
4438
4439 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4440 }
4441
4442 #ifndef _SYS_SYSPROTO_H_
4443 struct fhlinkat_args {
4444 fhandle_t *fhp;
4445 int tofd;
4446 const char *to;
4447 };
4448 #endif
4449 int
4450 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4451 {
4452
4453 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4454 }
4455
4456 static int
4457 kern_fhlinkat(struct thread *td, int fd, const char *path,
4458 enum uio_seg pathseg, fhandle_t *fhp)
4459 {
4460 fhandle_t fh;
4461 struct mount *mp;
4462 struct vnode *vp;
4463 int error;
4464
4465 error = priv_check(td, PRIV_VFS_GETFH);
4466 if (error != 0)
4467 return (error);
4468 error = copyin(fhp, &fh, sizeof(fh));
4469 if (error != 0)
4470 return (error);
4471 do {
4472 bwillwrite();
4473 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4474 return (ESTALE);
4475 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4476 vfs_unbusy(mp);
4477 if (error != 0)
4478 return (error);
4479 VOP_UNLOCK(vp);
4480 error = kern_linkat_vp(td, vp, fd, path, pathseg);
4481 } while (error == EAGAIN || error == ERELOOKUP);
4482 return (error);
4483 }
4484
4485 #ifndef _SYS_SYSPROTO_H_
4486 struct fhreadlink_args {
4487 fhandle_t *fhp;
4488 char *buf;
4489 size_t bufsize;
4490 };
4491 #endif
4492 int
4493 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4494 {
4495 fhandle_t fh;
4496 struct mount *mp;
4497 struct vnode *vp;
4498 int error;
4499
4500 error = priv_check(td, PRIV_VFS_GETFH);
4501 if (error != 0)
4502 return (error);
4503 if (uap->bufsize > IOSIZE_MAX)
4504 return (EINVAL);
4505 error = copyin(uap->fhp, &fh, sizeof(fh));
4506 if (error != 0)
4507 return (error);
4508 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4509 return (ESTALE);
4510 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4511 vfs_unbusy(mp);
4512 if (error != 0)
4513 return (error);
4514 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4515 vput(vp);
4516 return (error);
4517 }
4518
4519 /*
4520 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4521 * open descriptor.
4522 *
4523 * warning: do not remove the priv_check() call or this becomes one giant
4524 * security hole.
4525 */
4526 #ifndef _SYS_SYSPROTO_H_
4527 struct fhopen_args {
4528 const struct fhandle *u_fhp;
4529 int flags;
4530 };
4531 #endif
4532 int
4533 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4534 {
4535 return (kern_fhopen(td, uap->u_fhp, uap->flags));
4536 }
4537
4538 int
4539 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4540 {
4541 struct mount *mp;
4542 struct vnode *vp;
4543 struct fhandle fhp;
4544 struct file *fp;
4545 int fmode, error;
4546 int indx;
4547
4548 error = priv_check(td, PRIV_VFS_FHOPEN);
4549 if (error != 0)
4550 return (error);
4551 indx = -1;
4552 fmode = FFLAGS(flags);
4553 /* why not allow a non-read/write open for our lockd? */
4554 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4555 return (EINVAL);
4556 error = copyin(u_fhp, &fhp, sizeof(fhp));
4557 if (error != 0)
4558 return(error);
4559 /* find the mount point */
4560 mp = vfs_busyfs(&fhp.fh_fsid);
4561 if (mp == NULL)
4562 return (ESTALE);
4563 /* now give me my vnode, it gets returned to me locked */
4564 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4565 vfs_unbusy(mp);
4566 if (error != 0)
4567 return (error);
4568
4569 error = falloc_noinstall(td, &fp);
4570 if (error != 0) {
4571 vput(vp);
4572 return (error);
4573 }
4574 /*
4575 * An extra reference on `fp' has been held for us by
4576 * falloc_noinstall().
4577 */
4578
4579 #ifdef INVARIANTS
4580 td->td_dupfd = -1;
4581 #endif
4582 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4583 if (error != 0) {
4584 KASSERT(fp->f_ops == &badfileops,
4585 ("VOP_OPEN in fhopen() set f_ops"));
4586 KASSERT(td->td_dupfd < 0,
4587 ("fhopen() encountered fdopen()"));
4588
4589 vput(vp);
4590 goto bad;
4591 }
4592 #ifdef INVARIANTS
4593 td->td_dupfd = 0;
4594 #endif
4595 fp->f_vnode = vp;
4596 finit_vnode(fp, fmode, NULL, &vnops);
4597 VOP_UNLOCK(vp);
4598 if ((fmode & O_TRUNC) != 0) {
4599 error = fo_truncate(fp, 0, td->td_ucred, td);
4600 if (error != 0)
4601 goto bad;
4602 }
4603
4604 error = finstall(td, fp, &indx, fmode, NULL);
4605 bad:
4606 fdrop(fp, td);
4607 td->td_retval[0] = indx;
4608 return (error);
4609 }
4610
4611 /*
4612 * Stat an (NFS) file handle.
4613 */
4614 #ifndef _SYS_SYSPROTO_H_
4615 struct fhstat_args {
4616 struct fhandle *u_fhp;
4617 struct stat *sb;
4618 };
4619 #endif
4620 int
4621 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4622 {
4623 struct stat sb;
4624 struct fhandle fh;
4625 int error;
4626
4627 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4628 if (error != 0)
4629 return (error);
4630 error = kern_fhstat(td, fh, &sb);
4631 if (error == 0)
4632 error = copyout(&sb, uap->sb, sizeof(sb));
4633 return (error);
4634 }
4635
4636 int
4637 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4638 {
4639 struct mount *mp;
4640 struct vnode *vp;
4641 int error;
4642
4643 error = priv_check(td, PRIV_VFS_FHSTAT);
4644 if (error != 0)
4645 return (error);
4646 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4647 return (ESTALE);
4648 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4649 vfs_unbusy(mp);
4650 if (error != 0)
4651 return (error);
4652 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4653 vput(vp);
4654 return (error);
4655 }
4656
4657 /*
4658 * Implement fstatfs() for (NFS) file handles.
4659 */
4660 #ifndef _SYS_SYSPROTO_H_
4661 struct fhstatfs_args {
4662 struct fhandle *u_fhp;
4663 struct statfs *buf;
4664 };
4665 #endif
4666 int
4667 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4668 {
4669 struct statfs *sfp;
4670 fhandle_t fh;
4671 int error;
4672
4673 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4674 if (error != 0)
4675 return (error);
4676 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4677 error = kern_fhstatfs(td, fh, sfp);
4678 if (error == 0)
4679 error = copyout(sfp, uap->buf, sizeof(*sfp));
4680 free(sfp, M_STATFS);
4681 return (error);
4682 }
4683
4684 int
4685 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4686 {
4687 struct mount *mp;
4688 struct vnode *vp;
4689 int error;
4690
4691 error = priv_check(td, PRIV_VFS_FHSTATFS);
4692 if (error != 0)
4693 return (error);
4694 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4695 return (ESTALE);
4696 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4697 if (error != 0) {
4698 vfs_unbusy(mp);
4699 return (error);
4700 }
4701 vput(vp);
4702 error = prison_canseemount(td->td_ucred, mp);
4703 if (error != 0)
4704 goto out;
4705 #ifdef MAC
4706 error = mac_mount_check_stat(td->td_ucred, mp);
4707 if (error != 0)
4708 goto out;
4709 #endif
4710 error = VFS_STATFS(mp, buf);
4711 out:
4712 vfs_unbusy(mp);
4713 return (error);
4714 }
4715
4716 /*
4717 * Unlike madvise(2), we do not make a best effort to remember every
4718 * possible caching hint. Instead, we remember the last setting with
4719 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4720 * region of any current setting.
4721 */
4722 int
4723 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4724 int advice)
4725 {
4726 struct fadvise_info *fa, *new;
4727 struct file *fp;
4728 struct vnode *vp;
4729 off_t end;
4730 int error;
4731
4732 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4733 return (EINVAL);
4734 AUDIT_ARG_VALUE(advice);
4735 switch (advice) {
4736 case POSIX_FADV_SEQUENTIAL:
4737 case POSIX_FADV_RANDOM:
4738 case POSIX_FADV_NOREUSE:
4739 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4740 break;
4741 case POSIX_FADV_NORMAL:
4742 case POSIX_FADV_WILLNEED:
4743 case POSIX_FADV_DONTNEED:
4744 new = NULL;
4745 break;
4746 default:
4747 return (EINVAL);
4748 }
4749 /* XXX: CAP_POSIX_FADVISE? */
4750 AUDIT_ARG_FD(fd);
4751 error = fget(td, fd, &cap_no_rights, &fp);
4752 if (error != 0)
4753 goto out;
4754 AUDIT_ARG_FILE(td->td_proc, fp);
4755 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4756 error = ESPIPE;
4757 goto out;
4758 }
4759 if (fp->f_type != DTYPE_VNODE) {
4760 error = ENODEV;
4761 goto out;
4762 }
4763 vp = fp->f_vnode;
4764 if (vp->v_type != VREG) {
4765 error = ENODEV;
4766 goto out;
4767 }
4768 if (len == 0)
4769 end = OFF_MAX;
4770 else
4771 end = offset + len - 1;
4772 switch (advice) {
4773 case POSIX_FADV_SEQUENTIAL:
4774 case POSIX_FADV_RANDOM:
4775 case POSIX_FADV_NOREUSE:
4776 /*
4777 * Try to merge any existing non-standard region with
4778 * this new region if possible, otherwise create a new
4779 * non-standard region for this request.
4780 */
4781 mtx_pool_lock(mtxpool_sleep, fp);
4782 fa = fp->f_advice;
4783 if (fa != NULL && fa->fa_advice == advice &&
4784 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4785 (end != OFF_MAX && fa->fa_start == end + 1) ||
4786 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4787 if (offset < fa->fa_start)
4788 fa->fa_start = offset;
4789 if (end > fa->fa_end)
4790 fa->fa_end = end;
4791 } else {
4792 new->fa_advice = advice;
4793 new->fa_start = offset;
4794 new->fa_end = end;
4795 fp->f_advice = new;
4796 new = fa;
4797 }
4798 mtx_pool_unlock(mtxpool_sleep, fp);
4799 break;
4800 case POSIX_FADV_NORMAL:
4801 /*
4802 * If a the "normal" region overlaps with an existing
4803 * non-standard region, trim or remove the
4804 * non-standard region.
4805 */
4806 mtx_pool_lock(mtxpool_sleep, fp);
4807 fa = fp->f_advice;
4808 if (fa != NULL) {
4809 if (offset <= fa->fa_start && end >= fa->fa_end) {
4810 new = fa;
4811 fp->f_advice = NULL;
4812 } else if (offset <= fa->fa_start &&
4813 end >= fa->fa_start)
4814 fa->fa_start = end + 1;
4815 else if (offset <= fa->fa_end && end >= fa->fa_end)
4816 fa->fa_end = offset - 1;
4817 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4818 /*
4819 * If the "normal" region is a middle
4820 * portion of the existing
4821 * non-standard region, just remove
4822 * the whole thing rather than picking
4823 * one side or the other to
4824 * preserve.
4825 */
4826 new = fa;
4827 fp->f_advice = NULL;
4828 }
4829 }
4830 mtx_pool_unlock(mtxpool_sleep, fp);
4831 break;
4832 case POSIX_FADV_WILLNEED:
4833 case POSIX_FADV_DONTNEED:
4834 error = VOP_ADVISE(vp, offset, end, advice);
4835 break;
4836 }
4837 out:
4838 if (fp != NULL)
4839 fdrop(fp, td);
4840 free(new, M_FADVISE);
4841 return (error);
4842 }
4843
4844 int
4845 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4846 {
4847 int error;
4848
4849 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4850 uap->advice);
4851 return (kern_posix_error(td, error));
4852 }
4853
4854 int
4855 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4856 off_t *outoffp, size_t len, unsigned int flags)
4857 {
4858 struct file *infp, *outfp;
4859 struct vnode *invp, *outvp;
4860 int error;
4861 size_t retlen;
4862 void *rl_rcookie, *rl_wcookie;
4863 off_t savinoff, savoutoff;
4864
4865 infp = outfp = NULL;
4866 rl_rcookie = rl_wcookie = NULL;
4867 savinoff = -1;
4868 error = 0;
4869 retlen = 0;
4870
4871 if (flags != 0) {
4872 error = EINVAL;
4873 goto out;
4874 }
4875 if (len > SSIZE_MAX)
4876 /*
4877 * Although the len argument is size_t, the return argument
4878 * is ssize_t (which is signed). Therefore a size that won't
4879 * fit in ssize_t can't be returned.
4880 */
4881 len = SSIZE_MAX;
4882
4883 /* Get the file structures for the file descriptors. */
4884 error = fget_read(td, infd, &cap_read_rights, &infp);
4885 if (error != 0)
4886 goto out;
4887 if (infp->f_ops == &badfileops) {
4888 error = EBADF;
4889 goto out;
4890 }
4891 if (infp->f_vnode == NULL) {
4892 error = EINVAL;
4893 goto out;
4894 }
4895 error = fget_write(td, outfd, &cap_write_rights, &outfp);
4896 if (error != 0)
4897 goto out;
4898 if (outfp->f_ops == &badfileops) {
4899 error = EBADF;
4900 goto out;
4901 }
4902 if (outfp->f_vnode == NULL) {
4903 error = EINVAL;
4904 goto out;
4905 }
4906
4907 /* Set the offset pointers to the correct place. */
4908 if (inoffp == NULL)
4909 inoffp = &infp->f_offset;
4910 if (outoffp == NULL)
4911 outoffp = &outfp->f_offset;
4912 savinoff = *inoffp;
4913 savoutoff = *outoffp;
4914
4915 invp = infp->f_vnode;
4916 outvp = outfp->f_vnode;
4917 /* Sanity check the f_flag bits. */
4918 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4919 (infp->f_flag & FREAD) == 0) {
4920 error = EBADF;
4921 goto out;
4922 }
4923
4924 /* If len == 0, just return 0. */
4925 if (len == 0)
4926 goto out;
4927
4928 /*
4929 * If infp and outfp refer to the same file, the byte ranges cannot
4930 * overlap.
4931 */
4932 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4933 savoutoff) || (savinoff > savoutoff && savoutoff + len >
4934 savinoff))) {
4935 error = EINVAL;
4936 goto out;
4937 }
4938
4939 /* Range lock the byte ranges for both invp and outvp. */
4940 for (;;) {
4941 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4942 len);
4943 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4944 len);
4945 if (rl_rcookie != NULL)
4946 break;
4947 vn_rangelock_unlock(outvp, rl_wcookie);
4948 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4949 vn_rangelock_unlock(invp, rl_rcookie);
4950 }
4951
4952 retlen = len;
4953 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4954 flags, infp->f_cred, outfp->f_cred, td);
4955 out:
4956 if (rl_rcookie != NULL)
4957 vn_rangelock_unlock(invp, rl_rcookie);
4958 if (rl_wcookie != NULL)
4959 vn_rangelock_unlock(outvp, rl_wcookie);
4960 if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4961 *inoffp = savinoff;
4962 *outoffp = savoutoff;
4963 }
4964 if (outfp != NULL)
4965 fdrop(outfp, td);
4966 if (infp != NULL)
4967 fdrop(infp, td);
4968 td->td_retval[0] = retlen;
4969 return (error);
4970 }
4971
4972 int
4973 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4974 {
4975 off_t inoff, outoff, *inoffp, *outoffp;
4976 int error;
4977
4978 inoffp = outoffp = NULL;
4979 if (uap->inoffp != NULL) {
4980 error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4981 if (error != 0)
4982 return (error);
4983 inoffp = &inoff;
4984 }
4985 if (uap->outoffp != NULL) {
4986 error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4987 if (error != 0)
4988 return (error);
4989 outoffp = &outoff;
4990 }
4991 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4992 outoffp, uap->len, uap->flags);
4993 if (error == 0 && uap->inoffp != NULL)
4994 error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4995 if (error == 0 && uap->outoffp != NULL)
4996 error = copyout(outoffp, uap->outoffp, sizeof(off_t));
4997 return (error);
4998 }
Cache object: 640354bfbf8ffbbd6a57e8dabbe3efeb
|