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
at2cnpflags(u_int at_flags,u_int mask)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 return (res);
133 }
134
135 int
kern_sync(struct thread * td)136 kern_sync(struct thread *td)
137 {
138 struct mount *mp, *nmp;
139 int save;
140
141 mtx_lock(&mountlist_mtx);
142 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
143 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
144 nmp = TAILQ_NEXT(mp, mnt_list);
145 continue;
146 }
147 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
148 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
149 save = curthread_pflags_set(TDP_SYNCIO);
150 vfs_periodic(mp, MNT_NOWAIT);
151 VFS_SYNC(mp, MNT_NOWAIT);
152 curthread_pflags_restore(save);
153 vn_finished_write(mp);
154 }
155 mtx_lock(&mountlist_mtx);
156 nmp = TAILQ_NEXT(mp, mnt_list);
157 vfs_unbusy(mp);
158 }
159 mtx_unlock(&mountlist_mtx);
160 return (0);
161 }
162
163 /*
164 * Sync each mounted filesystem.
165 */
166 #ifndef _SYS_SYSPROTO_H_
167 struct sync_args {
168 int dummy;
169 };
170 #endif
171 /* ARGSUSED */
172 int
sys_sync(struct thread * td,struct sync_args * uap)173 sys_sync(struct thread *td, struct sync_args *uap)
174 {
175
176 return (kern_sync(td));
177 }
178
179 /*
180 * Change filesystem quotas.
181 */
182 #ifndef _SYS_SYSPROTO_H_
183 struct quotactl_args {
184 char *path;
185 int cmd;
186 int uid;
187 caddr_t arg;
188 };
189 #endif
190 int
sys_quotactl(struct thread * td,struct quotactl_args * uap)191 sys_quotactl(struct thread *td, struct quotactl_args *uap)
192 {
193 struct mount *mp;
194 struct nameidata nd;
195 int error;
196
197 AUDIT_ARG_CMD(uap->cmd);
198 AUDIT_ARG_UID(uap->uid);
199 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
200 return (EPERM);
201 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
202 uap->path, td);
203 if ((error = namei(&nd)) != 0)
204 return (error);
205 NDFREE(&nd, NDF_ONLY_PNBUF);
206 mp = nd.ni_vp->v_mount;
207 vfs_ref(mp);
208 vput(nd.ni_vp);
209 error = vfs_busy(mp, 0);
210 if (error != 0) {
211 vfs_rel(mp);
212 return (error);
213 }
214 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
215
216 /*
217 * Since quota on operation typically needs to open quota
218 * file, the Q_QUOTAON handler needs to unbusy the mount point
219 * before calling into namei. Otherwise, unmount might be
220 * started between two vfs_busy() invocations (first is our,
221 * second is from mount point cross-walk code in lookup()),
222 * causing deadlock.
223 *
224 * Require that Q_QUOTAON handles the vfs_busy() reference on
225 * its own, always returning with ubusied mount point.
226 */
227 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
228 (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
229 vfs_unbusy(mp);
230 vfs_rel(mp);
231 return (error);
232 }
233
234 /*
235 * Used by statfs conversion routines to scale the block size up if
236 * necessary so that all of the block counts are <= 'max_size'. Note
237 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
238 * value of 'n'.
239 */
240 void
statfs_scale_blocks(struct statfs * sf,long max_size)241 statfs_scale_blocks(struct statfs *sf, long max_size)
242 {
243 uint64_t count;
244 int shift;
245
246 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
247
248 /*
249 * Attempt to scale the block counts to give a more accurate
250 * overview to userland of the ratio of free space to used
251 * space. To do this, find the largest block count and compute
252 * a divisor that lets it fit into a signed integer <= max_size.
253 */
254 if (sf->f_bavail < 0)
255 count = -sf->f_bavail;
256 else
257 count = sf->f_bavail;
258 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
259 if (count <= max_size)
260 return;
261
262 count >>= flsl(max_size);
263 shift = 0;
264 while (count > 0) {
265 shift++;
266 count >>=1;
267 }
268
269 sf->f_bsize <<= shift;
270 sf->f_blocks >>= shift;
271 sf->f_bfree >>= shift;
272 sf->f_bavail >>= shift;
273 }
274
275 static int
kern_do_statfs(struct thread * td,struct mount * mp,struct statfs * buf)276 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
277 {
278 int error;
279
280 if (mp == NULL)
281 return (EBADF);
282 error = vfs_busy(mp, 0);
283 vfs_rel(mp);
284 if (error != 0)
285 return (error);
286 #ifdef MAC
287 error = mac_mount_check_stat(td->td_ucred, mp);
288 if (error != 0)
289 goto out;
290 #endif
291 error = VFS_STATFS(mp, buf);
292 if (error != 0)
293 goto out;
294 if (priv_check_cred_vfs_generation(td->td_ucred)) {
295 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
296 prison_enforce_statfs(td->td_ucred, mp, buf);
297 }
298 out:
299 vfs_unbusy(mp);
300 return (error);
301 }
302
303 /*
304 * Get filesystem statistics.
305 */
306 #ifndef _SYS_SYSPROTO_H_
307 struct statfs_args {
308 char *path;
309 struct statfs *buf;
310 };
311 #endif
312 int
sys_statfs(struct thread * td,struct statfs_args * uap)313 sys_statfs(struct thread *td, struct statfs_args *uap)
314 {
315 struct statfs *sfp;
316 int error;
317
318 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
319 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
320 if (error == 0)
321 error = copyout(sfp, uap->buf, sizeof(struct statfs));
322 free(sfp, M_STATFS);
323 return (error);
324 }
325
326 int
kern_statfs(struct thread * td,const char * path,enum uio_seg pathseg,struct statfs * buf)327 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
328 struct statfs *buf)
329 {
330 struct mount *mp;
331 struct nameidata nd;
332 int error;
333
334 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
335 pathseg, path, td);
336 error = namei(&nd);
337 if (error != 0)
338 return (error);
339 mp = nd.ni_vp->v_mount;
340 vfs_ref(mp);
341 NDFREE_NOTHING(&nd);
342 vput(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
sys_fstatfs(struct thread * td,struct fstatfs_args * uap)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
kern_fstatfs(struct thread * td,int fd,struct statfs * buf)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(td, fd, &cap_fstatfs_rights, &fp);
379 if (error != 0)
380 return (error);
381 vp = fp->f_vnode;
382 vn_lock(vp, LK_SHARED | LK_RETRY);
383 #ifdef AUDIT
384 AUDIT_ARG_VNODE1(vp);
385 #endif
386 mp = vp->v_mount;
387 if (mp != NULL)
388 vfs_ref(mp);
389 VOP_UNLOCK(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
sys_getfsstat(struct thread * td,struct getfsstat_args * uap)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
kern_getfsstat(struct thread * td,struct statfs ** buf,size_t bufsize,size_t * countp,enum uio_seg bufseg,int mode)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
freebsd4_statfs(struct thread * td,struct freebsd4_statfs_args * uap)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
freebsd4_fstatfs(struct thread * td,struct freebsd4_fstatfs_args * uap)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
freebsd4_getfsstat(struct thread * td,struct freebsd4_getfsstat_args * uap)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
freebsd4_fhstatfs(struct thread * td,struct freebsd4_fhstatfs_args * uap)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
freebsd4_cvtstatfs(struct statfs * nsp,struct ostatfs * osp)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
freebsd11_statfs(struct thread * td,struct freebsd11_statfs_args * uap)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
freebsd11_fstatfs(struct thread * td,struct freebsd11_fstatfs_args * uap)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
freebsd11_getfsstat(struct thread * td,struct freebsd11_getfsstat_args * uap)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
freebsd11_fhstatfs(struct thread * td,struct freebsd11_fhstatfs_args * uap)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
freebsd11_cvtstatfs(struct statfs * nsp,struct freebsd11_statfs * osp)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
sys_fchdir(struct thread * td,struct fchdir_args * uap)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(td, uap->fd, &cap_fchdir_rights,
895 &fp);
896 if (error != 0)
897 return (error);
898 vp = fp->f_vnode;
899 vrefact(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
sys_chdir(struct thread * td,struct chdir_args * uap)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
kern_chdir(struct thread * td,const char * path,enum uio_seg pathseg)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
sys_chroot(struct thread * td,struct chroot_args * uap)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
change_dir(struct vnode * vp,struct thread * td)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
flags_to_rights(int flags,cap_rights_t * rightsp)1024 flags_to_rights(int flags, cap_rights_t *rightsp)
1025 {
1026
1027 if (flags & O_EXEC) {
1028 cap_rights_set_one(rightsp, CAP_FEXECVE);
1029 } else {
1030 switch ((flags & O_ACCMODE)) {
1031 case O_RDONLY:
1032 cap_rights_set_one(rightsp, CAP_READ);
1033 break;
1034 case O_RDWR:
1035 cap_rights_set_one(rightsp, CAP_READ);
1036 /* FALLTHROUGH */
1037 case O_WRONLY:
1038 cap_rights_set_one(rightsp, CAP_WRITE);
1039 if (!(flags & (O_APPEND | O_TRUNC)))
1040 cap_rights_set_one(rightsp, CAP_SEEK);
1041 break;
1042 }
1043 }
1044
1045 if (flags & O_CREAT)
1046 cap_rights_set_one(rightsp, CAP_CREATE);
1047
1048 if (flags & O_TRUNC)
1049 cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1050
1051 if (flags & (O_SYNC | O_FSYNC))
1052 cap_rights_set_one(rightsp, CAP_FSYNC);
1053
1054 if (flags & (O_EXLOCK | O_SHLOCK))
1055 cap_rights_set_one(rightsp, CAP_FLOCK);
1056 }
1057
1058 /*
1059 * Check permissions, allocate an open file structure, and call the device
1060 * open routine if any.
1061 */
1062 #ifndef _SYS_SYSPROTO_H_
1063 struct open_args {
1064 char *path;
1065 int flags;
1066 int mode;
1067 };
1068 #endif
1069 int
sys_open(struct thread * td,struct open_args * uap)1070 sys_open(struct thread *td, struct open_args *uap)
1071 {
1072
1073 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1074 uap->flags, uap->mode));
1075 }
1076
1077 #ifndef _SYS_SYSPROTO_H_
1078 struct openat_args {
1079 int fd;
1080 char *path;
1081 int flag;
1082 int mode;
1083 };
1084 #endif
1085 int
sys_openat(struct thread * td,struct openat_args * uap)1086 sys_openat(struct thread *td, struct openat_args *uap)
1087 {
1088
1089 AUDIT_ARG_FD(uap->fd);
1090 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1091 uap->mode));
1092 }
1093
1094 int
kern_openat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int flags,int mode)1095 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1096 int flags, int mode)
1097 {
1098 struct proc *p = td->td_proc;
1099 struct filedesc *fdp;
1100 struct pwddesc *pdp;
1101 struct file *fp;
1102 struct vnode *vp;
1103 struct nameidata nd;
1104 cap_rights_t rights;
1105 int cmode, error, indx;
1106
1107 indx = -1;
1108 fdp = p->p_fd;
1109 pdp = p->p_pd;
1110
1111 AUDIT_ARG_FFLAGS(flags);
1112 AUDIT_ARG_MODE(mode);
1113 cap_rights_init_one(&rights, CAP_LOOKUP);
1114 flags_to_rights(flags, &rights);
1115 /*
1116 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1117 * may be specified.
1118 */
1119 if (flags & O_EXEC) {
1120 if (flags & O_ACCMODE)
1121 return (EINVAL);
1122 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1123 return (EINVAL);
1124 } else {
1125 flags = FFLAGS(flags);
1126 }
1127
1128 /*
1129 * Allocate a file structure. The descriptor to reference it
1130 * is allocated and used by finstall_refed() below.
1131 */
1132 error = falloc_noinstall(td, &fp);
1133 if (error != 0)
1134 return (error);
1135 /* Set the flags early so the finit in devfs can pick them up. */
1136 fp->f_flag = flags & FMASK;
1137 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1138 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1139 &rights, td);
1140 td->td_dupfd = -1; /* XXX check for fdopen */
1141 error = vn_open(&nd, &flags, cmode, fp);
1142 if (error != 0) {
1143 /*
1144 * If the vn_open replaced the method vector, something
1145 * wonderous happened deep below and we just pass it up
1146 * pretending we know what we do.
1147 */
1148 if (error == ENXIO && fp->f_ops != &badfileops)
1149 goto success;
1150
1151 /*
1152 * Handle special fdopen() case. bleh.
1153 *
1154 * Don't do this for relative (capability) lookups; we don't
1155 * understand exactly what would happen, and we don't think
1156 * that it ever should.
1157 */
1158 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1159 (error == ENODEV || error == ENXIO) &&
1160 td->td_dupfd >= 0) {
1161 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1162 &indx);
1163 if (error == 0)
1164 goto success;
1165 }
1166
1167 goto bad;
1168 }
1169 td->td_dupfd = 0;
1170 NDFREE(&nd, NDF_ONLY_PNBUF);
1171 vp = nd.ni_vp;
1172
1173 /*
1174 * Store the vnode, for any f_type. Typically, the vnode use
1175 * count is decremented by direct call to vn_closefile() for
1176 * files that switched type in the cdevsw fdopen() method.
1177 */
1178 fp->f_vnode = vp;
1179 /*
1180 * If the file wasn't claimed by devfs bind it to the normal
1181 * vnode operations here.
1182 */
1183 if (fp->f_ops == &badfileops) {
1184 KASSERT(vp->v_type != VFIFO,
1185 ("Unexpected fifo fp %p vp %p", fp, vp));
1186 finit_vnode(fp, flags, NULL, &vnops);
1187 }
1188
1189 VOP_UNLOCK(vp);
1190 if (flags & O_TRUNC) {
1191 error = fo_truncate(fp, 0, td->td_ucred, td);
1192 if (error != 0)
1193 goto bad;
1194 }
1195 success:
1196 /*
1197 * If we haven't already installed the FD (for dupfdopen), do so now.
1198 */
1199 if (indx == -1) {
1200 struct filecaps *fcaps;
1201
1202 #ifdef CAPABILITIES
1203 if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1204 fcaps = &nd.ni_filecaps;
1205 else
1206 #endif
1207 fcaps = NULL;
1208 error = finstall_refed(td, fp, &indx, flags, fcaps);
1209 /* On success finstall_refed() consumes fcaps. */
1210 if (error != 0) {
1211 filecaps_free(&nd.ni_filecaps);
1212 goto bad;
1213 }
1214 } else {
1215 filecaps_free(&nd.ni_filecaps);
1216 falloc_abort(td, fp);
1217 }
1218
1219 td->td_retval[0] = indx;
1220 return (0);
1221 bad:
1222 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1223 falloc_abort(td, fp);
1224 return (error);
1225 }
1226
1227 #ifdef COMPAT_43
1228 /*
1229 * Create a file.
1230 */
1231 #ifndef _SYS_SYSPROTO_H_
1232 struct ocreat_args {
1233 char *path;
1234 int mode;
1235 };
1236 #endif
1237 int
ocreat(struct thread * td,struct ocreat_args * uap)1238 ocreat(struct thread *td, struct ocreat_args *uap)
1239 {
1240
1241 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1242 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1243 }
1244 #endif /* COMPAT_43 */
1245
1246 /*
1247 * Create a special file.
1248 */
1249 #ifndef _SYS_SYSPROTO_H_
1250 struct mknodat_args {
1251 int fd;
1252 char *path;
1253 mode_t mode;
1254 dev_t dev;
1255 };
1256 #endif
1257 int
sys_mknodat(struct thread * td,struct mknodat_args * uap)1258 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1259 {
1260
1261 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1262 uap->dev));
1263 }
1264
1265 #if defined(COMPAT_FREEBSD11)
1266 int
freebsd11_mknod(struct thread * td,struct freebsd11_mknod_args * uap)1267 freebsd11_mknod(struct thread *td,
1268 struct freebsd11_mknod_args *uap)
1269 {
1270
1271 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1272 uap->mode, uap->dev));
1273 }
1274
1275 int
freebsd11_mknodat(struct thread * td,struct freebsd11_mknodat_args * uap)1276 freebsd11_mknodat(struct thread *td,
1277 struct freebsd11_mknodat_args *uap)
1278 {
1279
1280 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1281 uap->dev));
1282 }
1283 #endif /* COMPAT_FREEBSD11 */
1284
1285 int
kern_mknodat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int mode,dev_t dev)1286 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1287 int mode, dev_t dev)
1288 {
1289 struct vnode *vp;
1290 struct mount *mp;
1291 struct vattr vattr;
1292 struct nameidata nd;
1293 int error, whiteout = 0;
1294
1295 AUDIT_ARG_MODE(mode);
1296 AUDIT_ARG_DEV(dev);
1297 switch (mode & S_IFMT) {
1298 case S_IFCHR:
1299 case S_IFBLK:
1300 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1301 if (error == 0 && dev == VNOVAL)
1302 error = EINVAL;
1303 break;
1304 case S_IFWHT:
1305 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1306 break;
1307 case S_IFIFO:
1308 if (dev == 0)
1309 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1310 /* FALLTHROUGH */
1311 default:
1312 error = EINVAL;
1313 break;
1314 }
1315 if (error != 0)
1316 return (error);
1317 restart:
1318 bwillwrite();
1319 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1320 NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1321 td);
1322 if ((error = namei(&nd)) != 0)
1323 return (error);
1324 vp = nd.ni_vp;
1325 if (vp != NULL) {
1326 NDFREE(&nd, NDF_ONLY_PNBUF);
1327 if (vp == nd.ni_dvp)
1328 vrele(nd.ni_dvp);
1329 else
1330 vput(nd.ni_dvp);
1331 vrele(vp);
1332 return (EEXIST);
1333 } else {
1334 VATTR_NULL(&vattr);
1335 vattr.va_mode = (mode & ALLPERMS) &
1336 ~td->td_proc->p_pd->pd_cmask;
1337 vattr.va_rdev = dev;
1338 whiteout = 0;
1339
1340 switch (mode & S_IFMT) {
1341 case S_IFCHR:
1342 vattr.va_type = VCHR;
1343 break;
1344 case S_IFBLK:
1345 vattr.va_type = VBLK;
1346 break;
1347 case S_IFWHT:
1348 whiteout = 1;
1349 break;
1350 default:
1351 panic("kern_mknod: invalid mode");
1352 }
1353 }
1354 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1355 NDFREE(&nd, NDF_ONLY_PNBUF);
1356 vput(nd.ni_dvp);
1357 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1358 return (error);
1359 goto restart;
1360 }
1361 #ifdef MAC
1362 if (error == 0 && !whiteout)
1363 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1364 &nd.ni_cnd, &vattr);
1365 #endif
1366 if (error == 0) {
1367 if (whiteout)
1368 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1369 else {
1370 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1371 &nd.ni_cnd, &vattr);
1372 }
1373 }
1374 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1375 true);
1376 vn_finished_write(mp);
1377 NDFREE(&nd, NDF_ONLY_PNBUF);
1378 if (error == ERELOOKUP)
1379 goto restart;
1380 return (error);
1381 }
1382
1383 /*
1384 * Create a named pipe.
1385 */
1386 #ifndef _SYS_SYSPROTO_H_
1387 struct mkfifo_args {
1388 char *path;
1389 int mode;
1390 };
1391 #endif
1392 int
sys_mkfifo(struct thread * td,struct mkfifo_args * uap)1393 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1394 {
1395
1396 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1397 uap->mode));
1398 }
1399
1400 #ifndef _SYS_SYSPROTO_H_
1401 struct mkfifoat_args {
1402 int fd;
1403 char *path;
1404 mode_t mode;
1405 };
1406 #endif
1407 int
sys_mkfifoat(struct thread * td,struct mkfifoat_args * uap)1408 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1409 {
1410
1411 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1412 uap->mode));
1413 }
1414
1415 int
kern_mkfifoat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int mode)1416 kern_mkfifoat(struct thread *td, int fd, const char *path,
1417 enum uio_seg pathseg, int mode)
1418 {
1419 struct mount *mp;
1420 struct vattr vattr;
1421 struct nameidata nd;
1422 int error;
1423
1424 AUDIT_ARG_MODE(mode);
1425 restart:
1426 bwillwrite();
1427 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1428 NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1429 td);
1430 if ((error = namei(&nd)) != 0)
1431 return (error);
1432 if (nd.ni_vp != NULL) {
1433 NDFREE(&nd, NDF_ONLY_PNBUF);
1434 if (nd.ni_vp == nd.ni_dvp)
1435 vrele(nd.ni_dvp);
1436 else
1437 vput(nd.ni_dvp);
1438 vrele(nd.ni_vp);
1439 return (EEXIST);
1440 }
1441 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1442 NDFREE(&nd, NDF_ONLY_PNBUF);
1443 vput(nd.ni_dvp);
1444 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1445 return (error);
1446 goto restart;
1447 }
1448 VATTR_NULL(&vattr);
1449 vattr.va_type = VFIFO;
1450 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1451 #ifdef MAC
1452 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1453 &vattr);
1454 if (error != 0)
1455 goto out;
1456 #endif
1457 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1458 #ifdef MAC
1459 out:
1460 #endif
1461 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1462 vn_finished_write(mp);
1463 NDFREE(&nd, NDF_ONLY_PNBUF);
1464 if (error == ERELOOKUP)
1465 goto restart;
1466 return (error);
1467 }
1468
1469 /*
1470 * Make a hard file link.
1471 */
1472 #ifndef _SYS_SYSPROTO_H_
1473 struct link_args {
1474 char *path;
1475 char *link;
1476 };
1477 #endif
1478 int
sys_link(struct thread * td,struct link_args * uap)1479 sys_link(struct thread *td, struct link_args *uap)
1480 {
1481
1482 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1483 UIO_USERSPACE, FOLLOW));
1484 }
1485
1486 #ifndef _SYS_SYSPROTO_H_
1487 struct linkat_args {
1488 int fd1;
1489 char *path1;
1490 int fd2;
1491 char *path2;
1492 int flag;
1493 };
1494 #endif
1495 int
sys_linkat(struct thread * td,struct linkat_args * uap)1496 sys_linkat(struct thread *td, struct linkat_args *uap)
1497 {
1498 int flag;
1499
1500 flag = uap->flag;
1501 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH)) != 0)
1502 return (EINVAL);
1503
1504 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1505 UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW |
1506 AT_RESOLVE_BENEATH)));
1507 }
1508
1509 int hardlink_check_uid = 0;
1510 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1511 &hardlink_check_uid, 0,
1512 "Unprivileged processes cannot create hard links to files owned by other "
1513 "users");
1514 static int hardlink_check_gid = 0;
1515 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1516 &hardlink_check_gid, 0,
1517 "Unprivileged processes cannot create hard links to files owned by other "
1518 "groups");
1519
1520 static int
can_hardlink(struct vnode * vp,struct ucred * cred)1521 can_hardlink(struct vnode *vp, struct ucred *cred)
1522 {
1523 struct vattr va;
1524 int error;
1525
1526 if (!hardlink_check_uid && !hardlink_check_gid)
1527 return (0);
1528
1529 error = VOP_GETATTR(vp, &va, cred);
1530 if (error != 0)
1531 return (error);
1532
1533 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1534 error = priv_check_cred(cred, PRIV_VFS_LINK);
1535 if (error != 0)
1536 return (error);
1537 }
1538
1539 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1540 error = priv_check_cred(cred, PRIV_VFS_LINK);
1541 if (error != 0)
1542 return (error);
1543 }
1544
1545 return (0);
1546 }
1547
1548 int
kern_linkat(struct thread * td,int fd1,int fd2,const char * path1,const char * path2,enum uio_seg segflag,int follow)1549 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1550 const char *path2, enum uio_seg segflag, int follow)
1551 {
1552 struct nameidata nd;
1553 int error;
1554
1555 do {
1556 bwillwrite();
1557 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1558 path1, fd1, &cap_linkat_source_rights, td);
1559 if ((error = namei(&nd)) != 0)
1560 return (error);
1561 NDFREE(&nd, NDF_ONLY_PNBUF);
1562 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1563 } while (error == EAGAIN || error == ERELOOKUP);
1564 return (error);
1565 }
1566
1567 static int
kern_linkat_vp(struct thread * td,struct vnode * vp,int fd,const char * path,enum uio_seg segflag)1568 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1569 enum uio_seg segflag)
1570 {
1571 struct nameidata nd;
1572 struct mount *mp;
1573 int error;
1574
1575 if (vp->v_type == VDIR) {
1576 vrele(vp);
1577 return (EPERM); /* POSIX */
1578 }
1579 NDINIT_ATRIGHTS(&nd, CREATE,
1580 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1581 &cap_linkat_target_rights, td);
1582 if ((error = namei(&nd)) == 0) {
1583 if (nd.ni_vp != NULL) {
1584 NDFREE(&nd, NDF_ONLY_PNBUF);
1585 if (nd.ni_dvp == nd.ni_vp)
1586 vrele(nd.ni_dvp);
1587 else
1588 vput(nd.ni_dvp);
1589 vrele(nd.ni_vp);
1590 vrele(vp);
1591 return (EEXIST);
1592 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1593 /*
1594 * Cross-device link. No need to recheck
1595 * vp->v_type, since it cannot change, except
1596 * to VBAD.
1597 */
1598 NDFREE(&nd, NDF_ONLY_PNBUF);
1599 vput(nd.ni_dvp);
1600 vrele(vp);
1601 return (EXDEV);
1602 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1603 error = can_hardlink(vp, td->td_ucred);
1604 #ifdef MAC
1605 if (error == 0)
1606 error = mac_vnode_check_link(td->td_ucred,
1607 nd.ni_dvp, vp, &nd.ni_cnd);
1608 #endif
1609 if (error != 0) {
1610 vput(vp);
1611 vput(nd.ni_dvp);
1612 NDFREE(&nd, NDF_ONLY_PNBUF);
1613 return (error);
1614 }
1615 error = vn_start_write(vp, &mp, V_NOWAIT);
1616 if (error != 0) {
1617 vput(vp);
1618 vput(nd.ni_dvp);
1619 NDFREE(&nd, NDF_ONLY_PNBUF);
1620 error = vn_start_write(NULL, &mp,
1621 V_XSLEEP | PCATCH);
1622 if (error != 0)
1623 return (error);
1624 return (EAGAIN);
1625 }
1626 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1627 VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1628 vn_finished_write(mp);
1629 NDFREE(&nd, NDF_ONLY_PNBUF);
1630 vp = NULL;
1631 } else {
1632 vput(nd.ni_dvp);
1633 NDFREE(&nd, NDF_ONLY_PNBUF);
1634 vrele(vp);
1635 return (EAGAIN);
1636 }
1637 }
1638 if (vp != NULL)
1639 vrele(vp);
1640 return (error);
1641 }
1642
1643 /*
1644 * Make a symbolic link.
1645 */
1646 #ifndef _SYS_SYSPROTO_H_
1647 struct symlink_args {
1648 char *path;
1649 char *link;
1650 };
1651 #endif
1652 int
sys_symlink(struct thread * td,struct symlink_args * uap)1653 sys_symlink(struct thread *td, struct symlink_args *uap)
1654 {
1655
1656 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1657 UIO_USERSPACE));
1658 }
1659
1660 #ifndef _SYS_SYSPROTO_H_
1661 struct symlinkat_args {
1662 char *path;
1663 int fd;
1664 char *path2;
1665 };
1666 #endif
1667 int
sys_symlinkat(struct thread * td,struct symlinkat_args * uap)1668 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1669 {
1670
1671 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1672 UIO_USERSPACE));
1673 }
1674
1675 int
kern_symlinkat(struct thread * td,const char * path1,int fd,const char * path2,enum uio_seg segflg)1676 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1677 enum uio_seg segflg)
1678 {
1679 struct mount *mp;
1680 struct vattr vattr;
1681 const char *syspath;
1682 char *tmppath;
1683 struct nameidata nd;
1684 int error;
1685
1686 if (segflg == UIO_SYSSPACE) {
1687 syspath = path1;
1688 } else {
1689 tmppath = uma_zalloc(namei_zone, M_WAITOK);
1690 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1691 goto out;
1692 syspath = tmppath;
1693 }
1694 AUDIT_ARG_TEXT(syspath);
1695 restart:
1696 bwillwrite();
1697 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1698 NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1699 td);
1700 if ((error = namei(&nd)) != 0)
1701 goto out;
1702 if (nd.ni_vp) {
1703 NDFREE(&nd, NDF_ONLY_PNBUF);
1704 if (nd.ni_vp == nd.ni_dvp)
1705 vrele(nd.ni_dvp);
1706 else
1707 vput(nd.ni_dvp);
1708 vrele(nd.ni_vp);
1709 nd.ni_vp = NULL;
1710 error = EEXIST;
1711 goto out;
1712 }
1713 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1714 NDFREE(&nd, NDF_ONLY_PNBUF);
1715 vput(nd.ni_dvp);
1716 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1717 goto out;
1718 goto restart;
1719 }
1720 VATTR_NULL(&vattr);
1721 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1722 #ifdef MAC
1723 vattr.va_type = VLNK;
1724 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1725 &vattr);
1726 if (error != 0)
1727 goto out2;
1728 #endif
1729 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1730 #ifdef MAC
1731 out2:
1732 #endif
1733 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1734 vn_finished_write(mp);
1735 NDFREE(&nd, NDF_ONLY_PNBUF);
1736 if (error == ERELOOKUP)
1737 goto restart;
1738 out:
1739 if (segflg != UIO_SYSSPACE)
1740 uma_zfree(namei_zone, tmppath);
1741 return (error);
1742 }
1743
1744 /*
1745 * Delete a whiteout from the filesystem.
1746 */
1747 #ifndef _SYS_SYSPROTO_H_
1748 struct undelete_args {
1749 char *path;
1750 };
1751 #endif
1752 int
sys_undelete(struct thread * td,struct undelete_args * uap)1753 sys_undelete(struct thread *td, struct undelete_args *uap)
1754 {
1755 struct mount *mp;
1756 struct nameidata nd;
1757 int error;
1758
1759 restart:
1760 bwillwrite();
1761 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1762 UIO_USERSPACE, uap->path, td);
1763 error = namei(&nd);
1764 if (error != 0)
1765 return (error);
1766
1767 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1768 NDFREE(&nd, NDF_ONLY_PNBUF);
1769 if (nd.ni_vp == nd.ni_dvp)
1770 vrele(nd.ni_dvp);
1771 else
1772 vput(nd.ni_dvp);
1773 if (nd.ni_vp)
1774 vrele(nd.ni_vp);
1775 return (EEXIST);
1776 }
1777 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1778 NDFREE(&nd, NDF_ONLY_PNBUF);
1779 vput(nd.ni_dvp);
1780 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1781 return (error);
1782 goto restart;
1783 }
1784 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1785 NDFREE(&nd, NDF_ONLY_PNBUF);
1786 vput(nd.ni_dvp);
1787 vn_finished_write(mp);
1788 if (error == ERELOOKUP)
1789 goto restart;
1790 return (error);
1791 }
1792
1793 /*
1794 * Delete a name from the filesystem.
1795 */
1796 #ifndef _SYS_SYSPROTO_H_
1797 struct unlink_args {
1798 char *path;
1799 };
1800 #endif
1801 int
sys_unlink(struct thread * td,struct unlink_args * uap)1802 sys_unlink(struct thread *td, struct unlink_args *uap)
1803 {
1804
1805 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1806 0, 0));
1807 }
1808
1809 static int
kern_funlinkat_ex(struct thread * td,int dfd,const char * path,int fd,int flag,enum uio_seg pathseg,ino_t oldinum)1810 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1811 int flag, enum uio_seg pathseg, ino_t oldinum)
1812 {
1813
1814 if ((flag & ~AT_REMOVEDIR) != 0)
1815 return (EINVAL);
1816
1817 if ((flag & AT_REMOVEDIR) != 0)
1818 return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1819
1820 return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1821 }
1822
1823 #ifndef _SYS_SYSPROTO_H_
1824 struct unlinkat_args {
1825 int fd;
1826 char *path;
1827 int flag;
1828 };
1829 #endif
1830 int
sys_unlinkat(struct thread * td,struct unlinkat_args * uap)1831 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1832 {
1833
1834 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1835 UIO_USERSPACE, 0));
1836 }
1837
1838 #ifndef _SYS_SYSPROTO_H_
1839 struct funlinkat_args {
1840 int dfd;
1841 const char *path;
1842 int fd;
1843 int flag;
1844 };
1845 #endif
1846 int
sys_funlinkat(struct thread * td,struct funlinkat_args * uap)1847 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1848 {
1849
1850 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1851 UIO_USERSPACE, 0));
1852 }
1853
1854 int
kern_funlinkat(struct thread * td,int dfd,const char * path,int fd,enum uio_seg pathseg,int flag,ino_t oldinum)1855 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1856 enum uio_seg pathseg, int flag, ino_t oldinum)
1857 {
1858 struct mount *mp;
1859 struct file *fp;
1860 struct vnode *vp;
1861 struct nameidata nd;
1862 struct stat sb;
1863 int error;
1864
1865 fp = NULL;
1866 if (fd != FD_NONE) {
1867 error = getvnode(td, fd, &cap_no_rights, &fp);
1868 if (error != 0)
1869 return (error);
1870 }
1871
1872 restart:
1873 bwillwrite();
1874 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1875 at2cnpflags(flag, AT_RESOLVE_BENEATH),
1876 pathseg, path, dfd, &cap_unlinkat_rights, td);
1877 if ((error = namei(&nd)) != 0) {
1878 if (error == EINVAL)
1879 error = EPERM;
1880 goto fdout;
1881 }
1882 vp = nd.ni_vp;
1883 if (vp->v_type == VDIR && oldinum == 0) {
1884 error = EPERM; /* POSIX */
1885 } else if (oldinum != 0 &&
1886 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1887 sb.st_ino != oldinum) {
1888 error = EIDRM; /* Identifier removed */
1889 } else if (fp != NULL && fp->f_vnode != vp) {
1890 if (VN_IS_DOOMED(fp->f_vnode))
1891 error = EBADF;
1892 else
1893 error = EDEADLK;
1894 } else {
1895 /*
1896 * The root of a mounted filesystem cannot be deleted.
1897 *
1898 * XXX: can this only be a VDIR case?
1899 */
1900 if (vp->v_vflag & VV_ROOT)
1901 error = EBUSY;
1902 }
1903 if (error == 0) {
1904 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1905 NDFREE(&nd, NDF_ONLY_PNBUF);
1906 vput(nd.ni_dvp);
1907 if (vp == nd.ni_dvp)
1908 vrele(vp);
1909 else
1910 vput(vp);
1911 if ((error = vn_start_write(NULL, &mp,
1912 V_XSLEEP | PCATCH)) != 0) {
1913 goto fdout;
1914 }
1915 goto restart;
1916 }
1917 #ifdef MAC
1918 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1919 &nd.ni_cnd);
1920 if (error != 0)
1921 goto out;
1922 #endif
1923 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1924 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1925 #ifdef MAC
1926 out:
1927 #endif
1928 vn_finished_write(mp);
1929 }
1930 NDFREE(&nd, NDF_ONLY_PNBUF);
1931 vput(nd.ni_dvp);
1932 if (vp == nd.ni_dvp)
1933 vrele(vp);
1934 else
1935 vput(vp);
1936 if (error == ERELOOKUP)
1937 goto restart;
1938 fdout:
1939 if (fp != NULL)
1940 fdrop(fp, td);
1941 return (error);
1942 }
1943
1944 /*
1945 * Reposition read/write file offset.
1946 */
1947 #ifndef _SYS_SYSPROTO_H_
1948 struct lseek_args {
1949 int fd;
1950 int pad;
1951 off_t offset;
1952 int whence;
1953 };
1954 #endif
1955 int
sys_lseek(struct thread * td,struct lseek_args * uap)1956 sys_lseek(struct thread *td, struct lseek_args *uap)
1957 {
1958
1959 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1960 }
1961
1962 int
kern_lseek(struct thread * td,int fd,off_t offset,int whence)1963 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1964 {
1965 struct file *fp;
1966 int error;
1967
1968 AUDIT_ARG_FD(fd);
1969 error = fget(td, fd, &cap_seek_rights, &fp);
1970 if (error != 0)
1971 return (error);
1972 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1973 fo_seek(fp, offset, whence, td) : ESPIPE;
1974 fdrop(fp, td);
1975 return (error);
1976 }
1977
1978 #if defined(COMPAT_43)
1979 /*
1980 * Reposition read/write file offset.
1981 */
1982 #ifndef _SYS_SYSPROTO_H_
1983 struct olseek_args {
1984 int fd;
1985 long offset;
1986 int whence;
1987 };
1988 #endif
1989 int
olseek(struct thread * td,struct olseek_args * uap)1990 olseek(struct thread *td, struct olseek_args *uap)
1991 {
1992
1993 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1994 }
1995 #endif /* COMPAT_43 */
1996
1997 #if defined(COMPAT_FREEBSD6)
1998 /* Version with the 'pad' argument */
1999 int
freebsd6_lseek(struct thread * td,struct freebsd6_lseek_args * uap)2000 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2001 {
2002
2003 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2004 }
2005 #endif
2006
2007 /*
2008 * Check access permissions using passed credentials.
2009 */
2010 static int
vn_access(struct vnode * vp,int user_flags,struct ucred * cred,struct thread * td)2011 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2012 struct thread *td)
2013 {
2014 accmode_t accmode;
2015 int error;
2016
2017 /* Flags == 0 means only check for existence. */
2018 if (user_flags == 0)
2019 return (0);
2020
2021 accmode = 0;
2022 if (user_flags & R_OK)
2023 accmode |= VREAD;
2024 if (user_flags & W_OK)
2025 accmode |= VWRITE;
2026 if (user_flags & X_OK)
2027 accmode |= VEXEC;
2028 #ifdef MAC
2029 error = mac_vnode_check_access(cred, vp, accmode);
2030 if (error != 0)
2031 return (error);
2032 #endif
2033 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2034 error = VOP_ACCESS(vp, accmode, cred, td);
2035 return (error);
2036 }
2037
2038 /*
2039 * Check access permissions using "real" credentials.
2040 */
2041 #ifndef _SYS_SYSPROTO_H_
2042 struct access_args {
2043 char *path;
2044 int amode;
2045 };
2046 #endif
2047 int
sys_access(struct thread * td,struct access_args * uap)2048 sys_access(struct thread *td, struct access_args *uap)
2049 {
2050
2051 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2052 0, uap->amode));
2053 }
2054
2055 #ifndef _SYS_SYSPROTO_H_
2056 struct faccessat_args {
2057 int dirfd;
2058 char *path;
2059 int amode;
2060 int flag;
2061 }
2062 #endif
2063 int
sys_faccessat(struct thread * td,struct faccessat_args * uap)2064 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2065 {
2066
2067 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2068 uap->amode));
2069 }
2070
2071 int
kern_accessat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int flag,int amode)2072 kern_accessat(struct thread *td, int fd, const char *path,
2073 enum uio_seg pathseg, int flag, int amode)
2074 {
2075 struct ucred *cred, *usecred;
2076 struct vnode *vp;
2077 struct nameidata nd;
2078 int error;
2079
2080 if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH)) != 0)
2081 return (EINVAL);
2082 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2083 return (EINVAL);
2084
2085 /*
2086 * Create and modify a temporary credential instead of one that
2087 * is potentially shared (if we need one).
2088 */
2089 cred = td->td_ucred;
2090 if ((flag & AT_EACCESS) == 0 &&
2091 ((cred->cr_uid != cred->cr_ruid ||
2092 cred->cr_rgid != cred->cr_groups[0]))) {
2093 usecred = crdup(cred);
2094 usecred->cr_uid = cred->cr_ruid;
2095 usecred->cr_groups[0] = cred->cr_rgid;
2096 td->td_ucred = usecred;
2097 } else
2098 usecred = cred;
2099 AUDIT_ARG_VALUE(amode);
2100 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2101 AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH),
2102 pathseg, path, fd, &cap_fstat_rights, td);
2103 if ((error = namei(&nd)) != 0)
2104 goto out;
2105 vp = nd.ni_vp;
2106
2107 error = vn_access(vp, amode, usecred, td);
2108 NDFREE_NOTHING(&nd);
2109 vput(vp);
2110 out:
2111 if (usecred != cred) {
2112 td->td_ucred = cred;
2113 crfree(usecred);
2114 }
2115 return (error);
2116 }
2117
2118 /*
2119 * Check access permissions using "effective" credentials.
2120 */
2121 #ifndef _SYS_SYSPROTO_H_
2122 struct eaccess_args {
2123 char *path;
2124 int amode;
2125 };
2126 #endif
2127 int
sys_eaccess(struct thread * td,struct eaccess_args * uap)2128 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2129 {
2130
2131 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2132 AT_EACCESS, uap->amode));
2133 }
2134
2135 #if defined(COMPAT_43)
2136 /*
2137 * Get file status; this version follows links.
2138 */
2139 #ifndef _SYS_SYSPROTO_H_
2140 struct ostat_args {
2141 char *path;
2142 struct ostat *ub;
2143 };
2144 #endif
2145 int
ostat(struct thread * td,struct ostat_args * uap)2146 ostat(struct thread *td, struct ostat_args *uap)
2147 {
2148 struct stat sb;
2149 struct ostat osb;
2150 int error;
2151
2152 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2153 &sb, NULL);
2154 if (error != 0)
2155 return (error);
2156 cvtstat(&sb, &osb);
2157 return (copyout(&osb, uap->ub, sizeof (osb)));
2158 }
2159
2160 /*
2161 * Get file status; this version does not follow links.
2162 */
2163 #ifndef _SYS_SYSPROTO_H_
2164 struct olstat_args {
2165 char *path;
2166 struct ostat *ub;
2167 };
2168 #endif
2169 int
olstat(struct thread * td,struct olstat_args * uap)2170 olstat(struct thread *td, struct olstat_args *uap)
2171 {
2172 struct stat sb;
2173 struct ostat osb;
2174 int error;
2175
2176 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2177 UIO_USERSPACE, &sb, NULL);
2178 if (error != 0)
2179 return (error);
2180 cvtstat(&sb, &osb);
2181 return (copyout(&osb, uap->ub, sizeof (osb)));
2182 }
2183
2184 /*
2185 * Convert from an old to a new stat structure.
2186 * XXX: many values are blindly truncated.
2187 */
2188 void
cvtstat(struct stat * st,struct ostat * ost)2189 cvtstat(struct stat *st, struct ostat *ost)
2190 {
2191
2192 bzero(ost, sizeof(*ost));
2193 ost->st_dev = st->st_dev;
2194 ost->st_ino = st->st_ino;
2195 ost->st_mode = st->st_mode;
2196 ost->st_nlink = st->st_nlink;
2197 ost->st_uid = st->st_uid;
2198 ost->st_gid = st->st_gid;
2199 ost->st_rdev = st->st_rdev;
2200 ost->st_size = MIN(st->st_size, INT32_MAX);
2201 ost->st_atim = st->st_atim;
2202 ost->st_mtim = st->st_mtim;
2203 ost->st_ctim = st->st_ctim;
2204 ost->st_blksize = st->st_blksize;
2205 ost->st_blocks = st->st_blocks;
2206 ost->st_flags = st->st_flags;
2207 ost->st_gen = st->st_gen;
2208 }
2209 #endif /* COMPAT_43 */
2210
2211 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2212 int ino64_trunc_error;
2213 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2214 &ino64_trunc_error, 0,
2215 "Error on truncation of device, file or inode number, or link count");
2216
2217 int
freebsd11_cvtstat(struct stat * st,struct freebsd11_stat * ost)2218 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2219 {
2220
2221 ost->st_dev = st->st_dev;
2222 if (ost->st_dev != st->st_dev) {
2223 switch (ino64_trunc_error) {
2224 default:
2225 /*
2226 * Since dev_t is almost raw, don't clamp to the
2227 * maximum for case 2, but ignore the error.
2228 */
2229 break;
2230 case 1:
2231 return (EOVERFLOW);
2232 }
2233 }
2234 ost->st_ino = st->st_ino;
2235 if (ost->st_ino != st->st_ino) {
2236 switch (ino64_trunc_error) {
2237 default:
2238 case 0:
2239 break;
2240 case 1:
2241 return (EOVERFLOW);
2242 case 2:
2243 ost->st_ino = UINT32_MAX;
2244 break;
2245 }
2246 }
2247 ost->st_mode = st->st_mode;
2248 ost->st_nlink = st->st_nlink;
2249 if (ost->st_nlink != st->st_nlink) {
2250 switch (ino64_trunc_error) {
2251 default:
2252 case 0:
2253 break;
2254 case 1:
2255 return (EOVERFLOW);
2256 case 2:
2257 ost->st_nlink = UINT16_MAX;
2258 break;
2259 }
2260 }
2261 ost->st_uid = st->st_uid;
2262 ost->st_gid = st->st_gid;
2263 ost->st_rdev = st->st_rdev;
2264 if (ost->st_rdev != st->st_rdev) {
2265 switch (ino64_trunc_error) {
2266 default:
2267 break;
2268 case 1:
2269 return (EOVERFLOW);
2270 }
2271 }
2272 ost->st_atim = st->st_atim;
2273 ost->st_mtim = st->st_mtim;
2274 ost->st_ctim = st->st_ctim;
2275 ost->st_size = st->st_size;
2276 ost->st_blocks = st->st_blocks;
2277 ost->st_blksize = st->st_blksize;
2278 ost->st_flags = st->st_flags;
2279 ost->st_gen = st->st_gen;
2280 ost->st_lspare = 0;
2281 ost->st_birthtim = st->st_birthtim;
2282 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2283 sizeof(*ost) - offsetof(struct freebsd11_stat,
2284 st_birthtim) - sizeof(ost->st_birthtim));
2285 return (0);
2286 }
2287
2288 int
freebsd11_stat(struct thread * td,struct freebsd11_stat_args * uap)2289 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2290 {
2291 struct stat sb;
2292 struct freebsd11_stat osb;
2293 int error;
2294
2295 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2296 &sb, NULL);
2297 if (error != 0)
2298 return (error);
2299 error = freebsd11_cvtstat(&sb, &osb);
2300 if (error == 0)
2301 error = copyout(&osb, uap->ub, sizeof(osb));
2302 return (error);
2303 }
2304
2305 int
freebsd11_lstat(struct thread * td,struct freebsd11_lstat_args * uap)2306 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2307 {
2308 struct stat sb;
2309 struct freebsd11_stat osb;
2310 int error;
2311
2312 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2313 UIO_USERSPACE, &sb, NULL);
2314 if (error != 0)
2315 return (error);
2316 error = freebsd11_cvtstat(&sb, &osb);
2317 if (error == 0)
2318 error = copyout(&osb, uap->ub, sizeof(osb));
2319 return (error);
2320 }
2321
2322 int
freebsd11_fhstat(struct thread * td,struct freebsd11_fhstat_args * uap)2323 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2324 {
2325 struct fhandle fh;
2326 struct stat sb;
2327 struct freebsd11_stat osb;
2328 int error;
2329
2330 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2331 if (error != 0)
2332 return (error);
2333 error = kern_fhstat(td, fh, &sb);
2334 if (error != 0)
2335 return (error);
2336 error = freebsd11_cvtstat(&sb, &osb);
2337 if (error == 0)
2338 error = copyout(&osb, uap->sb, sizeof(osb));
2339 return (error);
2340 }
2341
2342 int
freebsd11_fstatat(struct thread * td,struct freebsd11_fstatat_args * uap)2343 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2344 {
2345 struct stat sb;
2346 struct freebsd11_stat osb;
2347 int error;
2348
2349 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2350 UIO_USERSPACE, &sb, NULL);
2351 if (error != 0)
2352 return (error);
2353 error = freebsd11_cvtstat(&sb, &osb);
2354 if (error == 0)
2355 error = copyout(&osb, uap->buf, sizeof(osb));
2356 return (error);
2357 }
2358 #endif /* COMPAT_FREEBSD11 */
2359
2360 /*
2361 * Get file status
2362 */
2363 #ifndef _SYS_SYSPROTO_H_
2364 struct fstatat_args {
2365 int fd;
2366 char *path;
2367 struct stat *buf;
2368 int flag;
2369 }
2370 #endif
2371 int
sys_fstatat(struct thread * td,struct fstatat_args * uap)2372 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2373 {
2374 struct stat sb;
2375 int error;
2376
2377 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2378 UIO_USERSPACE, &sb, NULL);
2379 if (error == 0)
2380 error = copyout(&sb, uap->buf, sizeof (sb));
2381 return (error);
2382 }
2383
2384 int
kern_statat(struct thread * td,int flag,int fd,const char * path,enum uio_seg pathseg,struct stat * sbp,void (* hook)(struct vnode * vp,struct stat * sbp))2385 kern_statat(struct thread *td, int flag, int fd, const char *path,
2386 enum uio_seg pathseg, struct stat *sbp,
2387 void (*hook)(struct vnode *vp, struct stat *sbp))
2388 {
2389 struct nameidata nd;
2390 int error;
2391
2392 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2393 return (EINVAL);
2394
2395 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2396 AT_SYMLINK_NOFOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2397 pathseg, path, fd, &cap_fstat_rights, td);
2398
2399 if ((error = namei(&nd)) != 0)
2400 return (error);
2401 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2402 if (error == 0) {
2403 if (__predict_false(hook != NULL))
2404 hook(nd.ni_vp, sbp);
2405 }
2406 NDFREE_NOTHING(&nd);
2407 vput(nd.ni_vp);
2408 #ifdef __STAT_TIME_T_EXT
2409 sbp->st_atim_ext = 0;
2410 sbp->st_mtim_ext = 0;
2411 sbp->st_ctim_ext = 0;
2412 sbp->st_btim_ext = 0;
2413 #endif
2414 #ifdef KTRACE
2415 if (KTRPOINT(td, KTR_STRUCT))
2416 ktrstat_error(sbp, error);
2417 #endif
2418 return (error);
2419 }
2420
2421 #if defined(COMPAT_FREEBSD11)
2422 /*
2423 * Implementation of the NetBSD [l]stat() functions.
2424 */
2425 void
freebsd11_cvtnstat(struct stat * sb,struct nstat * nsb)2426 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2427 {
2428
2429 bzero(nsb, sizeof(*nsb));
2430 nsb->st_dev = sb->st_dev;
2431 nsb->st_ino = sb->st_ino;
2432 nsb->st_mode = sb->st_mode;
2433 nsb->st_nlink = sb->st_nlink;
2434 nsb->st_uid = sb->st_uid;
2435 nsb->st_gid = sb->st_gid;
2436 nsb->st_rdev = sb->st_rdev;
2437 nsb->st_atim = sb->st_atim;
2438 nsb->st_mtim = sb->st_mtim;
2439 nsb->st_ctim = sb->st_ctim;
2440 nsb->st_size = sb->st_size;
2441 nsb->st_blocks = sb->st_blocks;
2442 nsb->st_blksize = sb->st_blksize;
2443 nsb->st_flags = sb->st_flags;
2444 nsb->st_gen = sb->st_gen;
2445 nsb->st_birthtim = sb->st_birthtim;
2446 }
2447
2448 #ifndef _SYS_SYSPROTO_H_
2449 struct freebsd11_nstat_args {
2450 char *path;
2451 struct nstat *ub;
2452 };
2453 #endif
2454 int
freebsd11_nstat(struct thread * td,struct freebsd11_nstat_args * uap)2455 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2456 {
2457 struct stat sb;
2458 struct nstat nsb;
2459 int error;
2460
2461 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2462 &sb, NULL);
2463 if (error != 0)
2464 return (error);
2465 freebsd11_cvtnstat(&sb, &nsb);
2466 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2467 }
2468
2469 /*
2470 * NetBSD lstat. Get file status; this version does not follow links.
2471 */
2472 #ifndef _SYS_SYSPROTO_H_
2473 struct freebsd11_nlstat_args {
2474 char *path;
2475 struct nstat *ub;
2476 };
2477 #endif
2478 int
freebsd11_nlstat(struct thread * td,struct freebsd11_nlstat_args * uap)2479 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2480 {
2481 struct stat sb;
2482 struct nstat nsb;
2483 int error;
2484
2485 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2486 UIO_USERSPACE, &sb, NULL);
2487 if (error != 0)
2488 return (error);
2489 freebsd11_cvtnstat(&sb, &nsb);
2490 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2491 }
2492 #endif /* COMPAT_FREEBSD11 */
2493
2494 /*
2495 * Get configurable pathname variables.
2496 */
2497 #ifndef _SYS_SYSPROTO_H_
2498 struct pathconf_args {
2499 char *path;
2500 int name;
2501 };
2502 #endif
2503 int
sys_pathconf(struct thread * td,struct pathconf_args * uap)2504 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2505 {
2506 long value;
2507 int error;
2508
2509 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2510 &value);
2511 if (error == 0)
2512 td->td_retval[0] = value;
2513 return (error);
2514 }
2515
2516 #ifndef _SYS_SYSPROTO_H_
2517 struct lpathconf_args {
2518 char *path;
2519 int name;
2520 };
2521 #endif
2522 int
sys_lpathconf(struct thread * td,struct lpathconf_args * uap)2523 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2524 {
2525 long value;
2526 int error;
2527
2528 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2529 NOFOLLOW, &value);
2530 if (error == 0)
2531 td->td_retval[0] = value;
2532 return (error);
2533 }
2534
2535 int
kern_pathconf(struct thread * td,const char * path,enum uio_seg pathseg,int name,u_long flags,long * valuep)2536 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2537 int name, u_long flags, long *valuep)
2538 {
2539 struct nameidata nd;
2540 int error;
2541
2542 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2543 pathseg, path, td);
2544 if ((error = namei(&nd)) != 0)
2545 return (error);
2546 NDFREE_NOTHING(&nd);
2547
2548 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2549 vput(nd.ni_vp);
2550 return (error);
2551 }
2552
2553 /*
2554 * Return target name of a symbolic link.
2555 */
2556 #ifndef _SYS_SYSPROTO_H_
2557 struct readlink_args {
2558 char *path;
2559 char *buf;
2560 size_t count;
2561 };
2562 #endif
2563 int
sys_readlink(struct thread * td,struct readlink_args * uap)2564 sys_readlink(struct thread *td, struct readlink_args *uap)
2565 {
2566
2567 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2568 uap->buf, UIO_USERSPACE, uap->count));
2569 }
2570 #ifndef _SYS_SYSPROTO_H_
2571 struct readlinkat_args {
2572 int fd;
2573 char *path;
2574 char *buf;
2575 size_t bufsize;
2576 };
2577 #endif
2578 int
sys_readlinkat(struct thread * td,struct readlinkat_args * uap)2579 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2580 {
2581
2582 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2583 uap->buf, UIO_USERSPACE, uap->bufsize));
2584 }
2585
2586 int
kern_readlinkat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,char * buf,enum uio_seg bufseg,size_t count)2587 kern_readlinkat(struct thread *td, int fd, const char *path,
2588 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2589 {
2590 struct vnode *vp;
2591 struct nameidata nd;
2592 int error;
2593
2594 if (count > IOSIZE_MAX)
2595 return (EINVAL);
2596
2597 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2598 pathseg, path, fd, td);
2599
2600 if ((error = namei(&nd)) != 0)
2601 return (error);
2602 NDFREE_NOTHING(&nd);
2603 vp = nd.ni_vp;
2604
2605 error = kern_readlink_vp(vp, buf, bufseg, count, td);
2606 vput(vp);
2607
2608 return (error);
2609 }
2610
2611 /*
2612 * Helper function to readlink from a vnode
2613 */
2614 static int
kern_readlink_vp(struct vnode * vp,char * buf,enum uio_seg bufseg,size_t count,struct thread * td)2615 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2616 struct thread *td)
2617 {
2618 struct iovec aiov;
2619 struct uio auio;
2620 int error;
2621
2622 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2623 #ifdef MAC
2624 error = mac_vnode_check_readlink(td->td_ucred, vp);
2625 if (error != 0)
2626 return (error);
2627 #endif
2628 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2629 return (EINVAL);
2630
2631 aiov.iov_base = buf;
2632 aiov.iov_len = count;
2633 auio.uio_iov = &aiov;
2634 auio.uio_iovcnt = 1;
2635 auio.uio_offset = 0;
2636 auio.uio_rw = UIO_READ;
2637 auio.uio_segflg = bufseg;
2638 auio.uio_td = td;
2639 auio.uio_resid = count;
2640 error = VOP_READLINK(vp, &auio, td->td_ucred);
2641 td->td_retval[0] = count - auio.uio_resid;
2642 return (error);
2643 }
2644
2645 /*
2646 * Common implementation code for chflags() and fchflags().
2647 */
2648 static int
setfflags(struct thread * td,struct vnode * vp,u_long flags)2649 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2650 {
2651 struct mount *mp;
2652 struct vattr vattr;
2653 int error;
2654
2655 /* We can't support the value matching VNOVAL. */
2656 if (flags == VNOVAL)
2657 return (EOPNOTSUPP);
2658
2659 /*
2660 * Prevent non-root users from setting flags on devices. When
2661 * a device is reused, users can retain ownership of the device
2662 * if they are allowed to set flags and programs assume that
2663 * chown can't fail when done as root.
2664 */
2665 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2666 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2667 if (error != 0)
2668 return (error);
2669 }
2670
2671 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2672 return (error);
2673 VATTR_NULL(&vattr);
2674 vattr.va_flags = flags;
2675 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2676 #ifdef MAC
2677 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2678 if (error == 0)
2679 #endif
2680 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2681 VOP_UNLOCK(vp);
2682 vn_finished_write(mp);
2683 return (error);
2684 }
2685
2686 /*
2687 * Change flags of a file given a path name.
2688 */
2689 #ifndef _SYS_SYSPROTO_H_
2690 struct chflags_args {
2691 const char *path;
2692 u_long flags;
2693 };
2694 #endif
2695 int
sys_chflags(struct thread * td,struct chflags_args * uap)2696 sys_chflags(struct thread *td, struct chflags_args *uap)
2697 {
2698
2699 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2700 uap->flags, 0));
2701 }
2702
2703 #ifndef _SYS_SYSPROTO_H_
2704 struct chflagsat_args {
2705 int fd;
2706 const char *path;
2707 u_long flags;
2708 int atflag;
2709 }
2710 #endif
2711 int
sys_chflagsat(struct thread * td,struct chflagsat_args * uap)2712 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2713 {
2714
2715 if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2716 return (EINVAL);
2717
2718 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2719 uap->flags, uap->atflag));
2720 }
2721
2722 /*
2723 * Same as chflags() but doesn't follow symlinks.
2724 */
2725 #ifndef _SYS_SYSPROTO_H_
2726 struct lchflags_args {
2727 const char *path;
2728 u_long flags;
2729 };
2730 #endif
2731 int
sys_lchflags(struct thread * td,struct lchflags_args * uap)2732 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2733 {
2734
2735 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2736 uap->flags, AT_SYMLINK_NOFOLLOW));
2737 }
2738
2739 static int
kern_chflagsat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,u_long flags,int atflag)2740 kern_chflagsat(struct thread *td, int fd, const char *path,
2741 enum uio_seg pathseg, u_long flags, int atflag)
2742 {
2743 struct nameidata nd;
2744 int error;
2745
2746 AUDIT_ARG_FFLAGS(flags);
2747 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2748 AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2749 &cap_fchflags_rights, td);
2750 if ((error = namei(&nd)) != 0)
2751 return (error);
2752 NDFREE_NOTHING(&nd);
2753 error = setfflags(td, nd.ni_vp, flags);
2754 vrele(nd.ni_vp);
2755 return (error);
2756 }
2757
2758 /*
2759 * Change flags of a file given a file descriptor.
2760 */
2761 #ifndef _SYS_SYSPROTO_H_
2762 struct fchflags_args {
2763 int fd;
2764 u_long flags;
2765 };
2766 #endif
2767 int
sys_fchflags(struct thread * td,struct fchflags_args * uap)2768 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2769 {
2770 struct file *fp;
2771 int error;
2772
2773 AUDIT_ARG_FD(uap->fd);
2774 AUDIT_ARG_FFLAGS(uap->flags);
2775 error = getvnode(td, uap->fd, &cap_fchflags_rights,
2776 &fp);
2777 if (error != 0)
2778 return (error);
2779 #ifdef AUDIT
2780 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2781 AUDIT_ARG_VNODE1(fp->f_vnode);
2782 VOP_UNLOCK(fp->f_vnode);
2783 #endif
2784 error = setfflags(td, fp->f_vnode, uap->flags);
2785 fdrop(fp, td);
2786 return (error);
2787 }
2788
2789 /*
2790 * Common implementation code for chmod(), lchmod() and fchmod().
2791 */
2792 int
setfmode(struct thread * td,struct ucred * cred,struct vnode * vp,int mode)2793 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2794 {
2795 struct mount *mp;
2796 struct vattr vattr;
2797 int error;
2798
2799 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2800 return (error);
2801 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2802 VATTR_NULL(&vattr);
2803 vattr.va_mode = mode & ALLPERMS;
2804 #ifdef MAC
2805 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2806 if (error == 0)
2807 #endif
2808 error = VOP_SETATTR(vp, &vattr, cred);
2809 VOP_UNLOCK(vp);
2810 vn_finished_write(mp);
2811 return (error);
2812 }
2813
2814 /*
2815 * Change mode of a file given path name.
2816 */
2817 #ifndef _SYS_SYSPROTO_H_
2818 struct chmod_args {
2819 char *path;
2820 int mode;
2821 };
2822 #endif
2823 int
sys_chmod(struct thread * td,struct chmod_args * uap)2824 sys_chmod(struct thread *td, struct chmod_args *uap)
2825 {
2826
2827 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2828 uap->mode, 0));
2829 }
2830
2831 #ifndef _SYS_SYSPROTO_H_
2832 struct fchmodat_args {
2833 int dirfd;
2834 char *path;
2835 mode_t mode;
2836 int flag;
2837 }
2838 #endif
2839 int
sys_fchmodat(struct thread * td,struct fchmodat_args * uap)2840 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2841 {
2842
2843 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2844 return (EINVAL);
2845
2846 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2847 uap->mode, uap->flag));
2848 }
2849
2850 /*
2851 * Change mode of a file given path name (don't follow links.)
2852 */
2853 #ifndef _SYS_SYSPROTO_H_
2854 struct lchmod_args {
2855 char *path;
2856 int mode;
2857 };
2858 #endif
2859 int
sys_lchmod(struct thread * td,struct lchmod_args * uap)2860 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2861 {
2862
2863 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2864 uap->mode, AT_SYMLINK_NOFOLLOW));
2865 }
2866
2867 int
kern_fchmodat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,mode_t mode,int flag)2868 kern_fchmodat(struct thread *td, int fd, const char *path,
2869 enum uio_seg pathseg, mode_t mode, int flag)
2870 {
2871 struct nameidata nd;
2872 int error;
2873
2874 AUDIT_ARG_MODE(mode);
2875 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2876 AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2877 &cap_fchmod_rights, td);
2878 if ((error = namei(&nd)) != 0)
2879 return (error);
2880 NDFREE_NOTHING(&nd);
2881 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2882 vrele(nd.ni_vp);
2883 return (error);
2884 }
2885
2886 /*
2887 * Change mode of a file given a file descriptor.
2888 */
2889 #ifndef _SYS_SYSPROTO_H_
2890 struct fchmod_args {
2891 int fd;
2892 int mode;
2893 };
2894 #endif
2895 int
sys_fchmod(struct thread * td,struct fchmod_args * uap)2896 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2897 {
2898 struct file *fp;
2899 int error;
2900
2901 AUDIT_ARG_FD(uap->fd);
2902 AUDIT_ARG_MODE(uap->mode);
2903
2904 error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2905 if (error != 0)
2906 return (error);
2907 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2908 fdrop(fp, td);
2909 return (error);
2910 }
2911
2912 /*
2913 * Common implementation for chown(), lchown(), and fchown()
2914 */
2915 int
setfown(struct thread * td,struct ucred * cred,struct vnode * vp,uid_t uid,gid_t gid)2916 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2917 gid_t gid)
2918 {
2919 struct mount *mp;
2920 struct vattr vattr;
2921 int error;
2922
2923 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2924 return (error);
2925 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2926 VATTR_NULL(&vattr);
2927 vattr.va_uid = uid;
2928 vattr.va_gid = gid;
2929 #ifdef MAC
2930 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2931 vattr.va_gid);
2932 if (error == 0)
2933 #endif
2934 error = VOP_SETATTR(vp, &vattr, cred);
2935 VOP_UNLOCK(vp);
2936 vn_finished_write(mp);
2937 return (error);
2938 }
2939
2940 /*
2941 * Set ownership given a path name.
2942 */
2943 #ifndef _SYS_SYSPROTO_H_
2944 struct chown_args {
2945 char *path;
2946 int uid;
2947 int gid;
2948 };
2949 #endif
2950 int
sys_chown(struct thread * td,struct chown_args * uap)2951 sys_chown(struct thread *td, struct chown_args *uap)
2952 {
2953
2954 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2955 uap->gid, 0));
2956 }
2957
2958 #ifndef _SYS_SYSPROTO_H_
2959 struct fchownat_args {
2960 int fd;
2961 const char * path;
2962 uid_t uid;
2963 gid_t gid;
2964 int flag;
2965 };
2966 #endif
2967 int
sys_fchownat(struct thread * td,struct fchownat_args * uap)2968 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2969 {
2970
2971 if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2972 return (EINVAL);
2973
2974 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2975 uap->gid, uap->flag));
2976 }
2977
2978 int
kern_fchownat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int uid,int gid,int flag)2979 kern_fchownat(struct thread *td, int fd, const char *path,
2980 enum uio_seg pathseg, int uid, int gid, int flag)
2981 {
2982 struct nameidata nd;
2983 int error;
2984
2985 AUDIT_ARG_OWNER(uid, gid);
2986 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2987 AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2988 &cap_fchown_rights, td);
2989
2990 if ((error = namei(&nd)) != 0)
2991 return (error);
2992 NDFREE_NOTHING(&nd);
2993 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2994 vrele(nd.ni_vp);
2995 return (error);
2996 }
2997
2998 /*
2999 * Set ownership given a path name, do not cross symlinks.
3000 */
3001 #ifndef _SYS_SYSPROTO_H_
3002 struct lchown_args {
3003 char *path;
3004 int uid;
3005 int gid;
3006 };
3007 #endif
3008 int
sys_lchown(struct thread * td,struct lchown_args * uap)3009 sys_lchown(struct thread *td, struct lchown_args *uap)
3010 {
3011
3012 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3013 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3014 }
3015
3016 /*
3017 * Set ownership given a file descriptor.
3018 */
3019 #ifndef _SYS_SYSPROTO_H_
3020 struct fchown_args {
3021 int fd;
3022 int uid;
3023 int gid;
3024 };
3025 #endif
3026 int
sys_fchown(struct thread * td,struct fchown_args * uap)3027 sys_fchown(struct thread *td, struct fchown_args *uap)
3028 {
3029 struct file *fp;
3030 int error;
3031
3032 AUDIT_ARG_FD(uap->fd);
3033 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3034 error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3035 if (error != 0)
3036 return (error);
3037 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3038 fdrop(fp, td);
3039 return (error);
3040 }
3041
3042 /*
3043 * Common implementation code for utimes(), lutimes(), and futimes().
3044 */
3045 static int
getutimes(const struct timeval * usrtvp,enum uio_seg tvpseg,struct timespec * tsp)3046 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3047 struct timespec *tsp)
3048 {
3049 struct timeval tv[2];
3050 const struct timeval *tvp;
3051 int error;
3052
3053 if (usrtvp == NULL) {
3054 vfs_timestamp(&tsp[0]);
3055 tsp[1] = tsp[0];
3056 } else {
3057 if (tvpseg == UIO_SYSSPACE) {
3058 tvp = usrtvp;
3059 } else {
3060 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3061 return (error);
3062 tvp = tv;
3063 }
3064
3065 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3066 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3067 return (EINVAL);
3068 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3069 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3070 }
3071 return (0);
3072 }
3073
3074 /*
3075 * Common implementation code for futimens(), utimensat().
3076 */
3077 #define UTIMENS_NULL 0x1
3078 #define UTIMENS_EXIT 0x2
3079 static int
getutimens(const struct timespec * usrtsp,enum uio_seg tspseg,struct timespec * tsp,int * retflags)3080 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3081 struct timespec *tsp, int *retflags)
3082 {
3083 struct timespec tsnow;
3084 int error;
3085
3086 vfs_timestamp(&tsnow);
3087 *retflags = 0;
3088 if (usrtsp == NULL) {
3089 tsp[0] = tsnow;
3090 tsp[1] = tsnow;
3091 *retflags |= UTIMENS_NULL;
3092 return (0);
3093 }
3094 if (tspseg == UIO_SYSSPACE) {
3095 tsp[0] = usrtsp[0];
3096 tsp[1] = usrtsp[1];
3097 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3098 return (error);
3099 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3100 *retflags |= UTIMENS_EXIT;
3101 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3102 *retflags |= UTIMENS_NULL;
3103 if (tsp[0].tv_nsec == UTIME_OMIT)
3104 tsp[0].tv_sec = VNOVAL;
3105 else if (tsp[0].tv_nsec == UTIME_NOW)
3106 tsp[0] = tsnow;
3107 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3108 return (EINVAL);
3109 if (tsp[1].tv_nsec == UTIME_OMIT)
3110 tsp[1].tv_sec = VNOVAL;
3111 else if (tsp[1].tv_nsec == UTIME_NOW)
3112 tsp[1] = tsnow;
3113 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3114 return (EINVAL);
3115
3116 return (0);
3117 }
3118
3119 /*
3120 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3121 * and utimensat().
3122 */
3123 static int
setutimes(struct thread * td,struct vnode * vp,const struct timespec * ts,int numtimes,int nullflag)3124 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3125 int numtimes, int nullflag)
3126 {
3127 struct mount *mp;
3128 struct vattr vattr;
3129 int error, setbirthtime;
3130
3131 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3132 return (error);
3133 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3134 setbirthtime = 0;
3135 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3136 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3137 setbirthtime = 1;
3138 VATTR_NULL(&vattr);
3139 vattr.va_atime = ts[0];
3140 vattr.va_mtime = ts[1];
3141 if (setbirthtime)
3142 vattr.va_birthtime = ts[1];
3143 if (numtimes > 2)
3144 vattr.va_birthtime = ts[2];
3145 if (nullflag)
3146 vattr.va_vaflags |= VA_UTIMES_NULL;
3147 #ifdef MAC
3148 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3149 vattr.va_mtime);
3150 #endif
3151 if (error == 0)
3152 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3153 VOP_UNLOCK(vp);
3154 vn_finished_write(mp);
3155 return (error);
3156 }
3157
3158 /*
3159 * Set the access and modification times of a file.
3160 */
3161 #ifndef _SYS_SYSPROTO_H_
3162 struct utimes_args {
3163 char *path;
3164 struct timeval *tptr;
3165 };
3166 #endif
3167 int
sys_utimes(struct thread * td,struct utimes_args * uap)3168 sys_utimes(struct thread *td, struct utimes_args *uap)
3169 {
3170
3171 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3172 uap->tptr, UIO_USERSPACE));
3173 }
3174
3175 #ifndef _SYS_SYSPROTO_H_
3176 struct futimesat_args {
3177 int fd;
3178 const char * path;
3179 const struct timeval * times;
3180 };
3181 #endif
3182 int
sys_futimesat(struct thread * td,struct futimesat_args * uap)3183 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3184 {
3185
3186 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3187 uap->times, UIO_USERSPACE));
3188 }
3189
3190 int
kern_utimesat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,struct timeval * tptr,enum uio_seg tptrseg)3191 kern_utimesat(struct thread *td, int fd, const char *path,
3192 enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3193 {
3194 struct nameidata nd;
3195 struct timespec ts[2];
3196 int error;
3197
3198 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3199 return (error);
3200 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3201 &cap_futimes_rights, td);
3202
3203 if ((error = namei(&nd)) != 0)
3204 return (error);
3205 NDFREE_NOTHING(&nd);
3206 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3207 vrele(nd.ni_vp);
3208 return (error);
3209 }
3210
3211 /*
3212 * Set the access and modification times of a file.
3213 */
3214 #ifndef _SYS_SYSPROTO_H_
3215 struct lutimes_args {
3216 char *path;
3217 struct timeval *tptr;
3218 };
3219 #endif
3220 int
sys_lutimes(struct thread * td,struct lutimes_args * uap)3221 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3222 {
3223
3224 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3225 UIO_USERSPACE));
3226 }
3227
3228 int
kern_lutimes(struct thread * td,const char * path,enum uio_seg pathseg,struct timeval * tptr,enum uio_seg tptrseg)3229 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3230 struct timeval *tptr, enum uio_seg tptrseg)
3231 {
3232 struct timespec ts[2];
3233 struct nameidata nd;
3234 int error;
3235
3236 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3237 return (error);
3238 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3239 if ((error = namei(&nd)) != 0)
3240 return (error);
3241 NDFREE_NOTHING(&nd);
3242 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3243 vrele(nd.ni_vp);
3244 return (error);
3245 }
3246
3247 /*
3248 * Set the access and modification times of a file.
3249 */
3250 #ifndef _SYS_SYSPROTO_H_
3251 struct futimes_args {
3252 int fd;
3253 struct timeval *tptr;
3254 };
3255 #endif
3256 int
sys_futimes(struct thread * td,struct futimes_args * uap)3257 sys_futimes(struct thread *td, struct futimes_args *uap)
3258 {
3259
3260 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3261 }
3262
3263 int
kern_futimes(struct thread * td,int fd,struct timeval * tptr,enum uio_seg tptrseg)3264 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3265 enum uio_seg tptrseg)
3266 {
3267 struct timespec ts[2];
3268 struct file *fp;
3269 int error;
3270
3271 AUDIT_ARG_FD(fd);
3272 error = getutimes(tptr, tptrseg, ts);
3273 if (error != 0)
3274 return (error);
3275 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3276 if (error != 0)
3277 return (error);
3278 #ifdef AUDIT
3279 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3280 AUDIT_ARG_VNODE1(fp->f_vnode);
3281 VOP_UNLOCK(fp->f_vnode);
3282 #endif
3283 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3284 fdrop(fp, td);
3285 return (error);
3286 }
3287
3288 int
sys_futimens(struct thread * td,struct futimens_args * uap)3289 sys_futimens(struct thread *td, struct futimens_args *uap)
3290 {
3291
3292 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3293 }
3294
3295 int
kern_futimens(struct thread * td,int fd,struct timespec * tptr,enum uio_seg tptrseg)3296 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3297 enum uio_seg tptrseg)
3298 {
3299 struct timespec ts[2];
3300 struct file *fp;
3301 int error, flags;
3302
3303 AUDIT_ARG_FD(fd);
3304 error = getutimens(tptr, tptrseg, ts, &flags);
3305 if (error != 0)
3306 return (error);
3307 if (flags & UTIMENS_EXIT)
3308 return (0);
3309 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3310 if (error != 0)
3311 return (error);
3312 #ifdef AUDIT
3313 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3314 AUDIT_ARG_VNODE1(fp->f_vnode);
3315 VOP_UNLOCK(fp->f_vnode);
3316 #endif
3317 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3318 fdrop(fp, td);
3319 return (error);
3320 }
3321
3322 int
sys_utimensat(struct thread * td,struct utimensat_args * uap)3323 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3324 {
3325
3326 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3327 uap->times, UIO_USERSPACE, uap->flag));
3328 }
3329
3330 int
kern_utimensat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,struct timespec * tptr,enum uio_seg tptrseg,int flag)3331 kern_utimensat(struct thread *td, int fd, const char *path,
3332 enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3333 int flag)
3334 {
3335 struct nameidata nd;
3336 struct timespec ts[2];
3337 int error, flags;
3338
3339 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
3340 return (EINVAL);
3341
3342 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3343 return (error);
3344 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3345 AT_RESOLVE_BENEATH) | AUDITVNODE1,
3346 pathseg, path, fd, &cap_futimes_rights, td);
3347 if ((error = namei(&nd)) != 0)
3348 return (error);
3349 /*
3350 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3351 * POSIX states:
3352 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3353 * "Search permission is denied by a component of the path prefix."
3354 */
3355 NDFREE_NOTHING(&nd);
3356 if ((flags & UTIMENS_EXIT) == 0)
3357 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3358 vrele(nd.ni_vp);
3359 return (error);
3360 }
3361
3362 /*
3363 * Truncate a file given its path name.
3364 */
3365 #ifndef _SYS_SYSPROTO_H_
3366 struct truncate_args {
3367 char *path;
3368 int pad;
3369 off_t length;
3370 };
3371 #endif
3372 int
sys_truncate(struct thread * td,struct truncate_args * uap)3373 sys_truncate(struct thread *td, struct truncate_args *uap)
3374 {
3375
3376 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3377 }
3378
3379 int
kern_truncate(struct thread * td,const char * path,enum uio_seg pathseg,off_t length)3380 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3381 off_t length)
3382 {
3383 struct mount *mp;
3384 struct vnode *vp;
3385 void *rl_cookie;
3386 struct vattr vattr;
3387 struct nameidata nd;
3388 int error;
3389
3390 if (length < 0)
3391 return (EINVAL);
3392 retry:
3393 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3394 if ((error = namei(&nd)) != 0)
3395 return (error);
3396 vp = nd.ni_vp;
3397 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3398 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3399 vn_rangelock_unlock(vp, rl_cookie);
3400 vrele(vp);
3401 return (error);
3402 }
3403 NDFREE(&nd, NDF_ONLY_PNBUF);
3404 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3405 if (vp->v_type == VDIR)
3406 error = EISDIR;
3407 #ifdef MAC
3408 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3409 }
3410 #endif
3411 else if ((error = vn_writechk(vp)) == 0 &&
3412 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3413 VATTR_NULL(&vattr);
3414 vattr.va_size = length;
3415 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3416 }
3417 VOP_UNLOCK(vp);
3418 vn_finished_write(mp);
3419 vn_rangelock_unlock(vp, rl_cookie);
3420 vrele(vp);
3421 if (error == ERELOOKUP)
3422 goto retry;
3423 return (error);
3424 }
3425
3426 #if defined(COMPAT_43)
3427 /*
3428 * Truncate a file given its path name.
3429 */
3430 #ifndef _SYS_SYSPROTO_H_
3431 struct otruncate_args {
3432 char *path;
3433 long length;
3434 };
3435 #endif
3436 int
otruncate(struct thread * td,struct otruncate_args * uap)3437 otruncate(struct thread *td, struct otruncate_args *uap)
3438 {
3439
3440 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3441 }
3442 #endif /* COMPAT_43 */
3443
3444 #if defined(COMPAT_FREEBSD6)
3445 /* Versions with the pad argument */
3446 int
freebsd6_truncate(struct thread * td,struct freebsd6_truncate_args * uap)3447 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3448 {
3449
3450 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3451 }
3452
3453 int
freebsd6_ftruncate(struct thread * td,struct freebsd6_ftruncate_args * uap)3454 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3455 {
3456
3457 return (kern_ftruncate(td, uap->fd, uap->length));
3458 }
3459 #endif
3460
3461 int
kern_fsync(struct thread * td,int fd,bool fullsync)3462 kern_fsync(struct thread *td, int fd, bool fullsync)
3463 {
3464 struct vnode *vp;
3465 struct mount *mp;
3466 struct file *fp;
3467 int error, lock_flags;
3468
3469 AUDIT_ARG_FD(fd);
3470 error = getvnode(td, fd, &cap_fsync_rights, &fp);
3471 if (error != 0)
3472 return (error);
3473 vp = fp->f_vnode;
3474 #if 0
3475 if (!fullsync)
3476 /* XXXKIB: compete outstanding aio writes */;
3477 #endif
3478 retry:
3479 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3480 if (error != 0)
3481 goto drop;
3482 if (MNT_SHARED_WRITES(mp) ||
3483 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3484 lock_flags = LK_SHARED;
3485 } else {
3486 lock_flags = LK_EXCLUSIVE;
3487 }
3488 vn_lock(vp, lock_flags | LK_RETRY);
3489 AUDIT_ARG_VNODE1(vp);
3490 if (vp->v_object != NULL) {
3491 VM_OBJECT_WLOCK(vp->v_object);
3492 vm_object_page_clean(vp->v_object, 0, 0, 0);
3493 VM_OBJECT_WUNLOCK(vp->v_object);
3494 }
3495 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3496 VOP_UNLOCK(vp);
3497 vn_finished_write(mp);
3498 if (error == ERELOOKUP)
3499 goto retry;
3500 drop:
3501 fdrop(fp, td);
3502 return (error);
3503 }
3504
3505 /*
3506 * Sync an open file.
3507 */
3508 #ifndef _SYS_SYSPROTO_H_
3509 struct fsync_args {
3510 int fd;
3511 };
3512 #endif
3513 int
sys_fsync(struct thread * td,struct fsync_args * uap)3514 sys_fsync(struct thread *td, struct fsync_args *uap)
3515 {
3516
3517 return (kern_fsync(td, uap->fd, true));
3518 }
3519
3520 int
sys_fdatasync(struct thread * td,struct fdatasync_args * uap)3521 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3522 {
3523
3524 return (kern_fsync(td, uap->fd, false));
3525 }
3526
3527 /*
3528 * Rename files. Source and destination must either both be directories, or
3529 * both not be directories. If target is a directory, it must be empty.
3530 */
3531 #ifndef _SYS_SYSPROTO_H_
3532 struct rename_args {
3533 char *from;
3534 char *to;
3535 };
3536 #endif
3537 int
sys_rename(struct thread * td,struct rename_args * uap)3538 sys_rename(struct thread *td, struct rename_args *uap)
3539 {
3540
3541 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3542 uap->to, UIO_USERSPACE));
3543 }
3544
3545 #ifndef _SYS_SYSPROTO_H_
3546 struct renameat_args {
3547 int oldfd;
3548 char *old;
3549 int newfd;
3550 char *new;
3551 };
3552 #endif
3553 int
sys_renameat(struct thread * td,struct renameat_args * uap)3554 sys_renameat(struct thread *td, struct renameat_args *uap)
3555 {
3556
3557 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3558 UIO_USERSPACE));
3559 }
3560
3561 #ifdef MAC
3562 static int
kern_renameat_mac(struct thread * td,int oldfd,const char * old,int newfd,const char * new,enum uio_seg pathseg,struct nameidata * fromnd)3563 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3564 const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3565 {
3566 int error;
3567
3568 NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3569 AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3570 if ((error = namei(fromnd)) != 0)
3571 return (error);
3572 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3573 fromnd->ni_vp, &fromnd->ni_cnd);
3574 VOP_UNLOCK(fromnd->ni_dvp);
3575 if (fromnd->ni_dvp != fromnd->ni_vp)
3576 VOP_UNLOCK(fromnd->ni_vp);
3577 if (error != 0) {
3578 NDFREE(fromnd, NDF_ONLY_PNBUF);
3579 vrele(fromnd->ni_dvp);
3580 vrele(fromnd->ni_vp);
3581 if (fromnd->ni_startdir)
3582 vrele(fromnd->ni_startdir);
3583 }
3584 return (error);
3585 }
3586 #endif
3587
3588 int
kern_renameat(struct thread * td,int oldfd,const char * old,int newfd,const char * new,enum uio_seg pathseg)3589 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3590 const char *new, enum uio_seg pathseg)
3591 {
3592 struct mount *mp = NULL;
3593 struct vnode *tvp, *fvp, *tdvp;
3594 struct nameidata fromnd, tond;
3595 u_int64_t tondflags;
3596 int error;
3597
3598 again:
3599 bwillwrite();
3600 #ifdef MAC
3601 if (mac_vnode_check_rename_from_enabled()) {
3602 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3603 &fromnd);
3604 if (error != 0)
3605 return (error);
3606 } else {
3607 #endif
3608 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3609 pathseg, old, oldfd, &cap_renameat_source_rights, td);
3610 if ((error = namei(&fromnd)) != 0)
3611 return (error);
3612 #ifdef MAC
3613 }
3614 #endif
3615 fvp = fromnd.ni_vp;
3616 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3617 if (fromnd.ni_vp->v_type == VDIR)
3618 tondflags |= WILLBEDIR;
3619 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3620 &cap_renameat_target_rights, td);
3621 if ((error = namei(&tond)) != 0) {
3622 /* Translate error code for rename("dir1", "dir2/."). */
3623 if (error == EISDIR && fvp->v_type == VDIR)
3624 error = EINVAL;
3625 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3626 vrele(fromnd.ni_dvp);
3627 vrele(fvp);
3628 goto out1;
3629 }
3630 tdvp = tond.ni_dvp;
3631 tvp = tond.ni_vp;
3632 error = vn_start_write(fvp, &mp, V_NOWAIT);
3633 if (error != 0) {
3634 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3635 NDFREE(&tond, NDF_ONLY_PNBUF);
3636 if (tvp != NULL)
3637 vput(tvp);
3638 if (tdvp == tvp)
3639 vrele(tdvp);
3640 else
3641 vput(tdvp);
3642 vrele(fromnd.ni_dvp);
3643 vrele(fvp);
3644 vrele(tond.ni_startdir);
3645 if (fromnd.ni_startdir != NULL)
3646 vrele(fromnd.ni_startdir);
3647 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3648 if (error != 0)
3649 return (error);
3650 goto again;
3651 }
3652 if (tvp != NULL) {
3653 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3654 error = ENOTDIR;
3655 goto out;
3656 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3657 error = EISDIR;
3658 goto out;
3659 }
3660 #ifdef CAPABILITIES
3661 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3662 /*
3663 * If the target already exists we require CAP_UNLINKAT
3664 * from 'newfd', when newfd was used for the lookup.
3665 */
3666 error = cap_check(&tond.ni_filecaps.fc_rights,
3667 &cap_unlinkat_rights);
3668 if (error != 0)
3669 goto out;
3670 }
3671 #endif
3672 }
3673 if (fvp == tdvp) {
3674 error = EINVAL;
3675 goto out;
3676 }
3677 /*
3678 * If the source is the same as the destination (that is, if they
3679 * are links to the same vnode), then there is nothing to do.
3680 */
3681 if (fvp == tvp)
3682 error = ERESTART;
3683 #ifdef MAC
3684 else
3685 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3686 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3687 #endif
3688 out:
3689 if (error == 0) {
3690 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3691 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3692 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3693 NDFREE(&tond, NDF_ONLY_PNBUF);
3694 } else {
3695 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3696 NDFREE(&tond, NDF_ONLY_PNBUF);
3697 if (tvp != NULL)
3698 vput(tvp);
3699 if (tdvp == tvp)
3700 vrele(tdvp);
3701 else
3702 vput(tdvp);
3703 vrele(fromnd.ni_dvp);
3704 vrele(fvp);
3705 }
3706 vrele(tond.ni_startdir);
3707 vn_finished_write(mp);
3708 out1:
3709 if (fromnd.ni_startdir)
3710 vrele(fromnd.ni_startdir);
3711 if (error == ERESTART)
3712 return (0);
3713 if (error == ERELOOKUP)
3714 goto again;
3715 return (error);
3716 }
3717
3718 /*
3719 * Make a directory file.
3720 */
3721 #ifndef _SYS_SYSPROTO_H_
3722 struct mkdir_args {
3723 char *path;
3724 int mode;
3725 };
3726 #endif
3727 int
sys_mkdir(struct thread * td,struct mkdir_args * uap)3728 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3729 {
3730
3731 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3732 uap->mode));
3733 }
3734
3735 #ifndef _SYS_SYSPROTO_H_
3736 struct mkdirat_args {
3737 int fd;
3738 char *path;
3739 mode_t mode;
3740 };
3741 #endif
3742 int
sys_mkdirat(struct thread * td,struct mkdirat_args * uap)3743 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3744 {
3745
3746 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3747 }
3748
3749 int
kern_mkdirat(struct thread * td,int fd,const char * path,enum uio_seg segflg,int mode)3750 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3751 int mode)
3752 {
3753 struct mount *mp;
3754 struct vattr vattr;
3755 struct nameidata nd;
3756 int error;
3757
3758 AUDIT_ARG_MODE(mode);
3759 restart:
3760 bwillwrite();
3761 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3762 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3763 segflg, path, fd, &cap_mkdirat_rights, td);
3764 if ((error = namei(&nd)) != 0)
3765 return (error);
3766 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3767 NDFREE(&nd, NDF_ONLY_PNBUF);
3768 vput(nd.ni_dvp);
3769 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3770 return (error);
3771 goto restart;
3772 }
3773 VATTR_NULL(&vattr);
3774 vattr.va_type = VDIR;
3775 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3776 #ifdef MAC
3777 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3778 &vattr);
3779 if (error != 0)
3780 goto out;
3781 #endif
3782 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3783 #ifdef MAC
3784 out:
3785 #endif
3786 NDFREE(&nd, NDF_ONLY_PNBUF);
3787 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3788 vn_finished_write(mp);
3789 if (error == ERELOOKUP)
3790 goto restart;
3791 return (error);
3792 }
3793
3794 /*
3795 * Remove a directory file.
3796 */
3797 #ifndef _SYS_SYSPROTO_H_
3798 struct rmdir_args {
3799 char *path;
3800 };
3801 #endif
3802 int
sys_rmdir(struct thread * td,struct rmdir_args * uap)3803 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3804 {
3805
3806 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3807 0));
3808 }
3809
3810 int
kern_frmdirat(struct thread * td,int dfd,const char * path,int fd,enum uio_seg pathseg,int flag)3811 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3812 enum uio_seg pathseg, int flag)
3813 {
3814 struct mount *mp;
3815 struct vnode *vp;
3816 struct file *fp;
3817 struct nameidata nd;
3818 cap_rights_t rights;
3819 int error;
3820
3821 fp = NULL;
3822 if (fd != FD_NONE) {
3823 error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP),
3824 &fp);
3825 if (error != 0)
3826 return (error);
3827 }
3828
3829 restart:
3830 bwillwrite();
3831 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3832 at2cnpflags(flag, AT_RESOLVE_BENEATH),
3833 pathseg, path, dfd, &cap_unlinkat_rights, td);
3834 if ((error = namei(&nd)) != 0)
3835 goto fdout;
3836 vp = nd.ni_vp;
3837 if (vp->v_type != VDIR) {
3838 error = ENOTDIR;
3839 goto out;
3840 }
3841 /*
3842 * No rmdir "." please.
3843 */
3844 if (nd.ni_dvp == vp) {
3845 error = EINVAL;
3846 goto out;
3847 }
3848 /*
3849 * The root of a mounted filesystem cannot be deleted.
3850 */
3851 if (vp->v_vflag & VV_ROOT) {
3852 error = EBUSY;
3853 goto out;
3854 }
3855
3856 if (fp != NULL && fp->f_vnode != vp) {
3857 if (VN_IS_DOOMED(fp->f_vnode))
3858 error = EBADF;
3859 else
3860 error = EDEADLK;
3861 goto out;
3862 }
3863
3864 #ifdef MAC
3865 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3866 &nd.ni_cnd);
3867 if (error != 0)
3868 goto out;
3869 #endif
3870 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3871 NDFREE(&nd, NDF_ONLY_PNBUF);
3872 vput(vp);
3873 if (nd.ni_dvp == vp)
3874 vrele(nd.ni_dvp);
3875 else
3876 vput(nd.ni_dvp);
3877 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3878 goto fdout;
3879 goto restart;
3880 }
3881 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3882 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3883 vn_finished_write(mp);
3884 out:
3885 NDFREE(&nd, NDF_ONLY_PNBUF);
3886 vput(vp);
3887 if (nd.ni_dvp == vp)
3888 vrele(nd.ni_dvp);
3889 else
3890 vput(nd.ni_dvp);
3891 if (error == ERELOOKUP)
3892 goto restart;
3893 fdout:
3894 if (fp != NULL)
3895 fdrop(fp, td);
3896 return (error);
3897 }
3898
3899 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3900 int
freebsd11_kern_getdirentries(struct thread * td,int fd,char * ubuf,u_int count,long * basep,void (* func)(struct freebsd11_dirent *))3901 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3902 long *basep, void (*func)(struct freebsd11_dirent *))
3903 {
3904 struct freebsd11_dirent dstdp;
3905 struct dirent *dp, *edp;
3906 char *dirbuf;
3907 off_t base;
3908 ssize_t resid, ucount;
3909 int error;
3910
3911 /* XXX arbitrary sanity limit on `count'. */
3912 count = min(count, 64 * 1024);
3913
3914 dirbuf = malloc(count, M_TEMP, M_WAITOK);
3915
3916 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3917 UIO_SYSSPACE);
3918 if (error != 0)
3919 goto done;
3920 if (basep != NULL)
3921 *basep = base;
3922
3923 ucount = 0;
3924 for (dp = (struct dirent *)dirbuf,
3925 edp = (struct dirent *)&dirbuf[count - resid];
3926 ucount < count && dp < edp; ) {
3927 if (dp->d_reclen == 0)
3928 break;
3929 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3930 if (dp->d_namlen >= sizeof(dstdp.d_name))
3931 continue;
3932 dstdp.d_type = dp->d_type;
3933 dstdp.d_namlen = dp->d_namlen;
3934 dstdp.d_fileno = dp->d_fileno; /* truncate */
3935 if (dstdp.d_fileno != dp->d_fileno) {
3936 switch (ino64_trunc_error) {
3937 default:
3938 case 0:
3939 break;
3940 case 1:
3941 error = EOVERFLOW;
3942 goto done;
3943 case 2:
3944 dstdp.d_fileno = UINT32_MAX;
3945 break;
3946 }
3947 }
3948 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3949 ((dp->d_namlen + 1 + 3) &~ 3);
3950 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3951 bzero(dstdp.d_name + dstdp.d_namlen,
3952 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3953 dstdp.d_namlen);
3954 MPASS(dstdp.d_reclen <= dp->d_reclen);
3955 MPASS(ucount + dstdp.d_reclen <= count);
3956 if (func != NULL)
3957 func(&dstdp);
3958 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3959 if (error != 0)
3960 break;
3961 dp = (struct dirent *)((char *)dp + dp->d_reclen);
3962 ucount += dstdp.d_reclen;
3963 }
3964
3965 done:
3966 free(dirbuf, M_TEMP);
3967 if (error == 0)
3968 td->td_retval[0] = ucount;
3969 return (error);
3970 }
3971 #endif /* COMPAT */
3972
3973 #ifdef COMPAT_43
3974 static void
ogetdirentries_cvt(struct freebsd11_dirent * dp)3975 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3976 {
3977 #if (BYTE_ORDER == LITTLE_ENDIAN)
3978 /*
3979 * The expected low byte of dp->d_namlen is our dp->d_type.
3980 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3981 */
3982 dp->d_type = dp->d_namlen;
3983 dp->d_namlen = 0;
3984 #else
3985 /*
3986 * The dp->d_type is the high byte of the expected dp->d_namlen,
3987 * so must be zero'ed.
3988 */
3989 dp->d_type = 0;
3990 #endif
3991 }
3992
3993 /*
3994 * Read a block of directory entries in a filesystem independent format.
3995 */
3996 #ifndef _SYS_SYSPROTO_H_
3997 struct ogetdirentries_args {
3998 int fd;
3999 char *buf;
4000 u_int count;
4001 long *basep;
4002 };
4003 #endif
4004 int
ogetdirentries(struct thread * td,struct ogetdirentries_args * uap)4005 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4006 {
4007 long loff;
4008 int error;
4009
4010 error = kern_ogetdirentries(td, uap, &loff);
4011 if (error == 0)
4012 error = copyout(&loff, uap->basep, sizeof(long));
4013 return (error);
4014 }
4015
4016 int
kern_ogetdirentries(struct thread * td,struct ogetdirentries_args * uap,long * ploff)4017 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4018 long *ploff)
4019 {
4020 long base;
4021 int error;
4022
4023 /* XXX arbitrary sanity limit on `count'. */
4024 if (uap->count > 64 * 1024)
4025 return (EINVAL);
4026
4027 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4028 &base, ogetdirentries_cvt);
4029
4030 if (error == 0 && uap->basep != NULL)
4031 error = copyout(&base, uap->basep, sizeof(long));
4032
4033 return (error);
4034 }
4035 #endif /* COMPAT_43 */
4036
4037 #if defined(COMPAT_FREEBSD11)
4038 #ifndef _SYS_SYSPROTO_H_
4039 struct freebsd11_getdirentries_args {
4040 int fd;
4041 char *buf;
4042 u_int count;
4043 long *basep;
4044 };
4045 #endif
4046 int
freebsd11_getdirentries(struct thread * td,struct freebsd11_getdirentries_args * uap)4047 freebsd11_getdirentries(struct thread *td,
4048 struct freebsd11_getdirentries_args *uap)
4049 {
4050 long base;
4051 int error;
4052
4053 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4054 &base, NULL);
4055
4056 if (error == 0 && uap->basep != NULL)
4057 error = copyout(&base, uap->basep, sizeof(long));
4058 return (error);
4059 }
4060
4061 int
freebsd11_getdents(struct thread * td,struct freebsd11_getdents_args * uap)4062 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4063 {
4064 struct freebsd11_getdirentries_args ap;
4065
4066 ap.fd = uap->fd;
4067 ap.buf = uap->buf;
4068 ap.count = uap->count;
4069 ap.basep = NULL;
4070 return (freebsd11_getdirentries(td, &ap));
4071 }
4072 #endif /* COMPAT_FREEBSD11 */
4073
4074 /*
4075 * Read a block of directory entries in a filesystem independent format.
4076 */
4077 int
sys_getdirentries(struct thread * td,struct getdirentries_args * uap)4078 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4079 {
4080 off_t base;
4081 int error;
4082
4083 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4084 NULL, UIO_USERSPACE);
4085 if (error != 0)
4086 return (error);
4087 if (uap->basep != NULL)
4088 error = copyout(&base, uap->basep, sizeof(off_t));
4089 return (error);
4090 }
4091
4092 int
kern_getdirentries(struct thread * td,int fd,char * buf,size_t count,off_t * basep,ssize_t * residp,enum uio_seg bufseg)4093 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4094 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4095 {
4096 struct vnode *vp;
4097 struct file *fp;
4098 struct uio auio;
4099 struct iovec aiov;
4100 off_t loff;
4101 int error, eofflag;
4102 off_t foffset;
4103
4104 AUDIT_ARG_FD(fd);
4105 if (count > IOSIZE_MAX)
4106 return (EINVAL);
4107 auio.uio_resid = count;
4108 error = getvnode(td, fd, &cap_read_rights, &fp);
4109 if (error != 0)
4110 return (error);
4111 if ((fp->f_flag & FREAD) == 0) {
4112 fdrop(fp, td);
4113 return (EBADF);
4114 }
4115 vp = fp->f_vnode;
4116 foffset = foffset_lock(fp, 0);
4117 unionread:
4118 if (vp->v_type != VDIR) {
4119 error = EINVAL;
4120 goto fail;
4121 }
4122 aiov.iov_base = buf;
4123 aiov.iov_len = count;
4124 auio.uio_iov = &aiov;
4125 auio.uio_iovcnt = 1;
4126 auio.uio_rw = UIO_READ;
4127 auio.uio_segflg = bufseg;
4128 auio.uio_td = td;
4129 vn_lock(vp, LK_SHARED | LK_RETRY);
4130 AUDIT_ARG_VNODE1(vp);
4131 loff = auio.uio_offset = foffset;
4132 #ifdef MAC
4133 error = mac_vnode_check_readdir(td->td_ucred, vp);
4134 if (error == 0)
4135 #endif
4136 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4137 NULL);
4138 foffset = auio.uio_offset;
4139 if (error != 0) {
4140 VOP_UNLOCK(vp);
4141 goto fail;
4142 }
4143 if (count == auio.uio_resid &&
4144 (vp->v_vflag & VV_ROOT) &&
4145 (vp->v_mount->mnt_flag & MNT_UNION)) {
4146 struct vnode *tvp = vp;
4147
4148 vp = vp->v_mount->mnt_vnodecovered;
4149 VREF(vp);
4150 fp->f_vnode = vp;
4151 foffset = 0;
4152 vput(tvp);
4153 goto unionread;
4154 }
4155 VOP_UNLOCK(vp);
4156 *basep = loff;
4157 if (residp != NULL)
4158 *residp = auio.uio_resid;
4159 td->td_retval[0] = count - auio.uio_resid;
4160 fail:
4161 foffset_unlock(fp, foffset, 0);
4162 fdrop(fp, td);
4163 return (error);
4164 }
4165
4166 /*
4167 * Set the mode mask for creation of filesystem nodes.
4168 */
4169 #ifndef _SYS_SYSPROTO_H_
4170 struct umask_args {
4171 int newmask;
4172 };
4173 #endif
4174 int
sys_umask(struct thread * td,struct umask_args * uap)4175 sys_umask(struct thread *td, struct umask_args *uap)
4176 {
4177 struct pwddesc *pdp;
4178
4179 pdp = td->td_proc->p_pd;
4180 PWDDESC_XLOCK(pdp);
4181 td->td_retval[0] = pdp->pd_cmask;
4182 pdp->pd_cmask = uap->newmask & ALLPERMS;
4183 PWDDESC_XUNLOCK(pdp);
4184 return (0);
4185 }
4186
4187 /*
4188 * Void all references to file by ripping underlying filesystem away from
4189 * vnode.
4190 */
4191 #ifndef _SYS_SYSPROTO_H_
4192 struct revoke_args {
4193 char *path;
4194 };
4195 #endif
4196 int
sys_revoke(struct thread * td,struct revoke_args * uap)4197 sys_revoke(struct thread *td, struct revoke_args *uap)
4198 {
4199 struct vnode *vp;
4200 struct vattr vattr;
4201 struct nameidata nd;
4202 int error;
4203
4204 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4205 uap->path, td);
4206 if ((error = namei(&nd)) != 0)
4207 return (error);
4208 vp = nd.ni_vp;
4209 NDFREE_NOTHING(&nd);
4210 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4211 error = EINVAL;
4212 goto out;
4213 }
4214 #ifdef MAC
4215 error = mac_vnode_check_revoke(td->td_ucred, vp);
4216 if (error != 0)
4217 goto out;
4218 #endif
4219 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4220 if (error != 0)
4221 goto out;
4222 if (td->td_ucred->cr_uid != vattr.va_uid) {
4223 error = priv_check(td, PRIV_VFS_ADMIN);
4224 if (error != 0)
4225 goto out;
4226 }
4227 if (devfs_usecount(vp) > 0)
4228 VOP_REVOKE(vp, REVOKEALL);
4229 out:
4230 vput(vp);
4231 return (error);
4232 }
4233
4234 /*
4235 * Convert a user file descriptor to a kernel file entry and check that, if it
4236 * is a capability, the correct rights are present. A reference on the file
4237 * entry is held upon returning.
4238 */
4239 int
getvnode(struct thread * td,int fd,cap_rights_t * rightsp,struct file ** fpp)4240 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4241 {
4242 struct file *fp;
4243 int error;
4244
4245 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4246 if (error != 0)
4247 return (error);
4248
4249 /*
4250 * The file could be not of the vnode type, or it may be not
4251 * yet fully initialized, in which case the f_vnode pointer
4252 * may be set, but f_ops is still badfileops. E.g.,
4253 * devfs_open() transiently create such situation to
4254 * facilitate csw d_fdopen().
4255 *
4256 * Dupfdopen() handling in kern_openat() installs the
4257 * half-baked file into the process descriptor table, allowing
4258 * other thread to dereference it. Guard against the race by
4259 * checking f_ops.
4260 */
4261 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4262 fdrop(fp, td);
4263 return (EINVAL);
4264 }
4265 *fpp = fp;
4266 return (0);
4267 }
4268
4269 /*
4270 * Get an (NFS) file handle.
4271 */
4272 #ifndef _SYS_SYSPROTO_H_
4273 struct lgetfh_args {
4274 char *fname;
4275 fhandle_t *fhp;
4276 };
4277 #endif
4278 int
sys_lgetfh(struct thread * td,struct lgetfh_args * uap)4279 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4280 {
4281
4282 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4283 UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4284 }
4285
4286 #ifndef _SYS_SYSPROTO_H_
4287 struct getfh_args {
4288 char *fname;
4289 fhandle_t *fhp;
4290 };
4291 #endif
4292 int
sys_getfh(struct thread * td,struct getfh_args * uap)4293 sys_getfh(struct thread *td, struct getfh_args *uap)
4294 {
4295
4296 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4297 uap->fhp, UIO_USERSPACE));
4298 }
4299
4300 /*
4301 * syscall for the rpc.lockd to use to translate an open descriptor into
4302 * a NFS file handle.
4303 *
4304 * warning: do not remove the priv_check() call or this becomes one giant
4305 * security hole.
4306 */
4307 #ifndef _SYS_SYSPROTO_H_
4308 struct getfhat_args {
4309 int fd;
4310 char *path;
4311 fhandle_t *fhp;
4312 int flags;
4313 };
4314 #endif
4315 int
sys_getfhat(struct thread * td,struct getfhat_args * uap)4316 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4317 {
4318
4319 if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4320 return (EINVAL);
4321 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4322 uap->fhp, UIO_USERSPACE));
4323 }
4324
4325 int
kern_getfhat(struct thread * td,int flags,int fd,const char * path,enum uio_seg pathseg,fhandle_t * fhp,enum uio_seg fhseg)4326 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4327 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4328 {
4329 struct nameidata nd;
4330 fhandle_t fh;
4331 struct vnode *vp;
4332 int error;
4333
4334 error = priv_check(td, PRIV_VFS_GETFH);
4335 if (error != 0)
4336 return (error);
4337 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4338 AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4339 fd, td);
4340 error = namei(&nd);
4341 if (error != 0)
4342 return (error);
4343 NDFREE_NOTHING(&nd);
4344 vp = nd.ni_vp;
4345 bzero(&fh, sizeof(fh));
4346 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4347 error = VOP_VPTOFH(vp, &fh.fh_fid);
4348 vput(vp);
4349 if (error == 0) {
4350 if (fhseg == UIO_USERSPACE)
4351 error = copyout(&fh, fhp, sizeof (fh));
4352 else
4353 memcpy(fhp, &fh, sizeof(fh));
4354 }
4355 return (error);
4356 }
4357
4358 #ifndef _SYS_SYSPROTO_H_
4359 struct fhlink_args {
4360 fhandle_t *fhp;
4361 const char *to;
4362 };
4363 #endif
4364 int
sys_fhlink(struct thread * td,struct fhlink_args * uap)4365 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4366 {
4367
4368 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4369 }
4370
4371 #ifndef _SYS_SYSPROTO_H_
4372 struct fhlinkat_args {
4373 fhandle_t *fhp;
4374 int tofd;
4375 const char *to;
4376 };
4377 #endif
4378 int
sys_fhlinkat(struct thread * td,struct fhlinkat_args * uap)4379 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4380 {
4381
4382 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4383 }
4384
4385 static int
kern_fhlinkat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,fhandle_t * fhp)4386 kern_fhlinkat(struct thread *td, int fd, const char *path,
4387 enum uio_seg pathseg, fhandle_t *fhp)
4388 {
4389 fhandle_t fh;
4390 struct mount *mp;
4391 struct vnode *vp;
4392 int error;
4393
4394 error = priv_check(td, PRIV_VFS_GETFH);
4395 if (error != 0)
4396 return (error);
4397 error = copyin(fhp, &fh, sizeof(fh));
4398 if (error != 0)
4399 return (error);
4400 do {
4401 bwillwrite();
4402 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4403 return (ESTALE);
4404 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4405 vfs_unbusy(mp);
4406 if (error != 0)
4407 return (error);
4408 VOP_UNLOCK(vp);
4409 error = kern_linkat_vp(td, vp, fd, path, pathseg);
4410 } while (error == EAGAIN || error == ERELOOKUP);
4411 return (error);
4412 }
4413
4414 #ifndef _SYS_SYSPROTO_H_
4415 struct fhreadlink_args {
4416 fhandle_t *fhp;
4417 char *buf;
4418 size_t bufsize;
4419 };
4420 #endif
4421 int
sys_fhreadlink(struct thread * td,struct fhreadlink_args * uap)4422 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4423 {
4424 fhandle_t fh;
4425 struct mount *mp;
4426 struct vnode *vp;
4427 int error;
4428
4429 error = priv_check(td, PRIV_VFS_GETFH);
4430 if (error != 0)
4431 return (error);
4432 if (uap->bufsize > IOSIZE_MAX)
4433 return (EINVAL);
4434 error = copyin(uap->fhp, &fh, sizeof(fh));
4435 if (error != 0)
4436 return (error);
4437 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4438 return (ESTALE);
4439 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4440 vfs_unbusy(mp);
4441 if (error != 0)
4442 return (error);
4443 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4444 vput(vp);
4445 return (error);
4446 }
4447
4448 /*
4449 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4450 * open descriptor.
4451 *
4452 * warning: do not remove the priv_check() call or this becomes one giant
4453 * security hole.
4454 */
4455 #ifndef _SYS_SYSPROTO_H_
4456 struct fhopen_args {
4457 const struct fhandle *u_fhp;
4458 int flags;
4459 };
4460 #endif
4461 int
sys_fhopen(struct thread * td,struct fhopen_args * uap)4462 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4463 {
4464 return (kern_fhopen(td, uap->u_fhp, uap->flags));
4465 }
4466
4467 int
kern_fhopen(struct thread * td,const struct fhandle * u_fhp,int flags)4468 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4469 {
4470 struct mount *mp;
4471 struct vnode *vp;
4472 struct fhandle fhp;
4473 struct file *fp;
4474 int fmode, error;
4475 int indx;
4476
4477 error = priv_check(td, PRIV_VFS_FHOPEN);
4478 if (error != 0)
4479 return (error);
4480 indx = -1;
4481 fmode = FFLAGS(flags);
4482 /* why not allow a non-read/write open for our lockd? */
4483 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4484 return (EINVAL);
4485 error = copyin(u_fhp, &fhp, sizeof(fhp));
4486 if (error != 0)
4487 return(error);
4488 /* find the mount point */
4489 mp = vfs_busyfs(&fhp.fh_fsid);
4490 if (mp == NULL)
4491 return (ESTALE);
4492 /* now give me my vnode, it gets returned to me locked */
4493 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4494 vfs_unbusy(mp);
4495 if (error != 0)
4496 return (error);
4497
4498 error = falloc_noinstall(td, &fp);
4499 if (error != 0) {
4500 vput(vp);
4501 return (error);
4502 }
4503 /*
4504 * An extra reference on `fp' has been held for us by
4505 * falloc_noinstall().
4506 */
4507
4508 #ifdef INVARIANTS
4509 td->td_dupfd = -1;
4510 #endif
4511 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4512 if (error != 0) {
4513 KASSERT(fp->f_ops == &badfileops,
4514 ("VOP_OPEN in fhopen() set f_ops"));
4515 KASSERT(td->td_dupfd < 0,
4516 ("fhopen() encountered fdopen()"));
4517
4518 vput(vp);
4519 goto bad;
4520 }
4521 #ifdef INVARIANTS
4522 td->td_dupfd = 0;
4523 #endif
4524 fp->f_vnode = vp;
4525 finit_vnode(fp, fmode, NULL, &vnops);
4526 VOP_UNLOCK(vp);
4527 if ((fmode & O_TRUNC) != 0) {
4528 error = fo_truncate(fp, 0, td->td_ucred, td);
4529 if (error != 0)
4530 goto bad;
4531 }
4532
4533 error = finstall(td, fp, &indx, fmode, NULL);
4534 bad:
4535 fdrop(fp, td);
4536 td->td_retval[0] = indx;
4537 return (error);
4538 }
4539
4540 /*
4541 * Stat an (NFS) file handle.
4542 */
4543 #ifndef _SYS_SYSPROTO_H_
4544 struct fhstat_args {
4545 struct fhandle *u_fhp;
4546 struct stat *sb;
4547 };
4548 #endif
4549 int
sys_fhstat(struct thread * td,struct fhstat_args * uap)4550 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4551 {
4552 struct stat sb;
4553 struct fhandle fh;
4554 int error;
4555
4556 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4557 if (error != 0)
4558 return (error);
4559 error = kern_fhstat(td, fh, &sb);
4560 if (error == 0)
4561 error = copyout(&sb, uap->sb, sizeof(sb));
4562 return (error);
4563 }
4564
4565 int
kern_fhstat(struct thread * td,struct fhandle fh,struct stat * sb)4566 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4567 {
4568 struct mount *mp;
4569 struct vnode *vp;
4570 int error;
4571
4572 error = priv_check(td, PRIV_VFS_FHSTAT);
4573 if (error != 0)
4574 return (error);
4575 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4576 return (ESTALE);
4577 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4578 vfs_unbusy(mp);
4579 if (error != 0)
4580 return (error);
4581 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4582 vput(vp);
4583 return (error);
4584 }
4585
4586 /*
4587 * Implement fstatfs() for (NFS) file handles.
4588 */
4589 #ifndef _SYS_SYSPROTO_H_
4590 struct fhstatfs_args {
4591 struct fhandle *u_fhp;
4592 struct statfs *buf;
4593 };
4594 #endif
4595 int
sys_fhstatfs(struct thread * td,struct fhstatfs_args * uap)4596 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4597 {
4598 struct statfs *sfp;
4599 fhandle_t fh;
4600 int error;
4601
4602 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4603 if (error != 0)
4604 return (error);
4605 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4606 error = kern_fhstatfs(td, fh, sfp);
4607 if (error == 0)
4608 error = copyout(sfp, uap->buf, sizeof(*sfp));
4609 free(sfp, M_STATFS);
4610 return (error);
4611 }
4612
4613 int
kern_fhstatfs(struct thread * td,fhandle_t fh,struct statfs * buf)4614 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4615 {
4616 struct mount *mp;
4617 struct vnode *vp;
4618 int error;
4619
4620 error = priv_check(td, PRIV_VFS_FHSTATFS);
4621 if (error != 0)
4622 return (error);
4623 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4624 return (ESTALE);
4625 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4626 if (error != 0) {
4627 vfs_unbusy(mp);
4628 return (error);
4629 }
4630 vput(vp);
4631 error = prison_canseemount(td->td_ucred, mp);
4632 if (error != 0)
4633 goto out;
4634 #ifdef MAC
4635 error = mac_mount_check_stat(td->td_ucred, mp);
4636 if (error != 0)
4637 goto out;
4638 #endif
4639 error = VFS_STATFS(mp, buf);
4640 out:
4641 vfs_unbusy(mp);
4642 return (error);
4643 }
4644
4645 /*
4646 * Unlike madvise(2), we do not make a best effort to remember every
4647 * possible caching hint. Instead, we remember the last setting with
4648 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4649 * region of any current setting.
4650 */
4651 int
kern_posix_fadvise(struct thread * td,int fd,off_t offset,off_t len,int advice)4652 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4653 int advice)
4654 {
4655 struct fadvise_info *fa, *new;
4656 struct file *fp;
4657 struct vnode *vp;
4658 off_t end;
4659 int error;
4660
4661 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4662 return (EINVAL);
4663 AUDIT_ARG_VALUE(advice);
4664 switch (advice) {
4665 case POSIX_FADV_SEQUENTIAL:
4666 case POSIX_FADV_RANDOM:
4667 case POSIX_FADV_NOREUSE:
4668 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4669 break;
4670 case POSIX_FADV_NORMAL:
4671 case POSIX_FADV_WILLNEED:
4672 case POSIX_FADV_DONTNEED:
4673 new = NULL;
4674 break;
4675 default:
4676 return (EINVAL);
4677 }
4678 /* XXX: CAP_POSIX_FADVISE? */
4679 AUDIT_ARG_FD(fd);
4680 error = fget(td, fd, &cap_no_rights, &fp);
4681 if (error != 0)
4682 goto out;
4683 AUDIT_ARG_FILE(td->td_proc, fp);
4684 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4685 error = ESPIPE;
4686 goto out;
4687 }
4688 if (fp->f_type != DTYPE_VNODE) {
4689 error = ENODEV;
4690 goto out;
4691 }
4692 vp = fp->f_vnode;
4693 if (vp->v_type != VREG) {
4694 error = ENODEV;
4695 goto out;
4696 }
4697 if (len == 0)
4698 end = OFF_MAX;
4699 else
4700 end = offset + len - 1;
4701 switch (advice) {
4702 case POSIX_FADV_SEQUENTIAL:
4703 case POSIX_FADV_RANDOM:
4704 case POSIX_FADV_NOREUSE:
4705 /*
4706 * Try to merge any existing non-standard region with
4707 * this new region if possible, otherwise create a new
4708 * non-standard region for this request.
4709 */
4710 mtx_pool_lock(mtxpool_sleep, fp);
4711 fa = fp->f_advice;
4712 if (fa != NULL && fa->fa_advice == advice &&
4713 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4714 (end != OFF_MAX && fa->fa_start == end + 1) ||
4715 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4716 if (offset < fa->fa_start)
4717 fa->fa_start = offset;
4718 if (end > fa->fa_end)
4719 fa->fa_end = end;
4720 } else {
4721 new->fa_advice = advice;
4722 new->fa_start = offset;
4723 new->fa_end = end;
4724 fp->f_advice = new;
4725 new = fa;
4726 }
4727 mtx_pool_unlock(mtxpool_sleep, fp);
4728 break;
4729 case POSIX_FADV_NORMAL:
4730 /*
4731 * If a the "normal" region overlaps with an existing
4732 * non-standard region, trim or remove the
4733 * non-standard region.
4734 */
4735 mtx_pool_lock(mtxpool_sleep, fp);
4736 fa = fp->f_advice;
4737 if (fa != NULL) {
4738 if (offset <= fa->fa_start && end >= fa->fa_end) {
4739 new = fa;
4740 fp->f_advice = NULL;
4741 } else if (offset <= fa->fa_start &&
4742 end >= fa->fa_start)
4743 fa->fa_start = end + 1;
4744 else if (offset <= fa->fa_end && end >= fa->fa_end)
4745 fa->fa_end = offset - 1;
4746 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4747 /*
4748 * If the "normal" region is a middle
4749 * portion of the existing
4750 * non-standard region, just remove
4751 * the whole thing rather than picking
4752 * one side or the other to
4753 * preserve.
4754 */
4755 new = fa;
4756 fp->f_advice = NULL;
4757 }
4758 }
4759 mtx_pool_unlock(mtxpool_sleep, fp);
4760 break;
4761 case POSIX_FADV_WILLNEED:
4762 case POSIX_FADV_DONTNEED:
4763 error = VOP_ADVISE(vp, offset, end, advice);
4764 break;
4765 }
4766 out:
4767 if (fp != NULL)
4768 fdrop(fp, td);
4769 free(new, M_FADVISE);
4770 return (error);
4771 }
4772
4773 int
sys_posix_fadvise(struct thread * td,struct posix_fadvise_args * uap)4774 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4775 {
4776 int error;
4777
4778 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4779 uap->advice);
4780 return (kern_posix_error(td, error));
4781 }
4782
4783 int
kern_copy_file_range(struct thread * td,int infd,off_t * inoffp,int outfd,off_t * outoffp,size_t len,unsigned int flags)4784 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4785 off_t *outoffp, size_t len, unsigned int flags)
4786 {
4787 struct file *infp, *outfp;
4788 struct vnode *invp, *outvp;
4789 int error;
4790 size_t retlen;
4791 void *rl_rcookie, *rl_wcookie;
4792 off_t savinoff, savoutoff;
4793
4794 infp = outfp = NULL;
4795 rl_rcookie = rl_wcookie = NULL;
4796 savinoff = -1;
4797 error = 0;
4798 retlen = 0;
4799
4800 if (flags != 0) {
4801 error = EINVAL;
4802 goto out;
4803 }
4804 if (len > SSIZE_MAX)
4805 /*
4806 * Although the len argument is size_t, the return argument
4807 * is ssize_t (which is signed). Therefore a size that won't
4808 * fit in ssize_t can't be returned.
4809 */
4810 len = SSIZE_MAX;
4811
4812 /* Get the file structures for the file descriptors. */
4813 error = fget_read(td, infd, &cap_read_rights, &infp);
4814 if (error != 0)
4815 goto out;
4816 if (infp->f_ops == &badfileops) {
4817 error = EBADF;
4818 goto out;
4819 }
4820 if (infp->f_vnode == NULL) {
4821 error = EINVAL;
4822 goto out;
4823 }
4824 error = fget_write(td, outfd, &cap_write_rights, &outfp);
4825 if (error != 0)
4826 goto out;
4827 if (outfp->f_ops == &badfileops) {
4828 error = EBADF;
4829 goto out;
4830 }
4831 if (outfp->f_vnode == NULL) {
4832 error = EINVAL;
4833 goto out;
4834 }
4835
4836 /* Set the offset pointers to the correct place. */
4837 if (inoffp == NULL)
4838 inoffp = &infp->f_offset;
4839 if (outoffp == NULL)
4840 outoffp = &outfp->f_offset;
4841 savinoff = *inoffp;
4842 savoutoff = *outoffp;
4843
4844 invp = infp->f_vnode;
4845 outvp = outfp->f_vnode;
4846 /* Sanity check the f_flag bits. */
4847 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4848 (infp->f_flag & FREAD) == 0) {
4849 error = EBADF;
4850 goto out;
4851 }
4852
4853 /* If len == 0, just return 0. */
4854 if (len == 0)
4855 goto out;
4856
4857 /*
4858 * If infp and outfp refer to the same file, the byte ranges cannot
4859 * overlap.
4860 */
4861 if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4862 savoutoff) || (savinoff > savoutoff && savoutoff + len >
4863 savinoff))) {
4864 error = EINVAL;
4865 goto out;
4866 }
4867
4868 /* Range lock the byte ranges for both invp and outvp. */
4869 for (;;) {
4870 rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4871 len);
4872 rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4873 len);
4874 if (rl_rcookie != NULL)
4875 break;
4876 vn_rangelock_unlock(outvp, rl_wcookie);
4877 rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4878 vn_rangelock_unlock(invp, rl_rcookie);
4879 }
4880
4881 retlen = len;
4882 error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4883 flags, infp->f_cred, outfp->f_cred, td);
4884 out:
4885 if (rl_rcookie != NULL)
4886 vn_rangelock_unlock(invp, rl_rcookie);
4887 if (rl_wcookie != NULL)
4888 vn_rangelock_unlock(outvp, rl_wcookie);
4889 if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4890 *inoffp = savinoff;
4891 *outoffp = savoutoff;
4892 }
4893 if (outfp != NULL)
4894 fdrop(outfp, td);
4895 if (infp != NULL)
4896 fdrop(infp, td);
4897 td->td_retval[0] = retlen;
4898 return (error);
4899 }
4900
4901 int
sys_copy_file_range(struct thread * td,struct copy_file_range_args * uap)4902 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4903 {
4904 off_t inoff, outoff, *inoffp, *outoffp;
4905 int error;
4906
4907 inoffp = outoffp = NULL;
4908 if (uap->inoffp != NULL) {
4909 error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4910 if (error != 0)
4911 return (error);
4912 inoffp = &inoff;
4913 }
4914 if (uap->outoffp != NULL) {
4915 error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4916 if (error != 0)
4917 return (error);
4918 outoffp = &outoff;
4919 }
4920 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4921 outoffp, uap->len, uap->flags);
4922 if (error == 0 && uap->inoffp != NULL)
4923 error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4924 if (error == 0 && uap->outoffp != NULL)
4925 error = copyout(outoffp, uap->outoffp, sizeof(off_t));
4926 return (error);
4927 }
4928