xref: /freebsd/sys/kern/vfs_syscalls.c (revision cf0f693f0361e76517cabc234321ce7870fa8f9f)
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, 0);
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 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
295 		prison_enforce_statfs(td->td_ucred, mp, buf);
296 	}
297 out:
298 	vfs_unbusy(mp);
299 	return (error);
300 }
301 
302 /*
303  * Get filesystem statistics.
304  */
305 #ifndef _SYS_SYSPROTO_H_
306 struct statfs_args {
307 	char *path;
308 	struct statfs *buf;
309 };
310 #endif
311 int
sys_statfs(struct thread * td,struct statfs_args * uap)312 sys_statfs(struct thread *td, struct statfs_args *uap)
313 {
314 	struct statfs *sfp;
315 	int error;
316 
317 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
318 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
319 	if (error == 0)
320 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
321 	free(sfp, M_STATFS);
322 	return (error);
323 }
324 
325 int
kern_statfs(struct thread * td,const char * path,enum uio_seg pathseg,struct statfs * buf)326 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
327     struct statfs *buf)
328 {
329 	struct mount *mp;
330 	struct nameidata nd;
331 	int error;
332 
333 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
334 	error = namei(&nd);
335 	if (error != 0)
336 		return (error);
337 	NDFREE_PNBUF(&nd);
338 	mp = vfs_ref_from_vp(nd.ni_vp);
339 	vrele(nd.ni_vp);
340 	return (kern_do_statfs(td, mp, buf));
341 }
342 
343 /*
344  * Get filesystem statistics.
345  */
346 #ifndef _SYS_SYSPROTO_H_
347 struct fstatfs_args {
348 	int fd;
349 	struct statfs *buf;
350 };
351 #endif
352 int
sys_fstatfs(struct thread * td,struct fstatfs_args * uap)353 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
354 {
355 	struct statfs *sfp;
356 	int error;
357 
358 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
359 	error = kern_fstatfs(td, uap->fd, sfp);
360 	if (error == 0)
361 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
362 	free(sfp, M_STATFS);
363 	return (error);
364 }
365 
366 int
kern_fstatfs(struct thread * td,int fd,struct statfs * buf)367 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
368 {
369 	struct file *fp;
370 	struct mount *mp;
371 	struct vnode *vp;
372 	int error;
373 
374 	AUDIT_ARG_FD(fd);
375 	error = getvnode_path(td, fd, &cap_fstatfs_rights, NULL, &fp);
376 	if (error != 0)
377 		return (error);
378 	vp = fp->f_vnode;
379 #ifdef AUDIT
380 	if (AUDITING_TD(td)) {
381 		vn_lock(vp, LK_SHARED | LK_RETRY);
382 		AUDIT_ARG_VNODE1(vp);
383 		VOP_UNLOCK(vp);
384 	}
385 #endif
386 	mp = vfs_ref_from_vp(vp);
387 	fdrop(fp, td);
388 	return (kern_do_statfs(td, mp, buf));
389 }
390 
391 /*
392  * Get statistics on all filesystems.
393  */
394 #ifndef _SYS_SYSPROTO_H_
395 struct getfsstat_args {
396 	struct statfs *buf;
397 	long bufsize;
398 	int mode;
399 };
400 #endif
401 int
sys_getfsstat(struct thread * td,struct getfsstat_args * uap)402 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
403 {
404 	size_t count;
405 	int error;
406 
407 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
408 		return (EINVAL);
409 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
410 	    UIO_USERSPACE, uap->mode);
411 	if (error == 0)
412 		td->td_retval[0] = count;
413 	return (error);
414 }
415 
416 /*
417  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
418  *	The caller is responsible for freeing memory which will be allocated
419  *	in '*buf'.
420  */
421 int
kern_getfsstat(struct thread * td,struct statfs ** buf,size_t bufsize,size_t * countp,enum uio_seg bufseg,int mode)422 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
423     size_t *countp, enum uio_seg bufseg, int mode)
424 {
425 	struct mount *mp, *nmp;
426 	struct statfs *sfsp, *sp, *sptmp, *tofree;
427 	size_t count, maxcount;
428 	int error;
429 
430 	switch (mode) {
431 	case MNT_WAIT:
432 	case MNT_NOWAIT:
433 		break;
434 	default:
435 		if (bufseg == UIO_SYSSPACE)
436 			*buf = NULL;
437 		return (EINVAL);
438 	}
439 restart:
440 	maxcount = bufsize / sizeof(struct statfs);
441 	if (bufsize == 0) {
442 		sfsp = NULL;
443 		tofree = NULL;
444 	} else if (bufseg == UIO_USERSPACE) {
445 		sfsp = *buf;
446 		tofree = NULL;
447 	} else /* if (bufseg == UIO_SYSSPACE) */ {
448 		count = 0;
449 		mtx_lock(&mountlist_mtx);
450 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
451 			count++;
452 		}
453 		mtx_unlock(&mountlist_mtx);
454 		if (maxcount > count)
455 			maxcount = count;
456 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
457 		    M_STATFS, M_WAITOK);
458 	}
459 
460 	count = 0;
461 
462 	/*
463 	 * If there is no target buffer they only want the count.
464 	 *
465 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
466 	 * code below.
467 	 */
468 	if (sfsp == NULL) {
469 		mtx_lock(&mountlist_mtx);
470 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
471 			if (prison_canseemount(td->td_ucred, mp) != 0) {
472 				nmp = TAILQ_NEXT(mp, mnt_list);
473 				continue;
474 			}
475 #ifdef MAC
476 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
477 				nmp = TAILQ_NEXT(mp, mnt_list);
478 				continue;
479 			}
480 #endif
481 			count++;
482 			nmp = TAILQ_NEXT(mp, mnt_list);
483 		}
484 		mtx_unlock(&mountlist_mtx);
485 		*countp = count;
486 		return (0);
487 	}
488 
489 	/*
490 	 * They want the entire thing.
491 	 *
492 	 * Short-circuit the corner case of no room for anything, avoids
493 	 * relocking below.
494 	 */
495 	if (maxcount < 1) {
496 		goto out;
497 	}
498 
499 	mtx_lock(&mountlist_mtx);
500 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
501 		if (prison_canseemount(td->td_ucred, mp) != 0) {
502 			nmp = TAILQ_NEXT(mp, mnt_list);
503 			continue;
504 		}
505 #ifdef MAC
506 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
507 			nmp = TAILQ_NEXT(mp, mnt_list);
508 			continue;
509 		}
510 #endif
511 		if (mode == MNT_WAIT) {
512 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
513 				/*
514 				 * If vfs_busy() failed, and MBF_NOWAIT
515 				 * wasn't passed, then the mp is gone.
516 				 * Furthermore, because of MBF_MNTLSTLOCK,
517 				 * the mountlist_mtx was dropped.  We have
518 				 * no other choice than to start over.
519 				 */
520 				mtx_unlock(&mountlist_mtx);
521 				free(tofree, M_STATFS);
522 				goto restart;
523 			}
524 		} else {
525 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
526 				nmp = TAILQ_NEXT(mp, mnt_list);
527 				continue;
528 			}
529 		}
530 		sp = &mp->mnt_stat;
531 		/*
532 		 * If MNT_NOWAIT is specified, do not refresh
533 		 * the fsstat cache.
534 		 */
535 		if (mode != MNT_NOWAIT) {
536 			error = VFS_STATFS(mp, sp);
537 			if (error != 0) {
538 				mtx_lock(&mountlist_mtx);
539 				nmp = TAILQ_NEXT(mp, mnt_list);
540 				vfs_unbusy(mp);
541 				continue;
542 			}
543 		}
544 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
545 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
546 			    M_WAITOK);
547 			*sptmp = *sp;
548 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
549 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
550 			sp = sptmp;
551 		} else
552 			sptmp = NULL;
553 		if (bufseg == UIO_SYSSPACE) {
554 			bcopy(sp, sfsp, sizeof(*sp));
555 			free(sptmp, M_STATFS);
556 		} else /* if (bufseg == UIO_USERSPACE) */ {
557 			error = copyout(sp, sfsp, sizeof(*sp));
558 			free(sptmp, M_STATFS);
559 			if (error != 0) {
560 				vfs_unbusy(mp);
561 				return (error);
562 			}
563 		}
564 		sfsp++;
565 		count++;
566 
567 		if (count == maxcount) {
568 			vfs_unbusy(mp);
569 			goto out;
570 		}
571 
572 		mtx_lock(&mountlist_mtx);
573 		nmp = TAILQ_NEXT(mp, mnt_list);
574 		vfs_unbusy(mp);
575 	}
576 	mtx_unlock(&mountlist_mtx);
577 out:
578 	*countp = count;
579 	return (0);
580 }
581 
582 #ifdef COMPAT_FREEBSD4
583 /*
584  * Get old format filesystem statistics.
585  */
586 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
587 
588 #ifndef _SYS_SYSPROTO_H_
589 struct freebsd4_statfs_args {
590 	char *path;
591 	struct ostatfs *buf;
592 };
593 #endif
594 int
freebsd4_statfs(struct thread * td,struct freebsd4_statfs_args * uap)595 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
596 {
597 	struct ostatfs osb;
598 	struct statfs *sfp;
599 	int error;
600 
601 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
602 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
603 	if (error == 0) {
604 		freebsd4_cvtstatfs(sfp, &osb);
605 		error = copyout(&osb, uap->buf, sizeof(osb));
606 	}
607 	free(sfp, M_STATFS);
608 	return (error);
609 }
610 
611 /*
612  * Get filesystem statistics.
613  */
614 #ifndef _SYS_SYSPROTO_H_
615 struct freebsd4_fstatfs_args {
616 	int fd;
617 	struct ostatfs *buf;
618 };
619 #endif
620 int
freebsd4_fstatfs(struct thread * td,struct freebsd4_fstatfs_args * uap)621 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
622 {
623 	struct ostatfs osb;
624 	struct statfs *sfp;
625 	int error;
626 
627 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
628 	error = kern_fstatfs(td, uap->fd, sfp);
629 	if (error == 0) {
630 		freebsd4_cvtstatfs(sfp, &osb);
631 		error = copyout(&osb, uap->buf, sizeof(osb));
632 	}
633 	free(sfp, M_STATFS);
634 	return (error);
635 }
636 
637 /*
638  * Get statistics on all filesystems.
639  */
640 #ifndef _SYS_SYSPROTO_H_
641 struct freebsd4_getfsstat_args {
642 	struct ostatfs *buf;
643 	long bufsize;
644 	int mode;
645 };
646 #endif
647 int
freebsd4_getfsstat(struct thread * td,struct freebsd4_getfsstat_args * uap)648 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
649 {
650 	struct statfs *buf, *sp;
651 	struct ostatfs osb;
652 	size_t count, size;
653 	int error;
654 
655 	if (uap->bufsize < 0)
656 		return (EINVAL);
657 	count = uap->bufsize / sizeof(struct ostatfs);
658 	if (count > SIZE_MAX / sizeof(struct statfs))
659 		return (EINVAL);
660 	size = count * sizeof(struct statfs);
661 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
662 	    uap->mode);
663 	if (error == 0)
664 		td->td_retval[0] = count;
665 	if (size != 0) {
666 		sp = buf;
667 		while (count != 0 && error == 0) {
668 			freebsd4_cvtstatfs(sp, &osb);
669 			error = copyout(&osb, uap->buf, sizeof(osb));
670 			sp++;
671 			uap->buf++;
672 			count--;
673 		}
674 		free(buf, M_STATFS);
675 	}
676 	return (error);
677 }
678 
679 /*
680  * Implement fstatfs() for (NFS) file handles.
681  */
682 #ifndef _SYS_SYSPROTO_H_
683 struct freebsd4_fhstatfs_args {
684 	struct fhandle *u_fhp;
685 	struct ostatfs *buf;
686 };
687 #endif
688 int
freebsd4_fhstatfs(struct thread * td,struct freebsd4_fhstatfs_args * uap)689 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
690 {
691 	struct ostatfs osb;
692 	struct statfs *sfp;
693 	fhandle_t fh;
694 	int error;
695 
696 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
697 	if (error != 0)
698 		return (error);
699 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
700 	error = kern_fhstatfs(td, fh, sfp);
701 	if (error == 0) {
702 		freebsd4_cvtstatfs(sfp, &osb);
703 		error = copyout(&osb, uap->buf, sizeof(osb));
704 	}
705 	free(sfp, M_STATFS);
706 	return (error);
707 }
708 
709 /*
710  * Convert a new format statfs structure to an old format statfs structure.
711  */
712 static void
freebsd4_cvtstatfs(struct statfs * nsp,struct ostatfs * osp)713 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
714 {
715 
716 	statfs_scale_blocks(nsp, LONG_MAX);
717 	bzero(osp, sizeof(*osp));
718 	osp->f_bsize = nsp->f_bsize;
719 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
720 	osp->f_blocks = nsp->f_blocks;
721 	osp->f_bfree = nsp->f_bfree;
722 	osp->f_bavail = nsp->f_bavail;
723 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
724 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
725 	osp->f_owner = nsp->f_owner;
726 	osp->f_type = nsp->f_type;
727 	osp->f_flags = nsp->f_flags;
728 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
729 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
730 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
731 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
732 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
733 	    MIN(MFSNAMELEN, OMFSNAMELEN));
734 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
735 	    MIN(MNAMELEN, OMNAMELEN));
736 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
737 	    MIN(MNAMELEN, OMNAMELEN));
738 	osp->f_fsid = nsp->f_fsid;
739 }
740 #endif /* COMPAT_FREEBSD4 */
741 
742 #if defined(COMPAT_FREEBSD11)
743 /*
744  * Get old format filesystem statistics.
745  */
746 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
747 
748 int
freebsd11_statfs(struct thread * td,struct freebsd11_statfs_args * uap)749 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
750 {
751 	struct freebsd11_statfs osb;
752 	struct statfs *sfp;
753 	int error;
754 
755 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
756 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
757 	if (error == 0) {
758 		freebsd11_cvtstatfs(sfp, &osb);
759 		error = copyout(&osb, uap->buf, sizeof(osb));
760 	}
761 	free(sfp, M_STATFS);
762 	return (error);
763 }
764 
765 /*
766  * Get filesystem statistics.
767  */
768 int
freebsd11_fstatfs(struct thread * td,struct freebsd11_fstatfs_args * uap)769 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
770 {
771 	struct freebsd11_statfs osb;
772 	struct statfs *sfp;
773 	int error;
774 
775 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
776 	error = kern_fstatfs(td, uap->fd, sfp);
777 	if (error == 0) {
778 		freebsd11_cvtstatfs(sfp, &osb);
779 		error = copyout(&osb, uap->buf, sizeof(osb));
780 	}
781 	free(sfp, M_STATFS);
782 	return (error);
783 }
784 
785 /*
786  * Get statistics on all filesystems.
787  */
788 int
freebsd11_getfsstat(struct thread * td,struct freebsd11_getfsstat_args * uap)789 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
790 {
791 	return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode));
792 }
793 
794 int
kern_freebsd11_getfsstat(struct thread * td,struct freebsd11_statfs * ubuf,long bufsize,int mode)795 kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf,
796     long bufsize, int mode)
797 {
798 	struct freebsd11_statfs osb;
799 	struct statfs *buf, *sp;
800 	size_t count, size;
801 	int error;
802 
803 	if (bufsize < 0)
804 		return (EINVAL);
805 
806 	count = bufsize / sizeof(struct ostatfs);
807 	size = count * sizeof(struct statfs);
808 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode);
809 	if (error == 0)
810 		td->td_retval[0] = count;
811 	if (size > 0) {
812 		sp = buf;
813 		while (count > 0 && error == 0) {
814 			freebsd11_cvtstatfs(sp, &osb);
815 			error = copyout(&osb, ubuf, sizeof(osb));
816 			sp++;
817 			ubuf++;
818 			count--;
819 		}
820 		free(buf, M_STATFS);
821 	}
822 	return (error);
823 }
824 
825 /*
826  * Implement fstatfs() for (NFS) file handles.
827  */
828 int
freebsd11_fhstatfs(struct thread * td,struct freebsd11_fhstatfs_args * uap)829 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
830 {
831 	struct freebsd11_statfs osb;
832 	struct statfs *sfp;
833 	fhandle_t fh;
834 	int error;
835 
836 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
837 	if (error)
838 		return (error);
839 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
840 	error = kern_fhstatfs(td, fh, sfp);
841 	if (error == 0) {
842 		freebsd11_cvtstatfs(sfp, &osb);
843 		error = copyout(&osb, uap->buf, sizeof(osb));
844 	}
845 	free(sfp, M_STATFS);
846 	return (error);
847 }
848 
849 /*
850  * Convert a new format statfs structure to an old format statfs structure.
851  */
852 static void
freebsd11_cvtstatfs(struct statfs * nsp,struct freebsd11_statfs * osp)853 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
854 {
855 
856 	bzero(osp, sizeof(*osp));
857 	osp->f_version = FREEBSD11_STATFS_VERSION;
858 	osp->f_type = nsp->f_type;
859 	osp->f_flags = nsp->f_flags;
860 	osp->f_bsize = nsp->f_bsize;
861 	osp->f_iosize = nsp->f_iosize;
862 	osp->f_blocks = nsp->f_blocks;
863 	osp->f_bfree = nsp->f_bfree;
864 	osp->f_bavail = nsp->f_bavail;
865 	osp->f_files = nsp->f_files;
866 	osp->f_ffree = nsp->f_ffree;
867 	osp->f_syncwrites = nsp->f_syncwrites;
868 	osp->f_asyncwrites = nsp->f_asyncwrites;
869 	osp->f_syncreads = nsp->f_syncreads;
870 	osp->f_asyncreads = nsp->f_asyncreads;
871 	osp->f_namemax = nsp->f_namemax;
872 	osp->f_owner = nsp->f_owner;
873 	osp->f_fsid = nsp->f_fsid;
874 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
875 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
876 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
877 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
878 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
879 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
880 }
881 #endif /* COMPAT_FREEBSD11 */
882 
883 /*
884  * Change current working directory to a given file descriptor.
885  */
886 #ifndef _SYS_SYSPROTO_H_
887 struct fchdir_args {
888 	int	fd;
889 };
890 #endif
891 int
sys_fchdir(struct thread * td,struct fchdir_args * uap)892 sys_fchdir(struct thread *td, struct fchdir_args *uap)
893 {
894 	struct vnode *vp, *tdp;
895 	struct mount *mp;
896 	struct file *fp;
897 	int error;
898 	uint8_t fdflags;
899 
900 	AUDIT_ARG_FD(uap->fd);
901 	error = getvnode_path(td, uap->fd, &cap_fchdir_rights, &fdflags,
902 	    &fp);
903 	if (error != 0)
904 		return (error);
905 	if ((fdflags & UF_RESOLVE_BENEATH) != 0) {
906 		fdrop(fp, td);
907 		return (ENOTCAPABLE);
908 	}
909 	vp = fp->f_vnode;
910 	vrefact(vp);
911 	fdrop(fp, td);
912 	vn_lock(vp, LK_SHARED | LK_RETRY);
913 	AUDIT_ARG_VNODE1(vp);
914 	error = change_dir(vp, td);
915 	while (!error && (mp = vp->v_mountedhere) != NULL) {
916 		if (vfs_busy(mp, 0))
917 			continue;
918 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
919 		vfs_unbusy(mp);
920 		if (error != 0)
921 			break;
922 		vput(vp);
923 		vp = tdp;
924 	}
925 	if (error != 0) {
926 		vput(vp);
927 		return (error);
928 	}
929 	VOP_UNLOCK(vp);
930 	pwd_chdir(td, vp);
931 	return (0);
932 }
933 
934 /*
935  * Change current working directory (``.'').
936  */
937 #ifndef _SYS_SYSPROTO_H_
938 struct chdir_args {
939 	char	*path;
940 };
941 #endif
942 int
sys_chdir(struct thread * td,struct chdir_args * uap)943 sys_chdir(struct thread *td, struct chdir_args *uap)
944 {
945 
946 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
947 }
948 
949 int
kern_chdir(struct thread * td,const char * path,enum uio_seg pathseg)950 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
951 {
952 	struct nameidata nd;
953 	int error;
954 
955 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
956 	    pathseg, path);
957 	if ((error = namei(&nd)) != 0)
958 		return (error);
959 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
960 		vput(nd.ni_vp);
961 		NDFREE_PNBUF(&nd);
962 		return (error);
963 	}
964 	VOP_UNLOCK(nd.ni_vp);
965 	NDFREE_PNBUF(&nd);
966 	pwd_chdir(td, nd.ni_vp);
967 	return (0);
968 }
969 
970 static int unprivileged_chroot = 0;
971 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
972     &unprivileged_chroot, 0,
973     "Unprivileged processes can use chroot(2)");
974 
975 /*
976  * Takes locked vnode, unlocks it before returning.
977  */
978 static int
kern_chroot(struct thread * td,struct vnode * vp)979 kern_chroot(struct thread *td, struct vnode *vp)
980 {
981 	struct proc *p;
982 	int error;
983 
984 	error = priv_check(td, PRIV_VFS_CHROOT);
985 	if (error != 0) {
986 		p = td->td_proc;
987 		if (unprivileged_chroot == 0) {
988 			error = EXTERROR(EPERM,
989 		    "security.bsd.unprivileged_chroot sysctl not enabled");
990 			goto e_vunlock;
991 		}
992 		if ((p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
993 			error = EXTERROR(EPERM,
994 			    "PROC_NO_NEW_PRIVS not enabled");
995 			goto e_vunlock;
996 		}
997 	}
998 
999 	error = change_dir(vp, td);
1000 	if (error != 0)
1001 		goto e_vunlock;
1002 #ifdef MAC
1003 	error = mac_vnode_check_chroot(td->td_ucred, vp);
1004 	if (error != 0)
1005 		goto e_vunlock;
1006 #endif
1007 	VOP_UNLOCK(vp);
1008 	error = pwd_chroot(td, vp);
1009 	vrele(vp);
1010 	return (error);
1011 e_vunlock:
1012 	vput(vp);
1013 	return (error);
1014 }
1015 
1016 /*
1017  * Change notion of root (``/'') directory.
1018  */
1019 #ifndef _SYS_SYSPROTO_H_
1020 struct chroot_args {
1021 	char	*path;
1022 };
1023 #endif
1024 int
sys_chroot(struct thread * td,struct chroot_args * uap)1025 sys_chroot(struct thread *td, struct chroot_args *uap)
1026 {
1027 	struct nameidata nd;
1028 	int error;
1029 
1030 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
1031 	    UIO_USERSPACE, uap->path);
1032 	error = namei(&nd);
1033 	if (error != 0)
1034 		return (error);
1035 	NDFREE_PNBUF(&nd);
1036 	error = kern_chroot(td, nd.ni_vp);
1037 	return (error);
1038 }
1039 
1040 /*
1041  * Change notion of root directory to a given file descriptor.
1042  */
1043 #ifndef _SYS_SYSPROTO_H_
1044 struct fchroot_args {
1045 	int	fd;
1046 };
1047 #endif
1048 int
sys_fchroot(struct thread * td,struct fchroot_args * uap)1049 sys_fchroot(struct thread *td, struct fchroot_args *uap)
1050 {
1051 	struct vnode *vp;
1052 	struct file *fp;
1053 	int error;
1054 	uint8_t fdflags;
1055 
1056 	error = getvnode_path(td, uap->fd, &cap_fchroot_rights, &fdflags, &fp);
1057 	if (error != 0)
1058 		return (error);
1059 	if ((fdflags & UF_RESOLVE_BENEATH) != 0) {
1060 		fdrop(fp, td);
1061 		return (ENOTCAPABLE);
1062 	}
1063 	vp = fp->f_vnode;
1064 	vrefact(vp);
1065 	fdrop(fp, td);
1066 	vn_lock(vp, LK_SHARED | LK_RETRY);
1067 	error = kern_chroot(td, vp);
1068 	return (error);
1069 }
1070 
1071 /*
1072  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1073  * instance.
1074  */
1075 int
change_dir(struct vnode * vp,struct thread * td)1076 change_dir(struct vnode *vp, struct thread *td)
1077 {
1078 #ifdef MAC
1079 	int error;
1080 #endif
1081 
1082 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1083 	if (vp->v_type != VDIR)
1084 		return (ENOTDIR);
1085 #ifdef MAC
1086 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1087 	if (error != 0)
1088 		return (error);
1089 #endif
1090 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1091 }
1092 
1093 static __inline void
flags_to_rights(int flags,cap_rights_t * rightsp)1094 flags_to_rights(int flags, cap_rights_t *rightsp)
1095 {
1096 	if (flags & O_EXEC) {
1097 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1098 		if (flags & O_PATH)
1099 			return;
1100 	} else {
1101 		switch ((flags & O_ACCMODE)) {
1102 		case O_RDONLY:
1103 			cap_rights_set_one(rightsp, CAP_READ);
1104 			break;
1105 		case O_RDWR:
1106 			cap_rights_set_one(rightsp, CAP_READ);
1107 			/* FALLTHROUGH */
1108 		case O_WRONLY:
1109 			cap_rights_set_one(rightsp, CAP_WRITE);
1110 			if (!(flags & (O_APPEND | O_TRUNC)))
1111 				cap_rights_set_one(rightsp, CAP_SEEK);
1112 			break;
1113 		}
1114 	}
1115 
1116 	if (flags & O_CREAT)
1117 		cap_rights_set_one(rightsp, CAP_CREATE);
1118 
1119 	if (flags & O_TRUNC)
1120 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1121 
1122 	if (flags & (O_SYNC | O_FSYNC))
1123 		cap_rights_set_one(rightsp, CAP_FSYNC);
1124 
1125 	if (flags & (O_EXLOCK | O_SHLOCK))
1126 		cap_rights_set_one(rightsp, CAP_FLOCK);
1127 }
1128 
1129 /*
1130  * Check permissions, allocate an open file structure, and call the device
1131  * open routine if any.
1132  */
1133 #ifndef _SYS_SYSPROTO_H_
1134 struct open_args {
1135 	char	*path;
1136 	int	flags;
1137 	int	mode;
1138 };
1139 #endif
1140 int
sys_open(struct thread * td,struct open_args * uap)1141 sys_open(struct thread *td, struct open_args *uap)
1142 {
1143 
1144 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1145 	    uap->flags, uap->mode));
1146 }
1147 
1148 #ifndef _SYS_SYSPROTO_H_
1149 struct openat_args {
1150 	int	fd;
1151 	char	*path;
1152 	int	flag;
1153 	int	mode;
1154 };
1155 #endif
1156 int
sys_openat(struct thread * td,struct openat_args * uap)1157 sys_openat(struct thread *td, struct openat_args *uap)
1158 {
1159 
1160 	AUDIT_ARG_FD(uap->fd);
1161 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1162 	    uap->mode));
1163 }
1164 
1165 /*
1166  * Validate open(2) flags and convert access mode flags (O_RDONLY etc.) to their
1167  * in-kernel representations (FREAD etc.).
1168  */
1169 static int
openflags(int * flagsp)1170 openflags(int *flagsp)
1171 {
1172 	int flags;
1173 
1174 	/*
1175 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1176 	 * may be specified.  On the other hand, for O_PATH any mode
1177 	 * except O_EXEC is ignored.
1178 	 */
1179 	flags = *flagsp;
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 out;
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 #ifdef MAC
1898 out2:
1899 #endif
1900 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1901 	vn_finished_write(mp);
1902 	NDFREE_PNBUF(&nd);
1903 	if (error == ERELOOKUP)
1904 		goto restart;
1905 out:
1906 	if (segflg != UIO_SYSSPACE)
1907 		uma_zfree(namei_zone, tmppath);
1908 	return (error);
1909 }
1910 
1911 /*
1912  * Delete a whiteout from the filesystem.
1913  */
1914 #ifndef _SYS_SYSPROTO_H_
1915 struct undelete_args {
1916 	char *path;
1917 };
1918 #endif
1919 int
sys_undelete(struct thread * td,struct undelete_args * uap)1920 sys_undelete(struct thread *td, struct undelete_args *uap)
1921 {
1922 	struct mount *mp;
1923 	struct nameidata nd;
1924 	int error;
1925 
1926 	NDPREINIT(&nd);
1927 restart:
1928 	bwillwrite();
1929 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1930 	    UIO_USERSPACE, uap->path);
1931 	error = namei(&nd);
1932 	if (error != 0)
1933 		return (error);
1934 
1935 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1936 		NDFREE_PNBUF(&nd);
1937 		if (nd.ni_vp == nd.ni_dvp)
1938 			vrele(nd.ni_dvp);
1939 		else
1940 			vput(nd.ni_dvp);
1941 		if (nd.ni_vp)
1942 			vrele(nd.ni_vp);
1943 		return (EEXIST);
1944 	}
1945 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1946 		NDFREE_PNBUF(&nd);
1947 		vput(nd.ni_dvp);
1948 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1949 			return (error);
1950 		goto restart;
1951 	}
1952 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1953 	NDFREE_PNBUF(&nd);
1954 	vput(nd.ni_dvp);
1955 	vn_finished_write(mp);
1956 	if (error == ERELOOKUP)
1957 		goto restart;
1958 	return (error);
1959 }
1960 
1961 /*
1962  * Delete a name from the filesystem.
1963  */
1964 #ifndef _SYS_SYSPROTO_H_
1965 struct unlink_args {
1966 	char	*path;
1967 };
1968 #endif
1969 int
sys_unlink(struct thread * td,struct unlink_args * uap)1970 sys_unlink(struct thread *td, struct unlink_args *uap)
1971 {
1972 
1973 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1974 	    0, 0));
1975 }
1976 
1977 static int
kern_funlinkat_ex(struct thread * td,int dfd,const char * path,int fd,int flag,enum uio_seg pathseg,ino_t oldinum)1978 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1979     int flag, enum uio_seg pathseg, ino_t oldinum)
1980 {
1981 
1982 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1983 		return (EINVAL);
1984 
1985 	if ((flag & AT_REMOVEDIR) != 0)
1986 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1987 
1988 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1989 }
1990 
1991 #ifndef _SYS_SYSPROTO_H_
1992 struct unlinkat_args {
1993 	int	fd;
1994 	char	*path;
1995 	int	flag;
1996 };
1997 #endif
1998 int
sys_unlinkat(struct thread * td,struct unlinkat_args * uap)1999 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
2000 {
2001 
2002 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
2003 	    UIO_USERSPACE, 0));
2004 }
2005 
2006 #ifndef _SYS_SYSPROTO_H_
2007 struct funlinkat_args {
2008 	int		dfd;
2009 	const char	*path;
2010 	int		fd;
2011 	int		flag;
2012 };
2013 #endif
2014 int
sys_funlinkat(struct thread * td,struct funlinkat_args * uap)2015 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
2016 {
2017 
2018 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
2019 	    UIO_USERSPACE, 0));
2020 }
2021 
2022 int
kern_funlinkat(struct thread * td,int dfd,const char * path,int fd,enum uio_seg pathseg,int flag,ino_t oldinum)2023 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
2024     enum uio_seg pathseg, int flag, ino_t oldinum)
2025 {
2026 	struct mount *mp;
2027 	struct file *fp;
2028 	struct vnode *vp;
2029 	struct nameidata nd;
2030 	struct stat sb;
2031 	int error;
2032 
2033 	fp = NULL;
2034 	if (fd != FD_NONE) {
2035 		error = getvnode_path(td, fd, &cap_no_rights, NULL, &fp);
2036 		if (error != 0)
2037 			return (error);
2038 	}
2039 
2040 	NDPREINIT(&nd);
2041 restart:
2042 	bwillwrite();
2043 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
2044 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
2045 	    pathseg, path, dfd, &cap_unlinkat_rights);
2046 	if ((error = namei(&nd)) != 0) {
2047 		if (error == EINVAL)
2048 			error = EPERM;
2049 		goto fdout;
2050 	}
2051 	vp = nd.ni_vp;
2052 	if (vp->v_type == VDIR && oldinum == 0) {
2053 		error = EPERM;		/* POSIX */
2054 	} else if (oldinum != 0 &&
2055 	    ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) &&
2056 	    sb.st_ino != oldinum) {
2057 		error = EIDRM;	/* Identifier removed */
2058 	} else if (fp != NULL && fp->f_vnode != vp) {
2059 		if (VN_IS_DOOMED(fp->f_vnode))
2060 			error = EBADF;
2061 		else
2062 			error = EDEADLK;
2063 	} else {
2064 		/*
2065 		 * The root of a mounted filesystem cannot be deleted.
2066 		 *
2067 		 * XXX: can this only be a VDIR case?
2068 		 */
2069 		if (vp->v_vflag & VV_ROOT)
2070 			error = EBUSY;
2071 	}
2072 	if (error == 0) {
2073 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
2074 			NDFREE_PNBUF(&nd);
2075 			vput(nd.ni_dvp);
2076 			if (vp == nd.ni_dvp)
2077 				vrele(vp);
2078 			else
2079 				vput(vp);
2080 			if ((error = vn_start_write(NULL, &mp,
2081 			    V_XSLEEP | V_PCATCH)) != 0) {
2082 				goto fdout;
2083 			}
2084 			goto restart;
2085 		}
2086 #ifdef MAC
2087 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
2088 		    &nd.ni_cnd);
2089 		if (error != 0)
2090 			goto out;
2091 #endif
2092 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
2093 #ifdef MAC
2094 out:
2095 #endif
2096 		vn_finished_write(mp);
2097 	}
2098 	NDFREE_PNBUF(&nd);
2099 	vput(nd.ni_dvp);
2100 	if (vp == nd.ni_dvp)
2101 		vrele(vp);
2102 	else
2103 		vput(vp);
2104 	if (error == ERELOOKUP)
2105 		goto restart;
2106 fdout:
2107 	if (fp != NULL)
2108 		fdrop(fp, td);
2109 	return (error);
2110 }
2111 
2112 /*
2113  * Reposition read/write file offset.
2114  */
2115 #ifndef _SYS_SYSPROTO_H_
2116 struct lseek_args {
2117 	int	fd;
2118 	int	pad;
2119 	off_t	offset;
2120 	int	whence;
2121 };
2122 #endif
2123 int
sys_lseek(struct thread * td,struct lseek_args * uap)2124 sys_lseek(struct thread *td, struct lseek_args *uap)
2125 {
2126 
2127 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2128 }
2129 
2130 int
kern_lseek(struct thread * td,int fd,off_t offset,int whence)2131 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2132 {
2133 	struct file *fp;
2134 	int error;
2135 
2136 	AUDIT_ARG_FD(fd);
2137 	error = fget(td, fd, &cap_seek_rights, &fp);
2138 	if (error != 0)
2139 		return (error);
2140 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2141 	    fo_seek(fp, offset, whence, td) : ESPIPE;
2142 	fdrop(fp, td);
2143 	return (error);
2144 }
2145 
2146 #if defined(COMPAT_43)
2147 /*
2148  * Reposition read/write file offset.
2149  */
2150 #ifndef _SYS_SYSPROTO_H_
2151 struct olseek_args {
2152 	int	fd;
2153 	long	offset;
2154 	int	whence;
2155 };
2156 #endif
2157 int
olseek(struct thread * td,struct olseek_args * uap)2158 olseek(struct thread *td, struct olseek_args *uap)
2159 {
2160 
2161 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2162 }
2163 #endif /* COMPAT_43 */
2164 
2165 #if defined(COMPAT_FREEBSD6)
2166 /* Version with the 'pad' argument */
2167 int
freebsd6_lseek(struct thread * td,struct freebsd6_lseek_args * uap)2168 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2169 {
2170 
2171 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2172 }
2173 #endif
2174 
2175 /*
2176  * Check access permissions using passed credentials.
2177  */
2178 static int
vn_access(struct vnode * vp,int user_flags,struct ucred * cred,struct thread * td)2179 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2180      struct thread *td)
2181 {
2182 	accmode_t accmode;
2183 	int error;
2184 
2185 	/* Flags == 0 means only check for existence. */
2186 	if (user_flags == 0)
2187 		return (0);
2188 
2189 	accmode = 0;
2190 	if (user_flags & R_OK)
2191 		accmode |= VREAD;
2192 	if (user_flags & W_OK)
2193 		accmode |= VWRITE;
2194 	if (user_flags & X_OK)
2195 		accmode |= VEXEC;
2196 #ifdef MAC
2197 	error = mac_vnode_check_access(cred, vp, accmode);
2198 	if (error != 0)
2199 		return (error);
2200 #endif
2201 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2202 		error = VOP_ACCESS(vp, accmode, cred, td);
2203 	return (error);
2204 }
2205 
2206 /*
2207  * Check access permissions using "real" credentials.
2208  */
2209 #ifndef _SYS_SYSPROTO_H_
2210 struct access_args {
2211 	char	*path;
2212 	int	amode;
2213 };
2214 #endif
2215 int
sys_access(struct thread * td,struct access_args * uap)2216 sys_access(struct thread *td, struct access_args *uap)
2217 {
2218 
2219 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2220 	    0, uap->amode));
2221 }
2222 
2223 #ifndef _SYS_SYSPROTO_H_
2224 struct faccessat_args {
2225 	int	dirfd;
2226 	char	*path;
2227 	int	amode;
2228 	int	flag;
2229 }
2230 #endif
2231 int
sys_faccessat(struct thread * td,struct faccessat_args * uap)2232 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2233 {
2234 
2235 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2236 	    uap->amode));
2237 }
2238 
2239 int
kern_accessat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int flag,int amode)2240 kern_accessat(struct thread *td, int fd, const char *path,
2241     enum uio_seg pathseg, int flag, int amode)
2242 {
2243 	struct ucred *cred, *usecred;
2244 	struct vnode *vp;
2245 	struct nameidata nd;
2246 	int error;
2247 
2248 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH |
2249 	    AT_SYMLINK_NOFOLLOW)) != 0)
2250 		return (EINVAL);
2251 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2252 		return (EINVAL);
2253 
2254 	/*
2255 	 * Create and modify a temporary credential instead of one that
2256 	 * is potentially shared (if we need one).
2257 	 */
2258 	cred = td->td_ucred;
2259 	if ((flag & AT_EACCESS) == 0 &&
2260 	    ((cred->cr_uid != cred->cr_ruid ||
2261 	    cred->cr_rgid != cred->cr_gid))) {
2262 		usecred = crdup(cred);
2263 		usecred->cr_uid = cred->cr_ruid;
2264 		usecred->cr_gid = cred->cr_rgid;
2265 		td->td_ucred = usecred;
2266 	} else
2267 		usecred = cred;
2268 	AUDIT_ARG_VALUE(amode);
2269 	NDINIT_ATRIGHTS(&nd, LOOKUP, LOCKSHARED | LOCKLEAF |
2270 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW |
2271 	    AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights);
2272 	if ((error = namei(&nd)) != 0)
2273 		goto out;
2274 	vp = nd.ni_vp;
2275 
2276 	error = vn_access(vp, amode, usecred, td);
2277 	NDFREE_PNBUF(&nd);
2278 	vput(vp);
2279 out:
2280 	if (usecred != cred) {
2281 		td->td_ucred = cred;
2282 		crfree(usecred);
2283 	}
2284 	return (error);
2285 }
2286 
2287 /*
2288  * Check access permissions using "effective" credentials.
2289  */
2290 #ifndef _SYS_SYSPROTO_H_
2291 struct eaccess_args {
2292 	char	*path;
2293 	int	amode;
2294 };
2295 #endif
2296 int
sys_eaccess(struct thread * td,struct eaccess_args * uap)2297 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2298 {
2299 
2300 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2301 	    AT_EACCESS, uap->amode));
2302 }
2303 
2304 #if defined(COMPAT_43)
2305 /*
2306  * Get file status; this version follows links.
2307  */
2308 #ifndef _SYS_SYSPROTO_H_
2309 struct ostat_args {
2310 	char	*path;
2311 	struct ostat *ub;
2312 };
2313 #endif
2314 int
ostat(struct thread * td,struct ostat_args * uap)2315 ostat(struct thread *td, struct ostat_args *uap)
2316 {
2317 	struct stat sb;
2318 	struct ostat osb;
2319 	int error;
2320 
2321 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2322 	if (error != 0)
2323 		return (error);
2324 	cvtstat(&sb, &osb);
2325 	return (copyout(&osb, uap->ub, sizeof (osb)));
2326 }
2327 
2328 /*
2329  * Get file status; this version does not follow links.
2330  */
2331 #ifndef _SYS_SYSPROTO_H_
2332 struct olstat_args {
2333 	char	*path;
2334 	struct ostat *ub;
2335 };
2336 #endif
2337 int
olstat(struct thread * td,struct olstat_args * uap)2338 olstat(struct thread *td, struct olstat_args *uap)
2339 {
2340 	struct stat sb;
2341 	struct ostat osb;
2342 	int error;
2343 
2344 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2345 	    UIO_USERSPACE, &sb);
2346 	if (error != 0)
2347 		return (error);
2348 	cvtstat(&sb, &osb);
2349 	return (copyout(&osb, uap->ub, sizeof (osb)));
2350 }
2351 
2352 /*
2353  * Convert from an old to a new stat structure.
2354  * XXX: many values are blindly truncated.
2355  */
2356 void
cvtstat(struct stat * st,struct ostat * ost)2357 cvtstat(struct stat *st, struct ostat *ost)
2358 {
2359 
2360 	bzero(ost, sizeof(*ost));
2361 	ost->st_dev = st->st_dev;
2362 	ost->st_ino = st->st_ino;
2363 	ost->st_mode = st->st_mode;
2364 	ost->st_nlink = st->st_nlink;
2365 	ost->st_uid = st->st_uid;
2366 	ost->st_gid = st->st_gid;
2367 	ost->st_rdev = st->st_rdev;
2368 	ost->st_size = MIN(st->st_size, INT32_MAX);
2369 	ost->st_atim = st->st_atim;
2370 	ost->st_mtim = st->st_mtim;
2371 	ost->st_ctim = st->st_ctim;
2372 	ost->st_blksize = st->st_blksize;
2373 	ost->st_blocks = st->st_blocks;
2374 	ost->st_flags = st->st_flags;
2375 	ost->st_gen = st->st_gen;
2376 }
2377 #endif /* COMPAT_43 */
2378 
2379 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2380 int ino64_trunc_error;
2381 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2382     &ino64_trunc_error, 0,
2383     "Error on truncation of device, file or inode number, or link count");
2384 
2385 int
freebsd11_cvtstat(struct stat * st,struct freebsd11_stat * ost)2386 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2387 {
2388 
2389 	ost->st_dev = st->st_dev;
2390 	if (ost->st_dev != st->st_dev) {
2391 		switch (ino64_trunc_error) {
2392 		default:
2393 			/*
2394 			 * Since dev_t is almost raw, don't clamp to the
2395 			 * maximum for case 2, but ignore the error.
2396 			 */
2397 			break;
2398 		case 1:
2399 			return (EOVERFLOW);
2400 		}
2401 	}
2402 	ost->st_ino = st->st_ino;
2403 	if (ost->st_ino != st->st_ino) {
2404 		switch (ino64_trunc_error) {
2405 		default:
2406 		case 0:
2407 			break;
2408 		case 1:
2409 			return (EOVERFLOW);
2410 		case 2:
2411 			ost->st_ino = UINT32_MAX;
2412 			break;
2413 		}
2414 	}
2415 	ost->st_mode = st->st_mode;
2416 	ost->st_nlink = st->st_nlink;
2417 	if (ost->st_nlink != st->st_nlink) {
2418 		switch (ino64_trunc_error) {
2419 		default:
2420 		case 0:
2421 			break;
2422 		case 1:
2423 			return (EOVERFLOW);
2424 		case 2:
2425 			ost->st_nlink = UINT16_MAX;
2426 			break;
2427 		}
2428 	}
2429 	ost->st_uid = st->st_uid;
2430 	ost->st_gid = st->st_gid;
2431 	ost->st_rdev = st->st_rdev;
2432 	if (ost->st_rdev != st->st_rdev) {
2433 		switch (ino64_trunc_error) {
2434 		default:
2435 			break;
2436 		case 1:
2437 			return (EOVERFLOW);
2438 		}
2439 	}
2440 	ost->st_atim = st->st_atim;
2441 	ost->st_mtim = st->st_mtim;
2442 	ost->st_ctim = st->st_ctim;
2443 	ost->st_size = st->st_size;
2444 	ost->st_blocks = st->st_blocks;
2445 	ost->st_blksize = st->st_blksize;
2446 	ost->st_flags = st->st_flags;
2447 	ost->st_gen = st->st_gen;
2448 	ost->st_lspare = 0;
2449 	ost->st_birthtim = st->st_birthtim;
2450 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2451 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2452 	    st_birthtim) - sizeof(ost->st_birthtim));
2453 	return (0);
2454 }
2455 
2456 int
freebsd11_stat(struct thread * td,struct freebsd11_stat_args * uap)2457 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2458 {
2459 	struct stat sb;
2460 	struct freebsd11_stat osb;
2461 	int error;
2462 
2463 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2464 	if (error != 0)
2465 		return (error);
2466 	error = freebsd11_cvtstat(&sb, &osb);
2467 	if (error == 0)
2468 		error = copyout(&osb, uap->ub, sizeof(osb));
2469 	return (error);
2470 }
2471 
2472 int
freebsd11_lstat(struct thread * td,struct freebsd11_lstat_args * uap)2473 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2474 {
2475 	struct stat sb;
2476 	struct freebsd11_stat osb;
2477 	int error;
2478 
2479 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2480 	    UIO_USERSPACE, &sb);
2481 	if (error != 0)
2482 		return (error);
2483 	error = freebsd11_cvtstat(&sb, &osb);
2484 	if (error == 0)
2485 		error = copyout(&osb, uap->ub, sizeof(osb));
2486 	return (error);
2487 }
2488 
2489 int
freebsd11_fhstat(struct thread * td,struct freebsd11_fhstat_args * uap)2490 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2491 {
2492 	struct fhandle fh;
2493 	struct stat sb;
2494 	struct freebsd11_stat osb;
2495 	int error;
2496 
2497 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2498 	if (error != 0)
2499 		return (error);
2500 	error = kern_fhstat(td, fh, &sb);
2501 	if (error != 0)
2502 		return (error);
2503 	error = freebsd11_cvtstat(&sb, &osb);
2504 	if (error == 0)
2505 		error = copyout(&osb, uap->sb, sizeof(osb));
2506 	return (error);
2507 }
2508 
2509 int
freebsd11_fstatat(struct thread * td,struct freebsd11_fstatat_args * uap)2510 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2511 {
2512 	struct stat sb;
2513 	struct freebsd11_stat osb;
2514 	int error;
2515 
2516 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2517 	    UIO_USERSPACE, &sb);
2518 	if (error != 0)
2519 		return (error);
2520 	error = freebsd11_cvtstat(&sb, &osb);
2521 	if (error == 0)
2522 		error = copyout(&osb, uap->buf, sizeof(osb));
2523 	return (error);
2524 }
2525 #endif	/* COMPAT_FREEBSD11 */
2526 
2527 /*
2528  * Get file status
2529  */
2530 #ifndef _SYS_SYSPROTO_H_
2531 struct fstatat_args {
2532 	int	fd;
2533 	char	*path;
2534 	struct stat	*buf;
2535 	int	flag;
2536 }
2537 #endif
2538 int
sys_fstatat(struct thread * td,struct fstatat_args * uap)2539 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2540 {
2541 	struct stat sb;
2542 	int error;
2543 
2544 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2545 	    UIO_USERSPACE, &sb);
2546 	if (error == 0)
2547 		error = copyout(&sb, uap->buf, sizeof (sb));
2548 	return (error);
2549 }
2550 
2551 int
kern_statat(struct thread * td,int flag,int fd,const char * path,enum uio_seg pathseg,struct stat * sbp)2552 kern_statat(struct thread *td, int flag, int fd, const char *path,
2553     enum uio_seg pathseg, struct stat *sbp)
2554 {
2555 	struct nameidata nd;
2556 	int error;
2557 
2558 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2559 	    AT_EMPTY_PATH)) != 0)
2560 		return (EINVAL);
2561 
2562 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2563 	    AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2564 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights);
2565 
2566 	if ((error = namei(&nd)) != 0) {
2567 		if (error == ENOTDIR &&
2568 		    (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2569 			error = kern_fstat(td, fd, sbp);
2570 		return (error);
2571 	}
2572 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED);
2573 	NDFREE_PNBUF(&nd);
2574 	vput(nd.ni_vp);
2575 #ifdef __STAT_TIME_T_EXT
2576 	sbp->st_atim_ext = 0;
2577 	sbp->st_mtim_ext = 0;
2578 	sbp->st_ctim_ext = 0;
2579 	sbp->st_btim_ext = 0;
2580 #endif
2581 #ifdef KTRACE
2582 	if (KTRPOINT(td, KTR_STRUCT))
2583 		ktrstat_error(sbp, error);
2584 #endif
2585 	return (error);
2586 }
2587 
2588 #if defined(COMPAT_FREEBSD11)
2589 /*
2590  * Implementation of the NetBSD [l]stat() functions.
2591  */
2592 int
freebsd11_cvtnstat(struct stat * sb,struct nstat * nsb)2593 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2594 {
2595 	struct freebsd11_stat sb11;
2596 	int error;
2597 
2598 	error = freebsd11_cvtstat(sb, &sb11);
2599 	if (error != 0)
2600 		return (error);
2601 
2602 	bzero(nsb, sizeof(*nsb));
2603 	CP(sb11, *nsb, st_dev);
2604 	CP(sb11, *nsb, st_ino);
2605 	CP(sb11, *nsb, st_mode);
2606 	CP(sb11, *nsb, st_nlink);
2607 	CP(sb11, *nsb, st_uid);
2608 	CP(sb11, *nsb, st_gid);
2609 	CP(sb11, *nsb, st_rdev);
2610 	CP(sb11, *nsb, st_atim);
2611 	CP(sb11, *nsb, st_mtim);
2612 	CP(sb11, *nsb, st_ctim);
2613 	CP(sb11, *nsb, st_size);
2614 	CP(sb11, *nsb, st_blocks);
2615 	CP(sb11, *nsb, st_blksize);
2616 	CP(sb11, *nsb, st_flags);
2617 	CP(sb11, *nsb, st_gen);
2618 	CP(sb11, *nsb, st_birthtim);
2619 	return (0);
2620 }
2621 
2622 #ifndef _SYS_SYSPROTO_H_
2623 struct freebsd11_nstat_args {
2624 	char	*path;
2625 	struct nstat *ub;
2626 };
2627 #endif
2628 int
freebsd11_nstat(struct thread * td,struct freebsd11_nstat_args * uap)2629 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2630 {
2631 	struct stat sb;
2632 	struct nstat nsb;
2633 	int error;
2634 
2635 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2636 	if (error != 0)
2637 		return (error);
2638 	error = freebsd11_cvtnstat(&sb, &nsb);
2639 	if (error == 0)
2640 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2641 	return (error);
2642 }
2643 
2644 /*
2645  * NetBSD lstat.  Get file status; this version does not follow links.
2646  */
2647 #ifndef _SYS_SYSPROTO_H_
2648 struct freebsd11_nlstat_args {
2649 	char	*path;
2650 	struct nstat *ub;
2651 };
2652 #endif
2653 int
freebsd11_nlstat(struct thread * td,struct freebsd11_nlstat_args * uap)2654 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2655 {
2656 	struct stat sb;
2657 	struct nstat nsb;
2658 	int error;
2659 
2660 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2661 	    UIO_USERSPACE, &sb);
2662 	if (error != 0)
2663 		return (error);
2664 	error = freebsd11_cvtnstat(&sb, &nsb);
2665 	if (error == 0)
2666 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2667 	return (error);
2668 }
2669 #endif /* COMPAT_FREEBSD11 */
2670 
2671 /*
2672  * Get configurable pathname variables.
2673  */
2674 #ifndef _SYS_SYSPROTO_H_
2675 struct pathconf_args {
2676 	char	*path;
2677 	int	name;
2678 };
2679 #endif
2680 int
sys_pathconf(struct thread * td,struct pathconf_args * uap)2681 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2682 {
2683 	long value;
2684 	int error;
2685 
2686 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2687 	    &value);
2688 	if (error == 0)
2689 		td->td_retval[0] = value;
2690 	return (error);
2691 }
2692 
2693 #ifndef _SYS_SYSPROTO_H_
2694 struct lpathconf_args {
2695 	char	*path;
2696 	int	name;
2697 };
2698 #endif
2699 int
sys_lpathconf(struct thread * td,struct lpathconf_args * uap)2700 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2701 {
2702 	long value;
2703 	int error;
2704 
2705 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2706 	    NOFOLLOW, &value);
2707 	if (error == 0)
2708 		td->td_retval[0] = value;
2709 	return (error);
2710 }
2711 
2712 int
kern_pathconf(struct thread * td,const char * path,enum uio_seg pathseg,int name,u_long flags,long * valuep)2713 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2714     int name, u_long flags, long *valuep)
2715 {
2716 	struct nameidata nd;
2717 	int error;
2718 
2719 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2720 	       pathseg, path);
2721 	if ((error = namei(&nd)) != 0)
2722 		return (error);
2723 	NDFREE_PNBUF(&nd);
2724 
2725 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2726 	vput(nd.ni_vp);
2727 	return (error);
2728 }
2729 
2730 /*
2731  * Return target name of a symbolic link.
2732  */
2733 #ifndef _SYS_SYSPROTO_H_
2734 struct readlink_args {
2735 	char	*path;
2736 	char	*buf;
2737 	size_t	count;
2738 };
2739 #endif
2740 int
sys_readlink(struct thread * td,struct readlink_args * uap)2741 sys_readlink(struct thread *td, struct readlink_args *uap)
2742 {
2743 
2744 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2745 	    uap->buf, UIO_USERSPACE, uap->count));
2746 }
2747 #ifndef _SYS_SYSPROTO_H_
2748 struct readlinkat_args {
2749 	int	fd;
2750 	char	*path;
2751 	char	*buf;
2752 	size_t	bufsize;
2753 };
2754 #endif
2755 int
sys_readlinkat(struct thread * td,struct readlinkat_args * uap)2756 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2757 {
2758 
2759 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2760 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2761 }
2762 
2763 int
kern_readlinkat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,char * buf,enum uio_seg bufseg,size_t count)2764 kern_readlinkat(struct thread *td, int fd, const char *path,
2765     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2766 {
2767 	struct vnode *vp;
2768 	struct nameidata nd;
2769 	int error;
2770 
2771 	if (count > IOSIZE_MAX)
2772 		return (EINVAL);
2773 
2774 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2775 	    EMPTYPATH, pathseg, path, fd);
2776 
2777 	if ((error = namei(&nd)) != 0)
2778 		return (error);
2779 	NDFREE_PNBUF(&nd);
2780 	vp = nd.ni_vp;
2781 
2782 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2783 	vput(vp);
2784 
2785 	return (error);
2786 }
2787 
2788 /*
2789  * Helper function to readlink from a vnode
2790  */
2791 static int
kern_readlink_vp(struct vnode * vp,char * buf,enum uio_seg bufseg,size_t count,struct thread * td)2792 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2793     struct thread *td)
2794 {
2795 	struct iovec aiov;
2796 	struct uio auio;
2797 	int error;
2798 
2799 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2800 #ifdef MAC
2801 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2802 	if (error != 0)
2803 		return (error);
2804 #endif
2805 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2806 		return (EINVAL);
2807 
2808 	aiov.iov_base = buf;
2809 	aiov.iov_len = count;
2810 	auio.uio_iov = &aiov;
2811 	auio.uio_iovcnt = 1;
2812 	auio.uio_offset = 0;
2813 	auio.uio_rw = UIO_READ;
2814 	auio.uio_segflg = bufseg;
2815 	auio.uio_td = td;
2816 	auio.uio_resid = count;
2817 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2818 	td->td_retval[0] = count - auio.uio_resid;
2819 	return (error);
2820 }
2821 
2822 /*
2823  * Common implementation code for chflags() and fchflags().
2824  */
2825 static int
setfflags(struct thread * td,struct vnode * vp,u_long flags)2826 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2827 {
2828 	struct mount *mp;
2829 	struct vattr vattr;
2830 	int error;
2831 
2832 	/* We can't support the value matching VNOVAL. */
2833 	if (flags == VNOVAL)
2834 		return (EOPNOTSUPP);
2835 
2836 	/*
2837 	 * Prevent non-root users from setting flags on devices.  When
2838 	 * a device is reused, users can retain ownership of the device
2839 	 * if they are allowed to set flags and programs assume that
2840 	 * chown can't fail when done as root.
2841 	 */
2842 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2843 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2844 		if (error != 0)
2845 			return (error);
2846 	}
2847 
2848 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2849 		return (error);
2850 	VATTR_NULL(&vattr);
2851 	vattr.va_flags = flags;
2852 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2853 #ifdef MAC
2854 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2855 	if (error == 0)
2856 #endif
2857 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2858 	VOP_UNLOCK(vp);
2859 	vn_finished_write(mp);
2860 	return (error);
2861 }
2862 
2863 /*
2864  * Change flags of a file given a path name.
2865  */
2866 #ifndef _SYS_SYSPROTO_H_
2867 struct chflags_args {
2868 	const char *path;
2869 	u_long	flags;
2870 };
2871 #endif
2872 int
sys_chflags(struct thread * td,struct chflags_args * uap)2873 sys_chflags(struct thread *td, struct chflags_args *uap)
2874 {
2875 
2876 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2877 	    uap->flags, 0));
2878 }
2879 
2880 #ifndef _SYS_SYSPROTO_H_
2881 struct chflagsat_args {
2882 	int	fd;
2883 	const char *path;
2884 	u_long	flags;
2885 	int	atflag;
2886 }
2887 #endif
2888 int
sys_chflagsat(struct thread * td,struct chflagsat_args * uap)2889 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2890 {
2891 
2892 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2893 	    uap->flags, uap->atflag));
2894 }
2895 
2896 /*
2897  * Same as chflags() but doesn't follow symlinks.
2898  */
2899 #ifndef _SYS_SYSPROTO_H_
2900 struct lchflags_args {
2901 	const char *path;
2902 	u_long flags;
2903 };
2904 #endif
2905 int
sys_lchflags(struct thread * td,struct lchflags_args * uap)2906 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2907 {
2908 
2909 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2910 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2911 }
2912 
2913 static int
kern_chflagsat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,u_long flags,int atflag)2914 kern_chflagsat(struct thread *td, int fd, const char *path,
2915     enum uio_seg pathseg, u_long flags, int atflag)
2916 {
2917 	struct nameidata nd;
2918 	int error;
2919 
2920 	if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2921 	    AT_EMPTY_PATH)) != 0)
2922 		return (EINVAL);
2923 
2924 	AUDIT_ARG_FFLAGS(flags);
2925 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2926 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2927 	    fd, &cap_fchflags_rights);
2928 	if ((error = namei(&nd)) != 0)
2929 		return (error);
2930 	NDFREE_PNBUF(&nd);
2931 	error = setfflags(td, nd.ni_vp, flags);
2932 	vrele(nd.ni_vp);
2933 	return (error);
2934 }
2935 
2936 /*
2937  * Change flags of a file given a file descriptor.
2938  */
2939 #ifndef _SYS_SYSPROTO_H_
2940 struct fchflags_args {
2941 	int	fd;
2942 	u_long	flags;
2943 };
2944 #endif
2945 int
sys_fchflags(struct thread * td,struct fchflags_args * uap)2946 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2947 {
2948 	struct file *fp;
2949 	int error;
2950 
2951 	AUDIT_ARG_FD(uap->fd);
2952 	AUDIT_ARG_FFLAGS(uap->flags);
2953 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2954 	    &fp);
2955 	if (error != 0)
2956 		return (error);
2957 #ifdef AUDIT
2958 	if (AUDITING_TD(td)) {
2959 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2960 		AUDIT_ARG_VNODE1(fp->f_vnode);
2961 		VOP_UNLOCK(fp->f_vnode);
2962 	}
2963 #endif
2964 	error = setfflags(td, fp->f_vnode, uap->flags);
2965 	fdrop(fp, td);
2966 	return (error);
2967 }
2968 
2969 /*
2970  * Common implementation code for chmod(), lchmod() and fchmod().
2971  */
2972 int
setfmode(struct thread * td,struct ucred * cred,struct vnode * vp,int mode)2973 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2974 {
2975 	struct mount *mp;
2976 	struct vattr vattr;
2977 	int error;
2978 
2979 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2980 		return (error);
2981 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2982 	VATTR_NULL(&vattr);
2983 	vattr.va_mode = mode & ALLPERMS;
2984 #ifdef MAC
2985 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2986 	if (error == 0)
2987 #endif
2988 		error = VOP_SETATTR(vp, &vattr, cred);
2989 	VOP_UNLOCK(vp);
2990 	vn_finished_write(mp);
2991 	return (error);
2992 }
2993 
2994 /*
2995  * Change mode of a file given path name.
2996  */
2997 #ifndef _SYS_SYSPROTO_H_
2998 struct chmod_args {
2999 	char	*path;
3000 	int	mode;
3001 };
3002 #endif
3003 int
sys_chmod(struct thread * td,struct chmod_args * uap)3004 sys_chmod(struct thread *td, struct chmod_args *uap)
3005 {
3006 
3007 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3008 	    uap->mode, 0));
3009 }
3010 
3011 #ifndef _SYS_SYSPROTO_H_
3012 struct fchmodat_args {
3013 	int	dirfd;
3014 	char	*path;
3015 	mode_t	mode;
3016 	int	flag;
3017 }
3018 #endif
3019 int
sys_fchmodat(struct thread * td,struct fchmodat_args * uap)3020 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
3021 {
3022 
3023 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
3024 	    uap->mode, uap->flag));
3025 }
3026 
3027 /*
3028  * Change mode of a file given path name (don't follow links.)
3029  */
3030 #ifndef _SYS_SYSPROTO_H_
3031 struct lchmod_args {
3032 	char	*path;
3033 	int	mode;
3034 };
3035 #endif
3036 int
sys_lchmod(struct thread * td,struct lchmod_args * uap)3037 sys_lchmod(struct thread *td, struct lchmod_args *uap)
3038 {
3039 
3040 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3041 	    uap->mode, AT_SYMLINK_NOFOLLOW));
3042 }
3043 
3044 int
kern_fchmodat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,mode_t mode,int flag)3045 kern_fchmodat(struct thread *td, int fd, const char *path,
3046     enum uio_seg pathseg, mode_t mode, int flag)
3047 {
3048 	struct nameidata nd;
3049 	int error;
3050 
3051 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3052 	    AT_EMPTY_PATH)) != 0)
3053 		return (EINVAL);
3054 
3055 	AUDIT_ARG_MODE(mode);
3056 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3057 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3058 	    fd, &cap_fchmod_rights);
3059 	if ((error = namei(&nd)) != 0)
3060 		return (error);
3061 	NDFREE_PNBUF(&nd);
3062 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
3063 	vrele(nd.ni_vp);
3064 	return (error);
3065 }
3066 
3067 /*
3068  * Change mode of a file given a file descriptor.
3069  */
3070 #ifndef _SYS_SYSPROTO_H_
3071 struct fchmod_args {
3072 	int	fd;
3073 	int	mode;
3074 };
3075 #endif
3076 int
sys_fchmod(struct thread * td,struct fchmod_args * uap)3077 sys_fchmod(struct thread *td, struct fchmod_args *uap)
3078 {
3079 	struct file *fp;
3080 	int error;
3081 
3082 	AUDIT_ARG_FD(uap->fd);
3083 	AUDIT_ARG_MODE(uap->mode);
3084 
3085 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
3086 	if (error != 0)
3087 		return (error);
3088 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
3089 	fdrop(fp, td);
3090 	return (error);
3091 }
3092 
3093 /*
3094  * Common implementation for chown(), lchown(), and fchown()
3095  */
3096 int
setfown(struct thread * td,struct ucred * cred,struct vnode * vp,uid_t uid,gid_t gid)3097 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
3098     gid_t gid)
3099 {
3100 	struct mount *mp;
3101 	struct vattr vattr;
3102 	int error;
3103 
3104 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3105 		return (error);
3106 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3107 	VATTR_NULL(&vattr);
3108 	vattr.va_uid = uid;
3109 	vattr.va_gid = gid;
3110 #ifdef MAC
3111 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3112 	    vattr.va_gid);
3113 	if (error == 0)
3114 #endif
3115 		error = VOP_SETATTR(vp, &vattr, cred);
3116 	VOP_UNLOCK(vp);
3117 	vn_finished_write(mp);
3118 	return (error);
3119 }
3120 
3121 /*
3122  * Set ownership given a path name.
3123  */
3124 #ifndef _SYS_SYSPROTO_H_
3125 struct chown_args {
3126 	char	*path;
3127 	int	uid;
3128 	int	gid;
3129 };
3130 #endif
3131 int
sys_chown(struct thread * td,struct chown_args * uap)3132 sys_chown(struct thread *td, struct chown_args *uap)
3133 {
3134 
3135 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3136 	    uap->gid, 0));
3137 }
3138 
3139 #ifndef _SYS_SYSPROTO_H_
3140 struct fchownat_args {
3141 	int fd;
3142 	const char * path;
3143 	uid_t uid;
3144 	gid_t gid;
3145 	int flag;
3146 };
3147 #endif
3148 int
sys_fchownat(struct thread * td,struct fchownat_args * uap)3149 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3150 {
3151 
3152 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3153 	    uap->gid, uap->flag));
3154 }
3155 
3156 int
kern_fchownat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,int uid,int gid,int flag)3157 kern_fchownat(struct thread *td, int fd, const char *path,
3158     enum uio_seg pathseg, int uid, int gid, int flag)
3159 {
3160 	struct nameidata nd;
3161 	int error;
3162 
3163 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3164 	    AT_EMPTY_PATH)) != 0)
3165 		return (EINVAL);
3166 
3167 	AUDIT_ARG_OWNER(uid, gid);
3168 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3169 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3170 	    fd, &cap_fchown_rights);
3171 
3172 	if ((error = namei(&nd)) != 0)
3173 		return (error);
3174 	NDFREE_PNBUF(&nd);
3175 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3176 	vrele(nd.ni_vp);
3177 	return (error);
3178 }
3179 
3180 /*
3181  * Set ownership given a path name, do not cross symlinks.
3182  */
3183 #ifndef _SYS_SYSPROTO_H_
3184 struct lchown_args {
3185 	char	*path;
3186 	int	uid;
3187 	int	gid;
3188 };
3189 #endif
3190 int
sys_lchown(struct thread * td,struct lchown_args * uap)3191 sys_lchown(struct thread *td, struct lchown_args *uap)
3192 {
3193 
3194 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3195 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3196 }
3197 
3198 /*
3199  * Set ownership given a file descriptor.
3200  */
3201 #ifndef _SYS_SYSPROTO_H_
3202 struct fchown_args {
3203 	int	fd;
3204 	int	uid;
3205 	int	gid;
3206 };
3207 #endif
3208 int
sys_fchown(struct thread * td,struct fchown_args * uap)3209 sys_fchown(struct thread *td, struct fchown_args *uap)
3210 {
3211 	struct file *fp;
3212 	int error;
3213 
3214 	AUDIT_ARG_FD(uap->fd);
3215 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3216 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3217 	if (error != 0)
3218 		return (error);
3219 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3220 	fdrop(fp, td);
3221 	return (error);
3222 }
3223 
3224 /*
3225  * Common implementation code for utimes(), lutimes(), and futimes().
3226  */
3227 static int
getutimes(const struct timeval * usrtvp,enum uio_seg tvpseg,struct timespec * tsp)3228 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3229     struct timespec *tsp)
3230 {
3231 	struct timeval tv[2];
3232 	const struct timeval *tvp;
3233 	int error;
3234 
3235 	if (usrtvp == NULL) {
3236 		vfs_timestamp(&tsp[0]);
3237 		tsp[1] = tsp[0];
3238 	} else {
3239 		if (tvpseg == UIO_SYSSPACE) {
3240 			tvp = usrtvp;
3241 		} else {
3242 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3243 				return (error);
3244 			tvp = tv;
3245 		}
3246 
3247 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3248 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3249 			return (EINVAL);
3250 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3251 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3252 	}
3253 	return (0);
3254 }
3255 
3256 /*
3257  * Common implementation code for futimens(), utimensat().
3258  */
3259 #define	UTIMENS_NULL	0x1
3260 #define	UTIMENS_EXIT	0x2
3261 static int
getutimens(const struct timespec * usrtsp,enum uio_seg tspseg,struct timespec * tsp,int * retflags)3262 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3263     struct timespec *tsp, int *retflags)
3264 {
3265 	struct timespec tsnow;
3266 	int error;
3267 
3268 	vfs_timestamp(&tsnow);
3269 	*retflags = 0;
3270 	if (usrtsp == NULL) {
3271 		tsp[0] = tsnow;
3272 		tsp[1] = tsnow;
3273 		*retflags |= UTIMENS_NULL;
3274 		return (0);
3275 	}
3276 	if (tspseg == UIO_SYSSPACE) {
3277 		tsp[0] = usrtsp[0];
3278 		tsp[1] = usrtsp[1];
3279 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3280 		return (error);
3281 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3282 		*retflags |= UTIMENS_EXIT;
3283 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3284 		*retflags |= UTIMENS_NULL;
3285 	if (tsp[0].tv_nsec == UTIME_OMIT)
3286 		tsp[0].tv_sec = VNOVAL;
3287 	else if (tsp[0].tv_nsec == UTIME_NOW)
3288 		tsp[0] = tsnow;
3289 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3290 		return (EINVAL);
3291 	if (tsp[1].tv_nsec == UTIME_OMIT)
3292 		tsp[1].tv_sec = VNOVAL;
3293 	else if (tsp[1].tv_nsec == UTIME_NOW)
3294 		tsp[1] = tsnow;
3295 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3296 		return (EINVAL);
3297 
3298 	return (0);
3299 }
3300 
3301 /*
3302  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3303  * and utimensat().
3304  */
3305 static int
setutimes(struct thread * td,struct vnode * vp,const struct timespec * ts,int numtimes,int nullflag)3306 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3307     int numtimes, int nullflag)
3308 {
3309 	struct mount *mp;
3310 	struct vattr vattr;
3311 	int error;
3312 	bool setbirthtime;
3313 
3314 	setbirthtime = false;
3315 	vattr.va_birthtime.tv_sec = VNOVAL;
3316 	vattr.va_birthtime.tv_nsec = 0;
3317 
3318 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3319 		return (error);
3320 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3321 	if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3322 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3323 		setbirthtime = true;
3324 	VATTR_NULL(&vattr);
3325 	vattr.va_atime = ts[0];
3326 	vattr.va_mtime = ts[1];
3327 	if (setbirthtime)
3328 		vattr.va_birthtime = ts[1];
3329 	if (numtimes > 2)
3330 		vattr.va_birthtime = ts[2];
3331 	if (nullflag)
3332 		vattr.va_vaflags |= VA_UTIMES_NULL;
3333 #ifdef MAC
3334 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3335 	    vattr.va_mtime);
3336 #endif
3337 	if (error == 0)
3338 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3339 	VOP_UNLOCK(vp);
3340 	vn_finished_write(mp);
3341 	return (error);
3342 }
3343 
3344 /*
3345  * Set the access and modification times of a file.
3346  */
3347 #ifndef _SYS_SYSPROTO_H_
3348 struct utimes_args {
3349 	char	*path;
3350 	struct	timeval *tptr;
3351 };
3352 #endif
3353 int
sys_utimes(struct thread * td,struct utimes_args * uap)3354 sys_utimes(struct thread *td, struct utimes_args *uap)
3355 {
3356 
3357 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3358 	    uap->tptr, UIO_USERSPACE));
3359 }
3360 
3361 #ifndef _SYS_SYSPROTO_H_
3362 struct futimesat_args {
3363 	int fd;
3364 	const char * path;
3365 	const struct timeval * times;
3366 };
3367 #endif
3368 int
sys_futimesat(struct thread * td,struct futimesat_args * uap)3369 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3370 {
3371 
3372 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3373 	    uap->times, UIO_USERSPACE));
3374 }
3375 
3376 int
kern_utimesat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,const struct timeval * tptr,enum uio_seg tptrseg)3377 kern_utimesat(struct thread *td, int fd, const char *path,
3378     enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg)
3379 {
3380 	struct nameidata nd;
3381 	struct timespec ts[2];
3382 	int error;
3383 
3384 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3385 		return (error);
3386 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3387 	    &cap_futimes_rights);
3388 
3389 	if ((error = namei(&nd)) != 0)
3390 		return (error);
3391 	NDFREE_PNBUF(&nd);
3392 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3393 	vrele(nd.ni_vp);
3394 	return (error);
3395 }
3396 
3397 /*
3398  * Set the access and modification times of a file.
3399  */
3400 #ifndef _SYS_SYSPROTO_H_
3401 struct lutimes_args {
3402 	char	*path;
3403 	struct	timeval *tptr;
3404 };
3405 #endif
3406 int
sys_lutimes(struct thread * td,struct lutimes_args * uap)3407 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3408 {
3409 
3410 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3411 	    UIO_USERSPACE));
3412 }
3413 
3414 int
kern_lutimes(struct thread * td,const char * path,enum uio_seg pathseg,const struct timeval * tptr,enum uio_seg tptrseg)3415 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3416     const struct timeval *tptr, enum uio_seg tptrseg)
3417 {
3418 	struct timespec ts[2];
3419 	struct nameidata nd;
3420 	int error;
3421 
3422 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3423 		return (error);
3424 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path);
3425 	if ((error = namei(&nd)) != 0)
3426 		return (error);
3427 	NDFREE_PNBUF(&nd);
3428 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3429 	vrele(nd.ni_vp);
3430 	return (error);
3431 }
3432 
3433 /*
3434  * Set the access and modification times of a file.
3435  */
3436 #ifndef _SYS_SYSPROTO_H_
3437 struct futimes_args {
3438 	int	fd;
3439 	struct	timeval *tptr;
3440 };
3441 #endif
3442 int
sys_futimes(struct thread * td,struct futimes_args * uap)3443 sys_futimes(struct thread *td, struct futimes_args *uap)
3444 {
3445 
3446 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3447 }
3448 
3449 int
kern_futimes(struct thread * td,int fd,const struct timeval * tptr,enum uio_seg tptrseg)3450 kern_futimes(struct thread *td, int fd, const struct timeval *tptr,
3451     enum uio_seg tptrseg)
3452 {
3453 	struct timespec ts[2];
3454 	struct file *fp;
3455 	int error;
3456 
3457 	AUDIT_ARG_FD(fd);
3458 	error = getutimes(tptr, tptrseg, ts);
3459 	if (error != 0)
3460 		return (error);
3461 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3462 	if (error != 0)
3463 		return (error);
3464 #ifdef AUDIT
3465 	if (AUDITING_TD(td)) {
3466 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3467 		AUDIT_ARG_VNODE1(fp->f_vnode);
3468 		VOP_UNLOCK(fp->f_vnode);
3469 	}
3470 #endif
3471 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3472 	fdrop(fp, td);
3473 	return (error);
3474 }
3475 
3476 int
sys_futimens(struct thread * td,struct futimens_args * uap)3477 sys_futimens(struct thread *td, struct futimens_args *uap)
3478 {
3479 
3480 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3481 }
3482 
3483 int
kern_futimens(struct thread * td,int fd,const struct timespec * tptr,enum uio_seg tptrseg)3484 kern_futimens(struct thread *td, int fd, const struct timespec *tptr,
3485     enum uio_seg tptrseg)
3486 {
3487 	struct timespec ts[2];
3488 	struct file *fp;
3489 	int error, flags;
3490 
3491 	AUDIT_ARG_FD(fd);
3492 	error = getutimens(tptr, tptrseg, ts, &flags);
3493 	if (error != 0)
3494 		return (error);
3495 	if (flags & UTIMENS_EXIT)
3496 		return (0);
3497 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3498 	if (error != 0)
3499 		return (error);
3500 #ifdef AUDIT
3501 	if (AUDITING_TD(td)) {
3502 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3503 		AUDIT_ARG_VNODE1(fp->f_vnode);
3504 		VOP_UNLOCK(fp->f_vnode);
3505 	}
3506 #endif
3507 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3508 	fdrop(fp, td);
3509 	return (error);
3510 }
3511 
3512 int
sys_utimensat(struct thread * td,struct utimensat_args * uap)3513 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3514 {
3515 
3516 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3517 	    uap->times, UIO_USERSPACE, uap->flag));
3518 }
3519 
3520 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)3521 kern_utimensat(struct thread *td, int fd, const char *path,
3522     enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg,
3523     int flag)
3524 {
3525 	struct nameidata nd;
3526 	struct timespec ts[2];
3527 	int error, flags;
3528 
3529 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3530 	    AT_EMPTY_PATH)) != 0)
3531 		return (EINVAL);
3532 
3533 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3534 		return (error);
3535 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3536 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3537 	    pathseg, path, fd, &cap_futimes_rights);
3538 	if ((error = namei(&nd)) != 0)
3539 		return (error);
3540 	/*
3541 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3542 	 * POSIX states:
3543 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3544 	 * "Search permission is denied by a component of the path prefix."
3545 	 */
3546 	NDFREE_PNBUF(&nd);
3547 	if ((flags & UTIMENS_EXIT) == 0)
3548 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3549 	vrele(nd.ni_vp);
3550 	return (error);
3551 }
3552 
3553 /*
3554  * Truncate a file given its path name.
3555  */
3556 #ifndef _SYS_SYSPROTO_H_
3557 struct truncate_args {
3558 	char	*path;
3559 	int	pad;
3560 	off_t	length;
3561 };
3562 #endif
3563 int
sys_truncate(struct thread * td,struct truncate_args * uap)3564 sys_truncate(struct thread *td, struct truncate_args *uap)
3565 {
3566 
3567 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3568 }
3569 
3570 int
kern_truncate(struct thread * td,const char * path,enum uio_seg pathseg,off_t length)3571 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3572     off_t length)
3573 {
3574 	struct mount *mp;
3575 	struct vnode *vp;
3576 	void *rl_cookie;
3577 	struct nameidata nd;
3578 	int error;
3579 
3580 	if (length < 0)
3581 		return (EINVAL);
3582 	NDPREINIT(&nd);
3583 retry:
3584 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
3585 	if ((error = namei(&nd)) != 0)
3586 		return (error);
3587 	vp = nd.ni_vp;
3588 	NDFREE_PNBUF(&nd);
3589 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3590 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3591 		vn_rangelock_unlock(vp, rl_cookie);
3592 		vrele(vp);
3593 		return (error);
3594 	}
3595 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3596 	if (vp->v_type == VDIR) {
3597 		error = EISDIR;
3598 		goto out;
3599 	}
3600 #ifdef MAC
3601 	error = mac_vnode_check_write(td->td_ucred, NOCRED, vp);
3602 	if (error != 0)
3603 		goto out;
3604 #endif
3605 	error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td);
3606 	if (error != 0)
3607 		goto out;
3608 
3609 	error = vn_truncate_locked(vp, length, false, td->td_ucred);
3610 out:
3611 	VOP_UNLOCK(vp);
3612 	vn_finished_write(mp);
3613 	vn_rangelock_unlock(vp, rl_cookie);
3614 	vrele(vp);
3615 	if (error == ERELOOKUP)
3616 		goto retry;
3617 	return (error);
3618 }
3619 
3620 #if defined(COMPAT_43)
3621 /*
3622  * Truncate a file given its path name.
3623  */
3624 #ifndef _SYS_SYSPROTO_H_
3625 struct otruncate_args {
3626 	char	*path;
3627 	long	length;
3628 };
3629 #endif
3630 int
otruncate(struct thread * td,struct otruncate_args * uap)3631 otruncate(struct thread *td, struct otruncate_args *uap)
3632 {
3633 
3634 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3635 }
3636 #endif /* COMPAT_43 */
3637 
3638 #if defined(COMPAT_FREEBSD6)
3639 /* Versions with the pad argument */
3640 int
freebsd6_truncate(struct thread * td,struct freebsd6_truncate_args * uap)3641 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3642 {
3643 
3644 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3645 }
3646 
3647 int
freebsd6_ftruncate(struct thread * td,struct freebsd6_ftruncate_args * uap)3648 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3649 {
3650 
3651 	return (kern_ftruncate(td, uap->fd, uap->length));
3652 }
3653 #endif
3654 
3655 int
kern_fsync(struct thread * td,int fd,bool fullsync)3656 kern_fsync(struct thread *td, int fd, bool fullsync)
3657 {
3658 	struct vnode *vp;
3659 	struct mount *mp;
3660 	struct file *fp;
3661 	int error;
3662 
3663 	AUDIT_ARG_FD(fd);
3664 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3665 	if (error != 0)
3666 		return (error);
3667 	vp = fp->f_vnode;
3668 #if 0
3669 	if (!fullsync)
3670 		/* XXXKIB: compete outstanding aio writes */;
3671 #endif
3672 retry:
3673 	error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3674 	if (error != 0)
3675 		goto drop;
3676 	vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3677 	AUDIT_ARG_VNODE1(vp);
3678 	vnode_pager_clean_async(vp);
3679 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3680 	VOP_UNLOCK(vp);
3681 	vn_finished_write(mp);
3682 	if (error == ERELOOKUP)
3683 		goto retry;
3684 drop:
3685 	fdrop(fp, td);
3686 	return (error);
3687 }
3688 
3689 /*
3690  * Sync an open file.
3691  */
3692 #ifndef _SYS_SYSPROTO_H_
3693 struct fsync_args {
3694 	int	fd;
3695 };
3696 #endif
3697 int
sys_fsync(struct thread * td,struct fsync_args * uap)3698 sys_fsync(struct thread *td, struct fsync_args *uap)
3699 {
3700 
3701 	return (kern_fsync(td, uap->fd, true));
3702 }
3703 
3704 int
sys_fdatasync(struct thread * td,struct fdatasync_args * uap)3705 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3706 {
3707 
3708 	return (kern_fsync(td, uap->fd, false));
3709 }
3710 
3711 /*
3712  * Rename files.  Source and destination must either both be directories, or
3713  * both not be directories.  If target is a directory, it must be empty.
3714  */
3715 #ifndef _SYS_SYSPROTO_H_
3716 struct rename_args {
3717 	char	*from;
3718 	char	*to;
3719 };
3720 #endif
3721 int
sys_rename(struct thread * td,struct rename_args * uap)3722 sys_rename(struct thread *td, struct rename_args *uap)
3723 {
3724 
3725 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3726 	    uap->to, UIO_USERSPACE));
3727 }
3728 
3729 #ifndef _SYS_SYSPROTO_H_
3730 struct renameat_args {
3731 	int	oldfd;
3732 	char	*old;
3733 	int	newfd;
3734 	char	*new;
3735 };
3736 #endif
3737 int
sys_renameat(struct thread * td,struct renameat_args * uap)3738 sys_renameat(struct thread *td, struct renameat_args *uap)
3739 {
3740 
3741 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3742 	    UIO_USERSPACE));
3743 }
3744 
3745 #ifdef MAC
3746 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)3747 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3748     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3749 {
3750 	int error;
3751 
3752 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3753 	    pathseg, old, oldfd, &cap_renameat_source_rights);
3754 	if ((error = namei(fromnd)) != 0)
3755 		return (error);
3756 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3757 	    fromnd->ni_vp, &fromnd->ni_cnd);
3758 	VOP_UNLOCK(fromnd->ni_dvp);
3759 	if (fromnd->ni_dvp != fromnd->ni_vp)
3760 		VOP_UNLOCK(fromnd->ni_vp);
3761 	if (error != 0) {
3762 		NDFREE_PNBUF(fromnd);
3763 		vrele(fromnd->ni_dvp);
3764 		vrele(fromnd->ni_vp);
3765 	}
3766 	return (error);
3767 }
3768 #endif
3769 
3770 int
kern_renameat(struct thread * td,int oldfd,const char * old,int newfd,const char * new,enum uio_seg pathseg)3771 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3772     const char *new, enum uio_seg pathseg)
3773 {
3774 	struct mount *mp, *tmp;
3775 	struct vnode *tvp, *fvp, *tdvp;
3776 	struct nameidata fromnd, tond;
3777 	uint64_t tondflags;
3778 	int error;
3779 	short irflag;
3780 
3781 again:
3782 	tmp = mp = NULL;
3783 	bwillwrite();
3784 #ifdef MAC
3785 	if (mac_vnode_check_rename_from_enabled()) {
3786 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3787 		    &fromnd);
3788 		if (error != 0)
3789 			return (error);
3790 	} else {
3791 #endif
3792 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | AUDITVNODE1,
3793 	    pathseg, old, oldfd, &cap_renameat_source_rights);
3794 	if ((error = namei(&fromnd)) != 0)
3795 		return (error);
3796 #ifdef MAC
3797 	}
3798 #endif
3799 	fvp = fromnd.ni_vp;
3800 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | AUDITVNODE2;
3801 	if (fromnd.ni_vp->v_type == VDIR)
3802 		tondflags |= WILLBEDIR;
3803 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3804 	    &cap_renameat_target_rights);
3805 	if ((error = namei(&tond)) != 0) {
3806 		/* Translate error code for rename("dir1", "dir2/."). */
3807 		if (error == EISDIR && fvp->v_type == VDIR)
3808 			error = EINVAL;
3809 		NDFREE_PNBUF(&fromnd);
3810 		vrele(fromnd.ni_dvp);
3811 		vrele(fvp);
3812 		goto out1;
3813 	}
3814 	tdvp = tond.ni_dvp;
3815 	tvp = tond.ni_vp;
3816 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3817 	if (error != 0) {
3818 again1:
3819 		NDFREE_PNBUF(&fromnd);
3820 		NDFREE_PNBUF(&tond);
3821 		if (tvp != NULL)
3822 			vput(tvp);
3823 		if (tdvp == tvp)
3824 			vrele(tdvp);
3825 		else
3826 			vput(tdvp);
3827 		vrele(fromnd.ni_dvp);
3828 		vrele(fvp);
3829 		if (tmp != NULL) {
3830 			lockmgr(&tmp->mnt_renamelock, LK_EXCLUSIVE, NULL);
3831 			lockmgr(&tmp->mnt_renamelock, LK_RELEASE, NULL);
3832 			vfs_rel(tmp);
3833 			tmp = NULL;
3834 		}
3835 		error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH);
3836 		if (error != 0)
3837 			return (error);
3838 		goto again;
3839 	}
3840 	error = VOP_GETWRITEMOUNT(tdvp, &tmp);
3841 	if (error != 0 || tmp == NULL)
3842 		goto again1;
3843 	error = lockmgr(&tmp->mnt_renamelock, LK_EXCLUSIVE | LK_NOWAIT, NULL);
3844 	if (error != 0) {
3845 		vn_finished_write(mp);
3846 		goto again1;
3847 	}
3848 	irflag = vn_irflag_read(fvp);
3849 	if (((irflag & VIRF_NAMEDATTR) != 0 && tdvp != fromnd.ni_dvp) ||
3850 	    (irflag & VIRF_NAMEDDIR) != 0) {
3851 		error = EINVAL;
3852 		goto out;
3853 	}
3854 	if (tvp != NULL) {
3855 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3856 			error = ENOTDIR;
3857 			goto out;
3858 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3859 			error = EISDIR;
3860 			goto out;
3861 		}
3862 #ifdef CAPABILITIES
3863 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3864 			/*
3865 			 * If the target already exists we require CAP_UNLINKAT
3866 			 * from 'newfd', when newfd was used for the lookup.
3867 			 */
3868 			error = cap_check(&tond.ni_filecaps.fc_rights,
3869 			    &cap_unlinkat_rights);
3870 			if (error != 0)
3871 				goto out;
3872 		}
3873 #endif
3874 	}
3875 	if (fvp == tdvp) {
3876 		error = EINVAL;
3877 		goto out;
3878 	}
3879 	/*
3880 	 * If the source is the same as the destination (that is, if they
3881 	 * are links to the same vnode), then there is nothing to do.
3882 	 */
3883 	if (fvp == tvp)
3884 		error = ERESTART;
3885 #ifdef MAC
3886 	else
3887 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3888 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3889 #endif
3890 out:
3891 	if (error == 0) {
3892 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3893 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3894 		NDFREE_PNBUF(&fromnd);
3895 		NDFREE_PNBUF(&tond);
3896 	} else {
3897 		NDFREE_PNBUF(&fromnd);
3898 		NDFREE_PNBUF(&tond);
3899 		if (tvp != NULL)
3900 			vput(tvp);
3901 		if (tdvp == tvp)
3902 			vrele(tdvp);
3903 		else
3904 			vput(tdvp);
3905 		vrele(fromnd.ni_dvp);
3906 		vrele(fvp);
3907 	}
3908 	lockmgr(&tmp->mnt_renamelock, LK_RELEASE, 0);
3909 	vfs_rel(tmp);
3910 	vn_finished_write(mp);
3911 out1:
3912 	if (error == ERESTART)
3913 		return (0);
3914 	if (error == ERELOOKUP)
3915 		goto again;
3916 	return (error);
3917 }
3918 
3919 /*
3920  * Make a directory file.
3921  */
3922 #ifndef _SYS_SYSPROTO_H_
3923 struct mkdir_args {
3924 	char	*path;
3925 	int	mode;
3926 };
3927 #endif
3928 int
sys_mkdir(struct thread * td,struct mkdir_args * uap)3929 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3930 {
3931 
3932 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3933 	    uap->mode));
3934 }
3935 
3936 #ifndef _SYS_SYSPROTO_H_
3937 struct mkdirat_args {
3938 	int	fd;
3939 	char	*path;
3940 	mode_t	mode;
3941 };
3942 #endif
3943 int
sys_mkdirat(struct thread * td,struct mkdirat_args * uap)3944 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3945 {
3946 
3947 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3948 }
3949 
3950 int
kern_mkdirat(struct thread * td,int fd,const char * path,enum uio_seg segflg,int mode)3951 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3952     int mode)
3953 {
3954 	struct mount *mp;
3955 	struct vattr vattr;
3956 	struct nameidata nd;
3957 	int error;
3958 
3959 	AUDIT_ARG_MODE(mode);
3960 	NDPREINIT(&nd);
3961 restart:
3962 	bwillwrite();
3963 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 |
3964 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3965 	    segflg, path, fd, &cap_mkdirat_rights);
3966 	if ((error = namei(&nd)) != 0)
3967 		return (error);
3968 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3969 		NDFREE_PNBUF(&nd);
3970 		vput(nd.ni_dvp);
3971 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
3972 			return (error);
3973 		goto restart;
3974 	}
3975 	if ((vn_irflag_read(nd.ni_dvp) & VIRF_NAMEDDIR) != 0) {
3976 		error = EINVAL;
3977 		goto out;
3978 	}
3979 	VATTR_NULL(&vattr);
3980 	vattr.va_type = VDIR;
3981 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3982 #ifdef MAC
3983 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3984 	    &vattr);
3985 	if (error != 0)
3986 		goto out;
3987 #endif
3988 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3989 out:
3990 	NDFREE_PNBUF(&nd);
3991 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3992 	vn_finished_write(mp);
3993 	if (error == ERELOOKUP)
3994 		goto restart;
3995 	return (error);
3996 }
3997 
3998 /*
3999  * Remove a directory file.
4000  */
4001 #ifndef _SYS_SYSPROTO_H_
4002 struct rmdir_args {
4003 	char	*path;
4004 };
4005 #endif
4006 int
sys_rmdir(struct thread * td,struct rmdir_args * uap)4007 sys_rmdir(struct thread *td, struct rmdir_args *uap)
4008 {
4009 
4010 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
4011 	    0));
4012 }
4013 
4014 int
kern_frmdirat(struct thread * td,int dfd,const char * path,int fd,enum uio_seg pathseg,int flag)4015 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
4016     enum uio_seg pathseg, int flag)
4017 {
4018 	struct mount *mp;
4019 	struct vnode *vp;
4020 	struct file *fp;
4021 	struct nameidata nd;
4022 	cap_rights_t rights;
4023 	int error;
4024 
4025 	fp = NULL;
4026 	if (fd != FD_NONE) {
4027 		error = getvnode(td, fd, cap_rights_init_one(&rights,
4028 		    CAP_LOOKUP), &fp);
4029 		if (error != 0)
4030 			return (error);
4031 	}
4032 
4033 	NDPREINIT(&nd);
4034 restart:
4035 	bwillwrite();
4036 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
4037 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
4038 	    pathseg, path, dfd, &cap_unlinkat_rights);
4039 	if ((error = namei(&nd)) != 0)
4040 		goto fdout;
4041 	vp = nd.ni_vp;
4042 	if (vp->v_type != VDIR) {
4043 		error = ENOTDIR;
4044 		goto out;
4045 	}
4046 	/*
4047 	 * No rmdir "." please.
4048 	 */
4049 	if (nd.ni_dvp == vp) {
4050 		error = EINVAL;
4051 		goto out;
4052 	}
4053 	/*
4054 	 * The root of a mounted filesystem cannot be deleted.
4055 	 */
4056 	if (vp->v_vflag & VV_ROOT) {
4057 		error = EBUSY;
4058 		goto out;
4059 	}
4060 
4061 	if (fp != NULL && fp->f_vnode != vp) {
4062 		if (VN_IS_DOOMED(fp->f_vnode))
4063 			error = EBADF;
4064 		else
4065 			error = EDEADLK;
4066 		goto out;
4067 	}
4068 
4069 #ifdef MAC
4070 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
4071 	    &nd.ni_cnd);
4072 	if (error != 0)
4073 		goto out;
4074 #endif
4075 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
4076 		NDFREE_PNBUF(&nd);
4077 		vput(vp);
4078 		if (nd.ni_dvp == vp)
4079 			vrele(nd.ni_dvp);
4080 		else
4081 			vput(nd.ni_dvp);
4082 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
4083 			goto fdout;
4084 		goto restart;
4085 	}
4086 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
4087 	vn_finished_write(mp);
4088 out:
4089 	NDFREE_PNBUF(&nd);
4090 	vput(vp);
4091 	if (nd.ni_dvp == vp)
4092 		vrele(nd.ni_dvp);
4093 	else
4094 		vput(nd.ni_dvp);
4095 	if (error == ERELOOKUP)
4096 		goto restart;
4097 fdout:
4098 	if (fp != NULL)
4099 		fdrop(fp, td);
4100 	return (error);
4101 }
4102 
4103 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
4104 int
freebsd11_kern_getdirentries(struct thread * td,int fd,char * ubuf,u_int count,long * basep,void (* func)(struct freebsd11_dirent *))4105 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
4106     long *basep, void (*func)(struct freebsd11_dirent *))
4107 {
4108 	struct freebsd11_dirent dstdp;
4109 	struct dirent *dp, *edp;
4110 	char *dirbuf;
4111 	off_t base;
4112 	ssize_t resid, ucount;
4113 	int error;
4114 
4115 	/* XXX arbitrary sanity limit on `count'. */
4116 	count = min(count, 64 * 1024);
4117 
4118 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
4119 
4120 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
4121 	    UIO_SYSSPACE);
4122 	if (error != 0)
4123 		goto done;
4124 	if (basep != NULL)
4125 		*basep = base;
4126 
4127 	ucount = 0;
4128 	for (dp = (struct dirent *)dirbuf,
4129 	    edp = (struct dirent *)&dirbuf[count - resid];
4130 	    ucount < count && dp < edp; ) {
4131 		if (dp->d_reclen == 0)
4132 			break;
4133 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
4134 		if (dp->d_namlen >= sizeof(dstdp.d_name))
4135 			continue;
4136 		dstdp.d_type = dp->d_type;
4137 		dstdp.d_namlen = dp->d_namlen;
4138 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
4139 		if (dstdp.d_fileno != dp->d_fileno) {
4140 			switch (ino64_trunc_error) {
4141 			default:
4142 			case 0:
4143 				break;
4144 			case 1:
4145 				error = EOVERFLOW;
4146 				goto done;
4147 			case 2:
4148 				dstdp.d_fileno = UINT32_MAX;
4149 				break;
4150 			}
4151 		}
4152 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
4153 		    ((dp->d_namlen + 1 + 3) &~ 3);
4154 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4155 		bzero(dstdp.d_name + dstdp.d_namlen,
4156 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4157 		    dstdp.d_namlen);
4158 		MPASS(dstdp.d_reclen <= dp->d_reclen);
4159 		MPASS(ucount + dstdp.d_reclen <= count);
4160 		if (func != NULL)
4161 			func(&dstdp);
4162 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4163 		if (error != 0)
4164 			break;
4165 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
4166 		ucount += dstdp.d_reclen;
4167 	}
4168 
4169 done:
4170 	free(dirbuf, M_TEMP);
4171 	if (error == 0)
4172 		td->td_retval[0] = ucount;
4173 	return (error);
4174 }
4175 #endif /* COMPAT */
4176 
4177 #ifdef COMPAT_43
4178 static void
ogetdirentries_cvt(struct freebsd11_dirent * dp)4179 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4180 {
4181 #if (BYTE_ORDER == LITTLE_ENDIAN)
4182 	/*
4183 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4184 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4185 	 */
4186 	dp->d_type = dp->d_namlen;
4187 	dp->d_namlen = 0;
4188 #else
4189 	/*
4190 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4191 	 * so must be zero'ed.
4192 	 */
4193 	dp->d_type = 0;
4194 #endif
4195 }
4196 
4197 /*
4198  * Read a block of directory entries in a filesystem independent format.
4199  */
4200 #ifndef _SYS_SYSPROTO_H_
4201 struct ogetdirentries_args {
4202 	int	fd;
4203 	char	*buf;
4204 	u_int	count;
4205 	long	*basep;
4206 };
4207 #endif
4208 int
ogetdirentries(struct thread * td,struct ogetdirentries_args * uap)4209 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4210 {
4211 	long loff;
4212 	int error;
4213 
4214 	error = kern_ogetdirentries(td, uap, &loff);
4215 	if (error == 0)
4216 		error = copyout(&loff, uap->basep, sizeof(long));
4217 	return (error);
4218 }
4219 
4220 int
kern_ogetdirentries(struct thread * td,struct ogetdirentries_args * uap,long * ploff)4221 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4222     long *ploff)
4223 {
4224 	long base;
4225 	int error;
4226 
4227 	/* XXX arbitrary sanity limit on `count'. */
4228 	if (uap->count > 64 * 1024)
4229 		return (EINVAL);
4230 
4231 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4232 	    &base, ogetdirentries_cvt);
4233 
4234 	if (error == 0 && uap->basep != NULL)
4235 		error = copyout(&base, uap->basep, sizeof(long));
4236 
4237 	return (error);
4238 }
4239 #endif /* COMPAT_43 */
4240 
4241 #if defined(COMPAT_FREEBSD11)
4242 #ifndef _SYS_SYSPROTO_H_
4243 struct freebsd11_getdirentries_args {
4244 	int	fd;
4245 	char	*buf;
4246 	u_int	count;
4247 	long	*basep;
4248 };
4249 #endif
4250 int
freebsd11_getdirentries(struct thread * td,struct freebsd11_getdirentries_args * uap)4251 freebsd11_getdirentries(struct thread *td,
4252     struct freebsd11_getdirentries_args *uap)
4253 {
4254 	long base;
4255 	int error;
4256 
4257 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4258 	    &base, NULL);
4259 
4260 	if (error == 0 && uap->basep != NULL)
4261 		error = copyout(&base, uap->basep, sizeof(long));
4262 	return (error);
4263 }
4264 
4265 int
freebsd11_getdents(struct thread * td,struct freebsd11_getdents_args * uap)4266 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4267 {
4268 	struct freebsd11_getdirentries_args ap;
4269 
4270 	ap.fd = uap->fd;
4271 	ap.buf = uap->buf;
4272 	ap.count = uap->count;
4273 	ap.basep = NULL;
4274 	return (freebsd11_getdirentries(td, &ap));
4275 }
4276 #endif /* COMPAT_FREEBSD11 */
4277 
4278 /*
4279  * Read a block of directory entries in a filesystem independent format.
4280  */
4281 int
sys_getdirentries(struct thread * td,struct getdirentries_args * uap)4282 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4283 {
4284 	off_t base;
4285 	int error;
4286 
4287 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4288 	    NULL, UIO_USERSPACE);
4289 	if (error != 0)
4290 		return (error);
4291 	if (uap->basep != NULL)
4292 		error = copyout(&base, uap->basep, sizeof(off_t));
4293 	return (error);
4294 }
4295 
4296 int
kern_getdirentries(struct thread * td,int fd,char * buf,size_t count,off_t * basep,ssize_t * residp,enum uio_seg bufseg)4297 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4298     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4299 {
4300 	struct vnode *vp;
4301 	struct file *fp;
4302 	struct uio auio;
4303 	struct iovec aiov;
4304 	off_t loff;
4305 	int error, eofflag;
4306 	off_t foffset;
4307 
4308 	AUDIT_ARG_FD(fd);
4309 	if (count > IOSIZE_MAX)
4310 		return (EINVAL);
4311 	auio.uio_resid = count;
4312 	error = getvnode(td, fd, &cap_read_rights, &fp);
4313 	if (error != 0)
4314 		return (error);
4315 	if ((fp->f_flag & FREAD) == 0) {
4316 		fdrop(fp, td);
4317 		return (EBADF);
4318 	}
4319 	vp = fp->f_vnode;
4320 	foffset = foffset_lock(fp, 0);
4321 unionread:
4322 	if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) {
4323 		error = ENOENT;
4324 		goto fail;
4325 	}
4326 	aiov.iov_base = buf;
4327 	aiov.iov_len = count;
4328 	auio.uio_iov = &aiov;
4329 	auio.uio_iovcnt = 1;
4330 	auio.uio_rw = UIO_READ;
4331 	auio.uio_segflg = bufseg;
4332 	auio.uio_td = td;
4333 	vn_lock(vp, LK_SHARED | LK_RETRY);
4334 	/*
4335 	 * We want to return ENOTDIR for anything that is not VDIR, but
4336 	 * not for VBAD, and we can't check for VBAD while the vnode is
4337 	 * unlocked.
4338 	 */
4339 	if (vp->v_type != VDIR) {
4340 		if (vp->v_type == VBAD)
4341 			error = EBADF;
4342 		else
4343 			error = ENOTDIR;
4344 		VOP_UNLOCK(vp);
4345 		goto fail;
4346 	}
4347 	AUDIT_ARG_VNODE1(vp);
4348 	loff = auio.uio_offset = foffset;
4349 #ifdef MAC
4350 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4351 	if (error == 0)
4352 #endif
4353 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4354 		    NULL);
4355 	foffset = auio.uio_offset;
4356 	if (error != 0) {
4357 		VOP_UNLOCK(vp);
4358 		goto fail;
4359 	}
4360 	if (count == auio.uio_resid &&
4361 	    (vp->v_vflag & VV_ROOT) &&
4362 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4363 		struct vnode *tvp = vp;
4364 
4365 		vp = vp->v_mount->mnt_vnodecovered;
4366 		VREF(vp);
4367 		fp->f_vnode = vp;
4368 		foffset = 0;
4369 		vput(tvp);
4370 		goto unionread;
4371 	}
4372 	VOP_UNLOCK(vp);
4373 	*basep = loff;
4374 	if (residp != NULL)
4375 		*residp = auio.uio_resid;
4376 	td->td_retval[0] = count - auio.uio_resid;
4377 fail:
4378 	foffset_unlock(fp, foffset, 0);
4379 	fdrop(fp, td);
4380 	return (error);
4381 }
4382 
4383 /*
4384  * Set the mode mask for creation of filesystem nodes.
4385  */
4386 #ifndef _SYS_SYSPROTO_H_
4387 struct umask_args {
4388 	int	newmask;
4389 };
4390 #endif
4391 int
sys_umask(struct thread * td,struct umask_args * uap)4392 sys_umask(struct thread *td, struct umask_args *uap)
4393 {
4394 	struct pwddesc *pdp;
4395 
4396 	pdp = td->td_proc->p_pd;
4397 	PWDDESC_XLOCK(pdp);
4398 	td->td_retval[0] = pdp->pd_cmask;
4399 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4400 	PWDDESC_XUNLOCK(pdp);
4401 	return (0);
4402 }
4403 
4404 /*
4405  * Void all references to file by ripping underlying filesystem away from
4406  * vnode.
4407  */
4408 #ifndef _SYS_SYSPROTO_H_
4409 struct revoke_args {
4410 	char	*path;
4411 };
4412 #endif
4413 int
sys_revoke(struct thread * td,struct revoke_args * uap)4414 sys_revoke(struct thread *td, struct revoke_args *uap)
4415 {
4416 	struct vnode *vp;
4417 	struct vattr vattr;
4418 	struct nameidata nd;
4419 	int error;
4420 
4421 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4422 	    uap->path);
4423 	if ((error = namei(&nd)) != 0)
4424 		return (error);
4425 	vp = nd.ni_vp;
4426 	NDFREE_PNBUF(&nd);
4427 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4428 		error = EINVAL;
4429 		goto out;
4430 	}
4431 #ifdef MAC
4432 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4433 	if (error != 0)
4434 		goto out;
4435 #endif
4436 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4437 	if (error != 0)
4438 		goto out;
4439 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4440 		error = priv_check(td, PRIV_VFS_ADMIN);
4441 		if (error != 0)
4442 			goto out;
4443 	}
4444 	if (devfs_usecount(vp) > 0)
4445 		VOP_REVOKE(vp, REVOKEALL);
4446 out:
4447 	vput(vp);
4448 	return (error);
4449 }
4450 
4451 /*
4452  * This variant of getvnode() allows O_PATH files.  Caller should
4453  * ensure that returned file and vnode are only used for compatible
4454  * semantics.
4455  */
4456 int
getvnode_path(struct thread * td,int fd,const cap_rights_t * rightsp,uint8_t * flagsp,struct file ** fpp)4457 getvnode_path(struct thread *td, int fd, const cap_rights_t *rightsp,
4458     uint8_t *flagsp, struct file **fpp)
4459 {
4460 	struct file *fp;
4461 	int error;
4462 
4463 	error = fget_unlocked_flags(td, fd, rightsp, flagsp, &fp);
4464 	if (error != 0)
4465 		return (error);
4466 
4467 	/*
4468 	 * The file could be not of the vnode type, or it may be not
4469 	 * yet fully initialized, in which case the f_vnode pointer
4470 	 * may be set, but f_ops is still badfileops.  E.g.,
4471 	 * devfs_open() transiently create such situation to
4472 	 * facilitate csw d_fdopen().
4473 	 *
4474 	 * Dupfdopen() handling in kern_openat() installs the
4475 	 * half-baked file into the process descriptor table, allowing
4476 	 * other thread to dereference it. Guard against the race by
4477 	 * checking f_ops.
4478 	 */
4479 	if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) {
4480 		fdrop(fp, td);
4481 		*fpp = NULL;
4482 		return (EINVAL);
4483 	}
4484 
4485 	*fpp = fp;
4486 	return (0);
4487 }
4488 
4489 /*
4490  * Convert a user file descriptor to a kernel file entry and check
4491  * that, if it is a capability, the correct rights are present.
4492  * A reference on the file entry is held upon returning.
4493  */
4494 int
getvnode(struct thread * td,int fd,const cap_rights_t * rightsp,struct file ** fpp)4495 getvnode(struct thread *td, int fd, const cap_rights_t *rightsp,
4496     struct file **fpp)
4497 {
4498 	int error;
4499 
4500 	error = getvnode_path(td, fd, rightsp, NULL, fpp);
4501 	if (__predict_false(error != 0))
4502 		return (error);
4503 
4504 	/*
4505 	 * Filter out O_PATH file descriptors, most getvnode() callers
4506 	 * do not call fo_ methods.
4507 	 */
4508 	if (__predict_false((*fpp)->f_ops == &path_fileops)) {
4509 		fdrop(*fpp, td);
4510 		*fpp = NULL;
4511 		error = EBADF;
4512 	}
4513 
4514 	return (error);
4515 }
4516 
4517 /*
4518  * Get an (NFS) file handle.
4519  */
4520 #ifndef _SYS_SYSPROTO_H_
4521 struct lgetfh_args {
4522 	char *fname;
4523 	fhandle_t *fhp;
4524 };
4525 #endif
4526 int
sys_lgetfh(struct thread * td,struct lgetfh_args * uap)4527 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4528 {
4529 
4530 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4531 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4532 }
4533 
4534 #ifndef _SYS_SYSPROTO_H_
4535 struct getfh_args {
4536 	char *fname;
4537 	fhandle_t *fhp;
4538 };
4539 #endif
4540 int
sys_getfh(struct thread * td,struct getfh_args * uap)4541 sys_getfh(struct thread *td, struct getfh_args *uap)
4542 {
4543 
4544 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4545 	    uap->fhp, UIO_USERSPACE));
4546 }
4547 
4548 /*
4549  * syscall for the rpc.lockd to use to translate an open descriptor into
4550  * a NFS file handle.
4551  *
4552  * warning: do not remove the priv_check() call or this becomes one giant
4553  * security hole.
4554  */
4555 #ifndef _SYS_SYSPROTO_H_
4556 struct getfhat_args {
4557 	int fd;
4558 	char *path;
4559 	fhandle_t *fhp;
4560 	int flags;
4561 };
4562 #endif
4563 int
sys_getfhat(struct thread * td,struct getfhat_args * uap)4564 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4565 {
4566 
4567 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4568 	    uap->fhp, UIO_USERSPACE));
4569 }
4570 
4571 int
kern_getfhat(struct thread * td,int flags,int fd,const char * path,enum uio_seg pathseg,fhandle_t * fhp,enum uio_seg fhseg)4572 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4573     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4574 {
4575 	struct nameidata nd;
4576 	fhandle_t fh;
4577 	struct vnode *vp;
4578 	int error;
4579 
4580 	if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4581 		return (EINVAL);
4582 	error = priv_check(td, PRIV_VFS_GETFH);
4583 	if (error != 0)
4584 		return (error);
4585 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4586 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4587 	    fd);
4588 	error = namei(&nd);
4589 	if (error != 0)
4590 		return (error);
4591 	NDFREE_PNBUF(&nd);
4592 	vp = nd.ni_vp;
4593 	bzero(&fh, sizeof(fh));
4594 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4595 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4596 	vput(vp);
4597 	if (error == 0) {
4598 		if (fhseg == UIO_USERSPACE)
4599 			error = copyout(&fh, fhp, sizeof (fh));
4600 		else
4601 			memcpy(fhp, &fh, sizeof(fh));
4602 	}
4603 	return (error);
4604 }
4605 
4606 #ifndef _SYS_SYSPROTO_H_
4607 struct fhlink_args {
4608 	fhandle_t *fhp;
4609 	const char *to;
4610 };
4611 #endif
4612 int
sys_fhlink(struct thread * td,struct fhlink_args * uap)4613 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4614 {
4615 
4616 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4617 }
4618 
4619 #ifndef _SYS_SYSPROTO_H_
4620 struct fhlinkat_args {
4621 	fhandle_t *fhp;
4622 	int tofd;
4623 	const char *to;
4624 };
4625 #endif
4626 int
sys_fhlinkat(struct thread * td,struct fhlinkat_args * uap)4627 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4628 {
4629 
4630 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4631 }
4632 
4633 static int
kern_fhlinkat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,fhandle_t * fhp)4634 kern_fhlinkat(struct thread *td, int fd, const char *path,
4635     enum uio_seg pathseg, fhandle_t *fhp)
4636 {
4637 	fhandle_t fh;
4638 	struct mount *mp;
4639 	struct vnode *vp;
4640 	int error;
4641 
4642 	error = priv_check(td, PRIV_VFS_GETFH);
4643 	if (error != 0)
4644 		return (error);
4645 	error = copyin(fhp, &fh, sizeof(fh));
4646 	if (error != 0)
4647 		return (error);
4648 	do {
4649 		bwillwrite();
4650 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4651 			return (ESTALE);
4652 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4653 		vfs_unbusy(mp);
4654 		if (error != 0)
4655 			return (error);
4656 		VOP_UNLOCK(vp);
4657 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4658 	} while (error == EAGAIN || error == ERELOOKUP);
4659 	return (error);
4660 }
4661 
4662 #ifndef _SYS_SYSPROTO_H_
4663 struct fhreadlink_args {
4664 	fhandle_t *fhp;
4665 	char *buf;
4666 	size_t bufsize;
4667 };
4668 #endif
4669 int
sys_fhreadlink(struct thread * td,struct fhreadlink_args * uap)4670 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4671 {
4672 	fhandle_t fh;
4673 	struct mount *mp;
4674 	struct vnode *vp;
4675 	int error;
4676 
4677 	error = priv_check(td, PRIV_VFS_GETFH);
4678 	if (error != 0)
4679 		return (error);
4680 	if (uap->bufsize > IOSIZE_MAX)
4681 		return (EINVAL);
4682 	error = copyin(uap->fhp, &fh, sizeof(fh));
4683 	if (error != 0)
4684 		return (error);
4685 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4686 		return (ESTALE);
4687 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4688 	vfs_unbusy(mp);
4689 	if (error != 0)
4690 		return (error);
4691 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4692 	vput(vp);
4693 	return (error);
4694 }
4695 
4696 /*
4697  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4698  * open descriptor.
4699  *
4700  * warning: do not remove the priv_check() call or this becomes one giant
4701  * security hole.
4702  */
4703 #ifndef _SYS_SYSPROTO_H_
4704 struct fhopen_args {
4705 	const struct fhandle *u_fhp;
4706 	int flags;
4707 };
4708 #endif
4709 int
sys_fhopen(struct thread * td,struct fhopen_args * uap)4710 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4711 {
4712 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4713 }
4714 
4715 int
kern_fhopen(struct thread * td,const struct fhandle * u_fhp,int flags)4716 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4717 {
4718 	struct mount *mp;
4719 	struct vnode *vp;
4720 	struct fhandle fhp;
4721 	struct file *fp;
4722 	int error, indx;
4723 	bool named_attr;
4724 
4725 	error = priv_check(td, PRIV_VFS_FHOPEN);
4726 	if (error != 0)
4727 		return (error);
4728 
4729 	indx = -1;
4730 	if ((flags & O_CREAT) != 0)
4731 		return (EINVAL);
4732 	error = openflags(&flags);
4733 	if (error != 0)
4734 		return (error);
4735 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4736 	if (error != 0)
4737 		return (error);
4738 	/* find the mount point */
4739 	mp = vfs_busyfs(&fhp.fh_fsid);
4740 	if (mp == NULL)
4741 		return (ESTALE);
4742 	/* now give me my vnode, it gets returned to me locked */
4743 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4744 	vfs_unbusy(mp);
4745 	if (error != 0)
4746 		return (error);
4747 
4748 	/*
4749 	 * Check to see if the file handle refers to a named attribute
4750 	 * directory or attribute.  If it does, the O_NAMEDATTR flag
4751 	 * must have been specified.
4752 	 */
4753 	named_attr = (vn_irflag_read(vp) &
4754 	    (VIRF_NAMEDDIR | VIRF_NAMEDATTR)) != 0;
4755 	if ((named_attr && (flags & O_NAMEDATTR) == 0) ||
4756 	    (!named_attr && (flags & O_NAMEDATTR) != 0)) {
4757 		vput(vp);
4758 		return (ENOATTR);
4759 	}
4760 
4761 	error = falloc_noinstall(td, &fp);
4762 	if (error != 0) {
4763 		vput(vp);
4764 		return (error);
4765 	}
4766 	/* Set the flags early so the finit in devfs can pick them up. */
4767 	fp->f_flag = flags & FMASK;
4768 
4769 #ifdef INVARIANTS
4770 	td->td_dupfd = -1;
4771 #endif
4772 	error = vn_open_vnode(vp, flags, td->td_ucred, td, fp);
4773 	if (error != 0) {
4774 		KASSERT(fp->f_ops == &badfileops,
4775 		    ("VOP_OPEN in fhopen() set f_ops"));
4776 		KASSERT(td->td_dupfd < 0,
4777 		    ("fhopen() encountered fdopen()"));
4778 
4779 		vput(vp);
4780 		goto bad;
4781 	}
4782 #ifdef INVARIANTS
4783 	td->td_dupfd = 0;
4784 #endif
4785 	finit_open(fp, vp, flags);
4786 	VOP_UNLOCK(vp);
4787 	if ((flags & O_TRUNC) != 0) {
4788 		error = fo_truncate(fp, 0, td->td_ucred, td);
4789 		if (error != 0)
4790 			goto bad;
4791 	}
4792 
4793 	error = finstall(td, fp, &indx, flags, NULL);
4794 bad:
4795 	fdrop(fp, td);
4796 	td->td_retval[0] = indx;
4797 	return (error);
4798 }
4799 
4800 /*
4801  * Stat an (NFS) file handle.
4802  */
4803 #ifndef _SYS_SYSPROTO_H_
4804 struct fhstat_args {
4805 	struct fhandle *u_fhp;
4806 	struct stat *sb;
4807 };
4808 #endif
4809 int
sys_fhstat(struct thread * td,struct fhstat_args * uap)4810 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4811 {
4812 	struct stat sb;
4813 	struct fhandle fh;
4814 	int error;
4815 
4816 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4817 	if (error != 0)
4818 		return (error);
4819 	error = kern_fhstat(td, fh, &sb);
4820 	if (error == 0)
4821 		error = copyout(&sb, uap->sb, sizeof(sb));
4822 	return (error);
4823 }
4824 
4825 int
kern_fhstat(struct thread * td,struct fhandle fh,struct stat * sb)4826 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4827 {
4828 	struct mount *mp;
4829 	struct vnode *vp;
4830 	int error;
4831 
4832 	error = priv_check(td, PRIV_VFS_FHSTAT);
4833 	if (error != 0)
4834 		return (error);
4835 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4836 		return (ESTALE);
4837 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4838 	vfs_unbusy(mp);
4839 	if (error != 0)
4840 		return (error);
4841 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED);
4842 	vput(vp);
4843 	return (error);
4844 }
4845 
4846 /*
4847  * Implement fstatfs() for (NFS) file handles.
4848  */
4849 #ifndef _SYS_SYSPROTO_H_
4850 struct fhstatfs_args {
4851 	struct fhandle *u_fhp;
4852 	struct statfs *buf;
4853 };
4854 #endif
4855 int
sys_fhstatfs(struct thread * td,struct fhstatfs_args * uap)4856 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4857 {
4858 	struct statfs *sfp;
4859 	fhandle_t fh;
4860 	int error;
4861 
4862 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4863 	if (error != 0)
4864 		return (error);
4865 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4866 	error = kern_fhstatfs(td, fh, sfp);
4867 	if (error == 0)
4868 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4869 	free(sfp, M_STATFS);
4870 	return (error);
4871 }
4872 
4873 int
kern_fhstatfs(struct thread * td,fhandle_t fh,struct statfs * buf)4874 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4875 {
4876 	struct mount *mp;
4877 	struct vnode *vp;
4878 	int error;
4879 
4880 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4881 	if (error != 0)
4882 		return (error);
4883 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4884 		return (ESTALE);
4885 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4886 	if (error != 0) {
4887 		vfs_unbusy(mp);
4888 		return (error);
4889 	}
4890 	vput(vp);
4891 	error = prison_canseemount(td->td_ucred, mp);
4892 	if (error != 0)
4893 		goto out;
4894 #ifdef MAC
4895 	error = mac_mount_check_stat(td->td_ucred, mp);
4896 	if (error != 0)
4897 		goto out;
4898 #endif
4899 	error = VFS_STATFS(mp, buf);
4900 out:
4901 	vfs_unbusy(mp);
4902 	return (error);
4903 }
4904 
4905 /*
4906  * Unlike madvise(2), we do not make a best effort to remember every
4907  * possible caching hint.  Instead, we remember the last setting with
4908  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4909  * region of any current setting.
4910  */
4911 int
kern_posix_fadvise(struct thread * td,int fd,off_t offset,off_t len,int advice)4912 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4913     int advice)
4914 {
4915 	struct fadvise_info *fa, *new;
4916 	struct file *fp;
4917 	struct vnode *vp;
4918 	off_t end;
4919 	int error;
4920 
4921 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4922 		return (EINVAL);
4923 	AUDIT_ARG_VALUE(advice);
4924 	switch (advice) {
4925 	case POSIX_FADV_SEQUENTIAL:
4926 	case POSIX_FADV_RANDOM:
4927 	case POSIX_FADV_NOREUSE:
4928 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4929 		break;
4930 	case POSIX_FADV_NORMAL:
4931 	case POSIX_FADV_WILLNEED:
4932 	case POSIX_FADV_DONTNEED:
4933 		new = NULL;
4934 		break;
4935 	default:
4936 		return (EINVAL);
4937 	}
4938 	/* XXX: CAP_POSIX_FADVISE? */
4939 	AUDIT_ARG_FD(fd);
4940 	error = fget(td, fd, &cap_no_rights, &fp);
4941 	if (error != 0)
4942 		goto out;
4943 	AUDIT_ARG_FILE(td->td_proc, fp);
4944 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4945 		error = ESPIPE;
4946 		goto out;
4947 	}
4948 	if (fp->f_type != DTYPE_VNODE) {
4949 		error = ENODEV;
4950 		goto out;
4951 	}
4952 	vp = fp->f_vnode;
4953 	if (vp->v_type != VREG) {
4954 		error = ENODEV;
4955 		goto out;
4956 	}
4957 	if (len == 0)
4958 		end = OFF_MAX;
4959 	else
4960 		end = offset + len - 1;
4961 	switch (advice) {
4962 	case POSIX_FADV_SEQUENTIAL:
4963 	case POSIX_FADV_RANDOM:
4964 	case POSIX_FADV_NOREUSE:
4965 		/*
4966 		 * Try to merge any existing non-standard region with
4967 		 * this new region if possible, otherwise create a new
4968 		 * non-standard region for this request.
4969 		 */
4970 		mtx_pool_lock(mtxpool_sleep, fp);
4971 		fa = fp->f_advice;
4972 		if (fa != NULL && fa->fa_advice == advice &&
4973 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4974 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4975 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4976 			if (offset < fa->fa_start)
4977 				fa->fa_start = offset;
4978 			if (end > fa->fa_end)
4979 				fa->fa_end = end;
4980 		} else {
4981 			new->fa_advice = advice;
4982 			new->fa_start = offset;
4983 			new->fa_end = end;
4984 			fp->f_advice = new;
4985 			new = fa;
4986 		}
4987 		mtx_pool_unlock(mtxpool_sleep, fp);
4988 		break;
4989 	case POSIX_FADV_NORMAL:
4990 		/*
4991 		 * If a the "normal" region overlaps with an existing
4992 		 * non-standard region, trim or remove the
4993 		 * non-standard region.
4994 		 */
4995 		mtx_pool_lock(mtxpool_sleep, fp);
4996 		fa = fp->f_advice;
4997 		if (fa != NULL) {
4998 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4999 				new = fa;
5000 				fp->f_advice = NULL;
5001 			} else if (offset <= fa->fa_start &&
5002 			    end >= fa->fa_start)
5003 				fa->fa_start = end + 1;
5004 			else if (offset <= fa->fa_end && end >= fa->fa_end)
5005 				fa->fa_end = offset - 1;
5006 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
5007 				/*
5008 				 * If the "normal" region is a middle
5009 				 * portion of the existing
5010 				 * non-standard region, just remove
5011 				 * the whole thing rather than picking
5012 				 * one side or the other to
5013 				 * preserve.
5014 				 */
5015 				new = fa;
5016 				fp->f_advice = NULL;
5017 			}
5018 		}
5019 		mtx_pool_unlock(mtxpool_sleep, fp);
5020 		break;
5021 	case POSIX_FADV_WILLNEED:
5022 	case POSIX_FADV_DONTNEED:
5023 		error = VOP_ADVISE(vp, offset, end, advice);
5024 		break;
5025 	}
5026 out:
5027 	if (fp != NULL)
5028 		fdrop(fp, td);
5029 	free(new, M_FADVISE);
5030 	return (error);
5031 }
5032 
5033 int
sys_posix_fadvise(struct thread * td,struct posix_fadvise_args * uap)5034 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
5035 {
5036 	int error;
5037 
5038 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
5039 	    uap->advice);
5040 	return (kern_posix_error(td, error));
5041 }
5042 
5043 int
kern_copy_file_range(struct thread * td,int infd,off_t * inoffp,int outfd,off_t * outoffp,size_t len,unsigned int flags)5044 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
5045     off_t *outoffp, size_t len, unsigned int flags)
5046 {
5047 	struct file *infp, *infp1, *outfp, *outfp1;
5048 	struct vnode *invp, *outvp;
5049 	int error;
5050 	size_t retlen;
5051 	void *rl_rcookie, *rl_wcookie;
5052 	off_t inoff, outoff, savinoff, savoutoff;
5053 	bool foffsets_locked;
5054 
5055 	infp = outfp = NULL;
5056 	rl_rcookie = rl_wcookie = NULL;
5057 	foffsets_locked = false;
5058 	error = 0;
5059 	retlen = 0;
5060 
5061 	if (flags != 0) {
5062 		error = EINVAL;
5063 		goto out;
5064 	}
5065 	if (len > SSIZE_MAX)
5066 		/*
5067 		 * Although the len argument is size_t, the return argument
5068 		 * is ssize_t (which is signed).  Therefore a size that won't
5069 		 * fit in ssize_t can't be returned.
5070 		 */
5071 		len = SSIZE_MAX;
5072 
5073 	/* Get the file structures for the file descriptors. */
5074 	error = fget_read(td, infd,
5075 	    inoffp != NULL ? &cap_pread_rights : &cap_read_rights, &infp);
5076 	if (error != 0)
5077 		goto out;
5078 	if (infp->f_ops == &badfileops) {
5079 		error = EBADF;
5080 		goto out;
5081 	}
5082 	if (infp->f_vnode == NULL) {
5083 		error = EINVAL;
5084 		goto out;
5085 	}
5086 	error = fget_write(td, outfd,
5087 	    outoffp != NULL ? &cap_pwrite_rights : &cap_write_rights, &outfp);
5088 	if (error != 0)
5089 		goto out;
5090 	if (outfp->f_ops == &badfileops) {
5091 		error = EBADF;
5092 		goto out;
5093 	}
5094 	if (outfp->f_vnode == NULL) {
5095 		error = EINVAL;
5096 		goto out;
5097 	}
5098 
5099 	/*
5100 	 * Figure out which file offsets we're reading from and writing to.
5101 	 * If the offsets come from the file descriptions, we need to lock them,
5102 	 * and locking both offsets requires a loop to avoid deadlocks.
5103 	 */
5104 	infp1 = outfp1 = NULL;
5105 	if (inoffp != NULL)
5106 		inoff = *inoffp;
5107 	else
5108 		infp1 = infp;
5109 	if (outoffp != NULL)
5110 		outoff = *outoffp;
5111 	else
5112 		outfp1 = outfp;
5113 	if (infp1 != NULL || outfp1 != NULL) {
5114 		if (infp1 == outfp1) {
5115 			/*
5116 			 * Overlapping ranges are not allowed.  A more thorough
5117 			 * check appears below, but we must not lock the same
5118 			 * offset twice.
5119 			 */
5120 			error = EINVAL;
5121 			goto out;
5122 		}
5123 		foffset_lock_pair(infp1, &inoff, outfp1, &outoff, 0);
5124 		foffsets_locked = true;
5125 	}
5126 	savinoff = inoff;
5127 	savoutoff = outoff;
5128 
5129 	invp = infp->f_vnode;
5130 	outvp = outfp->f_vnode;
5131 	/* Sanity check the f_flag bits. */
5132 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
5133 	    (infp->f_flag & FREAD) == 0) {
5134 		error = EBADF;
5135 		goto out;
5136 	}
5137 
5138 	/* If len == 0, just return 0. */
5139 	if (len == 0)
5140 		goto out;
5141 
5142 	/*
5143 	 * Make sure that the ranges we check and lock below are valid.  Note
5144 	 * that len is clamped to SSIZE_MAX above.
5145 	 */
5146 	if (inoff < 0 || outoff < 0) {
5147 		error = EINVAL;
5148 		goto out;
5149 	}
5150 
5151 	/*
5152 	 * If infp and outfp refer to the same file, the byte ranges cannot
5153 	 * overlap.
5154 	 */
5155 	if (invp == outvp) {
5156 		if ((inoff <= outoff && inoff + len > outoff) ||
5157 		    (inoff > outoff && outoff + len > inoff)) {
5158 			error = EINVAL;
5159 			goto out;
5160 		}
5161 		rangelock_may_recurse(&invp->v_rl);
5162 	}
5163 
5164 	/* Range lock the byte ranges for both invp and outvp. */
5165 	for (;;) {
5166 		rl_wcookie = vn_rangelock_wlock(outvp, outoff, outoff + len);
5167 		rl_rcookie = vn_rangelock_tryrlock(invp, inoff, inoff + len);
5168 		if (rl_rcookie != NULL)
5169 			break;
5170 		vn_rangelock_unlock(outvp, rl_wcookie);
5171 		rl_rcookie = vn_rangelock_rlock(invp, inoff, inoff + len);
5172 		vn_rangelock_unlock(invp, rl_rcookie);
5173 	}
5174 
5175 	retlen = len;
5176 	error = vn_copy_file_range(invp, &inoff, outvp, &outoff, &retlen,
5177 	    flags, infp->f_cred, outfp->f_cred, td);
5178 out:
5179 	if (rl_rcookie != NULL)
5180 		vn_rangelock_unlock(invp, rl_rcookie);
5181 	if (rl_wcookie != NULL)
5182 		vn_rangelock_unlock(outvp, rl_wcookie);
5183 	if (foffsets_locked) {
5184 		if (error == EINTR || error == ERESTART) {
5185 			inoff = savinoff;
5186 			outoff = savoutoff;
5187 		}
5188 		if (inoffp == NULL)
5189 			foffset_unlock(infp, inoff, 0);
5190 		else
5191 			*inoffp = inoff;
5192 		if (outoffp == NULL)
5193 			foffset_unlock(outfp, outoff, 0);
5194 		else
5195 			*outoffp = outoff;
5196 	}
5197 	if (outfp != NULL)
5198 		fdrop(outfp, td);
5199 	if (infp != NULL)
5200 		fdrop(infp, td);
5201 	td->td_retval[0] = retlen;
5202 	return (error);
5203 }
5204 
5205 int
sys_copy_file_range(struct thread * td,struct copy_file_range_args * uap)5206 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
5207 {
5208 	off_t inoff, outoff, *inoffp, *outoffp;
5209 	int error;
5210 
5211 	inoffp = outoffp = NULL;
5212 	if (uap->inoffp != NULL) {
5213 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
5214 		if (error != 0)
5215 			return (error);
5216 		inoffp = &inoff;
5217 	}
5218 	if (uap->outoffp != NULL) {
5219 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
5220 		if (error != 0)
5221 			return (error);
5222 		outoffp = &outoff;
5223 	}
5224 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
5225 	    outoffp, uap->len, uap->flags);
5226 	if (error == 0 && uap->inoffp != NULL)
5227 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
5228 	if (error == 0 && uap->outoffp != NULL)
5229 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
5230 	return (error);
5231 }
5232