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
37 #include "opt_capsicum.h"
38 #include "opt_ktrace.h"
39
40 #define EXTERR_CATEGORY EXTERR_CAT_VFSSYSCALL
41 #include <sys/systm.h>
42 #ifdef COMPAT_FREEBSD11
43 #include <sys/abi_compat.h>
44 #endif
45 #include <sys/bio.h>
46 #include <sys/buf.h>
47 #include <sys/capsicum.h>
48 #include <sys/disk.h>
49 #include <sys/dirent.h>
50 #include <sys/exterrvar.h>
51 #include <sys/fcntl.h>
52 #include <sys/file.h>
53 #include <sys/filedesc.h>
54 #include <sys/filio.h>
55 #include <sys/jail.h>
56 #include <sys/kernel.h>
57 #ifdef KTRACE
58 #include <sys/ktrace.h>
59 #endif
60 #include <sys/limits.h>
61 #include <sys/linker.h>
62 #include <sys/malloc.h>
63 #include <sys/mount.h>
64 #include <sys/mutex.h>
65 #include <sys/namei.h>
66 #include <sys/priv.h>
67 #include <sys/proc.h>
68 #include <sys/rwlock.h>
69 #include <sys/sdt.h>
70 #include <sys/stat.h>
71 #include <sys/stdarg.h>
72 #include <sys/sx.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #include <sys/sysproto.h>
76 #include <sys/unistd.h>
77 #include <sys/vnode.h>
78
79 #include <security/audit/audit.h>
80 #include <security/mac/mac_framework.h>
81
82 #include <vm/vm.h>
83 #include <vm/vm_object.h>
84 #include <vm/vm_page.h>
85 #include <vm/vnode_pager.h>
86 #include <vm/uma.h>
87
88 #include <fs/devfs/devfs.h>
89
90 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
91
92 static int kern_chflagsat(struct thread *td, int fd, const char *path,
93 enum uio_seg pathseg, u_long flags, int atflag);
94 static int setfflags(struct thread *td, struct vnode *, u_long);
95 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
96 static int getutimens(const struct timespec *, enum uio_seg,
97 struct timespec *, int *);
98 static int setutimes(struct thread *td, struct vnode *,
99 const struct timespec *, int, int);
100 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
101 struct thread *td);
102 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
103 enum uio_seg pathseg, fhandle_t *fhp);
104 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
105 size_t count, struct thread *td);
106 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
107 const char *path, enum uio_seg segflag);
108
109 uint64_t
at2cnpflags(u_int at_flags,u_int mask)110 at2cnpflags(u_int at_flags, u_int mask)
111 {
112 uint64_t res;
113
114 MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
115 (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
116
117 res = 0;
118 at_flags &= mask;
119 if ((at_flags & AT_RESOLVE_BENEATH) != 0)
120 res |= RBENEATH;
121 if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
122 res |= FOLLOW;
123 /* NOFOLLOW is pseudo flag */
124 if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
125 res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
126 FOLLOW;
127 }
128 if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
129 res |= EMPTYPATH;
130 return (res);
131 }
132
133 int
kern_sync(struct thread * td)134 kern_sync(struct thread *td)
135 {
136 struct mount *mp, *nmp;
137 int save;
138
139 mtx_lock(&mountlist_mtx);
140 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
141 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
142 nmp = TAILQ_NEXT(mp, mnt_list);
143 continue;
144 }
145 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
146 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
147 save = curthread_pflags_set(TDP_SYNCIO);
148 vfs_periodic(mp, MNT_NOWAIT);
149 VFS_SYNC(mp, MNT_NOWAIT);
150 curthread_pflags_restore(save);
151 vn_finished_write(mp);
152 }
153 mtx_lock(&mountlist_mtx);
154 nmp = TAILQ_NEXT(mp, mnt_list);
155 vfs_unbusy(mp);
156 }
157 mtx_unlock(&mountlist_mtx);
158 return (0);
159 }
160
161 /*
162 * Sync each mounted filesystem.
163 */
164 #ifndef _SYS_SYSPROTO_H_
165 struct sync_args {
166 int dummy;
167 };
168 #endif
169 /* ARGSUSED */
170 int
sys_sync(struct thread * td,struct sync_args * uap)171 sys_sync(struct thread *td, struct sync_args *uap)
172 {
173
174 return (kern_sync(td));
175 }
176
177 /*
178 * Change filesystem quotas.
179 */
180 #ifndef _SYS_SYSPROTO_H_
181 struct quotactl_args {
182 char *path;
183 int cmd;
184 int uid;
185 caddr_t arg;
186 };
187 #endif
188 int
sys_quotactl(struct thread * td,struct quotactl_args * uap)189 sys_quotactl(struct thread *td, struct quotactl_args *uap)
190 {
191 struct mount *mp;
192 struct nameidata nd;
193 int error;
194 bool mp_busy;
195
196 AUDIT_ARG_CMD(uap->cmd);
197 AUDIT_ARG_UID(uap->uid);
198 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
199 return (EPERM);
200 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
201 uap->path);
202 if ((error = namei(&nd)) != 0)
203 return (error);
204 NDFREE_PNBUF(&nd);
205 mp = nd.ni_vp->v_mount;
206 vfs_ref(mp);
207 vput(nd.ni_vp);
208 error = vfs_busy(mp, 0);
209 if (error != 0) {
210 vfs_rel(mp);
211 return (error);
212 }
213 mp_busy = true;
214 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
215
216 /*
217 * Since quota on/off operations typically need to open quota
218 * files, the implementation may need to unbusy the mount point
219 * before calling into namei. Otherwise, unmount might be
220 * started between two vfs_busy() invocations (first is ours,
221 * second is from mount point cross-walk code in lookup()),
222 * causing deadlock.
223 *
224 * Avoid unbusying mp if the implementation indicates it has
225 * already done so.
226 */
227 if (mp_busy)
228 vfs_unbusy(mp);
229 vfs_rel(mp);
230 return (error);
231 }
232
233 /*
234 * Used by statfs conversion routines to scale the block size up if
235 * necessary so that all of the block counts are <= 'max_size'. Note
236 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
237 * value of 'n'.
238 */
239 void
statfs_scale_blocks(struct statfs * sf,long max_size)240 statfs_scale_blocks(struct statfs *sf, long max_size)
241 {
242 uint64_t count;
243 int shift;
244
245 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
246
247 /*
248 * Attempt to scale the block counts to give a more accurate
249 * overview to userland of the ratio of free space to used
250 * space. To do this, find the largest block count and compute
251 * a divisor that lets it fit into a signed integer <= max_size.
252 */
253 if (sf->f_bavail < 0)
254 count = -sf->f_bavail;
255 else
256 count = sf->f_bavail;
257 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
258 if (count <= max_size)
259 return;
260
261 count >>= flsl(max_size);
262 shift = 0;
263 while (count > 0) {
264 shift++;
265 count >>=1;
266 }
267
268 sf->f_bsize <<= shift;
269 sf->f_blocks >>= shift;
270 sf->f_bfree >>= shift;
271 sf->f_bavail >>= shift;
272 }
273
274 static int
kern_do_statfs(struct thread * td,struct mount * mp,struct statfs * buf)275 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
276 {
277 int error;
278
279 if (mp == NULL)
280 return (EBADF);
281 error = vfs_busy(mp, MBF_PCATCH);
282 vfs_rel(mp);
283 if (error != 0)
284 return (error);
285 #ifdef MAC
286 error = mac_mount_check_stat(td->td_ucred, mp);
287 if (error != 0)
288 goto out;
289 #endif
290 error = VFS_STATFS(mp, buf);
291 if (error != 0)
292 goto out;
293 if (priv_check_cred_vfs_generation(td->td_ucred))
294 prison_enforce_statfs(td->td_ucred, mp, buf);
295 out:
296 vfs_unbusy(mp);
297 return (error);
298 }
299
300 /*
301 * Get filesystem statistics.
302 */
303 #ifndef _SYS_SYSPROTO_H_
304 struct statfs_args {
305 char *path;
306 struct statfs *buf;
307 };
308 #endif
309 int
sys_statfs(struct thread * td,struct statfs_args * uap)310 sys_statfs(struct thread *td, struct statfs_args *uap)
311 {
312 struct statfs *sfp;
313 int error;
314
315 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
316 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
317 if (error == 0)
318 error = copyout(sfp, uap->buf, sizeof(struct statfs));
319 free(sfp, M_STATFS);
320 return (error);
321 }
322
323 int
kern_statfs(struct thread * td,const char * path,enum uio_seg pathseg,struct statfs * buf)324 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
325 struct statfs *buf)
326 {
327 struct mount *mp;
328 struct nameidata nd;
329 int error;
330
331 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
332 error = namei(&nd);
333 if (error != 0)
334 return (error);
335 NDFREE_PNBUF(&nd);
336 mp = vfs_ref_from_vp(nd.ni_vp);
337 vrele(nd.ni_vp);
338 return (kern_do_statfs(td, mp, buf));
339 }
340
341 /*
342 * Get filesystem statistics.
343 */
344 #ifndef _SYS_SYSPROTO_H_
345 struct fstatfs_args {
346 int fd;
347 struct statfs *buf;
348 };
349 #endif
350 int
sys_fstatfs(struct thread * td,struct fstatfs_args * uap)351 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
352 {
353 struct statfs *sfp;
354 int error;
355
356 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
357 error = kern_fstatfs(td, uap->fd, sfp);
358 if (error == 0)
359 error = copyout(sfp, uap->buf, sizeof(struct statfs));
360 free(sfp, M_STATFS);
361 return (error);
362 }
363
364 int
kern_fstatfs(struct thread * td,int fd,struct statfs * buf)365 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
366 {
367 struct file *fp;
368 struct mount *mp;
369 struct vnode *vp;
370 int error;
371
372 AUDIT_ARG_FD(fd);
373 error = getvnode_path(td, fd, &cap_fstatfs_rights, NULL, &fp);
374 if (error != 0)
375 return (error);
376 vp = fp->f_vnode;
377 #ifdef AUDIT
378 if (AUDITING_TD(td)) {
379 vn_lock(vp, LK_SHARED | LK_RETRY);
380 AUDIT_ARG_VNODE1(vp);
381 VOP_UNLOCK(vp);
382 }
383 #endif
384 mp = vfs_ref_from_vp(vp);
385 fdrop(fp, td);
386 return (kern_do_statfs(td, mp, buf));
387 }
388
389 /*
390 * Get statistics on all filesystems.
391 */
392 #ifndef _SYS_SYSPROTO_H_
393 struct getfsstat_args {
394 struct statfs *buf;
395 long bufsize;
396 int mode;
397 };
398 #endif
399 int
sys_getfsstat(struct thread * td,struct getfsstat_args * uap)400 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
401 {
402 size_t count;
403 int error;
404
405 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
406 return (EINVAL);
407 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
408 UIO_USERSPACE, uap->mode);
409 if (error == 0)
410 td->td_retval[0] = count;
411 return (error);
412 }
413
414 /*
415 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
416 * The caller is responsible for freeing memory which will be allocated
417 * in '*buf'.
418 */
419 int
kern_getfsstat(struct thread * td,struct statfs ** buf,size_t bufsize,size_t * countp,enum uio_seg bufseg,int mode)420 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
421 size_t *countp, enum uio_seg bufseg, int mode)
422 {
423 struct mount *mp, *nmp;
424 struct statfs *sfsp, *sp, *sptmp, *tofree;
425 size_t count, maxcount;
426 int error;
427
428 switch (mode) {
429 case MNT_WAIT:
430 case MNT_NOWAIT:
431 break;
432 default:
433 if (bufseg == UIO_SYSSPACE)
434 *buf = NULL;
435 return (EINVAL);
436 }
437 restart:
438 maxcount = bufsize / sizeof(struct statfs);
439 if (bufsize == 0) {
440 sfsp = NULL;
441 tofree = NULL;
442 } else if (bufseg == UIO_USERSPACE) {
443 sfsp = *buf;
444 tofree = NULL;
445 } else /* if (bufseg == UIO_SYSSPACE) */ {
446 count = 0;
447 mtx_lock(&mountlist_mtx);
448 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
449 count++;
450 }
451 mtx_unlock(&mountlist_mtx);
452 if (maxcount > count)
453 maxcount = count;
454 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
455 M_STATFS, M_WAITOK);
456 }
457
458 count = 0;
459
460 /*
461 * If there is no target buffer they only want the count.
462 *
463 * This could be TAILQ_FOREACH but it is open-coded to match the original
464 * code below.
465 */
466 if (sfsp == NULL) {
467 mtx_lock(&mountlist_mtx);
468 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
469 if (prison_canseemount(td->td_ucred, mp) != 0) {
470 nmp = TAILQ_NEXT(mp, mnt_list);
471 continue;
472 }
473 #ifdef MAC
474 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
475 nmp = TAILQ_NEXT(mp, mnt_list);
476 continue;
477 }
478 #endif
479 count++;
480 nmp = TAILQ_NEXT(mp, mnt_list);
481 }
482 mtx_unlock(&mountlist_mtx);
483 *countp = count;
484 return (0);
485 }
486
487 /*
488 * They want the entire thing.
489 *
490 * Short-circuit the corner case of no room for anything, avoids
491 * relocking below.
492 */
493 if (maxcount < 1) {
494 goto out;
495 }
496
497 mtx_lock(&mountlist_mtx);
498 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
499 if (prison_canseemount(td->td_ucred, mp) != 0) {
500 nmp = TAILQ_NEXT(mp, mnt_list);
501 continue;
502 }
503 #ifdef MAC
504 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
505 nmp = TAILQ_NEXT(mp, mnt_list);
506 continue;
507 }
508 #endif
509 if (mode == MNT_WAIT) {
510 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
511 /*
512 * If vfs_busy() failed, and MBF_NOWAIT
513 * wasn't passed, then the mp is gone.
514 * Furthermore, because of MBF_MNTLSTLOCK,
515 * the mountlist_mtx was dropped. We have
516 * no other choice than to start over.
517 */
518 mtx_unlock(&mountlist_mtx);
519 free(tofree, M_STATFS);
520 goto restart;
521 }
522 } else {
523 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
524 nmp = TAILQ_NEXT(mp, mnt_list);
525 continue;
526 }
527 }
528 sp = &mp->mnt_stat;
529 /*
530 * If MNT_NOWAIT is specified, do not refresh
531 * the fsstat cache.
532 */
533 if (mode != MNT_NOWAIT) {
534 error = VFS_STATFS(mp, sp);
535 if (error != 0) {
536 mtx_lock(&mountlist_mtx);
537 nmp = TAILQ_NEXT(mp, mnt_list);
538 vfs_unbusy(mp);
539 continue;
540 }
541 }
542 if (priv_check_cred_vfs_generation(td->td_ucred)) {
543 sptmp = malloc(sizeof(struct statfs), M_STATFS,
544 M_WAITOK);
545 *sptmp = *sp;
546 prison_enforce_statfs(td->td_ucred, mp, sptmp);
547 sp = sptmp;
548 } else
549 sptmp = NULL;
550 if (bufseg == UIO_SYSSPACE) {
551 bcopy(sp, sfsp, sizeof(*sp));
552 free(sptmp, M_STATFS);
553 } else /* if (bufseg == UIO_USERSPACE) */ {
554 error = copyout(sp, sfsp, sizeof(*sp));
555 free(sptmp, M_STATFS);
556 if (error != 0) {
557 vfs_unbusy(mp);
558 return (error);
559 }
560 }
561 sfsp++;
562 count++;
563
564 if (count == maxcount) {
565 vfs_unbusy(mp);
566 goto out;
567 }
568
569 mtx_lock(&mountlist_mtx);
570 nmp = TAILQ_NEXT(mp, mnt_list);
571 vfs_unbusy(mp);
572 }
573 mtx_unlock(&mountlist_mtx);
574 out:
575 *countp = count;
576 return (0);
577 }
578
579 #ifdef COMPAT_FREEBSD4
580 /*
581 * Get old format filesystem statistics.
582 */
583 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
584
585 #ifndef _SYS_SYSPROTO_H_
586 struct freebsd4_statfs_args {
587 char *path;
588 struct ostatfs *buf;
589 };
590 #endif
591 int
freebsd4_statfs(struct thread * td,struct freebsd4_statfs_args * uap)592 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
593 {
594 struct ostatfs osb;
595 struct statfs *sfp;
596 int error;
597
598 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
599 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
600 if (error == 0) {
601 freebsd4_cvtstatfs(sfp, &osb);
602 error = copyout(&osb, uap->buf, sizeof(osb));
603 }
604 free(sfp, M_STATFS);
605 return (error);
606 }
607
608 /*
609 * Get filesystem statistics.
610 */
611 #ifndef _SYS_SYSPROTO_H_
612 struct freebsd4_fstatfs_args {
613 int fd;
614 struct ostatfs *buf;
615 };
616 #endif
617 int
freebsd4_fstatfs(struct thread * td,struct freebsd4_fstatfs_args * uap)618 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
619 {
620 struct ostatfs osb;
621 struct statfs *sfp;
622 int error;
623
624 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
625 error = kern_fstatfs(td, uap->fd, sfp);
626 if (error == 0) {
627 freebsd4_cvtstatfs(sfp, &osb);
628 error = copyout(&osb, uap->buf, sizeof(osb));
629 }
630 free(sfp, M_STATFS);
631 return (error);
632 }
633
634 /*
635 * Get statistics on all filesystems.
636 */
637 #ifndef _SYS_SYSPROTO_H_
638 struct freebsd4_getfsstat_args {
639 struct ostatfs *buf;
640 long bufsize;
641 int mode;
642 };
643 #endif
644 int
freebsd4_getfsstat(struct thread * td,struct freebsd4_getfsstat_args * uap)645 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
646 {
647 struct statfs *buf, *sp;
648 struct ostatfs osb;
649 size_t count, size;
650 int error;
651
652 if (uap->bufsize < 0)
653 return (EINVAL);
654 count = uap->bufsize / sizeof(struct ostatfs);
655 if (count > SIZE_MAX / sizeof(struct statfs))
656 return (EINVAL);
657 size = count * sizeof(struct statfs);
658 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
659 uap->mode);
660 if (error == 0)
661 td->td_retval[0] = count;
662 if (size != 0) {
663 sp = buf;
664 while (count != 0 && error == 0) {
665 freebsd4_cvtstatfs(sp, &osb);
666 error = copyout(&osb, uap->buf, sizeof(osb));
667 sp++;
668 uap->buf++;
669 count--;
670 }
671 free(buf, M_STATFS);
672 }
673 return (error);
674 }
675
676 /*
677 * Implement fstatfs() for (NFS) file handles.
678 */
679 #ifndef _SYS_SYSPROTO_H_
680 struct freebsd4_fhstatfs_args {
681 struct fhandle *u_fhp;
682 struct ostatfs *buf;
683 };
684 #endif
685 int
freebsd4_fhstatfs(struct thread * td,struct freebsd4_fhstatfs_args * uap)686 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
687 {
688 struct ostatfs osb;
689 struct statfs *sfp;
690 fhandle_t fh;
691 int error;
692
693 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
694 if (error != 0)
695 return (error);
696 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
697 error = kern_fhstatfs(td, fh, sfp);
698 if (error == 0) {
699 freebsd4_cvtstatfs(sfp, &osb);
700 error = copyout(&osb, uap->buf, sizeof(osb));
701 }
702 free(sfp, M_STATFS);
703 return (error);
704 }
705
706 /*
707 * Convert a new format statfs structure to an old format statfs structure.
708 */
709 static void
freebsd4_cvtstatfs(struct statfs * nsp,struct ostatfs * osp)710 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
711 {
712
713 statfs_scale_blocks(nsp, LONG_MAX);
714 bzero(osp, sizeof(*osp));
715 osp->f_bsize = nsp->f_bsize;
716 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
717 osp->f_blocks = nsp->f_blocks;
718 osp->f_bfree = nsp->f_bfree;
719 osp->f_bavail = nsp->f_bavail;
720 osp->f_files = MIN(nsp->f_files, LONG_MAX);
721 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
722 osp->f_owner = nsp->f_owner;
723 osp->f_type = nsp->f_type;
724 osp->f_flags = nsp->f_flags;
725 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
726 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
727 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
728 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
729 strlcpy(osp->f_fstypename, nsp->f_fstypename,
730 MIN(MFSNAMELEN, OMFSNAMELEN));
731 strlcpy(osp->f_mntonname, nsp->f_mntonname,
732 MIN(MNAMELEN, OMNAMELEN));
733 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
734 MIN(MNAMELEN, OMNAMELEN));
735 osp->f_fsid = nsp->f_fsid;
736 }
737 #endif /* COMPAT_FREEBSD4 */
738
739 #if defined(COMPAT_FREEBSD11)
740 /*
741 * Get old format filesystem statistics.
742 */
743 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
744
745 int
freebsd11_statfs(struct thread * td,struct freebsd11_statfs_args * uap)746 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
747 {
748 struct freebsd11_statfs osb;
749 struct statfs *sfp;
750 int error;
751
752 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
753 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
754 if (error == 0) {
755 freebsd11_cvtstatfs(sfp, &osb);
756 error = copyout(&osb, uap->buf, sizeof(osb));
757 }
758 free(sfp, M_STATFS);
759 return (error);
760 }
761
762 /*
763 * Get filesystem statistics.
764 */
765 int
freebsd11_fstatfs(struct thread * td,struct freebsd11_fstatfs_args * uap)766 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
767 {
768 struct freebsd11_statfs osb;
769 struct statfs *sfp;
770 int error;
771
772 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
773 error = kern_fstatfs(td, uap->fd, sfp);
774 if (error == 0) {
775 freebsd11_cvtstatfs(sfp, &osb);
776 error = copyout(&osb, uap->buf, sizeof(osb));
777 }
778 free(sfp, M_STATFS);
779 return (error);
780 }
781
782 /*
783 * Get statistics on all filesystems.
784 */
785 int
freebsd11_getfsstat(struct thread * td,struct freebsd11_getfsstat_args * uap)786 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
787 {
788 return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode));
789 }
790
791 int
kern_freebsd11_getfsstat(struct thread * td,struct freebsd11_statfs * ubuf,long bufsize,int mode)792 kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf,
793 long bufsize, int mode)
794 {
795 struct freebsd11_statfs osb;
796 struct statfs *buf, *sp;
797 size_t count, size;
798 int error;
799
800 if (bufsize < 0)
801 return (EINVAL);
802
803 count = bufsize / sizeof(struct ostatfs);
804 size = count * sizeof(struct statfs);
805 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode);
806 if (error == 0)
807 td->td_retval[0] = count;
808 if (size > 0) {
809 sp = buf;
810 while (count > 0 && error == 0) {
811 freebsd11_cvtstatfs(sp, &osb);
812 error = copyout(&osb, ubuf, sizeof(osb));
813 sp++;
814 ubuf++;
815 count--;
816 }
817 free(buf, M_STATFS);
818 }
819 return (error);
820 }
821
822 /*
823 * Implement fstatfs() for (NFS) file handles.
824 */
825 int
freebsd11_fhstatfs(struct thread * td,struct freebsd11_fhstatfs_args * uap)826 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
827 {
828 struct freebsd11_statfs osb;
829 struct statfs *sfp;
830 fhandle_t fh;
831 int error;
832
833 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
834 if (error)
835 return (error);
836 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
837 error = kern_fhstatfs(td, fh, sfp);
838 if (error == 0) {
839 freebsd11_cvtstatfs(sfp, &osb);
840 error = copyout(&osb, uap->buf, sizeof(osb));
841 }
842 free(sfp, M_STATFS);
843 return (error);
844 }
845
846 /*
847 * Convert a new format statfs structure to an old format statfs structure.
848 */
849 static void
freebsd11_cvtstatfs(struct statfs * nsp,struct freebsd11_statfs * osp)850 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
851 {
852
853 bzero(osp, sizeof(*osp));
854 osp->f_version = FREEBSD11_STATFS_VERSION;
855 osp->f_type = nsp->f_type;
856 osp->f_flags = nsp->f_flags;
857 osp->f_bsize = nsp->f_bsize;
858 osp->f_iosize = nsp->f_iosize;
859 osp->f_blocks = nsp->f_blocks;
860 osp->f_bfree = nsp->f_bfree;
861 osp->f_bavail = nsp->f_bavail;
862 osp->f_files = nsp->f_files;
863 osp->f_ffree = nsp->f_ffree;
864 osp->f_syncwrites = nsp->f_syncwrites;
865 osp->f_asyncwrites = nsp->f_asyncwrites;
866 osp->f_syncreads = nsp->f_syncreads;
867 osp->f_asyncreads = nsp->f_asyncreads;
868 osp->f_namemax = nsp->f_namemax;
869 osp->f_owner = nsp->f_owner;
870 osp->f_fsid = nsp->f_fsid;
871 strlcpy(osp->f_fstypename, nsp->f_fstypename,
872 MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
873 strlcpy(osp->f_mntonname, nsp->f_mntonname,
874 MIN(MNAMELEN, sizeof(osp->f_mntonname)));
875 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
876 MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
877 }
878 #endif /* COMPAT_FREEBSD11 */
879
880 /*
881 * Change current working directory to a given file descriptor.
882 */
883 #ifndef _SYS_SYSPROTO_H_
884 struct fchdir_args {
885 int fd;
886 };
887 #endif
888 int
sys_fchdir(struct thread * td,struct fchdir_args * uap)889 sys_fchdir(struct thread *td, struct fchdir_args *uap)
890 {
891 struct vnode *vp, *tdp;
892 struct mount *mp;
893 struct file *fp;
894 int error;
895 uint8_t fdflags;
896
897 AUDIT_ARG_FD(uap->fd);
898 error = getvnode_path(td, uap->fd, &cap_fchdir_rights, &fdflags,
899 &fp);
900 if (error != 0)
901 return (error);
902 if ((fdflags & UF_RESOLVE_BENEATH) != 0) {
903 fdrop(fp, td);
904 return (ENOTCAPABLE);
905 }
906 vp = fp->f_vnode;
907 vrefact(vp);
908 fdrop(fp, td);
909 vn_lock(vp, LK_SHARED | LK_RETRY);
910 AUDIT_ARG_VNODE1(vp);
911 error = change_dir(vp, td);
912 while (!error && (mp = vp->v_mountedhere) != NULL) {
913 if (vfs_busy(mp, 0))
914 continue;
915 error = VFS_ROOT(mp, LK_SHARED, &tdp);
916 vfs_unbusy(mp);
917 if (error != 0)
918 break;
919 vput(vp);
920 vp = tdp;
921 }
922 if (error != 0) {
923 vput(vp);
924 return (error);
925 }
926 VOP_UNLOCK(vp);
927 pwd_chdir(td, vp);
928 return (0);
929 }
930
931 /*
932 * Change current working directory (``.'').
933 */
934 #ifndef _SYS_SYSPROTO_H_
935 struct chdir_args {
936 char *path;
937 };
938 #endif
939 int
sys_chdir(struct thread * td,struct chdir_args * uap)940 sys_chdir(struct thread *td, struct chdir_args *uap)
941 {
942
943 return (kern_chdir(td, uap->path, UIO_USERSPACE));
944 }
945
946 int
kern_chdir(struct thread * td,const char * path,enum uio_seg pathseg)947 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
948 {
949 struct nameidata nd;
950 int error;
951
952 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
953 pathseg, path);
954 if ((error = namei(&nd)) != 0)
955 return (error);
956 if ((error = change_dir(nd.ni_vp, td)) != 0) {
957 vput(nd.ni_vp);
958 NDFREE_PNBUF(&nd);
959 return (error);
960 }
961 VOP_UNLOCK(nd.ni_vp);
962 NDFREE_PNBUF(&nd);
963 pwd_chdir(td, nd.ni_vp);
964 return (0);
965 }
966
967 static int unprivileged_chroot = 0;
968 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
969 &unprivileged_chroot, 0,
970 "Unprivileged processes can use chroot(2)");
971
972 /*
973 * Takes locked vnode, unlocks it before returning.
974 */
975 static int
kern_chroot(struct thread * td,struct vnode * vp)976 kern_chroot(struct thread *td, struct vnode *vp)
977 {
978 struct proc *p;
979 int error;
980
981 error = priv_check(td, PRIV_VFS_CHROOT);
982 if (error != 0) {
983 p = td->td_proc;
984 if (unprivileged_chroot == 0) {
985 error = EXTERROR(EPERM,
986 "security.bsd.unprivileged_chroot sysctl not enabled");
987 goto e_vunlock;
988 }
989 if ((p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
990 error = EXTERROR(EPERM,
991 "PROC_NO_NEW_PRIVS not enabled");
992 goto e_vunlock;
993 }
994 }
995
996 error = change_dir(vp, td);
997 if (error != 0)
998 goto e_vunlock;
999 #ifdef MAC
1000 error = mac_vnode_check_chroot(td->td_ucred, vp);
1001 if (error != 0)
1002 goto e_vunlock;
1003 #endif
1004 VOP_UNLOCK(vp);
1005 error = pwd_chroot(td, vp);
1006 vrele(vp);
1007 return (error);
1008 e_vunlock:
1009 vput(vp);
1010 return (error);
1011 }
1012
1013 /*
1014 * Change notion of root (``/'') directory.
1015 */
1016 #ifndef _SYS_SYSPROTO_H_
1017 struct chroot_args {
1018 char *path;
1019 };
1020 #endif
1021 int
sys_chroot(struct thread * td,struct chroot_args * uap)1022 sys_chroot(struct thread *td, struct chroot_args *uap)
1023 {
1024 struct nameidata nd;
1025 int error;
1026
1027 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
1028 UIO_USERSPACE, uap->path);
1029 error = namei(&nd);
1030 if (error != 0)
1031 return (error);
1032 NDFREE_PNBUF(&nd);
1033 error = kern_chroot(td, nd.ni_vp);
1034 return (error);
1035 }
1036
1037 /*
1038 * Change notion of root directory to a given file descriptor.
1039 */
1040 #ifndef _SYS_SYSPROTO_H_
1041 struct fchroot_args {
1042 int fd;
1043 };
1044 #endif
1045 int
sys_fchroot(struct thread * td,struct fchroot_args * uap)1046 sys_fchroot(struct thread *td, struct fchroot_args *uap)
1047 {
1048 struct vnode *vp;
1049 struct file *fp;
1050 int error;
1051 uint8_t fdflags;
1052
1053 error = getvnode_path(td, uap->fd, &cap_fchroot_rights, &fdflags, &fp);
1054 if (error != 0)
1055 return (error);
1056 if ((fdflags & UF_RESOLVE_BENEATH) != 0) {
1057 fdrop(fp, td);
1058 return (ENOTCAPABLE);
1059 }
1060 vp = fp->f_vnode;
1061 vrefact(vp);
1062 fdrop(fp, td);
1063 vn_lock(vp, LK_SHARED | LK_RETRY);
1064 error = kern_chroot(td, vp);
1065 return (error);
1066 }
1067
1068 /*
1069 * Common routine for chroot and chdir. Callers must provide a locked vnode
1070 * instance.
1071 */
1072 int
change_dir(struct vnode * vp,struct thread * td)1073 change_dir(struct vnode *vp, struct thread *td)
1074 {
1075 #ifdef MAC
1076 int error;
1077 #endif
1078
1079 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1080 if (vp->v_type != VDIR)
1081 return (ENOTDIR);
1082 #ifdef MAC
1083 error = mac_vnode_check_chdir(td->td_ucred, vp);
1084 if (error != 0)
1085 return (error);
1086 #endif
1087 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1088 }
1089
1090 static __inline void
flags_to_rights(int flags,cap_rights_t * rightsp)1091 flags_to_rights(int flags, cap_rights_t *rightsp)
1092 {
1093 if (flags & O_EXEC) {
1094 cap_rights_set_one(rightsp, CAP_FEXECVE);
1095 if (flags & O_PATH)
1096 return;
1097 } else {
1098 switch ((flags & O_ACCMODE)) {
1099 case O_RDONLY:
1100 cap_rights_set_one(rightsp, CAP_READ);
1101 break;
1102 case O_RDWR:
1103 cap_rights_set_one(rightsp, CAP_READ);
1104 /* FALLTHROUGH */
1105 case O_WRONLY:
1106 cap_rights_set_one(rightsp, CAP_WRITE);
1107 if (!(flags & (O_APPEND | O_TRUNC)))
1108 cap_rights_set_one(rightsp, CAP_SEEK);
1109 break;
1110 }
1111 }
1112
1113 if (flags & O_CREAT)
1114 cap_rights_set_one(rightsp, CAP_CREATE);
1115
1116 if (flags & O_TRUNC)
1117 cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1118
1119 if (flags & (O_SYNC | O_FSYNC | O_DSYNC))
1120 cap_rights_set_one(rightsp, CAP_FSYNC);
1121
1122 if (flags & (O_EXLOCK | O_SHLOCK))
1123 cap_rights_set_one(rightsp, CAP_FLOCK);
1124 }
1125
1126 /*
1127 * Check permissions, allocate an open file structure, and call the device
1128 * open routine if any.
1129 */
1130 #ifndef _SYS_SYSPROTO_H_
1131 struct open_args {
1132 char *path;
1133 int flags;
1134 int mode;
1135 };
1136 #endif
1137 int
sys_open(struct thread * td,struct open_args * uap)1138 sys_open(struct thread *td, struct open_args *uap)
1139 {
1140
1141 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1142 uap->flags, uap->mode));
1143 }
1144
1145 #ifndef _SYS_SYSPROTO_H_
1146 struct openat_args {
1147 int fd;
1148 char *path;
1149 int flag;
1150 int mode;
1151 };
1152 #endif
1153 int
sys_openat(struct thread * td,struct openat_args * uap)1154 sys_openat(struct thread *td, struct openat_args *uap)
1155 {
1156
1157 AUDIT_ARG_FD(uap->fd);
1158 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1159 uap->mode));
1160 }
1161
1162 /*
1163 * Validate open(2) flags and convert access mode flags (O_RDONLY etc.) to their
1164 * in-kernel representations (FREAD etc.).
1165 */
1166 static int
openflags(int * flagsp)1167 openflags(int *flagsp)
1168 {
1169 int flags;
1170
1171 flags = *flagsp;
1172 if ((flags & ~FUSERALLOWED) != 0)
1173 return (EINVAL);
1174
1175 /*
1176 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1177 * may be specified. On the other hand, for O_PATH any mode
1178 * except O_EXEC is ignored.
1179 */
1180 if ((flags & O_PATH) != 0) {
1181 flags &= ~O_ACCMODE;
1182 } else if ((flags & O_EXEC) != 0) {
1183 if ((flags & O_ACCMODE) != 0)
1184 return (EINVAL);
1185 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1186 return (EINVAL);
1187 } else {
1188 flags = FFLAGS(flags);
1189 }
1190 *flagsp = flags;
1191 return (0);
1192 }
1193
1194 static void
finit_open(struct file * fp,struct vnode * vp,int flags)1195 finit_open(struct file *fp, struct vnode *vp, int flags)
1196 {
1197 /*
1198 * Store the vnode, for any f_type. Typically, the vnode use count is
1199 * decremented by a direct call to vnops.fo_close() for files that
1200 * switched type.
1201 */
1202 fp->f_vnode = vp;
1203
1204 /*
1205 * If the file wasn't claimed by devfs or fifofs, bind it to the normal
1206 * vnode operations here.
1207 */
1208 if (fp->f_ops == &badfileops) {
1209 KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1210 ("Unexpected fifo fp %p vp %p", fp, vp));
1211 if ((flags & O_PATH) != 0) {
1212 finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1213 DTYPE_VNODE, NULL, &path_fileops);
1214 } else {
1215 finit_vnode(fp, flags, NULL, &vnops);
1216 }
1217 }
1218 }
1219
1220 /*
1221 * If fpp != NULL, opened file is not installed into the file
1222 * descriptor table, instead it is returned in *fpp. This is
1223 * incompatible with fdopen(), in which case we return EINVAL.
1224 */
1225 static int
openatfp(struct thread * td,int dirfd,const char * path,enum uio_seg pathseg,int flags,int mode,struct file ** fpp)1226 openatfp(struct thread *td, int dirfd, const char *path,
1227 enum uio_seg pathseg, int flags, int mode, struct file **fpp)
1228 {
1229 struct proc *p;
1230 struct filedesc *fdp;
1231 struct pwddesc *pdp;
1232 struct file *fp;
1233 struct vnode *vp;
1234 struct filecaps *fcaps;
1235 struct nameidata nd;
1236 cap_rights_t rights;
1237 int cmode, error, indx;
1238
1239 indx = -1;
1240 p = td->td_proc;
1241 fdp = p->p_fd;
1242 pdp = p->p_pd;
1243
1244 AUDIT_ARG_FFLAGS(flags);
1245 AUDIT_ARG_MODE(mode);
1246 cap_rights_init_one(&rights, CAP_LOOKUP);
1247 flags_to_rights(flags, &rights);
1248
1249 error = openflags(&flags);
1250 if (error != 0)
1251 return (error);
1252
1253 /*
1254 * Allocate a file structure. The descriptor to reference it
1255 * is allocated and used by finstall_refed() below.
1256 */
1257 error = falloc_noinstall(td, &fp);
1258 if (error != 0)
1259 return (error);
1260 /* Set the flags early so the finit in devfs can pick them up. */
1261 fp->f_flag = flags & FMASK;
1262 cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1263 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | WANTIOCTLCAPS,
1264 pathseg, path, dirfd, &rights);
1265 td->td_dupfd = -1; /* XXX check for fdopen */
1266 error = vn_open_cred(&nd, &flags, cmode, VN_OPEN_WANTIOCTLCAPS,
1267 td->td_ucred, fp);
1268 if (error != 0) {
1269 /*
1270 * If the vn_open replaced the method vector, something
1271 * wonderous happened deep below and we just pass it up
1272 * pretending we know what we do.
1273 */
1274 if (error == ENXIO && fp->f_ops != &badfileops) {
1275 MPASS((flags & O_PATH) == 0);
1276 goto success;
1277 }
1278
1279 /*
1280 * Handle special fdopen() case. bleh.
1281 *
1282 * Don't do this for relative (capability) lookups; we don't
1283 * understand exactly what would happen, and we don't think
1284 * that it ever should.
1285 */
1286 if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1287 (error == ENODEV || error == ENXIO) &&
1288 td->td_dupfd >= 0) {
1289 MPASS(fpp == NULL);
1290 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1291 &indx);
1292 if (error == 0)
1293 goto success;
1294 }
1295
1296 goto bad;
1297 }
1298 td->td_dupfd = 0;
1299 NDFREE_PNBUF(&nd);
1300 vp = nd.ni_vp;
1301
1302 finit_open(fp, vp, flags);
1303 VOP_UNLOCK(vp);
1304 if (flags & O_TRUNC) {
1305 error = fo_truncate(fp, 0, td->td_ucred, td);
1306 if (error != 0)
1307 goto bad;
1308 }
1309 success:
1310 if (fpp != NULL) {
1311 MPASS(error == 0);
1312 NDFREE_IOCTLCAPS(&nd);
1313 *fpp = fp;
1314 return (0);
1315 }
1316
1317 /*
1318 * If we haven't already installed the FD (for dupfdopen), do so now.
1319 */
1320 if (indx == -1) {
1321 #ifdef CAPABILITIES
1322 if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1323 fcaps = &nd.ni_filecaps;
1324 else
1325 #endif
1326 fcaps = NULL;
1327 if ((nd.ni_resflags & NIRES_BENEATH) != 0)
1328 flags |= O_RESOLVE_BENEATH;
1329 else
1330 flags &= ~O_RESOLVE_BENEATH;
1331 error = finstall_refed(td, fp, &indx, flags, fcaps);
1332 /* On success finstall_refed() consumes fcaps. */
1333 if (error != 0) {
1334 goto bad;
1335 }
1336 } else {
1337 NDFREE_IOCTLCAPS(&nd);
1338 falloc_abort(td, fp);
1339 }
1340
1341 td->td_retval[0] = indx;
1342 return (0);
1343 bad:
1344 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1345 NDFREE_IOCTLCAPS(&nd);
1346 falloc_abort(td, fp);
1347 return (error);
1348 }
1349
1350 int
kern_openat(struct thread * td,int dirfd,const char * path,enum uio_seg pathseg,int flags,int mode)1351 kern_openat(struct thread *td, int dirfd, const char *path,
1352 enum uio_seg pathseg, int flags, int mode)
1353 {
1354 return (openatfp(td, dirfd, path, pathseg, flags, mode, NULL));
1355 }
1356
1357 int
kern_openatfp(struct thread * td,int dirfd,const char * path,enum uio_seg pathseg,int flags,int mode,struct file ** fpp)1358 kern_openatfp(struct thread *td, int dirfd, const char *path,
1359 enum uio_seg pathseg, int flags, int mode, struct file **fpp)
1360 {
1361 int error, old_dupfd;
1362
1363 old_dupfd = td->td_dupfd;
1364 td->td_dupfd = -1;
1365 error = openatfp(td, dirfd, path, pathseg, flags, mode, fpp);
1366 td->td_dupfd = old_dupfd;
1367 return (error);
1368 }
1369
1370 #ifdef COMPAT_43
1371 /*
1372 * Create a file.
1373 */
1374 #ifndef _SYS_SYSPROTO_H_
1375 struct ocreat_args {
1376 char *path;
1377 int mode;
1378 };
1379 #endif
1380 int
ocreat(struct thread * td,struct ocreat_args * uap)1381 ocreat(struct thread *td, struct ocreat_args *uap)
1382 {
1383
1384 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1385 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1386 }
1387 #endif /* COMPAT_43 */
1388
1389 /*
1390 * Create a special file.
1391 */
1392 #ifndef _SYS_SYSPROTO_H_
1393 struct mknodat_args {
1394 int fd;
1395 char *path;
1396 mode_t mode;
1397 dev_t dev;
1398 };
1399 #endif
1400 int
sys_mknodat(struct thread * td,struct mknodat_args * uap)1401 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1402 {
1403
1404 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1405 uap->dev));
1406 }
1407
1408 #if defined(COMPAT_FREEBSD11)
1409 int
freebsd11_mknod(struct thread * td,struct freebsd11_mknod_args * uap)1410 freebsd11_mknod(struct thread *td,
1411 struct freebsd11_mknod_args *uap)
1412 {
1413
1414 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1415 uap->mode, uap->dev));
1416 }
1417
1418 int
freebsd11_mknodat(struct thread * td,struct freebsd11_mknodat_args * uap)1419 freebsd11_mknodat(struct thread *td,
1420 struct freebsd11_mknodat_args *uap)
1421 {
1422
1423 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1424 uap->dev));
1425 }
1426 #endif /* COMPAT_FREEBSD11 */
1427
1428 int
kern_mknodat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int mode,dev_t dev)1429 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1430 int mode, dev_t dev)
1431 {
1432 struct vnode *vp;
1433 struct mount *mp;
1434 struct vattr vattr;
1435 struct nameidata nd;
1436 int error, whiteout = 0;
1437
1438 AUDIT_ARG_MODE(mode);
1439 AUDIT_ARG_DEV(dev);
1440 switch (mode & S_IFMT) {
1441 case S_IFCHR:
1442 case S_IFBLK:
1443 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1444 if (error == 0 && dev == VNOVAL)
1445 error = EINVAL;
1446 break;
1447 case S_IFWHT:
1448 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1449 break;
1450 case S_IFIFO:
1451 if (dev == 0)
1452 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1453 /* FALLTHROUGH */
1454 default:
1455 error = EINVAL;
1456 break;
1457 }
1458 if (error != 0)
1459 return (error);
1460 NDPREINIT(&nd);
1461 restart:
1462 bwillwrite();
1463 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE,
1464 pathseg, path, fd, &cap_mknodat_rights);
1465 if ((error = namei(&nd)) != 0)
1466 return (error);
1467 vp = nd.ni_vp;
1468 if (vp != NULL) {
1469 NDFREE_PNBUF(&nd);
1470 if (vp == nd.ni_dvp)
1471 vrele(nd.ni_dvp);
1472 else
1473 vput(nd.ni_dvp);
1474 vrele(vp);
1475 return (EEXIST);
1476 } else if ((vn_irflag_read(nd.ni_dvp) & VIRF_NAMEDDIR) != 0) {
1477 NDFREE_PNBUF(&nd);
1478 vput(nd.ni_dvp);
1479 return (EINVAL);
1480 } else {
1481 VATTR_NULL(&vattr);
1482 vattr.va_mode = (mode & ALLPERMS) &
1483 ~td->td_proc->p_pd->pd_cmask;
1484 vattr.va_rdev = dev;
1485 whiteout = 0;
1486
1487 switch (mode & S_IFMT) {
1488 case S_IFCHR:
1489 vattr.va_type = VCHR;
1490 break;
1491 case S_IFBLK:
1492 vattr.va_type = VBLK;
1493 break;
1494 case S_IFWHT:
1495 whiteout = 1;
1496 break;
1497 default:
1498 panic("kern_mknod: invalid mode");
1499 }
1500 }
1501 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1502 NDFREE_PNBUF(&nd);
1503 vput(nd.ni_dvp);
1504 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1505 return (error);
1506 goto restart;
1507 }
1508 #ifdef MAC
1509 if (error == 0 && !whiteout)
1510 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1511 &nd.ni_cnd, &vattr);
1512 #endif
1513 if (error == 0) {
1514 if (whiteout)
1515 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1516 else {
1517 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1518 &nd.ni_cnd, &vattr);
1519 }
1520 }
1521 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1522 true);
1523 vn_finished_write(mp);
1524 NDFREE_PNBUF(&nd);
1525 if (error == ERELOOKUP)
1526 goto restart;
1527 return (error);
1528 }
1529
1530 /*
1531 * Create a named pipe.
1532 */
1533 #ifndef _SYS_SYSPROTO_H_
1534 struct mkfifo_args {
1535 char *path;
1536 int mode;
1537 };
1538 #endif
1539 int
sys_mkfifo(struct thread * td,struct mkfifo_args * uap)1540 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1541 {
1542
1543 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1544 uap->mode));
1545 }
1546
1547 #ifndef _SYS_SYSPROTO_H_
1548 struct mkfifoat_args {
1549 int fd;
1550 char *path;
1551 mode_t mode;
1552 };
1553 #endif
1554 int
sys_mkfifoat(struct thread * td,struct mkfifoat_args * uap)1555 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1556 {
1557
1558 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1559 uap->mode));
1560 }
1561
1562 int
kern_mkfifoat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int mode)1563 kern_mkfifoat(struct thread *td, int fd, const char *path,
1564 enum uio_seg pathseg, int mode)
1565 {
1566 struct mount *mp;
1567 struct vattr vattr;
1568 struct nameidata nd;
1569 int error;
1570
1571 AUDIT_ARG_MODE(mode);
1572 NDPREINIT(&nd);
1573 restart:
1574 bwillwrite();
1575 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE,
1576 pathseg, path, fd, &cap_mkfifoat_rights);
1577 if ((error = namei(&nd)) != 0)
1578 return (error);
1579 if (nd.ni_vp != NULL) {
1580 NDFREE_PNBUF(&nd);
1581 if (nd.ni_vp == nd.ni_dvp)
1582 vrele(nd.ni_dvp);
1583 else
1584 vput(nd.ni_dvp);
1585 vrele(nd.ni_vp);
1586 return (EEXIST);
1587 }
1588 if ((vn_irflag_read(nd.ni_dvp) & VIRF_NAMEDDIR) != 0) {
1589 NDFREE_PNBUF(&nd);
1590 vput(nd.ni_dvp);
1591 return (EINVAL);
1592 }
1593 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1594 NDFREE_PNBUF(&nd);
1595 vput(nd.ni_dvp);
1596 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1597 return (error);
1598 goto restart;
1599 }
1600 VATTR_NULL(&vattr);
1601 vattr.va_type = VFIFO;
1602 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1603 #ifdef MAC
1604 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1605 &vattr);
1606 if (error != 0)
1607 goto out;
1608 #endif
1609 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1610 #ifdef MAC
1611 out:
1612 #endif
1613 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1614 vn_finished_write(mp);
1615 NDFREE_PNBUF(&nd);
1616 if (error == ERELOOKUP)
1617 goto restart;
1618 return (error);
1619 }
1620
1621 /*
1622 * Make a hard file link.
1623 */
1624 #ifndef _SYS_SYSPROTO_H_
1625 struct link_args {
1626 char *path;
1627 char *link;
1628 };
1629 #endif
1630 int
sys_link(struct thread * td,struct link_args * uap)1631 sys_link(struct thread *td, struct link_args *uap)
1632 {
1633
1634 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1635 UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1636 }
1637
1638 #ifndef _SYS_SYSPROTO_H_
1639 struct linkat_args {
1640 int fd1;
1641 char *path1;
1642 int fd2;
1643 char *path2;
1644 int flag;
1645 };
1646 #endif
1647 int
sys_linkat(struct thread * td,struct linkat_args * uap)1648 sys_linkat(struct thread *td, struct linkat_args *uap)
1649 {
1650
1651 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1652 UIO_USERSPACE, uap->flag));
1653 }
1654
1655 int hardlink_check_uid = 0;
1656 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1657 &hardlink_check_uid, 0,
1658 "Unprivileged processes cannot create hard links to files owned by other "
1659 "users");
1660 static int hardlink_check_gid = 0;
1661 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1662 &hardlink_check_gid, 0,
1663 "Unprivileged processes cannot create hard links to files owned by other "
1664 "groups");
1665
1666 static int
can_hardlink(struct vnode * vp,struct ucred * cred)1667 can_hardlink(struct vnode *vp, struct ucred *cred)
1668 {
1669 struct vattr va;
1670 int error;
1671
1672 if (!hardlink_check_uid && !hardlink_check_gid)
1673 return (0);
1674
1675 error = VOP_GETATTR(vp, &va, cred);
1676 if (error != 0)
1677 return (error);
1678
1679 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1680 error = priv_check_cred(cred, PRIV_VFS_LINK);
1681 if (error != 0)
1682 return (error);
1683 }
1684
1685 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1686 error = priv_check_cred(cred, PRIV_VFS_LINK);
1687 if (error != 0)
1688 return (error);
1689 }
1690
1691 return (0);
1692 }
1693
1694 int
kern_linkat(struct thread * td,int fd1,int fd2,const char * path1,const char * path2,enum uio_seg segflag,int flag)1695 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1696 const char *path2, enum uio_seg segflag, int flag)
1697 {
1698 struct nameidata nd;
1699 int error;
1700
1701 if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1702 AT_EMPTY_PATH)) != 0)
1703 return (EINVAL);
1704
1705 NDPREINIT(&nd);
1706 do {
1707 bwillwrite();
1708 NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1709 AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1710 segflag, path1, fd1, &cap_linkat_source_rights);
1711 if ((error = namei(&nd)) != 0)
1712 return (error);
1713 NDFREE_PNBUF(&nd);
1714 if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1715 error = priv_check(td, PRIV_VFS_FHOPEN);
1716 if (error != 0) {
1717 vrele(nd.ni_vp);
1718 return (error);
1719 }
1720 }
1721 error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1722 } while (error == EAGAIN || error == ERELOOKUP);
1723 return (error);
1724 }
1725
1726 static int
kern_linkat_vp(struct thread * td,struct vnode * vp,int fd,const char * path,enum uio_seg segflag)1727 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1728 enum uio_seg segflag)
1729 {
1730 struct nameidata nd;
1731 struct mount *mp;
1732 int error;
1733
1734 if (vp->v_type == VDIR) {
1735 vrele(vp);
1736 return (EPERM); /* POSIX */
1737 }
1738 if ((vn_irflag_read(vp) & (VIRF_NAMEDDIR | VIRF_NAMEDATTR)) != 0) {
1739 vrele(vp);
1740 return (EINVAL);
1741 }
1742 NDINIT_ATRIGHTS(&nd, CREATE,
1743 LOCKPARENT | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1744 &cap_linkat_target_rights);
1745 if ((error = namei(&nd)) == 0) {
1746 if (nd.ni_vp != NULL) {
1747 NDFREE_PNBUF(&nd);
1748 if (nd.ni_dvp == nd.ni_vp)
1749 vrele(nd.ni_dvp);
1750 else
1751 vput(nd.ni_dvp);
1752 vrele(nd.ni_vp);
1753 vrele(vp);
1754 return (EEXIST);
1755 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1756 /*
1757 * Cross-device link. No need to recheck
1758 * vp->v_type, since it cannot change, except
1759 * to VBAD.
1760 */
1761 NDFREE_PNBUF(&nd);
1762 vput(nd.ni_dvp);
1763 vrele(vp);
1764 return (EXDEV);
1765 } else if (vn_lock(vp, LK_EXCLUSIVE) == 0) {
1766 error = can_hardlink(vp, td->td_ucred);
1767 #ifdef MAC
1768 if (error == 0)
1769 error = mac_vnode_check_link(td->td_ucred,
1770 nd.ni_dvp, vp, &nd.ni_cnd);
1771 #endif
1772 if (error != 0) {
1773 vput(vp);
1774 vput(nd.ni_dvp);
1775 NDFREE_PNBUF(&nd);
1776 return (error);
1777 }
1778 error = vn_start_write(vp, &mp, V_NOWAIT);
1779 if (error != 0) {
1780 vput(vp);
1781 vput(nd.ni_dvp);
1782 NDFREE_PNBUF(&nd);
1783 error = vn_start_write(NULL, &mp,
1784 V_XSLEEP | V_PCATCH);
1785 if (error != 0)
1786 return (error);
1787 return (EAGAIN);
1788 }
1789 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1790 VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1791 vn_finished_write(mp);
1792 NDFREE_PNBUF(&nd);
1793 vp = NULL;
1794 } else {
1795 vput(nd.ni_dvp);
1796 NDFREE_PNBUF(&nd);
1797 vrele(vp);
1798 return (EAGAIN);
1799 }
1800 }
1801 if (vp != NULL)
1802 vrele(vp);
1803 return (error);
1804 }
1805
1806 /*
1807 * Make a symbolic link.
1808 */
1809 #ifndef _SYS_SYSPROTO_H_
1810 struct symlink_args {
1811 char *path;
1812 char *link;
1813 };
1814 #endif
1815 int
sys_symlink(struct thread * td,struct symlink_args * uap)1816 sys_symlink(struct thread *td, struct symlink_args *uap)
1817 {
1818
1819 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1820 UIO_USERSPACE));
1821 }
1822
1823 #ifndef _SYS_SYSPROTO_H_
1824 struct symlinkat_args {
1825 char *path;
1826 int fd;
1827 char *path2;
1828 };
1829 #endif
1830 int
sys_symlinkat(struct thread * td,struct symlinkat_args * uap)1831 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1832 {
1833
1834 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1835 UIO_USERSPACE));
1836 }
1837
1838 int
kern_symlinkat(struct thread * td,const char * path1,int fd,const char * path2,enum uio_seg segflg)1839 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1840 enum uio_seg segflg)
1841 {
1842 struct mount *mp;
1843 struct vattr vattr;
1844 const char *syspath;
1845 char *tmppath;
1846 struct nameidata nd;
1847 int error;
1848
1849 if (segflg == UIO_SYSSPACE) {
1850 syspath = path1;
1851 } else {
1852 tmppath = uma_zalloc(namei_zone, M_WAITOK);
1853 if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1854 goto out;
1855 syspath = tmppath;
1856 }
1857 AUDIT_ARG_TEXT(syspath);
1858 NDPREINIT(&nd);
1859 restart:
1860 bwillwrite();
1861 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, segflg,
1862 path2, fd, &cap_symlinkat_rights);
1863 if ((error = namei(&nd)) != 0)
1864 goto out;
1865 if (nd.ni_vp) {
1866 NDFREE_PNBUF(&nd);
1867 if (nd.ni_vp == nd.ni_dvp)
1868 vrele(nd.ni_dvp);
1869 else
1870 vput(nd.ni_dvp);
1871 vrele(nd.ni_vp);
1872 nd.ni_vp = NULL;
1873 error = EEXIST;
1874 goto out;
1875 }
1876 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1877 NDFREE_PNBUF(&nd);
1878 vput(nd.ni_dvp);
1879 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1880 goto out;
1881 goto restart;
1882 }
1883 if ((vn_irflag_read(nd.ni_dvp) & VIRF_NAMEDDIR) != 0) {
1884 error = EINVAL;
1885 goto out2;
1886 }
1887 VATTR_NULL(&vattr);
1888 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1889 #ifdef MAC
1890 vattr.va_type = VLNK;
1891 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1892 &vattr);
1893 if (error != 0)
1894 goto out2;
1895 #endif
1896 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1897 out2:
1898 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1899 vn_finished_write(mp);
1900 NDFREE_PNBUF(&nd);
1901 if (error == ERELOOKUP)
1902 goto restart;
1903 out:
1904 if (segflg != UIO_SYSSPACE)
1905 uma_zfree(namei_zone, tmppath);
1906 return (error);
1907 }
1908
1909 /*
1910 * Delete a whiteout from the filesystem.
1911 */
1912 #ifndef _SYS_SYSPROTO_H_
1913 struct undelete_args {
1914 char *path;
1915 };
1916 #endif
1917 int
sys_undelete(struct thread * td,struct undelete_args * uap)1918 sys_undelete(struct thread *td, struct undelete_args *uap)
1919 {
1920 struct mount *mp;
1921 struct nameidata nd;
1922 int error;
1923
1924 NDPREINIT(&nd);
1925 restart:
1926 bwillwrite();
1927 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1928 UIO_USERSPACE, uap->path);
1929 error = namei(&nd);
1930 if (error != 0)
1931 return (error);
1932
1933 if (nd.ni_vp != NULL || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1934 NDFREE_PNBUF(&nd);
1935 if (nd.ni_vp == nd.ni_dvp)
1936 vrele(nd.ni_dvp);
1937 else
1938 vput(nd.ni_dvp);
1939 if (nd.ni_vp)
1940 vrele(nd.ni_vp);
1941 return (EEXIST);
1942 }
1943 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1944 NDFREE_PNBUF(&nd);
1945 vput(nd.ni_dvp);
1946 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1947 return (error);
1948 goto restart;
1949 }
1950 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1951 NDFREE_PNBUF(&nd);
1952 vput(nd.ni_dvp);
1953 vn_finished_write(mp);
1954 if (error == ERELOOKUP)
1955 goto restart;
1956 return (error);
1957 }
1958
1959 /*
1960 * Delete a name from the filesystem.
1961 */
1962 #ifndef _SYS_SYSPROTO_H_
1963 struct unlink_args {
1964 char *path;
1965 };
1966 #endif
1967 int
sys_unlink(struct thread * td,struct unlink_args * uap)1968 sys_unlink(struct thread *td, struct unlink_args *uap)
1969 {
1970
1971 return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1972 0, 0));
1973 }
1974
1975 static int
kern_funlinkat_ex(struct thread * td,int dfd,const char * path,int fd,int flags,enum uio_seg pathseg,ino_t oldinum)1976 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1977 int flags, enum uio_seg pathseg, ino_t oldinum)
1978 {
1979
1980 if ((flags & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1981 return (EINVAL);
1982
1983 if ((flags & AT_REMOVEDIR) != 0)
1984 return (kern_frmdirat(td, dfd, path, fd, pathseg,
1985 flags & ~AT_REMOVEDIR));
1986
1987 return (kern_funlinkat(td, dfd, path, fd, pathseg, flags, 0));
1988 }
1989
1990 #ifndef _SYS_SYSPROTO_H_
1991 struct unlinkat_args {
1992 int fd;
1993 char *path;
1994 int flag;
1995 };
1996 #endif
1997 int
sys_unlinkat(struct thread * td,struct unlinkat_args * uap)1998 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1999 {
2000
2001 return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
2002 UIO_USERSPACE, 0));
2003 }
2004
2005 #ifndef _SYS_SYSPROTO_H_
2006 struct funlinkat_args {
2007 int dfd;
2008 const char *path;
2009 int fd;
2010 int flag;
2011 };
2012 #endif
2013 int
sys_funlinkat(struct thread * td,struct funlinkat_args * uap)2014 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
2015 {
2016
2017 return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
2018 UIO_USERSPACE, 0));
2019 }
2020
2021 int
kern_funlinkat(struct thread * td,int dfd,const char * path,int fd,enum uio_seg pathseg,int flag,ino_t oldinum)2022 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
2023 enum uio_seg pathseg, int flag, ino_t oldinum)
2024 {
2025 struct mount *mp;
2026 struct file *fp;
2027 struct vnode *vp;
2028 struct nameidata nd;
2029 struct stat sb;
2030 int error;
2031
2032 fp = NULL;
2033 if (fd != FD_NONE) {
2034 error = getvnode_path(td, fd, &cap_no_rights, NULL, &fp);
2035 if (error != 0)
2036 return (error);
2037 }
2038
2039 NDPREINIT(&nd);
2040 restart:
2041 bwillwrite();
2042 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
2043 at2cnpflags(flag, AT_RESOLVE_BENEATH),
2044 pathseg, path, dfd, &cap_unlinkat_rights);
2045 if ((error = namei(&nd)) != 0) {
2046 if (error == EINVAL)
2047 error = EPERM;
2048 goto fdout;
2049 }
2050 vp = nd.ni_vp;
2051 if (vp->v_type == VDIR && oldinum == 0) {
2052 error = EPERM; /* POSIX */
2053 } else if (oldinum != 0 &&
2054 ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) &&
2055 sb.st_ino != oldinum) {
2056 error = EIDRM; /* Identifier removed */
2057 } else if (fp != NULL && fp->f_vnode != vp) {
2058 if (VN_IS_DOOMED(fp->f_vnode))
2059 error = EBADF;
2060 else
2061 error = EDEADLK;
2062 } else {
2063 /*
2064 * The root of a mounted filesystem cannot be deleted.
2065 *
2066 * XXX: can this only be a VDIR case?
2067 */
2068 if (vp->v_vflag & VV_ROOT)
2069 error = EBUSY;
2070 }
2071 if (error == 0) {
2072 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
2073 NDFREE_PNBUF(&nd);
2074 vput(nd.ni_dvp);
2075 if (vp == nd.ni_dvp)
2076 vrele(vp);
2077 else
2078 vput(vp);
2079 if ((error = vn_start_write(NULL, &mp,
2080 V_XSLEEP | V_PCATCH)) != 0) {
2081 goto fdout;
2082 }
2083 goto restart;
2084 }
2085 #ifdef MAC
2086 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
2087 &nd.ni_cnd);
2088 if (error != 0)
2089 goto out;
2090 #endif
2091 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
2092 #ifdef MAC
2093 out:
2094 #endif
2095 vn_finished_write(mp);
2096 }
2097 NDFREE_PNBUF(&nd);
2098 vput(nd.ni_dvp);
2099 if (vp == nd.ni_dvp)
2100 vrele(vp);
2101 else
2102 vput(vp);
2103 if (error == ERELOOKUP)
2104 goto restart;
2105 fdout:
2106 if (fp != NULL)
2107 fdrop(fp, td);
2108 return (error);
2109 }
2110
2111 /*
2112 * Reposition read/write file offset.
2113 */
2114 #ifndef _SYS_SYSPROTO_H_
2115 struct lseek_args {
2116 int fd;
2117 int pad;
2118 off_t offset;
2119 int whence;
2120 };
2121 #endif
2122 int
sys_lseek(struct thread * td,struct lseek_args * uap)2123 sys_lseek(struct thread *td, struct lseek_args *uap)
2124 {
2125
2126 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2127 }
2128
2129 int
kern_lseek(struct thread * td,int fd,off_t offset,int whence)2130 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2131 {
2132 struct file *fp;
2133 int error;
2134
2135 AUDIT_ARG_FD(fd);
2136 error = fget(td, fd, &cap_seek_rights, &fp);
2137 if (error != 0)
2138 return (error);
2139 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2140 fo_seek(fp, offset, whence, td) : ESPIPE;
2141 fdrop(fp, td);
2142 return (error);
2143 }
2144
2145 #if defined(COMPAT_43)
2146 /*
2147 * Reposition read/write file offset.
2148 */
2149 #ifndef _SYS_SYSPROTO_H_
2150 struct olseek_args {
2151 int fd;
2152 long offset;
2153 int whence;
2154 };
2155 #endif
2156 int
olseek(struct thread * td,struct olseek_args * uap)2157 olseek(struct thread *td, struct olseek_args *uap)
2158 {
2159
2160 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2161 }
2162 #endif /* COMPAT_43 */
2163
2164 #if defined(COMPAT_FREEBSD6)
2165 /* Version with the 'pad' argument */
2166 int
freebsd6_lseek(struct thread * td,struct freebsd6_lseek_args * uap)2167 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2168 {
2169
2170 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2171 }
2172 #endif
2173
2174 /*
2175 * Check access permissions using passed credentials.
2176 */
2177 static int
vn_access(struct vnode * vp,int user_flags,struct ucred * cred,struct thread * td)2178 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2179 struct thread *td)
2180 {
2181 accmode_t accmode;
2182 int error;
2183
2184 /* Flags == 0 means only check for existence. */
2185 if (user_flags == 0)
2186 return (0);
2187
2188 accmode = 0;
2189 if (user_flags & R_OK)
2190 accmode |= VREAD;
2191 if (user_flags & W_OK)
2192 accmode |= VWRITE;
2193 if (user_flags & X_OK)
2194 accmode |= VEXEC;
2195 #ifdef MAC
2196 error = mac_vnode_check_access(cred, vp, accmode);
2197 if (error != 0)
2198 return (error);
2199 #endif
2200 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2201 error = VOP_ACCESS(vp, accmode, cred, td);
2202 return (error);
2203 }
2204
2205 /*
2206 * Check access permissions using "real" credentials.
2207 */
2208 #ifndef _SYS_SYSPROTO_H_
2209 struct access_args {
2210 char *path;
2211 int amode;
2212 };
2213 #endif
2214 int
sys_access(struct thread * td,struct access_args * uap)2215 sys_access(struct thread *td, struct access_args *uap)
2216 {
2217
2218 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2219 0, uap->amode));
2220 }
2221
2222 #ifndef _SYS_SYSPROTO_H_
2223 struct faccessat_args {
2224 int dirfd;
2225 char *path;
2226 int amode;
2227 int flag;
2228 }
2229 #endif
2230 int
sys_faccessat(struct thread * td,struct faccessat_args * uap)2231 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2232 {
2233
2234 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2235 uap->amode));
2236 }
2237
2238 int
kern_accessat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int flag,int amode)2239 kern_accessat(struct thread *td, int fd, const char *path,
2240 enum uio_seg pathseg, int flag, int amode)
2241 {
2242 struct ucred *cred, *usecred;
2243 struct vnode *vp;
2244 struct nameidata nd;
2245 int error;
2246
2247 if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH |
2248 AT_SYMLINK_NOFOLLOW)) != 0)
2249 return (EINVAL);
2250 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2251 return (EINVAL);
2252
2253 /*
2254 * Create and modify a temporary credential instead of one that
2255 * is potentially shared (if we need one).
2256 */
2257 cred = td->td_ucred;
2258 if ((flag & AT_EACCESS) == 0 &&
2259 ((cred->cr_uid != cred->cr_ruid ||
2260 cred->cr_rgid != cred->cr_gid))) {
2261 usecred = crdup(cred);
2262 usecred->cr_uid = cred->cr_ruid;
2263 usecred->cr_gid = cred->cr_rgid;
2264 td->td_ucred = usecred;
2265 } else
2266 usecred = cred;
2267 AUDIT_ARG_VALUE(amode);
2268 NDINIT_ATRIGHTS(&nd, LOOKUP, LOCKSHARED | LOCKLEAF |
2269 AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW |
2270 AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights);
2271 if ((error = namei(&nd)) != 0)
2272 goto out;
2273 vp = nd.ni_vp;
2274
2275 error = vn_access(vp, amode, usecred, td);
2276 NDFREE_PNBUF(&nd);
2277 vput(vp);
2278 out:
2279 if (usecred != cred) {
2280 td->td_ucred = cred;
2281 crfree(usecred);
2282 }
2283 return (error);
2284 }
2285
2286 /*
2287 * Check access permissions using "effective" credentials.
2288 */
2289 #ifndef _SYS_SYSPROTO_H_
2290 struct eaccess_args {
2291 char *path;
2292 int amode;
2293 };
2294 #endif
2295 int
sys_eaccess(struct thread * td,struct eaccess_args * uap)2296 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2297 {
2298
2299 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2300 AT_EACCESS, uap->amode));
2301 }
2302
2303 #if defined(COMPAT_43)
2304 /*
2305 * Get file status; this version follows links.
2306 */
2307 #ifndef _SYS_SYSPROTO_H_
2308 struct ostat_args {
2309 char *path;
2310 struct ostat *ub;
2311 };
2312 #endif
2313 int
ostat(struct thread * td,struct ostat_args * uap)2314 ostat(struct thread *td, struct ostat_args *uap)
2315 {
2316 struct stat sb;
2317 struct ostat osb;
2318 int error;
2319
2320 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2321 if (error != 0)
2322 return (error);
2323 cvtstat(&sb, &osb);
2324 return (copyout(&osb, uap->ub, sizeof (osb)));
2325 }
2326
2327 /*
2328 * Get file status; this version does not follow links.
2329 */
2330 #ifndef _SYS_SYSPROTO_H_
2331 struct olstat_args {
2332 char *path;
2333 struct ostat *ub;
2334 };
2335 #endif
2336 int
olstat(struct thread * td,struct olstat_args * uap)2337 olstat(struct thread *td, struct olstat_args *uap)
2338 {
2339 struct stat sb;
2340 struct ostat osb;
2341 int error;
2342
2343 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2344 UIO_USERSPACE, &sb);
2345 if (error != 0)
2346 return (error);
2347 cvtstat(&sb, &osb);
2348 return (copyout(&osb, uap->ub, sizeof (osb)));
2349 }
2350
2351 /*
2352 * Convert from an old to a new stat structure.
2353 * XXX: many values are blindly truncated.
2354 */
2355 void
cvtstat(struct stat * st,struct ostat * ost)2356 cvtstat(struct stat *st, struct ostat *ost)
2357 {
2358
2359 bzero(ost, sizeof(*ost));
2360 ost->st_dev = st->st_dev;
2361 ost->st_ino = st->st_ino;
2362 ost->st_mode = st->st_mode;
2363 ost->st_nlink = st->st_nlink;
2364 ost->st_uid = st->st_uid;
2365 ost->st_gid = st->st_gid;
2366 ost->st_rdev = st->st_rdev;
2367 ost->st_size = MIN(st->st_size, INT32_MAX);
2368 ost->st_atim = st->st_atim;
2369 ost->st_mtim = st->st_mtim;
2370 ost->st_ctim = st->st_ctim;
2371 ost->st_blksize = st->st_blksize;
2372 ost->st_blocks = st->st_blocks;
2373 ost->st_flags = st->st_flags;
2374 ost->st_gen = st->st_gen;
2375 }
2376 #endif /* COMPAT_43 */
2377
2378 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2379 int ino64_trunc_error;
2380 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2381 &ino64_trunc_error, 0,
2382 "Error on truncation of device, file or inode number, or link count");
2383
2384 int
freebsd11_cvtstat(struct stat * st,struct freebsd11_stat * ost)2385 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2386 {
2387
2388 ost->st_dev = st->st_dev;
2389 if (ost->st_dev != st->st_dev) {
2390 switch (ino64_trunc_error) {
2391 default:
2392 /*
2393 * Since dev_t is almost raw, don't clamp to the
2394 * maximum for case 2, but ignore the error.
2395 */
2396 break;
2397 case 1:
2398 return (EOVERFLOW);
2399 }
2400 }
2401 ost->st_ino = st->st_ino;
2402 if (ost->st_ino != st->st_ino) {
2403 switch (ino64_trunc_error) {
2404 default:
2405 case 0:
2406 break;
2407 case 1:
2408 return (EOVERFLOW);
2409 case 2:
2410 ost->st_ino = UINT32_MAX;
2411 break;
2412 }
2413 }
2414 ost->st_mode = st->st_mode;
2415 ost->st_nlink = st->st_nlink;
2416 if (ost->st_nlink != st->st_nlink) {
2417 switch (ino64_trunc_error) {
2418 default:
2419 case 0:
2420 break;
2421 case 1:
2422 return (EOVERFLOW);
2423 case 2:
2424 ost->st_nlink = UINT16_MAX;
2425 break;
2426 }
2427 }
2428 ost->st_uid = st->st_uid;
2429 ost->st_gid = st->st_gid;
2430 ost->st_rdev = st->st_rdev;
2431 if (ost->st_rdev != st->st_rdev) {
2432 switch (ino64_trunc_error) {
2433 default:
2434 break;
2435 case 1:
2436 return (EOVERFLOW);
2437 }
2438 }
2439 ost->st_atim = st->st_atim;
2440 ost->st_mtim = st->st_mtim;
2441 ost->st_ctim = st->st_ctim;
2442 ost->st_size = st->st_size;
2443 ost->st_blocks = st->st_blocks;
2444 ost->st_blksize = st->st_blksize;
2445 ost->st_flags = st->st_flags;
2446 ost->st_gen = st->st_gen;
2447 ost->st_lspare = 0;
2448 ost->st_birthtim = st->st_birthtim;
2449 bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2450 sizeof(*ost) - offsetof(struct freebsd11_stat,
2451 st_birthtim) - sizeof(ost->st_birthtim));
2452 return (0);
2453 }
2454
2455 int
freebsd11_stat(struct thread * td,struct freebsd11_stat_args * uap)2456 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2457 {
2458 struct stat sb;
2459 struct freebsd11_stat osb;
2460 int error;
2461
2462 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2463 if (error != 0)
2464 return (error);
2465 error = freebsd11_cvtstat(&sb, &osb);
2466 if (error == 0)
2467 error = copyout(&osb, uap->ub, sizeof(osb));
2468 return (error);
2469 }
2470
2471 int
freebsd11_lstat(struct thread * td,struct freebsd11_lstat_args * uap)2472 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2473 {
2474 struct stat sb;
2475 struct freebsd11_stat osb;
2476 int error;
2477
2478 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2479 UIO_USERSPACE, &sb);
2480 if (error != 0)
2481 return (error);
2482 error = freebsd11_cvtstat(&sb, &osb);
2483 if (error == 0)
2484 error = copyout(&osb, uap->ub, sizeof(osb));
2485 return (error);
2486 }
2487
2488 int
freebsd11_fhstat(struct thread * td,struct freebsd11_fhstat_args * uap)2489 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2490 {
2491 struct fhandle fh;
2492 struct stat sb;
2493 struct freebsd11_stat osb;
2494 int error;
2495
2496 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2497 if (error != 0)
2498 return (error);
2499 error = kern_fhstat(td, fh, &sb);
2500 if (error != 0)
2501 return (error);
2502 error = freebsd11_cvtstat(&sb, &osb);
2503 if (error == 0)
2504 error = copyout(&osb, uap->sb, sizeof(osb));
2505 return (error);
2506 }
2507
2508 int
freebsd11_fstatat(struct thread * td,struct freebsd11_fstatat_args * uap)2509 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2510 {
2511 struct stat sb;
2512 struct freebsd11_stat osb;
2513 int error;
2514
2515 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2516 UIO_USERSPACE, &sb);
2517 if (error != 0)
2518 return (error);
2519 error = freebsd11_cvtstat(&sb, &osb);
2520 if (error == 0)
2521 error = copyout(&osb, uap->buf, sizeof(osb));
2522 return (error);
2523 }
2524 #endif /* COMPAT_FREEBSD11 */
2525
2526 /*
2527 * Get file status
2528 */
2529 #ifndef _SYS_SYSPROTO_H_
2530 struct fstatat_args {
2531 int fd;
2532 char *path;
2533 struct stat *buf;
2534 int flag;
2535 }
2536 #endif
2537 int
sys_fstatat(struct thread * td,struct fstatat_args * uap)2538 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2539 {
2540 struct stat sb;
2541 int error;
2542
2543 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2544 UIO_USERSPACE, &sb);
2545 if (error == 0)
2546 error = copyout(&sb, uap->buf, sizeof (sb));
2547 return (error);
2548 }
2549
2550 int
kern_statat(struct thread * td,int flag,int fd,const char * path,enum uio_seg pathseg,struct stat * sbp)2551 kern_statat(struct thread *td, int flag, int fd, const char *path,
2552 enum uio_seg pathseg, struct stat *sbp)
2553 {
2554 struct nameidata nd;
2555 int error;
2556
2557 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2558 AT_EMPTY_PATH)) != 0)
2559 return (EINVAL);
2560
2561 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2562 AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2563 AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights);
2564
2565 if ((error = namei(&nd)) != 0) {
2566 if (error == ENOTDIR &&
2567 (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2568 error = kern_fstat(td, fd, sbp);
2569 return (error);
2570 }
2571 error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED);
2572 NDFREE_PNBUF(&nd);
2573 vput(nd.ni_vp);
2574 #ifdef __STAT_TIME_T_EXT
2575 sbp->st_atim_ext = 0;
2576 sbp->st_mtim_ext = 0;
2577 sbp->st_ctim_ext = 0;
2578 sbp->st_btim_ext = 0;
2579 #endif
2580 #ifdef KTRACE
2581 if (KTRPOINT(td, KTR_STRUCT))
2582 ktrstat_error(sbp, error);
2583 #endif
2584 return (error);
2585 }
2586
2587 #if defined(COMPAT_FREEBSD11)
2588 /*
2589 * Implementation of the NetBSD [l]stat() functions.
2590 */
2591 int
freebsd11_cvtnstat(struct stat * sb,struct nstat * nsb)2592 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2593 {
2594 struct freebsd11_stat sb11;
2595 int error;
2596
2597 error = freebsd11_cvtstat(sb, &sb11);
2598 if (error != 0)
2599 return (error);
2600
2601 bzero(nsb, sizeof(*nsb));
2602 CP(sb11, *nsb, st_dev);
2603 CP(sb11, *nsb, st_ino);
2604 CP(sb11, *nsb, st_mode);
2605 CP(sb11, *nsb, st_nlink);
2606 CP(sb11, *nsb, st_uid);
2607 CP(sb11, *nsb, st_gid);
2608 CP(sb11, *nsb, st_rdev);
2609 CP(sb11, *nsb, st_atim);
2610 CP(sb11, *nsb, st_mtim);
2611 CP(sb11, *nsb, st_ctim);
2612 CP(sb11, *nsb, st_size);
2613 CP(sb11, *nsb, st_blocks);
2614 CP(sb11, *nsb, st_blksize);
2615 CP(sb11, *nsb, st_flags);
2616 CP(sb11, *nsb, st_gen);
2617 CP(sb11, *nsb, st_birthtim);
2618 return (0);
2619 }
2620
2621 #ifndef _SYS_SYSPROTO_H_
2622 struct freebsd11_nstat_args {
2623 char *path;
2624 struct nstat *ub;
2625 };
2626 #endif
2627 int
freebsd11_nstat(struct thread * td,struct freebsd11_nstat_args * uap)2628 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2629 {
2630 struct stat sb;
2631 struct nstat nsb;
2632 int error;
2633
2634 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2635 if (error != 0)
2636 return (error);
2637 error = freebsd11_cvtnstat(&sb, &nsb);
2638 if (error == 0)
2639 error = copyout(&nsb, uap->ub, sizeof (nsb));
2640 return (error);
2641 }
2642
2643 /*
2644 * NetBSD lstat. Get file status; this version does not follow links.
2645 */
2646 #ifndef _SYS_SYSPROTO_H_
2647 struct freebsd11_nlstat_args {
2648 char *path;
2649 struct nstat *ub;
2650 };
2651 #endif
2652 int
freebsd11_nlstat(struct thread * td,struct freebsd11_nlstat_args * uap)2653 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2654 {
2655 struct stat sb;
2656 struct nstat nsb;
2657 int error;
2658
2659 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2660 UIO_USERSPACE, &sb);
2661 if (error != 0)
2662 return (error);
2663 error = freebsd11_cvtnstat(&sb, &nsb);
2664 if (error == 0)
2665 error = copyout(&nsb, uap->ub, sizeof (nsb));
2666 return (error);
2667 }
2668 #endif /* COMPAT_FREEBSD11 */
2669
2670 /*
2671 * Get configurable pathname variables.
2672 */
2673 #ifndef _SYS_SYSPROTO_H_
2674 struct pathconf_args {
2675 char *path;
2676 int name;
2677 };
2678 #endif
2679 int
sys_pathconf(struct thread * td,struct pathconf_args * uap)2680 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2681 {
2682 long value;
2683 int error;
2684
2685 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2686 &value);
2687 if (error == 0)
2688 td->td_retval[0] = value;
2689 return (error);
2690 }
2691
2692 #ifndef _SYS_SYSPROTO_H_
2693 struct lpathconf_args {
2694 char *path;
2695 int name;
2696 };
2697 #endif
2698 int
sys_lpathconf(struct thread * td,struct lpathconf_args * uap)2699 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2700 {
2701 long value;
2702 int error;
2703
2704 error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2705 NOFOLLOW, &value);
2706 if (error == 0)
2707 td->td_retval[0] = value;
2708 return (error);
2709 }
2710
2711 int
kern_pathconf(struct thread * td,const char * path,enum uio_seg pathseg,int name,u_long flags,long * valuep)2712 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2713 int name, u_long flags, long *valuep)
2714 {
2715 struct nameidata nd;
2716 int error;
2717
2718 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2719 pathseg, path);
2720 if ((error = namei(&nd)) != 0)
2721 return (error);
2722 NDFREE_PNBUF(&nd);
2723
2724 error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2725 vput(nd.ni_vp);
2726 return (error);
2727 }
2728
2729 /*
2730 * Return target name of a symbolic link.
2731 */
2732 #ifndef _SYS_SYSPROTO_H_
2733 struct readlink_args {
2734 char *path;
2735 char *buf;
2736 size_t count;
2737 };
2738 #endif
2739 int
sys_readlink(struct thread * td,struct readlink_args * uap)2740 sys_readlink(struct thread *td, struct readlink_args *uap)
2741 {
2742
2743 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2744 uap->buf, UIO_USERSPACE, uap->count));
2745 }
2746 #ifndef _SYS_SYSPROTO_H_
2747 struct readlinkat_args {
2748 int fd;
2749 char *path;
2750 char *buf;
2751 size_t bufsize;
2752 };
2753 #endif
2754 int
sys_readlinkat(struct thread * td,struct readlinkat_args * uap)2755 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2756 {
2757
2758 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2759 uap->buf, UIO_USERSPACE, uap->bufsize));
2760 }
2761
2762 int
kern_readlinkat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,char * buf,enum uio_seg bufseg,size_t count)2763 kern_readlinkat(struct thread *td, int fd, const char *path,
2764 enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2765 {
2766 struct vnode *vp;
2767 struct nameidata nd;
2768 int error;
2769
2770 if (count > IOSIZE_MAX)
2771 return (EINVAL);
2772
2773 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2774 EMPTYPATH, pathseg, path, fd);
2775
2776 if ((error = namei(&nd)) != 0)
2777 return (error);
2778 NDFREE_PNBUF(&nd);
2779 vp = nd.ni_vp;
2780
2781 error = kern_readlink_vp(vp, buf, bufseg, count, td);
2782 vput(vp);
2783
2784 return (error);
2785 }
2786
2787 /*
2788 * Helper function to readlink from a vnode
2789 */
2790 static int
kern_readlink_vp(struct vnode * vp,char * buf,enum uio_seg bufseg,size_t count,struct thread * td)2791 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2792 struct thread *td)
2793 {
2794 struct iovec aiov;
2795 struct uio auio;
2796 int error;
2797
2798 ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2799 #ifdef MAC
2800 error = mac_vnode_check_readlink(td->td_ucred, vp);
2801 if (error != 0)
2802 return (error);
2803 #endif
2804 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2805 return (EINVAL);
2806
2807 aiov.iov_base = buf;
2808 aiov.iov_len = count;
2809 auio.uio_iov = &aiov;
2810 auio.uio_iovcnt = 1;
2811 auio.uio_offset = 0;
2812 auio.uio_rw = UIO_READ;
2813 auio.uio_segflg = bufseg;
2814 auio.uio_td = td;
2815 auio.uio_resid = count;
2816 error = VOP_READLINK(vp, &auio, td->td_ucred);
2817 td->td_retval[0] = count - auio.uio_resid;
2818 return (error);
2819 }
2820
2821 /*
2822 * Common implementation code for chflags() and fchflags().
2823 */
2824 static int
setfflags(struct thread * td,struct vnode * vp,u_long flags)2825 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2826 {
2827 struct mount *mp;
2828 struct vattr vattr;
2829 int error;
2830
2831 /* We can't support the value matching VNOVAL. */
2832 if (flags == VNOVAL)
2833 return (EOPNOTSUPP);
2834
2835 /*
2836 * Prevent non-root users from setting flags on devices. When
2837 * a device is reused, users can retain ownership of the device
2838 * if they are allowed to set flags and programs assume that
2839 * chown can't fail when done as root.
2840 */
2841 if (VN_ISDEV(vp)) {
2842 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2843 if (error != 0)
2844 return (error);
2845 }
2846
2847 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2848 return (error);
2849 VATTR_NULL(&vattr);
2850 vattr.va_flags = flags;
2851 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2852 #ifdef MAC
2853 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2854 if (error == 0)
2855 #endif
2856 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2857 VOP_UNLOCK(vp);
2858 vn_finished_write(mp);
2859 return (error);
2860 }
2861
2862 /*
2863 * Change flags of a file given a path name.
2864 */
2865 #ifndef _SYS_SYSPROTO_H_
2866 struct chflags_args {
2867 const char *path;
2868 u_long flags;
2869 };
2870 #endif
2871 int
sys_chflags(struct thread * td,struct chflags_args * uap)2872 sys_chflags(struct thread *td, struct chflags_args *uap)
2873 {
2874
2875 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2876 uap->flags, 0));
2877 }
2878
2879 #ifndef _SYS_SYSPROTO_H_
2880 struct chflagsat_args {
2881 int fd;
2882 const char *path;
2883 u_long flags;
2884 int atflag;
2885 }
2886 #endif
2887 int
sys_chflagsat(struct thread * td,struct chflagsat_args * uap)2888 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2889 {
2890
2891 return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2892 uap->flags, uap->atflag));
2893 }
2894
2895 /*
2896 * Same as chflags() but doesn't follow symlinks.
2897 */
2898 #ifndef _SYS_SYSPROTO_H_
2899 struct lchflags_args {
2900 const char *path;
2901 u_long flags;
2902 };
2903 #endif
2904 int
sys_lchflags(struct thread * td,struct lchflags_args * uap)2905 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2906 {
2907
2908 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2909 uap->flags, AT_SYMLINK_NOFOLLOW));
2910 }
2911
2912 static int
kern_chflagsat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,u_long flags,int atflag)2913 kern_chflagsat(struct thread *td, int fd, const char *path,
2914 enum uio_seg pathseg, u_long flags, int atflag)
2915 {
2916 struct nameidata nd;
2917 int error;
2918
2919 if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2920 AT_EMPTY_PATH)) != 0)
2921 return (EINVAL);
2922
2923 AUDIT_ARG_FFLAGS(flags);
2924 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2925 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2926 fd, &cap_fchflags_rights);
2927 if ((error = namei(&nd)) != 0)
2928 return (error);
2929 NDFREE_PNBUF(&nd);
2930 error = setfflags(td, nd.ni_vp, flags);
2931 vrele(nd.ni_vp);
2932 return (error);
2933 }
2934
2935 /*
2936 * Change flags of a file given a file descriptor.
2937 */
2938 #ifndef _SYS_SYSPROTO_H_
2939 struct fchflags_args {
2940 int fd;
2941 u_long flags;
2942 };
2943 #endif
2944 int
sys_fchflags(struct thread * td,struct fchflags_args * uap)2945 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2946 {
2947 struct file *fp;
2948 int error;
2949
2950 AUDIT_ARG_FD(uap->fd);
2951 AUDIT_ARG_FFLAGS(uap->flags);
2952 error = getvnode(td, uap->fd, &cap_fchflags_rights,
2953 &fp);
2954 if (error != 0)
2955 return (error);
2956 #ifdef AUDIT
2957 if (AUDITING_TD(td)) {
2958 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2959 AUDIT_ARG_VNODE1(fp->f_vnode);
2960 VOP_UNLOCK(fp->f_vnode);
2961 }
2962 #endif
2963 error = setfflags(td, fp->f_vnode, uap->flags);
2964 fdrop(fp, td);
2965 return (error);
2966 }
2967
2968 /*
2969 * Common implementation code for chmod(), lchmod() and fchmod().
2970 */
2971 int
setfmode(struct thread * td,struct ucred * cred,struct vnode * vp,int mode)2972 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2973 {
2974 struct mount *mp;
2975 struct vattr vattr;
2976 int error;
2977
2978 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2979 return (error);
2980 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2981 VATTR_NULL(&vattr);
2982 vattr.va_mode = mode & ALLPERMS;
2983 #ifdef MAC
2984 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2985 if (error == 0)
2986 #endif
2987 error = VOP_SETATTR(vp, &vattr, cred);
2988 VOP_UNLOCK(vp);
2989 vn_finished_write(mp);
2990 return (error);
2991 }
2992
2993 /*
2994 * Change mode of a file given path name.
2995 */
2996 #ifndef _SYS_SYSPROTO_H_
2997 struct chmod_args {
2998 char *path;
2999 int mode;
3000 };
3001 #endif
3002 int
sys_chmod(struct thread * td,struct chmod_args * uap)3003 sys_chmod(struct thread *td, struct chmod_args *uap)
3004 {
3005
3006 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3007 uap->mode, 0));
3008 }
3009
3010 #ifndef _SYS_SYSPROTO_H_
3011 struct fchmodat_args {
3012 int dirfd;
3013 char *path;
3014 mode_t mode;
3015 int flag;
3016 }
3017 #endif
3018 int
sys_fchmodat(struct thread * td,struct fchmodat_args * uap)3019 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
3020 {
3021
3022 return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
3023 uap->mode, uap->flag));
3024 }
3025
3026 /*
3027 * Change mode of a file given path name (don't follow links.)
3028 */
3029 #ifndef _SYS_SYSPROTO_H_
3030 struct lchmod_args {
3031 char *path;
3032 int mode;
3033 };
3034 #endif
3035 int
sys_lchmod(struct thread * td,struct lchmod_args * uap)3036 sys_lchmod(struct thread *td, struct lchmod_args *uap)
3037 {
3038
3039 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3040 uap->mode, AT_SYMLINK_NOFOLLOW));
3041 }
3042
3043 int
kern_fchmodat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,mode_t mode,int flag)3044 kern_fchmodat(struct thread *td, int fd, const char *path,
3045 enum uio_seg pathseg, mode_t mode, int flag)
3046 {
3047 struct nameidata nd;
3048 int error;
3049
3050 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3051 AT_EMPTY_PATH)) != 0)
3052 return (EINVAL);
3053
3054 AUDIT_ARG_MODE(mode);
3055 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3056 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3057 fd, &cap_fchmod_rights);
3058 if ((error = namei(&nd)) != 0)
3059 return (error);
3060 NDFREE_PNBUF(&nd);
3061 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
3062 vrele(nd.ni_vp);
3063 return (error);
3064 }
3065
3066 /*
3067 * Change mode of a file given a file descriptor.
3068 */
3069 #ifndef _SYS_SYSPROTO_H_
3070 struct fchmod_args {
3071 int fd;
3072 int mode;
3073 };
3074 #endif
3075 int
sys_fchmod(struct thread * td,struct fchmod_args * uap)3076 sys_fchmod(struct thread *td, struct fchmod_args *uap)
3077 {
3078 struct file *fp;
3079 int error;
3080
3081 AUDIT_ARG_FD(uap->fd);
3082 AUDIT_ARG_MODE(uap->mode);
3083
3084 error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
3085 if (error != 0)
3086 return (error);
3087 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
3088 fdrop(fp, td);
3089 return (error);
3090 }
3091
3092 /*
3093 * Common implementation for chown(), lchown(), and fchown()
3094 */
3095 int
setfown(struct thread * td,struct ucred * cred,struct vnode * vp,uid_t uid,gid_t gid)3096 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
3097 gid_t gid)
3098 {
3099 struct mount *mp;
3100 struct vattr vattr;
3101 int error;
3102
3103 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3104 return (error);
3105 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3106 VATTR_NULL(&vattr);
3107 vattr.va_uid = uid;
3108 vattr.va_gid = gid;
3109 #ifdef MAC
3110 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3111 vattr.va_gid);
3112 if (error == 0)
3113 #endif
3114 error = VOP_SETATTR(vp, &vattr, cred);
3115 VOP_UNLOCK(vp);
3116 vn_finished_write(mp);
3117 return (error);
3118 }
3119
3120 /*
3121 * Set ownership given a path name.
3122 */
3123 #ifndef _SYS_SYSPROTO_H_
3124 struct chown_args {
3125 char *path;
3126 int uid;
3127 int gid;
3128 };
3129 #endif
3130 int
sys_chown(struct thread * td,struct chown_args * uap)3131 sys_chown(struct thread *td, struct chown_args *uap)
3132 {
3133
3134 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3135 uap->gid, 0));
3136 }
3137
3138 #ifndef _SYS_SYSPROTO_H_
3139 struct fchownat_args {
3140 int fd;
3141 const char * path;
3142 uid_t uid;
3143 gid_t gid;
3144 int flag;
3145 };
3146 #endif
3147 int
sys_fchownat(struct thread * td,struct fchownat_args * uap)3148 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3149 {
3150
3151 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3152 uap->gid, uap->flag));
3153 }
3154
3155 int
kern_fchownat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int uid,int gid,int flag)3156 kern_fchownat(struct thread *td, int fd, const char *path,
3157 enum uio_seg pathseg, int uid, int gid, int flag)
3158 {
3159 struct nameidata nd;
3160 int error;
3161
3162 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3163 AT_EMPTY_PATH)) != 0)
3164 return (EINVAL);
3165
3166 AUDIT_ARG_OWNER(uid, gid);
3167 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3168 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3169 fd, &cap_fchown_rights);
3170
3171 if ((error = namei(&nd)) != 0)
3172 return (error);
3173 NDFREE_PNBUF(&nd);
3174 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3175 vrele(nd.ni_vp);
3176 return (error);
3177 }
3178
3179 /*
3180 * Set ownership given a path name, do not cross symlinks.
3181 */
3182 #ifndef _SYS_SYSPROTO_H_
3183 struct lchown_args {
3184 char *path;
3185 int uid;
3186 int gid;
3187 };
3188 #endif
3189 int
sys_lchown(struct thread * td,struct lchown_args * uap)3190 sys_lchown(struct thread *td, struct lchown_args *uap)
3191 {
3192
3193 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3194 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3195 }
3196
3197 /*
3198 * Set ownership given a file descriptor.
3199 */
3200 #ifndef _SYS_SYSPROTO_H_
3201 struct fchown_args {
3202 int fd;
3203 int uid;
3204 int gid;
3205 };
3206 #endif
3207 int
sys_fchown(struct thread * td,struct fchown_args * uap)3208 sys_fchown(struct thread *td, struct fchown_args *uap)
3209 {
3210 struct file *fp;
3211 int error;
3212
3213 AUDIT_ARG_FD(uap->fd);
3214 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3215 error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3216 if (error != 0)
3217 return (error);
3218 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3219 fdrop(fp, td);
3220 return (error);
3221 }
3222
3223 /*
3224 * Common implementation code for utimes(), lutimes(), and futimes().
3225 */
3226 static int
getutimes(const struct timeval * usrtvp,enum uio_seg tvpseg,struct timespec * tsp)3227 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3228 struct timespec *tsp)
3229 {
3230 struct timeval tv[2];
3231 const struct timeval *tvp;
3232 int error;
3233
3234 if (usrtvp == NULL) {
3235 vfs_timestamp(&tsp[0]);
3236 tsp[1] = tsp[0];
3237 } else {
3238 if (tvpseg == UIO_SYSSPACE) {
3239 tvp = usrtvp;
3240 } else {
3241 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3242 return (error);
3243 tvp = tv;
3244 }
3245
3246 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3247 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3248 return (EINVAL);
3249 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3250 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3251 }
3252 return (0);
3253 }
3254
3255 /*
3256 * Common implementation code for futimens(), utimensat().
3257 */
3258 #define UTIMENS_NULL 0x1
3259 #define UTIMENS_EXIT 0x2
3260 static int
getutimens(const struct timespec * usrtsp,enum uio_seg tspseg,struct timespec * tsp,int * retflags)3261 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3262 struct timespec *tsp, int *retflags)
3263 {
3264 struct timespec tsnow;
3265 int error;
3266
3267 vfs_timestamp(&tsnow);
3268 *retflags = 0;
3269 if (usrtsp == NULL) {
3270 tsp[0] = tsnow;
3271 tsp[1] = tsnow;
3272 *retflags |= UTIMENS_NULL;
3273 return (0);
3274 }
3275 if (tspseg == UIO_SYSSPACE) {
3276 tsp[0] = usrtsp[0];
3277 tsp[1] = usrtsp[1];
3278 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3279 return (error);
3280 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3281 *retflags |= UTIMENS_EXIT;
3282 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3283 *retflags |= UTIMENS_NULL;
3284 if (tsp[0].tv_nsec == UTIME_OMIT)
3285 tsp[0].tv_sec = VNOVAL;
3286 else if (tsp[0].tv_nsec == UTIME_NOW)
3287 tsp[0] = tsnow;
3288 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3289 return (EINVAL);
3290 if (tsp[1].tv_nsec == UTIME_OMIT)
3291 tsp[1].tv_sec = VNOVAL;
3292 else if (tsp[1].tv_nsec == UTIME_NOW)
3293 tsp[1] = tsnow;
3294 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3295 return (EINVAL);
3296
3297 return (0);
3298 }
3299
3300 /*
3301 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3302 * and utimensat().
3303 */
3304 static int
setutimes(struct thread * td,struct vnode * vp,const struct timespec * ts,int numtimes,int nullflag)3305 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3306 int numtimes, int nullflag)
3307 {
3308 struct mount *mp;
3309 struct vattr vattr;
3310 int error;
3311 bool setbirthtime;
3312
3313 setbirthtime = false;
3314 vattr.va_birthtime.tv_sec = VNOVAL;
3315 vattr.va_birthtime.tv_nsec = 0;
3316
3317 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3318 return (error);
3319 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3320 if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3321 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3322 setbirthtime = true;
3323 VATTR_NULL(&vattr);
3324 vattr.va_atime = ts[0];
3325 vattr.va_mtime = ts[1];
3326 if (setbirthtime)
3327 vattr.va_birthtime = ts[1];
3328 if (numtimes > 2)
3329 vattr.va_birthtime = ts[2];
3330 if (nullflag)
3331 vattr.va_vaflags |= VA_UTIMES_NULL;
3332 #ifdef MAC
3333 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3334 vattr.va_mtime);
3335 #endif
3336 if (error == 0)
3337 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3338 VOP_UNLOCK(vp);
3339 vn_finished_write(mp);
3340 return (error);
3341 }
3342
3343 /*
3344 * Set the access and modification times of a file.
3345 */
3346 #ifndef _SYS_SYSPROTO_H_
3347 struct utimes_args {
3348 char *path;
3349 struct timeval *tptr;
3350 };
3351 #endif
3352 int
sys_utimes(struct thread * td,struct utimes_args * uap)3353 sys_utimes(struct thread *td, struct utimes_args *uap)
3354 {
3355
3356 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3357 uap->tptr, UIO_USERSPACE));
3358 }
3359
3360 #ifndef _SYS_SYSPROTO_H_
3361 struct futimesat_args {
3362 int fd;
3363 const char * path;
3364 const struct timeval * times;
3365 };
3366 #endif
3367 int
sys_futimesat(struct thread * td,struct futimesat_args * uap)3368 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3369 {
3370
3371 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3372 uap->times, UIO_USERSPACE));
3373 }
3374
3375 int
kern_utimesat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,const struct timeval * tptr,enum uio_seg tptrseg)3376 kern_utimesat(struct thread *td, int fd, const char *path,
3377 enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg)
3378 {
3379 struct nameidata nd;
3380 struct timespec ts[2];
3381 int error;
3382
3383 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3384 return (error);
3385 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3386 &cap_futimes_rights);
3387
3388 if ((error = namei(&nd)) != 0)
3389 return (error);
3390 NDFREE_PNBUF(&nd);
3391 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3392 vrele(nd.ni_vp);
3393 return (error);
3394 }
3395
3396 /*
3397 * Set the access and modification times of a file.
3398 */
3399 #ifndef _SYS_SYSPROTO_H_
3400 struct lutimes_args {
3401 char *path;
3402 struct timeval *tptr;
3403 };
3404 #endif
3405 int
sys_lutimes(struct thread * td,struct lutimes_args * uap)3406 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3407 {
3408
3409 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3410 UIO_USERSPACE));
3411 }
3412
3413 int
kern_lutimes(struct thread * td,const char * path,enum uio_seg pathseg,const struct timeval * tptr,enum uio_seg tptrseg)3414 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3415 const struct timeval *tptr, enum uio_seg tptrseg)
3416 {
3417 struct timespec ts[2];
3418 struct nameidata nd;
3419 int error;
3420
3421 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3422 return (error);
3423 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path);
3424 if ((error = namei(&nd)) != 0)
3425 return (error);
3426 NDFREE_PNBUF(&nd);
3427 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3428 vrele(nd.ni_vp);
3429 return (error);
3430 }
3431
3432 /*
3433 * Set the access and modification times of a file.
3434 */
3435 #ifndef _SYS_SYSPROTO_H_
3436 struct futimes_args {
3437 int fd;
3438 struct timeval *tptr;
3439 };
3440 #endif
3441 int
sys_futimes(struct thread * td,struct futimes_args * uap)3442 sys_futimes(struct thread *td, struct futimes_args *uap)
3443 {
3444
3445 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3446 }
3447
3448 int
kern_futimes(struct thread * td,int fd,const struct timeval * tptr,enum uio_seg tptrseg)3449 kern_futimes(struct thread *td, int fd, const struct timeval *tptr,
3450 enum uio_seg tptrseg)
3451 {
3452 struct timespec ts[2];
3453 struct file *fp;
3454 int error;
3455
3456 AUDIT_ARG_FD(fd);
3457 error = getutimes(tptr, tptrseg, ts);
3458 if (error != 0)
3459 return (error);
3460 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3461 if (error != 0)
3462 return (error);
3463 #ifdef AUDIT
3464 if (AUDITING_TD(td)) {
3465 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3466 AUDIT_ARG_VNODE1(fp->f_vnode);
3467 VOP_UNLOCK(fp->f_vnode);
3468 }
3469 #endif
3470 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3471 fdrop(fp, td);
3472 return (error);
3473 }
3474
3475 int
sys_futimens(struct thread * td,struct futimens_args * uap)3476 sys_futimens(struct thread *td, struct futimens_args *uap)
3477 {
3478
3479 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3480 }
3481
3482 int
kern_futimens(struct thread * td,int fd,const struct timespec * tptr,enum uio_seg tptrseg)3483 kern_futimens(struct thread *td, int fd, const struct timespec *tptr,
3484 enum uio_seg tptrseg)
3485 {
3486 struct timespec ts[2];
3487 struct file *fp;
3488 int error, flags;
3489
3490 AUDIT_ARG_FD(fd);
3491 error = getutimens(tptr, tptrseg, ts, &flags);
3492 if (error != 0)
3493 return (error);
3494 if (flags & UTIMENS_EXIT)
3495 return (0);
3496 error = getvnode(td, fd, &cap_futimes_rights, &fp);
3497 if (error != 0)
3498 return (error);
3499 #ifdef AUDIT
3500 if (AUDITING_TD(td)) {
3501 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3502 AUDIT_ARG_VNODE1(fp->f_vnode);
3503 VOP_UNLOCK(fp->f_vnode);
3504 }
3505 #endif
3506 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3507 fdrop(fp, td);
3508 return (error);
3509 }
3510
3511 int
sys_utimensat(struct thread * td,struct utimensat_args * uap)3512 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3513 {
3514
3515 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3516 uap->times, UIO_USERSPACE, uap->flag));
3517 }
3518
3519 int
kern_utimensat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,const struct timespec * tptr,enum uio_seg tptrseg,int flag)3520 kern_utimensat(struct thread *td, int fd, const char *path,
3521 enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg,
3522 int flag)
3523 {
3524 struct nameidata nd;
3525 struct timespec ts[2];
3526 int error, flags;
3527
3528 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3529 AT_EMPTY_PATH)) != 0)
3530 return (EINVAL);
3531
3532 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3533 return (error);
3534 NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3535 AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3536 pathseg, path, fd, &cap_futimes_rights);
3537 if ((error = namei(&nd)) != 0)
3538 return (error);
3539 /*
3540 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3541 * POSIX states:
3542 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3543 * "Search permission is denied by a component of the path prefix."
3544 */
3545 NDFREE_PNBUF(&nd);
3546 if ((flags & UTIMENS_EXIT) == 0)
3547 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3548 vrele(nd.ni_vp);
3549 return (error);
3550 }
3551
3552 /*
3553 * Truncate a file given its path name.
3554 */
3555 #ifndef _SYS_SYSPROTO_H_
3556 struct truncate_args {
3557 char *path;
3558 int pad;
3559 off_t length;
3560 };
3561 #endif
3562 int
sys_truncate(struct thread * td,struct truncate_args * uap)3563 sys_truncate(struct thread *td, struct truncate_args *uap)
3564 {
3565
3566 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3567 }
3568
3569 int
kern_truncate(struct thread * td,const char * path,enum uio_seg pathseg,off_t length)3570 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3571 off_t length)
3572 {
3573 struct mount *mp;
3574 struct vnode *vp;
3575 void *rl_cookie;
3576 struct nameidata nd;
3577 int error;
3578
3579 if (length < 0)
3580 return (EINVAL);
3581 NDPREINIT(&nd);
3582 retry:
3583 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
3584 if ((error = namei(&nd)) != 0)
3585 return (error);
3586 vp = nd.ni_vp;
3587 NDFREE_PNBUF(&nd);
3588 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3589 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3590 vn_rangelock_unlock(vp, rl_cookie);
3591 vrele(vp);
3592 return (error);
3593 }
3594 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3595 if (vp->v_type == VDIR) {
3596 error = EISDIR;
3597 goto out;
3598 }
3599 #ifdef MAC
3600 error = mac_vnode_check_write(td->td_ucred, NOCRED, vp);
3601 if (error != 0)
3602 goto out;
3603 #endif
3604 error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td);
3605 if (error != 0)
3606 goto out;
3607
3608 error = vn_truncate_locked(vp, length, false, td->td_ucred);
3609 out:
3610 VOP_UNLOCK(vp);
3611 vn_finished_write(mp);
3612 vn_rangelock_unlock(vp, rl_cookie);
3613 vrele(vp);
3614 if (error == ERELOOKUP)
3615 goto retry;
3616 return (error);
3617 }
3618
3619 #if defined(COMPAT_43)
3620 /*
3621 * Truncate a file given its path name.
3622 */
3623 #ifndef _SYS_SYSPROTO_H_
3624 struct otruncate_args {
3625 char *path;
3626 long length;
3627 };
3628 #endif
3629 int
otruncate(struct thread * td,struct otruncate_args * uap)3630 otruncate(struct thread *td, struct otruncate_args *uap)
3631 {
3632
3633 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3634 }
3635 #endif /* COMPAT_43 */
3636
3637 #if defined(COMPAT_FREEBSD6)
3638 /* Versions with the pad argument */
3639 int
freebsd6_truncate(struct thread * td,struct freebsd6_truncate_args * uap)3640 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3641 {
3642
3643 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3644 }
3645
3646 int
freebsd6_ftruncate(struct thread * td,struct freebsd6_ftruncate_args * uap)3647 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3648 {
3649
3650 return (kern_ftruncate(td, uap->fd, uap->length));
3651 }
3652 #endif
3653
3654 int
kern_fsync(struct thread * td,int fd,bool fullsync)3655 kern_fsync(struct thread *td, int fd, bool fullsync)
3656 {
3657 struct vnode *vp;
3658 struct mount *mp;
3659 struct file *fp;
3660 int error;
3661
3662 AUDIT_ARG_FD(fd);
3663 error = getvnode(td, fd, &cap_fsync_rights, &fp);
3664 if (error != 0)
3665 return (error);
3666 vp = fp->f_vnode;
3667 #if 0
3668 if (!fullsync)
3669 /* XXXKIB: compete outstanding aio writes */;
3670 #endif
3671 retry:
3672 error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3673 if (error != 0)
3674 goto drop;
3675 vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3676 AUDIT_ARG_VNODE1(vp);
3677 vnode_pager_clean_async(vp);
3678 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3679 VOP_UNLOCK(vp);
3680 vn_finished_write(mp);
3681 if (error == ERELOOKUP)
3682 goto retry;
3683 drop:
3684 fdrop(fp, td);
3685 return (error);
3686 }
3687
3688 /*
3689 * Sync an open file.
3690 */
3691 #ifndef _SYS_SYSPROTO_H_
3692 struct fsync_args {
3693 int fd;
3694 };
3695 #endif
3696 int
sys_fsync(struct thread * td,struct fsync_args * uap)3697 sys_fsync(struct thread *td, struct fsync_args *uap)
3698 {
3699
3700 return (kern_fsync(td, uap->fd, true));
3701 }
3702
3703 int
sys_fdatasync(struct thread * td,struct fdatasync_args * uap)3704 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3705 {
3706
3707 return (kern_fsync(td, uap->fd, false));
3708 }
3709
3710 /*
3711 * Rename files. Source and destination must either both be directories, or
3712 * both not be directories. If target is a directory, it must be empty.
3713 */
3714 #ifndef _SYS_SYSPROTO_H_
3715 struct rename_args {
3716 char *from;
3717 char *to;
3718 };
3719 #endif
3720 int
sys_rename(struct thread * td,struct rename_args * uap)3721 sys_rename(struct thread *td, struct rename_args *uap)
3722 {
3723
3724 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3725 uap->to, UIO_USERSPACE, 0));
3726 }
3727
3728 #ifndef _SYS_SYSPROTO_H_
3729 struct renameat_args {
3730 int oldfd;
3731 char *old;
3732 int newfd;
3733 char *new;
3734 };
3735 #endif
3736 int
sys_renameat(struct thread * td,struct renameat_args * uap)3737 sys_renameat(struct thread *td, struct renameat_args *uap)
3738 {
3739
3740 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3741 UIO_USERSPACE, 0));
3742 }
3743
3744 int
sys_renameat2(struct thread * td,struct renameat2_args * uap)3745 sys_renameat2(struct thread *td, struct renameat2_args *uap)
3746 {
3747
3748 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3749 UIO_USERSPACE, uap->flags));
3750 }
3751
3752 #ifdef MAC
3753 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,int op,int ndflags)3754 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3755 const char *new, enum uio_seg pathseg, struct nameidata *fromnd, int op,
3756 int ndflags)
3757 {
3758 int error;
3759
3760 NDINIT_ATRIGHTS(fromnd, op, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3761 ndflags, pathseg, old, oldfd, &cap_renameat_source_rights);
3762 if ((error = namei(fromnd)) != 0)
3763 return (error);
3764 error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3765 fromnd->ni_vp, &fromnd->ni_cnd);
3766 VOP_UNLOCK(fromnd->ni_dvp);
3767 if (fromnd->ni_dvp != fromnd->ni_vp)
3768 VOP_UNLOCK(fromnd->ni_vp);
3769 if (error != 0) {
3770 NDFREE_PNBUF(fromnd);
3771 vrele(fromnd->ni_dvp);
3772 vrele(fromnd->ni_vp);
3773 }
3774 return (error);
3775 }
3776 #endif
3777
3778 int
kern_renameat(struct thread * td,int oldfd,const char * old,int newfd,const char * new,enum uio_seg pathseg,u_int flags)3779 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3780 const char *new, enum uio_seg pathseg, u_int flags)
3781 {
3782 struct mount *mp, *tmp;
3783 struct vnode *tvp, *fvp, *tdvp;
3784 struct nameidata fromnd, tond;
3785 uint64_t tondflags;
3786 int error, fndflags, op;
3787 short irflag;
3788 bool exchange;
3789
3790 if ((flags & ~(AT_RENAME_NOREPLACE | AT_RENAME_EXCHANGE)) != 0)
3791 return (EINVAL);
3792 if ((flags & (AT_RENAME_NOREPLACE | AT_RENAME_EXCHANGE)) ==
3793 (AT_RENAME_NOREPLACE | AT_RENAME_EXCHANGE))
3794 return (EINVAL);
3795 if ((flags & AT_RENAME_EXCHANGE) != 0) {
3796 op = RENAME;
3797 exchange = true;
3798 } else {
3799 op = DELETE;
3800 exchange = false;
3801 }
3802 fndflags = 0;
3803 again:
3804 tmp = mp = NULL;
3805 bwillwrite();
3806 #ifdef MAC
3807 if (mac_vnode_check_rename_from_enabled()) {
3808 error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3809 &fromnd, op, fndflags);
3810 if (error != 0)
3811 return (error);
3812 } else {
3813 #endif
3814 NDINIT_ATRIGHTS(&fromnd, op, WANTPARENT | AUDITVNODE1 | fndflags,
3815 pathseg, old, oldfd, &cap_renameat_source_rights);
3816 if ((error = namei(&fromnd)) != 0)
3817 return (error);
3818 #ifdef MAC
3819 }
3820 #endif
3821 fvp = fromnd.ni_vp;
3822 if (fromnd.ni_dvp->v_type != VDIR) {
3823 NDFREE_PNBUF(&fromnd);
3824 vrele(fromnd.ni_dvp);
3825 if (fvp != NULL)
3826 vrele(fromnd.ni_vp);
3827 return (EBUSY);
3828 }
3829 if (exchange && fvp == NULL) {
3830 NDFREE_PNBUF(&fromnd);
3831 vrele(fromnd.ni_dvp);
3832 return (ENOENT);
3833 }
3834 tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | AUDITVNODE2;
3835 if (fromnd.ni_vp->v_type == VDIR)
3836 tondflags |= WILLBEDIR;
3837 NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3838 &cap_renameat_target_rights);
3839 if ((error = namei(&tond)) != 0) {
3840 /* Translate error code for rename("dir1", "dir2/."). */
3841 if (error == EISDIR && fvp->v_type == VDIR)
3842 error = EINVAL;
3843 NDFREE_PNBUF(&fromnd);
3844 vrele(fromnd.ni_dvp);
3845 vrele(fvp);
3846 goto out1;
3847 }
3848 tdvp = tond.ni_dvp;
3849 tvp = tond.ni_vp;
3850 if (tdvp->v_type != VDIR) {
3851 /*
3852 * Rename of the root vnode of the mounted
3853 * filesystem. It is possible to get there with the
3854 * nullfs mount over the regular file.
3855 */
3856 error = EBUSY;
3857 goto out;
3858 }
3859 if (tvp != NULL && (flags & AT_RENAME_NOREPLACE) != 0) {
3860 /*
3861 * Often filesystems need to relock the vnodes in
3862 * VOP_RENAME(), which opens a window for invalidation
3863 * of this check. Then, not all filesystems might
3864 * implement AT_RENAME_NOREPLACE. This leads to
3865 * situation where sometimes EOPNOTSUPP might be
3866 * returned from the VOP due to race, while most of
3867 * the time this check works.
3868 */
3869 error = EEXIST;
3870 goto out;
3871 }
3872 if (exchange) {
3873 if (tvp == NULL) {
3874 error = ENOENT;
3875 goto out;
3876 }
3877 if (tvp->v_type == VDIR && fndflags == 0) {
3878 fndflags = WILLBEDIR;
3879 again2:
3880 NDFREE_PNBUF(&fromnd);
3881 vrele(fromnd.ni_dvp);
3882 vrele(fvp);
3883 NDFREE_PNBUF(&tond);
3884 VOP_VPUT_PAIR(tdvp, &tvp, true);
3885 error = sig_intr();
3886 if (error != 0)
3887 goto out1;
3888 goto again;
3889 }
3890 if (tvp->v_type != VDIR && fndflags != 0) {
3891 fndflags = 0;
3892 goto again2;
3893 }
3894 }
3895 error = vn_start_write(fvp, &mp, V_NOWAIT);
3896 if (error != 0) {
3897 again1:
3898 NDFREE_PNBUF(&fromnd);
3899 NDFREE_PNBUF(&tond);
3900 if (tvp != NULL)
3901 vput(tvp);
3902 if (tdvp == tvp)
3903 vrele(tdvp);
3904 else
3905 vput(tdvp);
3906 vrele(fromnd.ni_dvp);
3907 vrele(fvp);
3908 if (tmp != NULL) {
3909 lockmgr(&tmp->mnt_renamelock, LK_EXCLUSIVE, NULL);
3910 lockmgr(&tmp->mnt_renamelock, LK_RELEASE, NULL);
3911 vfs_rel(tmp);
3912 tmp = NULL;
3913 }
3914 error = sig_intr();
3915 if (error != 0)
3916 return (error);
3917 error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH);
3918 if (error != 0)
3919 return (error);
3920 goto again;
3921 }
3922 error = VOP_GETWRITEMOUNT(tdvp, &tmp);
3923 if (error != 0 || tmp == NULL)
3924 goto again1;
3925 error = lockmgr(&tmp->mnt_renamelock, LK_EXCLUSIVE | LK_NOWAIT, NULL);
3926 if (error != 0) {
3927 vn_finished_write(mp);
3928 goto again1;
3929 }
3930 irflag = vn_irflag_read(fvp);
3931 if (((irflag & VIRF_NAMEDATTR) != 0 && tdvp != fromnd.ni_dvp) ||
3932 (irflag & VIRF_NAMEDDIR) != 0) {
3933 error = EINVAL;
3934 goto out;
3935 }
3936 if (tvp != NULL) {
3937 if (!exchange) {
3938 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3939 error = ENOTDIR;
3940 goto out;
3941 } else if (fvp->v_type != VDIR &&
3942 tvp->v_type == VDIR) {
3943 error = EISDIR;
3944 goto out;
3945 }
3946 }
3947 #ifdef CAPABILITIES
3948 if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3949 /*
3950 * If the target already exists we require CAP_UNLINKAT
3951 * from 'newfd', when newfd was used for the lookup.
3952 */
3953 error = cap_check(&tond.ni_filecaps.fc_rights,
3954 &cap_unlinkat_rights);
3955 if (error != 0)
3956 goto out;
3957 }
3958 #endif
3959 }
3960 if (fvp == tdvp) {
3961 error = EINVAL;
3962 goto out;
3963 }
3964 /*
3965 * If the source is the same as the destination (that is, if they
3966 * are links to the same vnode), then there is nothing to do.
3967 */
3968 if (fvp == tvp)
3969 error = ERESTART;
3970 #ifdef MAC
3971 else
3972 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3973 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3974 #endif
3975 out:
3976 if (error == 0) {
3977 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3978 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd, flags);
3979 NDFREE_PNBUF(&fromnd);
3980 NDFREE_PNBUF(&tond);
3981 } else {
3982 NDFREE_PNBUF(&fromnd);
3983 NDFREE_PNBUF(&tond);
3984 if (tvp != NULL)
3985 vput(tvp);
3986 if (tdvp == tvp)
3987 vrele(tdvp);
3988 else
3989 vput(tdvp);
3990 vrele(fromnd.ni_dvp);
3991 vrele(fvp);
3992 }
3993 if (tmp != NULL) {
3994 lockmgr(&tmp->mnt_renamelock, LK_RELEASE, 0);
3995 vfs_rel(tmp);
3996 }
3997 if (mp != NULL)
3998 vn_finished_write(mp);
3999 out1:
4000 if (error == ERESTART)
4001 return (0);
4002 if (error == ERELOOKUP) {
4003 error = sig_intr();
4004 if (error == 0)
4005 goto again;
4006 }
4007 return (error);
4008 }
4009
4010 /*
4011 * Make a directory file.
4012 */
4013 #ifndef _SYS_SYSPROTO_H_
4014 struct mkdir_args {
4015 char *path;
4016 int mode;
4017 };
4018 #endif
4019 int
sys_mkdir(struct thread * td,struct mkdir_args * uap)4020 sys_mkdir(struct thread *td, struct mkdir_args *uap)
4021 {
4022
4023 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
4024 uap->mode));
4025 }
4026
4027 #ifndef _SYS_SYSPROTO_H_
4028 struct mkdirat_args {
4029 int fd;
4030 char *path;
4031 mode_t mode;
4032 };
4033 #endif
4034 int
sys_mkdirat(struct thread * td,struct mkdirat_args * uap)4035 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
4036 {
4037
4038 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
4039 }
4040
4041 int
kern_mkdirat(struct thread * td,int fd,const char * path,enum uio_seg segflg,int mode)4042 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
4043 int mode)
4044 {
4045 struct mount *mp;
4046 struct vattr vattr;
4047 struct nameidata nd;
4048 int error;
4049
4050 AUDIT_ARG_MODE(mode);
4051 NDPREINIT(&nd);
4052 restart:
4053 bwillwrite();
4054 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 |
4055 NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
4056 segflg, path, fd, &cap_mkdirat_rights);
4057 if ((error = namei(&nd)) != 0)
4058 return (error);
4059 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
4060 NDFREE_PNBUF(&nd);
4061 vput(nd.ni_dvp);
4062 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
4063 return (error);
4064 goto restart;
4065 }
4066 if ((vn_irflag_read(nd.ni_dvp) & VIRF_NAMEDDIR) != 0) {
4067 error = EINVAL;
4068 goto out;
4069 }
4070 VATTR_NULL(&vattr);
4071 vattr.va_type = VDIR;
4072 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
4073 #ifdef MAC
4074 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
4075 &vattr);
4076 if (error != 0)
4077 goto out;
4078 #endif
4079 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
4080 out:
4081 NDFREE_PNBUF(&nd);
4082 VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
4083 vn_finished_write(mp);
4084 if (error == ERELOOKUP)
4085 goto restart;
4086 return (error);
4087 }
4088
4089 /*
4090 * Remove a directory file.
4091 */
4092 #ifndef _SYS_SYSPROTO_H_
4093 struct rmdir_args {
4094 char *path;
4095 };
4096 #endif
4097 int
sys_rmdir(struct thread * td,struct rmdir_args * uap)4098 sys_rmdir(struct thread *td, struct rmdir_args *uap)
4099 {
4100
4101 return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
4102 0));
4103 }
4104
4105 int
kern_frmdirat(struct thread * td,int dfd,const char * path,int fd,enum uio_seg pathseg,int flag)4106 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
4107 enum uio_seg pathseg, int flag)
4108 {
4109 struct mount *mp;
4110 struct vnode *vp;
4111 struct file *fp;
4112 struct nameidata nd;
4113 cap_rights_t rights;
4114 int error;
4115
4116 fp = NULL;
4117 if (fd != FD_NONE) {
4118 error = getvnode(td, fd, cap_rights_init_one(&rights,
4119 CAP_LOOKUP), &fp);
4120 if (error != 0)
4121 return (error);
4122 }
4123
4124 NDPREINIT(&nd);
4125 restart:
4126 bwillwrite();
4127 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
4128 at2cnpflags(flag, AT_RESOLVE_BENEATH),
4129 pathseg, path, dfd, &cap_unlinkat_rights);
4130 if ((error = namei(&nd)) != 0)
4131 goto fdout;
4132 vp = nd.ni_vp;
4133 if (vp->v_type != VDIR) {
4134 error = ENOTDIR;
4135 goto out;
4136 }
4137 /*
4138 * No rmdir "." please.
4139 */
4140 if (nd.ni_dvp == vp) {
4141 error = EINVAL;
4142 goto out;
4143 }
4144 /*
4145 * The root of a mounted filesystem cannot be deleted.
4146 */
4147 if (vp->v_vflag & VV_ROOT) {
4148 error = EBUSY;
4149 goto out;
4150 }
4151
4152 if (fp != NULL && fp->f_vnode != vp) {
4153 if (VN_IS_DOOMED(fp->f_vnode))
4154 error = EBADF;
4155 else
4156 error = EDEADLK;
4157 goto out;
4158 }
4159
4160 #ifdef MAC
4161 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
4162 &nd.ni_cnd);
4163 if (error != 0)
4164 goto out;
4165 #endif
4166 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
4167 NDFREE_PNBUF(&nd);
4168 vput(vp);
4169 if (nd.ni_dvp == vp)
4170 vrele(nd.ni_dvp);
4171 else
4172 vput(nd.ni_dvp);
4173 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
4174 goto fdout;
4175 goto restart;
4176 }
4177 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
4178 vn_finished_write(mp);
4179 out:
4180 NDFREE_PNBUF(&nd);
4181 vput(vp);
4182 if (nd.ni_dvp == vp)
4183 vrele(nd.ni_dvp);
4184 else
4185 vput(nd.ni_dvp);
4186 if (error == ERELOOKUP)
4187 goto restart;
4188 fdout:
4189 if (fp != NULL)
4190 fdrop(fp, td);
4191 return (error);
4192 }
4193
4194 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
4195 int
freebsd11_kern_getdirentries(struct thread * td,int fd,char * ubuf,u_int count,long * basep,void (* func)(struct freebsd11_dirent *))4196 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
4197 long *basep, void (*func)(struct freebsd11_dirent *))
4198 {
4199 struct freebsd11_dirent dstdp;
4200 struct dirent *dp, *edp;
4201 char *dirbuf;
4202 off_t base;
4203 ssize_t resid, ucount;
4204 int error;
4205
4206 /* XXX arbitrary sanity limit on `count'. */
4207 count = min(count, 64 * 1024);
4208
4209 dirbuf = malloc(count, M_TEMP, M_WAITOK);
4210
4211 error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
4212 UIO_SYSSPACE);
4213 if (error != 0)
4214 goto done;
4215 if (basep != NULL)
4216 *basep = base;
4217
4218 ucount = 0;
4219 for (dp = (struct dirent *)dirbuf,
4220 edp = (struct dirent *)&dirbuf[count - resid];
4221 ucount < count && dp < edp; ) {
4222 if (dp->d_reclen == 0)
4223 break;
4224 MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
4225 if (dp->d_namlen >= sizeof(dstdp.d_name))
4226 continue;
4227 dstdp.d_type = dp->d_type;
4228 dstdp.d_namlen = dp->d_namlen;
4229 dstdp.d_fileno = dp->d_fileno; /* truncate */
4230 if (dstdp.d_fileno != dp->d_fileno) {
4231 switch (ino64_trunc_error) {
4232 default:
4233 case 0:
4234 break;
4235 case 1:
4236 error = EOVERFLOW;
4237 goto done;
4238 case 2:
4239 dstdp.d_fileno = UINT32_MAX;
4240 break;
4241 }
4242 }
4243 dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
4244 ((dp->d_namlen + 1 + 3) &~ 3);
4245 bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4246 bzero(dstdp.d_name + dstdp.d_namlen,
4247 dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4248 dstdp.d_namlen);
4249 MPASS(dstdp.d_reclen <= dp->d_reclen);
4250 MPASS(ucount + dstdp.d_reclen <= count);
4251 if (func != NULL)
4252 func(&dstdp);
4253 error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4254 if (error != 0)
4255 break;
4256 dp = (struct dirent *)((char *)dp + dp->d_reclen);
4257 ucount += dstdp.d_reclen;
4258 }
4259
4260 done:
4261 free(dirbuf, M_TEMP);
4262 if (error == 0)
4263 td->td_retval[0] = ucount;
4264 return (error);
4265 }
4266 #endif /* COMPAT */
4267
4268 #ifdef COMPAT_43
4269 static void
ogetdirentries_cvt(struct freebsd11_dirent * dp)4270 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4271 {
4272 #if (BYTE_ORDER == LITTLE_ENDIAN)
4273 /*
4274 * The expected low byte of dp->d_namlen is our dp->d_type.
4275 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4276 */
4277 dp->d_type = dp->d_namlen;
4278 dp->d_namlen = 0;
4279 #else
4280 /*
4281 * The dp->d_type is the high byte of the expected dp->d_namlen,
4282 * so must be zero'ed.
4283 */
4284 dp->d_type = 0;
4285 #endif
4286 }
4287
4288 /*
4289 * Read a block of directory entries in a filesystem independent format.
4290 */
4291 #ifndef _SYS_SYSPROTO_H_
4292 struct ogetdirentries_args {
4293 int fd;
4294 char *buf;
4295 u_int count;
4296 long *basep;
4297 };
4298 #endif
4299 int
ogetdirentries(struct thread * td,struct ogetdirentries_args * uap)4300 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4301 {
4302 long loff;
4303 int error;
4304
4305 error = kern_ogetdirentries(td, uap, &loff);
4306 if (error == 0)
4307 error = copyout(&loff, uap->basep, sizeof(long));
4308 return (error);
4309 }
4310
4311 int
kern_ogetdirentries(struct thread * td,struct ogetdirentries_args * uap,long * ploff)4312 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4313 long *ploff)
4314 {
4315 long base;
4316 int error;
4317
4318 /* XXX arbitrary sanity limit on `count'. */
4319 if (uap->count > 64 * 1024)
4320 return (EINVAL);
4321
4322 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4323 &base, ogetdirentries_cvt);
4324
4325 if (error == 0 && uap->basep != NULL)
4326 error = copyout(&base, uap->basep, sizeof(long));
4327
4328 return (error);
4329 }
4330 #endif /* COMPAT_43 */
4331
4332 #if defined(COMPAT_FREEBSD11)
4333 #ifndef _SYS_SYSPROTO_H_
4334 struct freebsd11_getdirentries_args {
4335 int fd;
4336 char *buf;
4337 u_int count;
4338 long *basep;
4339 };
4340 #endif
4341 int
freebsd11_getdirentries(struct thread * td,struct freebsd11_getdirentries_args * uap)4342 freebsd11_getdirentries(struct thread *td,
4343 struct freebsd11_getdirentries_args *uap)
4344 {
4345 long base;
4346 int error;
4347
4348 error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4349 &base, NULL);
4350
4351 if (error == 0 && uap->basep != NULL)
4352 error = copyout(&base, uap->basep, sizeof(long));
4353 return (error);
4354 }
4355
4356 int
freebsd11_getdents(struct thread * td,struct freebsd11_getdents_args * uap)4357 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4358 {
4359 struct freebsd11_getdirentries_args ap;
4360
4361 ap.fd = uap->fd;
4362 ap.buf = uap->buf;
4363 ap.count = uap->count;
4364 ap.basep = NULL;
4365 return (freebsd11_getdirentries(td, &ap));
4366 }
4367 #endif /* COMPAT_FREEBSD11 */
4368
4369 /*
4370 * Read a block of directory entries in a filesystem independent format.
4371 */
4372 int
sys_getdirentries(struct thread * td,struct getdirentries_args * uap)4373 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4374 {
4375 off_t base;
4376 int error;
4377
4378 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4379 NULL, UIO_USERSPACE);
4380 if (error != 0)
4381 return (error);
4382 if (uap->basep != NULL)
4383 error = copyout(&base, uap->basep, sizeof(off_t));
4384 return (error);
4385 }
4386
4387 int
kern_getdirentries(struct thread * td,int fd,char * buf,size_t count,off_t * basep,ssize_t * residp,enum uio_seg bufseg)4388 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4389 off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4390 {
4391 struct vnode *vp;
4392 struct file *fp;
4393 struct uio auio;
4394 struct iovec aiov;
4395 off_t loff;
4396 int error, eofflag;
4397 off_t foffset;
4398
4399 AUDIT_ARG_FD(fd);
4400 if (count > IOSIZE_MAX)
4401 return (EINVAL);
4402 auio.uio_resid = count;
4403 error = getvnode(td, fd, &cap_read_rights, &fp);
4404 if (error != 0)
4405 return (error);
4406 if ((fp->f_flag & FREAD) == 0) {
4407 fdrop(fp, td);
4408 return (EBADF);
4409 }
4410 vp = fp->f_vnode;
4411 foffset = foffset_lock(fp, 0);
4412 unionread:
4413 if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) {
4414 error = ENOENT;
4415 goto fail;
4416 }
4417 aiov.iov_base = buf;
4418 aiov.iov_len = count;
4419 auio.uio_iov = &aiov;
4420 auio.uio_iovcnt = 1;
4421 auio.uio_rw = UIO_READ;
4422 auio.uio_segflg = bufseg;
4423 auio.uio_td = td;
4424 vn_lock(vp, LK_SHARED | LK_RETRY);
4425 /*
4426 * We want to return ENOTDIR for anything that is not VDIR, but
4427 * not for VBAD, and we can't check for VBAD while the vnode is
4428 * unlocked.
4429 */
4430 if (vp->v_type != VDIR) {
4431 if (vp->v_type == VBAD)
4432 error = EBADF;
4433 else
4434 error = ENOTDIR;
4435 VOP_UNLOCK(vp);
4436 goto fail;
4437 }
4438 AUDIT_ARG_VNODE1(vp);
4439 loff = auio.uio_offset = foffset;
4440 #ifdef MAC
4441 error = mac_vnode_check_readdir(td->td_ucred, vp);
4442 if (error == 0)
4443 #endif
4444 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4445 NULL);
4446 foffset = auio.uio_offset;
4447 if (error != 0) {
4448 VOP_UNLOCK(vp);
4449 goto fail;
4450 }
4451 if (count == auio.uio_resid &&
4452 (vp->v_vflag & VV_ROOT) &&
4453 (vp->v_mount->mnt_flag & MNT_UNION)) {
4454 struct vnode *tvp = vp;
4455
4456 vp = vp->v_mount->mnt_vnodecovered;
4457 vref(vp);
4458 fp->f_vnode = vp;
4459 foffset = 0;
4460 vput(tvp);
4461 goto unionread;
4462 }
4463 VOP_UNLOCK(vp);
4464 *basep = loff;
4465 if (residp != NULL)
4466 *residp = auio.uio_resid;
4467 td->td_retval[0] = count - auio.uio_resid;
4468 fail:
4469 foffset_unlock(fp, foffset, 0);
4470 fdrop(fp, td);
4471 return (error);
4472 }
4473
4474 /*
4475 * Set the mode mask for creation of filesystem nodes.
4476 */
4477 #ifndef _SYS_SYSPROTO_H_
4478 struct umask_args {
4479 int newmask;
4480 };
4481 #endif
4482 int
sys_umask(struct thread * td,struct umask_args * uap)4483 sys_umask(struct thread *td, struct umask_args *uap)
4484 {
4485 struct pwddesc *pdp;
4486
4487 pdp = td->td_proc->p_pd;
4488 PWDDESC_XLOCK(pdp);
4489 td->td_retval[0] = pdp->pd_cmask;
4490 pdp->pd_cmask = uap->newmask & ALLPERMS;
4491 PWDDESC_XUNLOCK(pdp);
4492 return (0);
4493 }
4494
4495 /*
4496 * Void all references to file by ripping underlying filesystem away from
4497 * vnode.
4498 */
4499 #ifndef _SYS_SYSPROTO_H_
4500 struct revoke_args {
4501 char *path;
4502 };
4503 #endif
4504 int
sys_revoke(struct thread * td,struct revoke_args * uap)4505 sys_revoke(struct thread *td, struct revoke_args *uap)
4506 {
4507 struct vnode *vp;
4508 struct vattr vattr;
4509 struct nameidata nd;
4510 int error;
4511
4512 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4513 uap->path);
4514 if ((error = namei(&nd)) != 0)
4515 return (error);
4516 vp = nd.ni_vp;
4517 NDFREE_PNBUF(&nd);
4518 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4519 error = EINVAL;
4520 goto out;
4521 }
4522 #ifdef MAC
4523 error = mac_vnode_check_revoke(td->td_ucred, vp);
4524 if (error != 0)
4525 goto out;
4526 #endif
4527 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4528 if (error != 0)
4529 goto out;
4530 if (td->td_ucred->cr_uid != vattr.va_uid) {
4531 error = priv_check(td, PRIV_VFS_ADMIN);
4532 if (error != 0)
4533 goto out;
4534 }
4535 if (devfs_usecount(vp) > 0)
4536 VOP_REVOKE(vp, REVOKEALL);
4537 out:
4538 vput(vp);
4539 return (error);
4540 }
4541
4542 /*
4543 * This variant of getvnode() allows O_PATH files. Caller should
4544 * ensure that returned file and vnode are only used for compatible
4545 * semantics.
4546 */
4547 int
getvnode_path(struct thread * td,int fd,const cap_rights_t * rightsp,uint8_t * flagsp,struct file ** fpp)4548 getvnode_path(struct thread *td, int fd, const cap_rights_t *rightsp,
4549 uint8_t *flagsp, struct file **fpp)
4550 {
4551 struct file *fp;
4552 int error;
4553
4554 error = fget_unlocked_flags(td, fd, rightsp, flagsp, &fp);
4555 if (error != 0)
4556 return (error);
4557
4558 /*
4559 * The file could be not of the vnode type, or it may be not
4560 * yet fully initialized, in which case the f_vnode pointer
4561 * may be set, but f_ops is still badfileops. E.g.,
4562 * devfs_open() transiently create such situation to
4563 * facilitate csw d_fdopen().
4564 *
4565 * Dupfdopen() handling in kern_openat() installs the
4566 * half-baked file into the process descriptor table, allowing
4567 * other thread to dereference it. Guard against the race by
4568 * checking f_ops.
4569 */
4570 if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) {
4571 fdrop(fp, td);
4572 *fpp = NULL;
4573 return (EINVAL);
4574 }
4575
4576 *fpp = fp;
4577 return (0);
4578 }
4579
4580 /*
4581 * Convert a user file descriptor to a kernel file entry and check
4582 * that, if it is a capability, the correct rights are present.
4583 * A reference on the file entry is held upon returning.
4584 */
4585 int
getvnode(struct thread * td,int fd,const cap_rights_t * rightsp,struct file ** fpp)4586 getvnode(struct thread *td, int fd, const cap_rights_t *rightsp,
4587 struct file **fpp)
4588 {
4589 int error;
4590
4591 error = getvnode_path(td, fd, rightsp, NULL, fpp);
4592 if (__predict_false(error != 0))
4593 return (error);
4594
4595 /*
4596 * Filter out O_PATH file descriptors, most getvnode() callers
4597 * do not call fo_ methods.
4598 */
4599 if (__predict_false((*fpp)->f_ops == &path_fileops)) {
4600 fdrop(*fpp, td);
4601 *fpp = NULL;
4602 error = EBADF;
4603 }
4604
4605 return (error);
4606 }
4607
4608 /*
4609 * Get an (NFS) file handle.
4610 */
4611 #ifndef _SYS_SYSPROTO_H_
4612 struct lgetfh_args {
4613 char *fname;
4614 fhandle_t *fhp;
4615 };
4616 #endif
4617 int
sys_lgetfh(struct thread * td,struct lgetfh_args * uap)4618 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4619 {
4620
4621 return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4622 UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4623 }
4624
4625 #ifndef _SYS_SYSPROTO_H_
4626 struct getfh_args {
4627 char *fname;
4628 fhandle_t *fhp;
4629 };
4630 #endif
4631 int
sys_getfh(struct thread * td,struct getfh_args * uap)4632 sys_getfh(struct thread *td, struct getfh_args *uap)
4633 {
4634
4635 return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4636 uap->fhp, UIO_USERSPACE));
4637 }
4638
4639 /*
4640 * syscall for the rpc.lockd to use to translate an open descriptor into
4641 * a NFS file handle.
4642 *
4643 * warning: do not remove the priv_check() call or this becomes one giant
4644 * security hole.
4645 */
4646 #ifndef _SYS_SYSPROTO_H_
4647 struct getfhat_args {
4648 int fd;
4649 char *path;
4650 fhandle_t *fhp;
4651 int flags;
4652 };
4653 #endif
4654 int
sys_getfhat(struct thread * td,struct getfhat_args * uap)4655 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4656 {
4657
4658 return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4659 uap->fhp, UIO_USERSPACE));
4660 }
4661
4662 int
kern_getfhat(struct thread * td,int flags,int fd,const char * path,enum uio_seg pathseg,fhandle_t * fhp,enum uio_seg fhseg)4663 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4664 enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4665 {
4666 struct nameidata nd;
4667 fhandle_t fh;
4668 struct vnode *vp;
4669 int error;
4670
4671 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4672 return (EINVAL);
4673 error = priv_check(td, PRIV_VFS_GETFH);
4674 if (error != 0)
4675 return (error);
4676 NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4677 AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4678 fd);
4679 error = namei(&nd);
4680 if (error != 0)
4681 return (error);
4682 NDFREE_PNBUF(&nd);
4683 vp = nd.ni_vp;
4684 bzero(&fh, sizeof(fh));
4685 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4686 error = VOP_VPTOFH(vp, &fh.fh_fid);
4687 vput(vp);
4688 if (error == 0) {
4689 if (fhseg == UIO_USERSPACE)
4690 error = copyout(&fh, fhp, sizeof (fh));
4691 else
4692 memcpy(fhp, &fh, sizeof(fh));
4693 }
4694 return (error);
4695 }
4696
4697 #ifndef _SYS_SYSPROTO_H_
4698 struct fhlink_args {
4699 fhandle_t *fhp;
4700 const char *to;
4701 };
4702 #endif
4703 int
sys_fhlink(struct thread * td,struct fhlink_args * uap)4704 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4705 {
4706
4707 return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4708 }
4709
4710 #ifndef _SYS_SYSPROTO_H_
4711 struct fhlinkat_args {
4712 fhandle_t *fhp;
4713 int tofd;
4714 const char *to;
4715 };
4716 #endif
4717 int
sys_fhlinkat(struct thread * td,struct fhlinkat_args * uap)4718 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4719 {
4720
4721 return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4722 }
4723
4724 static int
kern_fhlinkat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,fhandle_t * fhp)4725 kern_fhlinkat(struct thread *td, int fd, const char *path,
4726 enum uio_seg pathseg, fhandle_t *fhp)
4727 {
4728 fhandle_t fh;
4729 struct mount *mp;
4730 struct vnode *vp;
4731 int error;
4732
4733 error = priv_check(td, PRIV_VFS_GETFH);
4734 if (error != 0)
4735 return (error);
4736 error = copyin(fhp, &fh, sizeof(fh));
4737 if (error != 0)
4738 return (error);
4739 do {
4740 bwillwrite();
4741 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4742 return (ESTALE);
4743 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4744 vfs_unbusy(mp);
4745 if (error != 0)
4746 return (error);
4747 VOP_UNLOCK(vp);
4748 error = kern_linkat_vp(td, vp, fd, path, pathseg);
4749 } while (error == EAGAIN || error == ERELOOKUP);
4750 return (error);
4751 }
4752
4753 #ifndef _SYS_SYSPROTO_H_
4754 struct fhreadlink_args {
4755 fhandle_t *fhp;
4756 char *buf;
4757 size_t bufsize;
4758 };
4759 #endif
4760 int
sys_fhreadlink(struct thread * td,struct fhreadlink_args * uap)4761 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4762 {
4763 fhandle_t fh;
4764 struct mount *mp;
4765 struct vnode *vp;
4766 int error;
4767
4768 error = priv_check(td, PRIV_VFS_GETFH);
4769 if (error != 0)
4770 return (error);
4771 if (uap->bufsize > IOSIZE_MAX)
4772 return (EINVAL);
4773 error = copyin(uap->fhp, &fh, sizeof(fh));
4774 if (error != 0)
4775 return (error);
4776 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4777 return (ESTALE);
4778 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4779 vfs_unbusy(mp);
4780 if (error != 0)
4781 return (error);
4782 error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4783 vput(vp);
4784 return (error);
4785 }
4786
4787 /*
4788 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4789 * open descriptor.
4790 *
4791 * warning: do not remove the priv_check() call or this becomes one giant
4792 * security hole.
4793 */
4794 #ifndef _SYS_SYSPROTO_H_
4795 struct fhopen_args {
4796 const struct fhandle *u_fhp;
4797 int flags;
4798 };
4799 #endif
4800 int
sys_fhopen(struct thread * td,struct fhopen_args * uap)4801 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4802 {
4803 return (kern_fhopen(td, uap->u_fhp, uap->flags));
4804 }
4805
4806 int
kern_fhopen(struct thread * td,const struct fhandle * u_fhp,int flags)4807 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4808 {
4809 struct mount *mp;
4810 struct vnode *vp;
4811 struct fhandle fhp;
4812 struct file *fp;
4813 int error, indx;
4814 bool named_attr;
4815
4816 error = priv_check(td, PRIV_VFS_FHOPEN);
4817 if (error != 0)
4818 return (error);
4819
4820 indx = -1;
4821 if ((flags & O_CREAT) != 0)
4822 return (EINVAL);
4823 error = openflags(&flags);
4824 if (error != 0)
4825 return (error);
4826 error = copyin(u_fhp, &fhp, sizeof(fhp));
4827 if (error != 0)
4828 return (error);
4829 /* find the mount point */
4830 mp = vfs_busyfs(&fhp.fh_fsid);
4831 if (mp == NULL)
4832 return (ESTALE);
4833 /* now give me my vnode, it gets returned to me locked */
4834 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4835 vfs_unbusy(mp);
4836 if (error != 0)
4837 return (error);
4838
4839 /*
4840 * Check to see if the file handle refers to a named attribute
4841 * directory or attribute. If it does, the O_NAMEDATTR flag
4842 * must have been specified.
4843 */
4844 named_attr = (vn_irflag_read(vp) &
4845 (VIRF_NAMEDDIR | VIRF_NAMEDATTR)) != 0;
4846 if ((named_attr && (flags & O_NAMEDATTR) == 0) ||
4847 (!named_attr && (flags & O_NAMEDATTR) != 0)) {
4848 vput(vp);
4849 return (ENOATTR);
4850 }
4851
4852 error = falloc_noinstall(td, &fp);
4853 if (error != 0) {
4854 vput(vp);
4855 return (error);
4856 }
4857 /* Set the flags early so the finit in devfs can pick them up. */
4858 fp->f_flag = flags & FMASK;
4859
4860 #ifdef INVARIANTS
4861 td->td_dupfd = -1;
4862 #endif
4863 error = vn_open_vnode(vp, flags, td->td_ucred, td, fp);
4864 if (error != 0) {
4865 KASSERT(fp->f_ops == &badfileops,
4866 ("VOP_OPEN in fhopen() set f_ops"));
4867 KASSERT(td->td_dupfd < 0,
4868 ("fhopen() encountered fdopen()"));
4869
4870 vput(vp);
4871 goto bad;
4872 }
4873 #ifdef INVARIANTS
4874 td->td_dupfd = 0;
4875 #endif
4876 finit_open(fp, vp, flags);
4877 VOP_UNLOCK(vp);
4878 if ((flags & O_TRUNC) != 0) {
4879 error = fo_truncate(fp, 0, td->td_ucred, td);
4880 if (error != 0)
4881 goto bad;
4882 }
4883
4884 error = finstall(td, fp, &indx, flags, NULL);
4885 bad:
4886 fdrop(fp, td);
4887 td->td_retval[0] = indx;
4888 return (error);
4889 }
4890
4891 /*
4892 * Stat an (NFS) file handle.
4893 */
4894 #ifndef _SYS_SYSPROTO_H_
4895 struct fhstat_args {
4896 struct fhandle *u_fhp;
4897 struct stat *sb;
4898 };
4899 #endif
4900 int
sys_fhstat(struct thread * td,struct fhstat_args * uap)4901 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4902 {
4903 struct stat sb;
4904 struct fhandle fh;
4905 int error;
4906
4907 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4908 if (error != 0)
4909 return (error);
4910 error = kern_fhstat(td, fh, &sb);
4911 if (error == 0)
4912 error = copyout(&sb, uap->sb, sizeof(sb));
4913 return (error);
4914 }
4915
4916 int
kern_fhstat(struct thread * td,struct fhandle fh,struct stat * sb)4917 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4918 {
4919 struct mount *mp;
4920 struct vnode *vp;
4921 int error;
4922
4923 error = priv_check(td, PRIV_VFS_FHSTAT);
4924 if (error != 0)
4925 return (error);
4926 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4927 return (ESTALE);
4928 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4929 vfs_unbusy(mp);
4930 if (error != 0)
4931 return (error);
4932 error = VOP_STAT(vp, sb, td->td_ucred, NOCRED);
4933 vput(vp);
4934 return (error);
4935 }
4936
4937 /*
4938 * Implement fstatfs() for (NFS) file handles.
4939 */
4940 #ifndef _SYS_SYSPROTO_H_
4941 struct fhstatfs_args {
4942 struct fhandle *u_fhp;
4943 struct statfs *buf;
4944 };
4945 #endif
4946 int
sys_fhstatfs(struct thread * td,struct fhstatfs_args * uap)4947 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4948 {
4949 struct statfs *sfp;
4950 fhandle_t fh;
4951 int error;
4952
4953 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4954 if (error != 0)
4955 return (error);
4956 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4957 error = kern_fhstatfs(td, fh, sfp);
4958 if (error == 0)
4959 error = copyout(sfp, uap->buf, sizeof(*sfp));
4960 free(sfp, M_STATFS);
4961 return (error);
4962 }
4963
4964 int
kern_fhstatfs(struct thread * td,fhandle_t fh,struct statfs * buf)4965 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4966 {
4967 struct mount *mp;
4968 struct vnode *vp;
4969 int error;
4970
4971 error = priv_check(td, PRIV_VFS_FHSTATFS);
4972 if (error != 0)
4973 return (error);
4974 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4975 return (ESTALE);
4976 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4977 if (error != 0) {
4978 vfs_unbusy(mp);
4979 return (error);
4980 }
4981 vput(vp);
4982 error = prison_canseemount(td->td_ucred, mp);
4983 if (error != 0)
4984 goto out;
4985 #ifdef MAC
4986 error = mac_mount_check_stat(td->td_ucred, mp);
4987 if (error != 0)
4988 goto out;
4989 #endif
4990 error = VFS_STATFS(mp, buf);
4991 out:
4992 vfs_unbusy(mp);
4993 return (error);
4994 }
4995
4996 /*
4997 * Unlike madvise(2), we do not make a best effort to remember every
4998 * possible caching hint. Instead, we remember the last setting with
4999 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
5000 * region of any current setting.
5001 */
5002 int
kern_posix_fadvise(struct thread * td,int fd,off_t offset,off_t len,int advice)5003 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
5004 int advice)
5005 {
5006 struct fadvise_info *fa, *new;
5007 struct file *fp;
5008 struct vnode *vp;
5009 off_t end;
5010 int error;
5011
5012 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
5013 return (EINVAL);
5014 AUDIT_ARG_VALUE(advice);
5015 switch (advice) {
5016 case POSIX_FADV_SEQUENTIAL:
5017 case POSIX_FADV_RANDOM:
5018 case POSIX_FADV_NOREUSE:
5019 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
5020 break;
5021 case POSIX_FADV_NORMAL:
5022 case POSIX_FADV_WILLNEED:
5023 case POSIX_FADV_DONTNEED:
5024 new = NULL;
5025 break;
5026 default:
5027 return (EINVAL);
5028 }
5029 /* XXX: CAP_POSIX_FADVISE? */
5030 AUDIT_ARG_FD(fd);
5031 error = fget(td, fd, &cap_no_rights, &fp);
5032 if (error != 0)
5033 goto out;
5034 AUDIT_ARG_FILE(td->td_proc, fp);
5035 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
5036 error = ESPIPE;
5037 goto out;
5038 }
5039 if (fp->f_type != DTYPE_VNODE) {
5040 error = ENODEV;
5041 goto out;
5042 }
5043 vp = fp->f_vnode;
5044 if (vp->v_type != VREG) {
5045 error = ENODEV;
5046 goto out;
5047 }
5048 if (len == 0)
5049 end = OFF_MAX;
5050 else
5051 end = offset + len - 1;
5052 switch (advice) {
5053 case POSIX_FADV_SEQUENTIAL:
5054 case POSIX_FADV_RANDOM:
5055 case POSIX_FADV_NOREUSE:
5056 /*
5057 * Try to merge any existing non-standard region with
5058 * this new region if possible, otherwise create a new
5059 * non-standard region for this request.
5060 */
5061 mtx_pool_lock(mtxpool_sleep, fp);
5062 fa = fp->f_advice;
5063 if (fa != NULL && fa->fa_advice == advice &&
5064 ((fa->fa_start <= end && fa->fa_end >= offset) ||
5065 (end != OFF_MAX && fa->fa_start == end + 1) ||
5066 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
5067 if (offset < fa->fa_start)
5068 fa->fa_start = offset;
5069 if (end > fa->fa_end)
5070 fa->fa_end = end;
5071 } else {
5072 new->fa_advice = advice;
5073 new->fa_start = offset;
5074 new->fa_end = end;
5075 fp->f_advice = new;
5076 new = fa;
5077 }
5078 mtx_pool_unlock(mtxpool_sleep, fp);
5079 break;
5080 case POSIX_FADV_NORMAL:
5081 /*
5082 * If a the "normal" region overlaps with an existing
5083 * non-standard region, trim or remove the
5084 * non-standard region.
5085 */
5086 mtx_pool_lock(mtxpool_sleep, fp);
5087 fa = fp->f_advice;
5088 if (fa != NULL) {
5089 if (offset <= fa->fa_start && end >= fa->fa_end) {
5090 new = fa;
5091 fp->f_advice = NULL;
5092 } else if (offset <= fa->fa_start &&
5093 end >= fa->fa_start)
5094 fa->fa_start = end + 1;
5095 else if (offset <= fa->fa_end && end >= fa->fa_end)
5096 fa->fa_end = offset - 1;
5097 else if (offset >= fa->fa_start && end <= fa->fa_end) {
5098 /*
5099 * If the "normal" region is a middle
5100 * portion of the existing
5101 * non-standard region, just remove
5102 * the whole thing rather than picking
5103 * one side or the other to
5104 * preserve.
5105 */
5106 new = fa;
5107 fp->f_advice = NULL;
5108 }
5109 }
5110 mtx_pool_unlock(mtxpool_sleep, fp);
5111 break;
5112 case POSIX_FADV_WILLNEED:
5113 case POSIX_FADV_DONTNEED:
5114 error = VOP_ADVISE(vp, offset, end, advice);
5115 break;
5116 }
5117 out:
5118 if (fp != NULL)
5119 fdrop(fp, td);
5120 free(new, M_FADVISE);
5121 return (error);
5122 }
5123
5124 int
sys_posix_fadvise(struct thread * td,struct posix_fadvise_args * uap)5125 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
5126 {
5127 int error;
5128
5129 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
5130 uap->advice);
5131 return (kern_posix_error(td, error));
5132 }
5133
5134 int
kern_copy_file_range(struct thread * td,int infd,off_t * inoffp,int outfd,off_t * outoffp,size_t len,unsigned int flags)5135 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
5136 off_t *outoffp, size_t len, unsigned int flags)
5137 {
5138 struct file *infp, *infp1, *outfp, *outfp1;
5139 struct vnode *invp, *outvp;
5140 int error;
5141 size_t retlen;
5142 void *rl_rcookie, *rl_wcookie;
5143 off_t inoff, outoff, savinoff, savoutoff;
5144 bool foffsets_locked, foffsets_set;
5145
5146 infp = outfp = NULL;
5147 rl_rcookie = rl_wcookie = NULL;
5148 foffsets_locked = false;
5149 foffsets_set = false;
5150 error = 0;
5151 retlen = 0;
5152
5153 if ((flags & ~COPY_FILE_RANGE_USERFLAGS) != 0) {
5154 error = EINVAL;
5155 goto out;
5156 }
5157 if (len > SSIZE_MAX)
5158 /*
5159 * Although the len argument is size_t, the return argument
5160 * is ssize_t (which is signed). Therefore a size that won't
5161 * fit in ssize_t can't be returned.
5162 */
5163 len = SSIZE_MAX;
5164
5165 /* Get the file structures for the file descriptors. */
5166 error = fget_read(td, infd,
5167 inoffp != NULL ? &cap_pread_rights : &cap_read_rights, &infp);
5168 if (error != 0)
5169 goto out;
5170 if (infp->f_ops == &badfileops) {
5171 error = EBADF;
5172 goto out;
5173 }
5174 if (infp->f_vnode == NULL) {
5175 error = EINVAL;
5176 goto out;
5177 }
5178 error = fget_write(td, outfd,
5179 outoffp != NULL ? &cap_pwrite_rights : &cap_write_rights, &outfp);
5180 if (error != 0)
5181 goto out;
5182 if (outfp->f_ops == &badfileops) {
5183 error = EBADF;
5184 goto out;
5185 }
5186 if (outfp->f_vnode == NULL) {
5187 error = EINVAL;
5188 goto out;
5189 }
5190
5191 /*
5192 * Figure out which file offsets we're reading from and writing to.
5193 * If the offsets come from the file descriptions, we need to lock them,
5194 * and locking both offsets requires a loop to avoid deadlocks.
5195 */
5196 infp1 = outfp1 = NULL;
5197 if (inoffp != NULL)
5198 inoff = *inoffp;
5199 else
5200 infp1 = infp;
5201 if (outoffp != NULL)
5202 outoff = *outoffp;
5203 else
5204 outfp1 = outfp;
5205 if (infp1 != NULL || outfp1 != NULL) {
5206 if (infp1 == outfp1) {
5207 /*
5208 * Overlapping ranges are not allowed. A more thorough
5209 * check appears below, but we must not lock the same
5210 * offset twice.
5211 */
5212 error = EINVAL;
5213 goto out;
5214 }
5215 foffset_lock_pair(infp1, &inoff, outfp1, &outoff, 0);
5216 foffsets_locked = true;
5217 } else {
5218 foffsets_set = true;
5219 }
5220 savinoff = inoff;
5221 savoutoff = outoff;
5222
5223 invp = infp->f_vnode;
5224 outvp = outfp->f_vnode;
5225 /* Sanity check the f_flag bits. */
5226 if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
5227 (infp->f_flag & FREAD) == 0) {
5228 error = EBADF;
5229 goto out;
5230 }
5231
5232 /* If len == 0, just return 0. */
5233 if (len == 0)
5234 goto out;
5235
5236 /*
5237 * Make sure that the ranges we check and lock below are valid. Note
5238 * that len is clamped to SSIZE_MAX above.
5239 */
5240 if (inoff < 0 || outoff < 0) {
5241 error = EINVAL;
5242 goto out;
5243 }
5244
5245 /*
5246 * If infp and outfp refer to the same file, the byte ranges cannot
5247 * overlap.
5248 */
5249 if (invp == outvp) {
5250 if ((inoff <= outoff && inoff + len > outoff) ||
5251 (inoff > outoff && outoff + len > inoff)) {
5252 error = EINVAL;
5253 goto out;
5254 }
5255 rangelock_may_recurse(&invp->v_rl);
5256 }
5257
5258 /* Range lock the byte ranges for both invp and outvp. */
5259 for (;;) {
5260 rl_wcookie = vn_rangelock_wlock(outvp, outoff, outoff + len);
5261 rl_rcookie = vn_rangelock_tryrlock(invp, inoff, inoff + len);
5262 if (rl_rcookie != NULL)
5263 break;
5264 vn_rangelock_unlock(outvp, rl_wcookie);
5265 rl_rcookie = vn_rangelock_rlock(invp, inoff, inoff + len);
5266 vn_rangelock_unlock(invp, rl_rcookie);
5267 }
5268
5269 retlen = len;
5270 error = vn_copy_file_range(invp, &inoff, outvp, &outoff, &retlen,
5271 flags, infp->f_cred, outfp->f_cred, td);
5272 out:
5273 if (rl_rcookie != NULL)
5274 vn_rangelock_unlock(invp, rl_rcookie);
5275 if (rl_wcookie != NULL)
5276 vn_rangelock_unlock(outvp, rl_wcookie);
5277 if ((foffsets_locked || foffsets_set) &&
5278 (error == EINTR || error == ERESTART)) {
5279 inoff = savinoff;
5280 outoff = savoutoff;
5281 }
5282 if (foffsets_locked) {
5283 if (inoffp == NULL)
5284 foffset_unlock(infp, inoff, 0);
5285 else
5286 *inoffp = inoff;
5287 if (outoffp == NULL)
5288 foffset_unlock(outfp, outoff, 0);
5289 else
5290 *outoffp = outoff;
5291 } else if (foffsets_set) {
5292 *inoffp = inoff;
5293 *outoffp = outoff;
5294 }
5295 if (outfp != NULL)
5296 fdrop(outfp, td);
5297 if (infp != NULL)
5298 fdrop(infp, td);
5299 td->td_retval[0] = retlen;
5300 return (error);
5301 }
5302
5303 int
sys_copy_file_range(struct thread * td,struct copy_file_range_args * uap)5304 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
5305 {
5306 off_t inoff, outoff, *inoffp, *outoffp;
5307 int error;
5308
5309 inoffp = outoffp = NULL;
5310 if (uap->inoffp != NULL) {
5311 error = copyin(uap->inoffp, &inoff, sizeof(off_t));
5312 if (error != 0)
5313 return (error);
5314 inoffp = &inoff;
5315 }
5316 if (uap->outoffp != NULL) {
5317 error = copyin(uap->outoffp, &outoff, sizeof(off_t));
5318 if (error != 0)
5319 return (error);
5320 outoffp = &outoff;
5321 }
5322 error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
5323 outoffp, uap->len, uap->flags);
5324 if (error == 0 && uap->inoffp != NULL)
5325 error = copyout(inoffp, uap->inoffp, sizeof(off_t));
5326 if (error == 0 && uap->outoffp != NULL)
5327 error = copyout(outoffp, uap->outoffp, sizeof(off_t));
5328 return (error);
5329 }
5330