xref: /trueos/sys/kern/vfs_syscalls.c (revision f7da1a0d21213ff4c813bf5d4544fbe28242a686)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_kdtrace.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capsicum.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <ufs/ufs/quota.h>
91 
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 
94 SDT_PROVIDER_DEFINE(vfs);
95 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
96 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
97 
98 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int kern_chflags(struct thread *td, const char *path,
101     enum uio_seg pathseg, u_long flags);
102 static int kern_chflagsat(struct thread *td, int fd, const char *path,
103     enum uio_seg pathseg, u_long flags, int atflag);
104 static int setfflags(struct thread *td, struct vnode *, u_long);
105 static int setutimes(struct thread *td, struct vnode *,
106     const struct timespec *, int, int);
107 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
108     struct thread *td);
109 
110 /*
111  * The module initialization routine for POSIX asynchronous I/O will
112  * set this to the version of AIO that it implements.  (Zero means
113  * that it is not implemented.)  This value is used here by pathconf()
114  * and in kern_descrip.c by fpathconf().
115  */
116 int async_io_version;
117 
118 /*
119  * Sync each mounted filesystem.
120  */
121 #ifndef _SYS_SYSPROTO_H_
122 struct sync_args {
123 	int     dummy;
124 };
125 #endif
126 /* ARGSUSED */
127 int
sys_sync(td,uap)128 sys_sync(td, uap)
129 	struct thread *td;
130 	struct sync_args *uap;
131 {
132 	struct mount *mp, *nmp;
133 	int save;
134 
135 	mtx_lock(&mountlist_mtx);
136 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
137 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
138 			nmp = TAILQ_NEXT(mp, mnt_list);
139 			continue;
140 		}
141 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
142 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
143 			save = curthread_pflags_set(TDP_SYNCIO);
144 			vfs_msync(mp, MNT_NOWAIT);
145 			VFS_SYNC(mp, MNT_NOWAIT);
146 			curthread_pflags_restore(save);
147 			vn_finished_write(mp);
148 		}
149 		mtx_lock(&mountlist_mtx);
150 		nmp = TAILQ_NEXT(mp, mnt_list);
151 		vfs_unbusy(mp);
152 	}
153 	mtx_unlock(&mountlist_mtx);
154 	return (0);
155 }
156 
157 /*
158  * Change filesystem quotas.
159  */
160 #ifndef _SYS_SYSPROTO_H_
161 struct quotactl_args {
162 	char *path;
163 	int cmd;
164 	int uid;
165 	caddr_t arg;
166 };
167 #endif
168 int
sys_quotactl(td,uap)169 sys_quotactl(td, uap)
170 	struct thread *td;
171 	register struct quotactl_args /* {
172 		char *path;
173 		int cmd;
174 		int uid;
175 		caddr_t arg;
176 	} */ *uap;
177 {
178 	struct mount *mp;
179 	struct nameidata nd;
180 	int error;
181 
182 	AUDIT_ARG_CMD(uap->cmd);
183 	AUDIT_ARG_UID(uap->uid);
184 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
185 		return (EPERM);
186 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
187 	    uap->path, td);
188 	if ((error = namei(&nd)) != 0)
189 		return (error);
190 	NDFREE(&nd, NDF_ONLY_PNBUF);
191 	mp = nd.ni_vp->v_mount;
192 	vfs_ref(mp);
193 	vput(nd.ni_vp);
194 	error = vfs_busy(mp, 0);
195 	vfs_rel(mp);
196 	if (error != 0)
197 		return (error);
198 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
199 
200 	/*
201 	 * Since quota on operation typically needs to open quota
202 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
203 	 * before calling into namei.  Otherwise, unmount might be
204 	 * started between two vfs_busy() invocations (first is our,
205 	 * second is from mount point cross-walk code in lookup()),
206 	 * causing deadlock.
207 	 *
208 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
209 	 * its own, always returning with ubusied mount point.
210 	 */
211 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
212 		vfs_unbusy(mp);
213 	return (error);
214 }
215 
216 /*
217  * Used by statfs conversion routines to scale the block size up if
218  * necessary so that all of the block counts are <= 'max_size'.  Note
219  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
220  * value of 'n'.
221  */
222 void
statfs_scale_blocks(struct statfs * sf,long max_size)223 statfs_scale_blocks(struct statfs *sf, long max_size)
224 {
225 	uint64_t count;
226 	int shift;
227 
228 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
229 
230 	/*
231 	 * Attempt to scale the block counts to give a more accurate
232 	 * overview to userland of the ratio of free space to used
233 	 * space.  To do this, find the largest block count and compute
234 	 * a divisor that lets it fit into a signed integer <= max_size.
235 	 */
236 	if (sf->f_bavail < 0)
237 		count = -sf->f_bavail;
238 	else
239 		count = sf->f_bavail;
240 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
241 	if (count <= max_size)
242 		return;
243 
244 	count >>= flsl(max_size);
245 	shift = 0;
246 	while (count > 0) {
247 		shift++;
248 		count >>=1;
249 	}
250 
251 	sf->f_bsize <<= shift;
252 	sf->f_blocks >>= shift;
253 	sf->f_bfree >>= shift;
254 	sf->f_bavail >>= shift;
255 }
256 
257 /*
258  * Get filesystem statistics.
259  */
260 #ifndef _SYS_SYSPROTO_H_
261 struct statfs_args {
262 	char *path;
263 	struct statfs *buf;
264 };
265 #endif
266 int
sys_statfs(td,uap)267 sys_statfs(td, uap)
268 	struct thread *td;
269 	register struct statfs_args /* {
270 		char *path;
271 		struct statfs *buf;
272 	} */ *uap;
273 {
274 	struct statfs sf;
275 	int error;
276 
277 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
278 	if (error == 0)
279 		error = copyout(&sf, uap->buf, sizeof(sf));
280 	return (error);
281 }
282 
283 int
kern_statfs(struct thread * td,char * path,enum uio_seg pathseg,struct statfs * buf)284 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
285     struct statfs *buf)
286 {
287 	struct mount *mp;
288 	struct statfs *sp, sb;
289 	struct nameidata nd;
290 	int error;
291 
292 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
293 	    pathseg, path, td);
294 	error = namei(&nd);
295 	if (error != 0)
296 		return (error);
297 	mp = nd.ni_vp->v_mount;
298 	vfs_ref(mp);
299 	NDFREE(&nd, NDF_ONLY_PNBUF);
300 	vput(nd.ni_vp);
301 	error = vfs_busy(mp, 0);
302 	vfs_rel(mp);
303 	if (error != 0)
304 		return (error);
305 #ifdef MAC
306 	error = mac_mount_check_stat(td->td_ucred, mp);
307 	if (error != 0)
308 		goto out;
309 #endif
310 	/*
311 	 * Set these in case the underlying filesystem fails to do so.
312 	 */
313 	sp = &mp->mnt_stat;
314 	sp->f_version = STATFS_VERSION;
315 	sp->f_namemax = NAME_MAX;
316 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
317 	error = VFS_STATFS(mp, sp);
318 	if (error != 0)
319 		goto out;
320 	if (priv_check(td, PRIV_VFS_GENERATION)) {
321 		bcopy(sp, &sb, sizeof(sb));
322 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
323 		prison_enforce_statfs(td->td_ucred, mp, &sb);
324 		sp = &sb;
325 	}
326 	*buf = *sp;
327 out:
328 	vfs_unbusy(mp);
329 	return (error);
330 }
331 
332 /*
333  * Get filesystem statistics.
334  */
335 #ifndef _SYS_SYSPROTO_H_
336 struct fstatfs_args {
337 	int fd;
338 	struct statfs *buf;
339 };
340 #endif
341 int
sys_fstatfs(td,uap)342 sys_fstatfs(td, uap)
343 	struct thread *td;
344 	register struct fstatfs_args /* {
345 		int fd;
346 		struct statfs *buf;
347 	} */ *uap;
348 {
349 	struct statfs sf;
350 	int error;
351 
352 	error = kern_fstatfs(td, uap->fd, &sf);
353 	if (error == 0)
354 		error = copyout(&sf, uap->buf, sizeof(sf));
355 	return (error);
356 }
357 
358 int
kern_fstatfs(struct thread * td,int fd,struct statfs * buf)359 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
360 {
361 	struct file *fp;
362 	struct mount *mp;
363 	struct statfs *sp, sb;
364 	struct vnode *vp;
365 	cap_rights_t rights;
366 	int error;
367 
368 	AUDIT_ARG_FD(fd);
369 	error = getvnode(td->td_proc->p_fd, fd,
370 	    cap_rights_init(&rights, CAP_FSTATFS), &fp);
371 	if (error != 0)
372 		return (error);
373 	vp = fp->f_vnode;
374 	vn_lock(vp, LK_SHARED | LK_RETRY);
375 #ifdef AUDIT
376 	AUDIT_ARG_VNODE1(vp);
377 #endif
378 	mp = vp->v_mount;
379 	if (mp)
380 		vfs_ref(mp);
381 	VOP_UNLOCK(vp, 0);
382 	fdrop(fp, td);
383 	if (mp == NULL) {
384 		error = EBADF;
385 		goto out;
386 	}
387 	error = vfs_busy(mp, 0);
388 	vfs_rel(mp);
389 	if (error != 0)
390 		return (error);
391 #ifdef MAC
392 	error = mac_mount_check_stat(td->td_ucred, mp);
393 	if (error != 0)
394 		goto out;
395 #endif
396 	/*
397 	 * Set these in case the underlying filesystem fails to do so.
398 	 */
399 	sp = &mp->mnt_stat;
400 	sp->f_version = STATFS_VERSION;
401 	sp->f_namemax = NAME_MAX;
402 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
403 	error = VFS_STATFS(mp, sp);
404 	if (error != 0)
405 		goto out;
406 	if (priv_check(td, PRIV_VFS_GENERATION)) {
407 		bcopy(sp, &sb, sizeof(sb));
408 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
409 		prison_enforce_statfs(td->td_ucred, mp, &sb);
410 		sp = &sb;
411 	}
412 	*buf = *sp;
413 out:
414 	if (mp)
415 		vfs_unbusy(mp);
416 	return (error);
417 }
418 
419 /*
420  * Get statistics on all filesystems.
421  */
422 #ifndef _SYS_SYSPROTO_H_
423 struct getfsstat_args {
424 	struct statfs *buf;
425 	long bufsize;
426 	int flags;
427 };
428 #endif
429 int
sys_getfsstat(td,uap)430 sys_getfsstat(td, uap)
431 	struct thread *td;
432 	register struct getfsstat_args /* {
433 		struct statfs *buf;
434 		long bufsize;
435 		int flags;
436 	} */ *uap;
437 {
438 
439 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
440 	    uap->flags));
441 }
442 
443 /*
444  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
445  *	The caller is responsible for freeing memory which will be allocated
446  *	in '*buf'.
447  */
448 int
kern_getfsstat(struct thread * td,struct statfs ** buf,size_t bufsize,enum uio_seg bufseg,int flags)449 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
450     enum uio_seg bufseg, int flags)
451 {
452 	struct mount *mp, *nmp;
453 	struct statfs *sfsp, *sp, sb;
454 	size_t count, maxcount;
455 	int error;
456 
457 	maxcount = bufsize / sizeof(struct statfs);
458 	if (bufsize == 0)
459 		sfsp = NULL;
460 	else if (bufseg == UIO_USERSPACE)
461 		sfsp = *buf;
462 	else /* if (bufseg == UIO_SYSSPACE) */ {
463 		count = 0;
464 		mtx_lock(&mountlist_mtx);
465 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
466 			count++;
467 		}
468 		mtx_unlock(&mountlist_mtx);
469 		if (maxcount > count)
470 			maxcount = count;
471 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
472 		    M_WAITOK);
473 	}
474 	count = 0;
475 	mtx_lock(&mountlist_mtx);
476 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
477 		if (prison_canseemount(td->td_ucred, mp) != 0) {
478 			nmp = TAILQ_NEXT(mp, mnt_list);
479 			continue;
480 		}
481 #ifdef MAC
482 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
483 			nmp = TAILQ_NEXT(mp, mnt_list);
484 			continue;
485 		}
486 #endif
487 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
488 			nmp = TAILQ_NEXT(mp, mnt_list);
489 			continue;
490 		}
491 		if (sfsp && count < maxcount) {
492 			sp = &mp->mnt_stat;
493 			/*
494 			 * Set these in case the underlying filesystem
495 			 * fails to do so.
496 			 */
497 			sp->f_version = STATFS_VERSION;
498 			sp->f_namemax = NAME_MAX;
499 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
500 			/*
501 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
502 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
503 			 * overrides MNT_WAIT.
504 			 */
505 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
506 			    (flags & MNT_WAIT)) &&
507 			    (error = VFS_STATFS(mp, sp))) {
508 				mtx_lock(&mountlist_mtx);
509 				nmp = TAILQ_NEXT(mp, mnt_list);
510 				vfs_unbusy(mp);
511 				continue;
512 			}
513 			if (priv_check(td, PRIV_VFS_GENERATION)) {
514 				bcopy(sp, &sb, sizeof(sb));
515 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
516 				prison_enforce_statfs(td->td_ucred, mp, &sb);
517 				sp = &sb;
518 			}
519 			if (bufseg == UIO_SYSSPACE)
520 				bcopy(sp, sfsp, sizeof(*sp));
521 			else /* if (bufseg == UIO_USERSPACE) */ {
522 				error = copyout(sp, sfsp, sizeof(*sp));
523 				if (error != 0) {
524 					vfs_unbusy(mp);
525 					return (error);
526 				}
527 			}
528 			sfsp++;
529 		}
530 		count++;
531 		mtx_lock(&mountlist_mtx);
532 		nmp = TAILQ_NEXT(mp, mnt_list);
533 		vfs_unbusy(mp);
534 	}
535 	mtx_unlock(&mountlist_mtx);
536 	if (sfsp && count > maxcount)
537 		td->td_retval[0] = maxcount;
538 	else
539 		td->td_retval[0] = count;
540 	return (0);
541 }
542 
543 #ifdef COMPAT_FREEBSD4
544 /*
545  * Get old format filesystem statistics.
546  */
547 static void cvtstatfs(struct statfs *, struct ostatfs *);
548 
549 #ifndef _SYS_SYSPROTO_H_
550 struct freebsd4_statfs_args {
551 	char *path;
552 	struct ostatfs *buf;
553 };
554 #endif
555 int
freebsd4_statfs(td,uap)556 freebsd4_statfs(td, uap)
557 	struct thread *td;
558 	struct freebsd4_statfs_args /* {
559 		char *path;
560 		struct ostatfs *buf;
561 	} */ *uap;
562 {
563 	struct ostatfs osb;
564 	struct statfs sf;
565 	int error;
566 
567 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
568 	if (error != 0)
569 		return (error);
570 	cvtstatfs(&sf, &osb);
571 	return (copyout(&osb, uap->buf, sizeof(osb)));
572 }
573 
574 /*
575  * Get filesystem statistics.
576  */
577 #ifndef _SYS_SYSPROTO_H_
578 struct freebsd4_fstatfs_args {
579 	int fd;
580 	struct ostatfs *buf;
581 };
582 #endif
583 int
freebsd4_fstatfs(td,uap)584 freebsd4_fstatfs(td, uap)
585 	struct thread *td;
586 	struct freebsd4_fstatfs_args /* {
587 		int fd;
588 		struct ostatfs *buf;
589 	} */ *uap;
590 {
591 	struct ostatfs osb;
592 	struct statfs sf;
593 	int error;
594 
595 	error = kern_fstatfs(td, uap->fd, &sf);
596 	if (error != 0)
597 		return (error);
598 	cvtstatfs(&sf, &osb);
599 	return (copyout(&osb, uap->buf, sizeof(osb)));
600 }
601 
602 /*
603  * Get statistics on all filesystems.
604  */
605 #ifndef _SYS_SYSPROTO_H_
606 struct freebsd4_getfsstat_args {
607 	struct ostatfs *buf;
608 	long bufsize;
609 	int flags;
610 };
611 #endif
612 int
freebsd4_getfsstat(td,uap)613 freebsd4_getfsstat(td, uap)
614 	struct thread *td;
615 	register struct freebsd4_getfsstat_args /* {
616 		struct ostatfs *buf;
617 		long bufsize;
618 		int flags;
619 	} */ *uap;
620 {
621 	struct statfs *buf, *sp;
622 	struct ostatfs osb;
623 	size_t count, size;
624 	int error;
625 
626 	count = uap->bufsize / sizeof(struct ostatfs);
627 	size = count * sizeof(struct statfs);
628 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
629 	if (size > 0) {
630 		count = td->td_retval[0];
631 		sp = buf;
632 		while (count > 0 && error == 0) {
633 			cvtstatfs(sp, &osb);
634 			error = copyout(&osb, uap->buf, sizeof(osb));
635 			sp++;
636 			uap->buf++;
637 			count--;
638 		}
639 		free(buf, M_TEMP);
640 	}
641 	return (error);
642 }
643 
644 /*
645  * Implement fstatfs() for (NFS) file handles.
646  */
647 #ifndef _SYS_SYSPROTO_H_
648 struct freebsd4_fhstatfs_args {
649 	struct fhandle *u_fhp;
650 	struct ostatfs *buf;
651 };
652 #endif
653 int
freebsd4_fhstatfs(td,uap)654 freebsd4_fhstatfs(td, uap)
655 	struct thread *td;
656 	struct freebsd4_fhstatfs_args /* {
657 		struct fhandle *u_fhp;
658 		struct ostatfs *buf;
659 	} */ *uap;
660 {
661 	struct ostatfs osb;
662 	struct statfs sf;
663 	fhandle_t fh;
664 	int error;
665 
666 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
667 	if (error != 0)
668 		return (error);
669 	error = kern_fhstatfs(td, fh, &sf);
670 	if (error != 0)
671 		return (error);
672 	cvtstatfs(&sf, &osb);
673 	return (copyout(&osb, uap->buf, sizeof(osb)));
674 }
675 
676 /*
677  * Convert a new format statfs structure to an old format statfs structure.
678  */
679 static void
cvtstatfs(nsp,osp)680 cvtstatfs(nsp, osp)
681 	struct statfs *nsp;
682 	struct ostatfs *osp;
683 {
684 
685 	statfs_scale_blocks(nsp, LONG_MAX);
686 	bzero(osp, sizeof(*osp));
687 	osp->f_bsize = nsp->f_bsize;
688 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
689 	osp->f_blocks = nsp->f_blocks;
690 	osp->f_bfree = nsp->f_bfree;
691 	osp->f_bavail = nsp->f_bavail;
692 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
693 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
694 	osp->f_owner = nsp->f_owner;
695 	osp->f_type = nsp->f_type;
696 	osp->f_flags = nsp->f_flags;
697 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
698 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
699 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
700 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
701 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
702 	    MIN(MFSNAMELEN, OMFSNAMELEN));
703 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
704 	    MIN(MNAMELEN, OMNAMELEN));
705 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
706 	    MIN(MNAMELEN, OMNAMELEN));
707 	osp->f_fsid = nsp->f_fsid;
708 }
709 #endif /* COMPAT_FREEBSD4 */
710 
711 /*
712  * Change current working directory to a given file descriptor.
713  */
714 #ifndef _SYS_SYSPROTO_H_
715 struct fchdir_args {
716 	int	fd;
717 };
718 #endif
719 int
sys_fchdir(td,uap)720 sys_fchdir(td, uap)
721 	struct thread *td;
722 	struct fchdir_args /* {
723 		int fd;
724 	} */ *uap;
725 {
726 	register struct filedesc *fdp = td->td_proc->p_fd;
727 	struct vnode *vp, *tdp, *vpold;
728 	struct mount *mp;
729 	struct file *fp;
730 	cap_rights_t rights;
731 	int error;
732 
733 	AUDIT_ARG_FD(uap->fd);
734 	error = getvnode(fdp, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
735 	    &fp);
736 	if (error != 0)
737 		return (error);
738 	vp = fp->f_vnode;
739 	VREF(vp);
740 	fdrop(fp, td);
741 	vn_lock(vp, LK_SHARED | LK_RETRY);
742 	AUDIT_ARG_VNODE1(vp);
743 	error = change_dir(vp, td);
744 	while (!error && (mp = vp->v_mountedhere) != NULL) {
745 		if (vfs_busy(mp, 0))
746 			continue;
747 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
748 		vfs_unbusy(mp);
749 		if (error != 0)
750 			break;
751 		vput(vp);
752 		vp = tdp;
753 	}
754 	if (error != 0) {
755 		vput(vp);
756 		return (error);
757 	}
758 	VOP_UNLOCK(vp, 0);
759 	FILEDESC_XLOCK(fdp);
760 	vpold = fdp->fd_cdir;
761 	fdp->fd_cdir = vp;
762 	FILEDESC_XUNLOCK(fdp);
763 	vrele(vpold);
764 	return (0);
765 }
766 
767 /*
768  * Change current working directory (``.'').
769  */
770 #ifndef _SYS_SYSPROTO_H_
771 struct chdir_args {
772 	char	*path;
773 };
774 #endif
775 int
sys_chdir(td,uap)776 sys_chdir(td, uap)
777 	struct thread *td;
778 	struct chdir_args /* {
779 		char *path;
780 	} */ *uap;
781 {
782 
783 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
784 }
785 
786 int
kern_chdir(struct thread * td,char * path,enum uio_seg pathseg)787 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
788 {
789 	register struct filedesc *fdp = td->td_proc->p_fd;
790 	struct nameidata nd;
791 	struct vnode *vp;
792 	int error;
793 
794 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
795 	    pathseg, path, td);
796 	if ((error = namei(&nd)) != 0)
797 		return (error);
798 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
799 		vput(nd.ni_vp);
800 		NDFREE(&nd, NDF_ONLY_PNBUF);
801 		return (error);
802 	}
803 	VOP_UNLOCK(nd.ni_vp, 0);
804 	NDFREE(&nd, NDF_ONLY_PNBUF);
805 	FILEDESC_XLOCK(fdp);
806 	vp = fdp->fd_cdir;
807 	fdp->fd_cdir = nd.ni_vp;
808 	FILEDESC_XUNLOCK(fdp);
809 	vrele(vp);
810 	return (0);
811 }
812 
813 /*
814  * Helper function for raised chroot(2) security function:  Refuse if
815  * any filedescriptors are open directories.
816  */
817 static int
chroot_refuse_vdir_fds(fdp)818 chroot_refuse_vdir_fds(fdp)
819 	struct filedesc *fdp;
820 {
821 	struct vnode *vp;
822 	struct file *fp;
823 	int fd;
824 
825 	FILEDESC_LOCK_ASSERT(fdp);
826 
827 	for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
828 		fp = fget_locked(fdp, fd);
829 		if (fp == NULL)
830 			continue;
831 		if (fp->f_type == DTYPE_VNODE) {
832 			vp = fp->f_vnode;
833 			if (vp->v_type == VDIR)
834 				return (EPERM);
835 		}
836 	}
837 	return (0);
838 }
839 
840 /*
841  * This sysctl determines if we will allow a process to chroot(2) if it
842  * has a directory open:
843  *	0: disallowed for all processes.
844  *	1: allowed for processes that were not already chroot(2)'ed.
845  *	2: allowed for all processes.
846  */
847 
848 static int chroot_allow_open_directories = 1;
849 
850 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
851      &chroot_allow_open_directories, 0,
852      "Allow a process to chroot(2) if it has a directory open");
853 
854 /*
855  * Change notion of root (``/'') directory.
856  */
857 #ifndef _SYS_SYSPROTO_H_
858 struct chroot_args {
859 	char	*path;
860 };
861 #endif
862 int
sys_chroot(td,uap)863 sys_chroot(td, uap)
864 	struct thread *td;
865 	struct chroot_args /* {
866 		char *path;
867 	} */ *uap;
868 {
869 	struct nameidata nd;
870 	int error;
871 
872 	error = priv_check(td, PRIV_VFS_CHROOT);
873 	if (error != 0)
874 		return (error);
875 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
876 	    UIO_USERSPACE, uap->path, td);
877 	error = namei(&nd);
878 	if (error != 0)
879 		goto error;
880 	error = change_dir(nd.ni_vp, td);
881 	if (error != 0)
882 		goto e_vunlock;
883 #ifdef MAC
884 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
885 	if (error != 0)
886 		goto e_vunlock;
887 #endif
888 	VOP_UNLOCK(nd.ni_vp, 0);
889 	error = change_root(nd.ni_vp, td);
890 	vrele(nd.ni_vp);
891 	NDFREE(&nd, NDF_ONLY_PNBUF);
892 	return (error);
893 e_vunlock:
894 	vput(nd.ni_vp);
895 error:
896 	NDFREE(&nd, NDF_ONLY_PNBUF);
897 	return (error);
898 }
899 
900 /*
901  * Common routine for chroot and chdir.  Callers must provide a locked vnode
902  * instance.
903  */
904 int
change_dir(vp,td)905 change_dir(vp, td)
906 	struct vnode *vp;
907 	struct thread *td;
908 {
909 #ifdef MAC
910 	int error;
911 #endif
912 
913 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
914 	if (vp->v_type != VDIR)
915 		return (ENOTDIR);
916 #ifdef MAC
917 	error = mac_vnode_check_chdir(td->td_ucred, vp);
918 	if (error != 0)
919 		return (error);
920 #endif
921 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
922 }
923 
924 /*
925  * Common routine for kern_chroot() and jail_attach().  The caller is
926  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
927  * authorize this operation.
928  */
929 int
change_root(vp,td)930 change_root(vp, td)
931 	struct vnode *vp;
932 	struct thread *td;
933 {
934 	struct filedesc *fdp;
935 	struct vnode *oldvp;
936 	int error;
937 
938 	fdp = td->td_proc->p_fd;
939 	FILEDESC_XLOCK(fdp);
940 	if (chroot_allow_open_directories == 0 ||
941 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
942 		error = chroot_refuse_vdir_fds(fdp);
943 		if (error != 0) {
944 			FILEDESC_XUNLOCK(fdp);
945 			return (error);
946 		}
947 	}
948 	oldvp = fdp->fd_rdir;
949 	fdp->fd_rdir = vp;
950 	VREF(fdp->fd_rdir);
951 	if (!fdp->fd_jdir) {
952 		fdp->fd_jdir = vp;
953 		VREF(fdp->fd_jdir);
954 	}
955 	FILEDESC_XUNLOCK(fdp);
956 	vrele(oldvp);
957 	return (0);
958 }
959 
960 static __inline void
flags_to_rights(int flags,cap_rights_t * rightsp)961 flags_to_rights(int flags, cap_rights_t *rightsp)
962 {
963 
964 	if (flags & O_EXEC) {
965 		cap_rights_set(rightsp, CAP_FEXECVE);
966 	} else {
967 		switch ((flags & O_ACCMODE)) {
968 		case O_RDONLY:
969 			cap_rights_set(rightsp, CAP_READ);
970 			break;
971 		case O_RDWR:
972 			cap_rights_set(rightsp, CAP_READ);
973 			/* FALLTHROUGH */
974 		case O_WRONLY:
975 			cap_rights_set(rightsp, CAP_WRITE);
976 			if (!(flags & (O_APPEND | O_TRUNC)))
977 				cap_rights_set(rightsp, CAP_SEEK);
978 			break;
979 		}
980 	}
981 
982 	if (flags & O_CREAT)
983 		cap_rights_set(rightsp, CAP_CREATE);
984 
985 	if (flags & O_TRUNC)
986 		cap_rights_set(rightsp, CAP_FTRUNCATE);
987 
988 	if (flags & (O_SYNC | O_FSYNC))
989 		cap_rights_set(rightsp, CAP_FSYNC);
990 
991 	if (flags & (O_EXLOCK | O_SHLOCK))
992 		cap_rights_set(rightsp, CAP_FLOCK);
993 }
994 
995 /*
996  * Check permissions, allocate an open file structure, and call the device
997  * open routine if any.
998  */
999 #ifndef _SYS_SYSPROTO_H_
1000 struct open_args {
1001 	char	*path;
1002 	int	flags;
1003 	int	mode;
1004 };
1005 #endif
1006 int
sys_open(td,uap)1007 sys_open(td, uap)
1008 	struct thread *td;
1009 	register struct open_args /* {
1010 		char *path;
1011 		int flags;
1012 		int mode;
1013 	} */ *uap;
1014 {
1015 
1016 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1017 }
1018 
1019 #ifndef _SYS_SYSPROTO_H_
1020 struct openat_args {
1021 	int	fd;
1022 	char	*path;
1023 	int	flag;
1024 	int	mode;
1025 };
1026 #endif
1027 int
sys_openat(struct thread * td,struct openat_args * uap)1028 sys_openat(struct thread *td, struct openat_args *uap)
1029 {
1030 
1031 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1032 	    uap->mode));
1033 }
1034 
1035 int
kern_open(struct thread * td,char * path,enum uio_seg pathseg,int flags,int mode)1036 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1037     int mode)
1038 {
1039 
1040 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1041 }
1042 
1043 int
kern_openat(struct thread * td,int fd,char * path,enum uio_seg pathseg,int flags,int mode)1044 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1045     int flags, int mode)
1046 {
1047 	struct proc *p = td->td_proc;
1048 	struct filedesc *fdp = p->p_fd;
1049 	struct file *fp;
1050 	struct vnode *vp;
1051 	struct nameidata nd;
1052 	cap_rights_t rights;
1053 	int cmode, error, indx;
1054 
1055 	indx = -1;
1056 
1057 	AUDIT_ARG_FFLAGS(flags);
1058 	AUDIT_ARG_MODE(mode);
1059 	/* XXX: audit dirfd */
1060 	cap_rights_init(&rights, CAP_LOOKUP);
1061 	flags_to_rights(flags, &rights);
1062 	/*
1063 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1064 	 * may be specified.
1065 	 */
1066 	if (flags & O_EXEC) {
1067 		if (flags & O_ACCMODE)
1068 			return (EINVAL);
1069 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1070 		return (EINVAL);
1071 	} else {
1072 		flags = FFLAGS(flags);
1073 	}
1074 
1075 	/*
1076 	 * Allocate the file descriptor, but don't install a descriptor yet.
1077 	 */
1078 	error = falloc_noinstall(td, &fp);
1079 	if (error != 0)
1080 		return (error);
1081 	/*
1082 	 * An extra reference on `fp' has been held for us by
1083 	 * falloc_noinstall().
1084 	 */
1085 	/* Set the flags early so the finit in devfs can pick them up. */
1086 	fp->f_flag = flags & FMASK;
1087 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1088 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1089 	    &rights, td);
1090 	td->td_dupfd = -1;		/* XXX check for fdopen */
1091 	error = vn_open(&nd, &flags, cmode, fp);
1092 	if (error != 0) {
1093 		/*
1094 		 * If the vn_open replaced the method vector, something
1095 		 * wonderous happened deep below and we just pass it up
1096 		 * pretending we know what we do.
1097 		 */
1098 		if (error == ENXIO && fp->f_ops != &badfileops)
1099 			goto success;
1100 
1101 		/*
1102 		 * Handle special fdopen() case. bleh.
1103 		 *
1104 		 * Don't do this for relative (capability) lookups; we don't
1105 		 * understand exactly what would happen, and we don't think
1106 		 * that it ever should.
1107 		 */
1108 		if (nd.ni_strictrelative == 0 &&
1109 		    (error == ENODEV || error == ENXIO) &&
1110 		    td->td_dupfd >= 0) {
1111 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1112 			    &indx);
1113 			if (error == 0)
1114 				goto success;
1115 		}
1116 
1117 		goto bad;
1118 	}
1119 	td->td_dupfd = 0;
1120 	NDFREE(&nd, NDF_ONLY_PNBUF);
1121 	vp = nd.ni_vp;
1122 
1123 	/*
1124 	 * Store the vnode, for any f_type. Typically, the vnode use
1125 	 * count is decremented by direct call to vn_closefile() for
1126 	 * files that switched type in the cdevsw fdopen() method.
1127 	 */
1128 	fp->f_vnode = vp;
1129 	/*
1130 	 * If the file wasn't claimed by devfs bind it to the normal
1131 	 * vnode operations here.
1132 	 */
1133 	if (fp->f_ops == &badfileops) {
1134 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1135 		fp->f_seqcount = 1;
1136 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1137 		    DTYPE_VNODE, vp, &vnops);
1138 	}
1139 
1140 	VOP_UNLOCK(vp, 0);
1141 	if (flags & O_TRUNC) {
1142 		error = fo_truncate(fp, 0, td->td_ucred, td);
1143 		if (error != 0)
1144 			goto bad;
1145 	}
1146 success:
1147 	/*
1148 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1149 	 */
1150 	if (indx == -1) {
1151 		struct filecaps *fcaps;
1152 
1153 #ifdef CAPABILITIES
1154 		if (nd.ni_strictrelative == 1)
1155 			fcaps = &nd.ni_filecaps;
1156 		else
1157 #endif
1158 			fcaps = NULL;
1159 		error = finstall(td, fp, &indx, flags, fcaps);
1160 		/* On success finstall() consumes fcaps. */
1161 		if (error != 0) {
1162 			filecaps_free(&nd.ni_filecaps);
1163 			goto bad;
1164 		}
1165 	} else {
1166 		filecaps_free(&nd.ni_filecaps);
1167 	}
1168 
1169 	/*
1170 	 * Release our private reference, leaving the one associated with
1171 	 * the descriptor table intact.
1172 	 */
1173 	fdrop(fp, td);
1174 	td->td_retval[0] = indx;
1175 	return (0);
1176 bad:
1177 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1178 	fdrop(fp, td);
1179 	return (error);
1180 }
1181 
1182 #ifdef COMPAT_43
1183 /*
1184  * Create a file.
1185  */
1186 #ifndef _SYS_SYSPROTO_H_
1187 struct ocreat_args {
1188 	char	*path;
1189 	int	mode;
1190 };
1191 #endif
1192 int
ocreat(td,uap)1193 ocreat(td, uap)
1194 	struct thread *td;
1195 	register struct ocreat_args /* {
1196 		char *path;
1197 		int mode;
1198 	} */ *uap;
1199 {
1200 
1201 	return (kern_open(td, uap->path, UIO_USERSPACE,
1202 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1203 }
1204 #endif /* COMPAT_43 */
1205 
1206 /*
1207  * Create a special file.
1208  */
1209 #ifndef _SYS_SYSPROTO_H_
1210 struct mknod_args {
1211 	char	*path;
1212 	int	mode;
1213 	int	dev;
1214 };
1215 #endif
1216 int
sys_mknod(td,uap)1217 sys_mknod(td, uap)
1218 	struct thread *td;
1219 	register struct mknod_args /* {
1220 		char *path;
1221 		int mode;
1222 		int dev;
1223 	} */ *uap;
1224 {
1225 
1226 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1227 }
1228 
1229 #ifndef _SYS_SYSPROTO_H_
1230 struct mknodat_args {
1231 	int	fd;
1232 	char	*path;
1233 	mode_t	mode;
1234 	dev_t	dev;
1235 };
1236 #endif
1237 int
sys_mknodat(struct thread * td,struct mknodat_args * uap)1238 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1239 {
1240 
1241 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1242 	    uap->dev));
1243 }
1244 
1245 int
kern_mknod(struct thread * td,char * path,enum uio_seg pathseg,int mode,int dev)1246 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1247     int dev)
1248 {
1249 
1250 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1251 }
1252 
1253 int
kern_mknodat(struct thread * td,int fd,char * path,enum uio_seg pathseg,int mode,int dev)1254 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1255     int mode, int dev)
1256 {
1257 	struct vnode *vp;
1258 	struct mount *mp;
1259 	struct vattr vattr;
1260 	struct nameidata nd;
1261 	cap_rights_t rights;
1262 	int error, whiteout = 0;
1263 
1264 	AUDIT_ARG_MODE(mode);
1265 	AUDIT_ARG_DEV(dev);
1266 	switch (mode & S_IFMT) {
1267 	case S_IFCHR:
1268 	case S_IFBLK:
1269 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1270 		break;
1271 	case S_IFMT:
1272 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1273 		break;
1274 	case S_IFWHT:
1275 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1276 		break;
1277 	case S_IFIFO:
1278 		if (dev == 0)
1279 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1280 		/* FALLTHROUGH */
1281 	default:
1282 		error = EINVAL;
1283 		break;
1284 	}
1285 	if (error != 0)
1286 		return (error);
1287 restart:
1288 	bwillwrite();
1289 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1290 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1291 	    td);
1292 	if ((error = namei(&nd)) != 0)
1293 		return (error);
1294 	vp = nd.ni_vp;
1295 	if (vp != NULL) {
1296 		NDFREE(&nd, NDF_ONLY_PNBUF);
1297 		if (vp == nd.ni_dvp)
1298 			vrele(nd.ni_dvp);
1299 		else
1300 			vput(nd.ni_dvp);
1301 		vrele(vp);
1302 		return (EEXIST);
1303 	} else {
1304 		VATTR_NULL(&vattr);
1305 		vattr.va_mode = (mode & ALLPERMS) &
1306 		    ~td->td_proc->p_fd->fd_cmask;
1307 		vattr.va_rdev = dev;
1308 		whiteout = 0;
1309 
1310 		switch (mode & S_IFMT) {
1311 		case S_IFMT:	/* used by badsect to flag bad sectors */
1312 			vattr.va_type = VBAD;
1313 			break;
1314 		case S_IFCHR:
1315 			vattr.va_type = VCHR;
1316 			break;
1317 		case S_IFBLK:
1318 			vattr.va_type = VBLK;
1319 			break;
1320 		case S_IFWHT:
1321 			whiteout = 1;
1322 			break;
1323 		default:
1324 			panic("kern_mknod: invalid mode");
1325 		}
1326 	}
1327 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1328 		NDFREE(&nd, NDF_ONLY_PNBUF);
1329 		vput(nd.ni_dvp);
1330 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1331 			return (error);
1332 		goto restart;
1333 	}
1334 #ifdef MAC
1335 	if (error == 0 && !whiteout)
1336 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1337 		    &nd.ni_cnd, &vattr);
1338 #endif
1339 	if (error == 0) {
1340 		if (whiteout)
1341 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1342 		else {
1343 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1344 						&nd.ni_cnd, &vattr);
1345 			if (error == 0)
1346 				vput(nd.ni_vp);
1347 		}
1348 	}
1349 	NDFREE(&nd, NDF_ONLY_PNBUF);
1350 	vput(nd.ni_dvp);
1351 	vn_finished_write(mp);
1352 	return (error);
1353 }
1354 
1355 /*
1356  * Create a named pipe.
1357  */
1358 #ifndef _SYS_SYSPROTO_H_
1359 struct mkfifo_args {
1360 	char	*path;
1361 	int	mode;
1362 };
1363 #endif
1364 int
sys_mkfifo(td,uap)1365 sys_mkfifo(td, uap)
1366 	struct thread *td;
1367 	register struct mkfifo_args /* {
1368 		char *path;
1369 		int mode;
1370 	} */ *uap;
1371 {
1372 
1373 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1374 }
1375 
1376 #ifndef _SYS_SYSPROTO_H_
1377 struct mkfifoat_args {
1378 	int	fd;
1379 	char	*path;
1380 	mode_t	mode;
1381 };
1382 #endif
1383 int
sys_mkfifoat(struct thread * td,struct mkfifoat_args * uap)1384 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1385 {
1386 
1387 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1388 	    uap->mode));
1389 }
1390 
1391 int
kern_mkfifo(struct thread * td,char * path,enum uio_seg pathseg,int mode)1392 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1393 {
1394 
1395 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1396 }
1397 
1398 int
kern_mkfifoat(struct thread * td,int fd,char * path,enum uio_seg pathseg,int mode)1399 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1400     int mode)
1401 {
1402 	struct mount *mp;
1403 	struct vattr vattr;
1404 	struct nameidata nd;
1405 	cap_rights_t rights;
1406 	int error;
1407 
1408 	AUDIT_ARG_MODE(mode);
1409 restart:
1410 	bwillwrite();
1411 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1412 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1413 	    td);
1414 	if ((error = namei(&nd)) != 0)
1415 		return (error);
1416 	if (nd.ni_vp != NULL) {
1417 		NDFREE(&nd, NDF_ONLY_PNBUF);
1418 		if (nd.ni_vp == nd.ni_dvp)
1419 			vrele(nd.ni_dvp);
1420 		else
1421 			vput(nd.ni_dvp);
1422 		vrele(nd.ni_vp);
1423 		return (EEXIST);
1424 	}
1425 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1426 		NDFREE(&nd, NDF_ONLY_PNBUF);
1427 		vput(nd.ni_dvp);
1428 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1429 			return (error);
1430 		goto restart;
1431 	}
1432 	VATTR_NULL(&vattr);
1433 	vattr.va_type = VFIFO;
1434 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1435 #ifdef MAC
1436 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1437 	    &vattr);
1438 	if (error != 0)
1439 		goto out;
1440 #endif
1441 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1442 	if (error == 0)
1443 		vput(nd.ni_vp);
1444 #ifdef MAC
1445 out:
1446 #endif
1447 	vput(nd.ni_dvp);
1448 	vn_finished_write(mp);
1449 	NDFREE(&nd, NDF_ONLY_PNBUF);
1450 	return (error);
1451 }
1452 
1453 /*
1454  * Make a hard file link.
1455  */
1456 #ifndef _SYS_SYSPROTO_H_
1457 struct link_args {
1458 	char	*path;
1459 	char	*link;
1460 };
1461 #endif
1462 int
sys_link(td,uap)1463 sys_link(td, uap)
1464 	struct thread *td;
1465 	register struct link_args /* {
1466 		char *path;
1467 		char *link;
1468 	} */ *uap;
1469 {
1470 
1471 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1472 }
1473 
1474 #ifndef _SYS_SYSPROTO_H_
1475 struct linkat_args {
1476 	int	fd1;
1477 	char	*path1;
1478 	int	fd2;
1479 	char	*path2;
1480 	int	flag;
1481 };
1482 #endif
1483 int
sys_linkat(struct thread * td,struct linkat_args * uap)1484 sys_linkat(struct thread *td, struct linkat_args *uap)
1485 {
1486 	int flag;
1487 
1488 	flag = uap->flag;
1489 	if (flag & ~AT_SYMLINK_FOLLOW)
1490 		return (EINVAL);
1491 
1492 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1493 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1494 }
1495 
1496 int hardlink_check_uid = 0;
1497 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1498     &hardlink_check_uid, 0,
1499     "Unprivileged processes cannot create hard links to files owned by other "
1500     "users");
1501 static int hardlink_check_gid = 0;
1502 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1503     &hardlink_check_gid, 0,
1504     "Unprivileged processes cannot create hard links to files owned by other "
1505     "groups");
1506 
1507 static int
can_hardlink(struct vnode * vp,struct ucred * cred)1508 can_hardlink(struct vnode *vp, struct ucred *cred)
1509 {
1510 	struct vattr va;
1511 	int error;
1512 
1513 	if (!hardlink_check_uid && !hardlink_check_gid)
1514 		return (0);
1515 
1516 	error = VOP_GETATTR(vp, &va, cred);
1517 	if (error != 0)
1518 		return (error);
1519 
1520 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1521 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1522 		if (error != 0)
1523 			return (error);
1524 	}
1525 
1526 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1527 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1528 		if (error != 0)
1529 			return (error);
1530 	}
1531 
1532 	return (0);
1533 }
1534 
1535 int
kern_link(struct thread * td,char * path,char * link,enum uio_seg segflg)1536 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1537 {
1538 
1539 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1540 }
1541 
1542 int
kern_linkat(struct thread * td,int fd1,int fd2,char * path1,char * path2,enum uio_seg segflg,int follow)1543 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1544     enum uio_seg segflg, int follow)
1545 {
1546 	struct vnode *vp;
1547 	struct mount *mp;
1548 	struct nameidata nd;
1549 	cap_rights_t rights;
1550 	int error;
1551 
1552 again:
1553 	bwillwrite();
1554 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1555 
1556 	if ((error = namei(&nd)) != 0)
1557 		return (error);
1558 	NDFREE(&nd, NDF_ONLY_PNBUF);
1559 	vp = nd.ni_vp;
1560 	if (vp->v_type == VDIR) {
1561 		vrele(vp);
1562 		return (EPERM);		/* POSIX */
1563 	}
1564 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2 |
1565 	    NOCACHE, segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT),
1566 	    td);
1567 	if ((error = namei(&nd)) == 0) {
1568 		if (nd.ni_vp != NULL) {
1569 			NDFREE(&nd, NDF_ONLY_PNBUF);
1570 			if (nd.ni_dvp == nd.ni_vp)
1571 				vrele(nd.ni_dvp);
1572 			else
1573 				vput(nd.ni_dvp);
1574 			vrele(nd.ni_vp);
1575 			vrele(vp);
1576 			return (EEXIST);
1577 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1578 			/*
1579 			 * Cross-device link.  No need to recheck
1580 			 * vp->v_type, since it cannot change, except
1581 			 * to VBAD.
1582 			 */
1583 			NDFREE(&nd, NDF_ONLY_PNBUF);
1584 			vput(nd.ni_dvp);
1585 			vrele(vp);
1586 			return (EXDEV);
1587 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1588 			error = can_hardlink(vp, td->td_ucred);
1589 #ifdef MAC
1590 			if (error == 0)
1591 				error = mac_vnode_check_link(td->td_ucred,
1592 				    nd.ni_dvp, vp, &nd.ni_cnd);
1593 #endif
1594 			if (error != 0) {
1595 				vput(vp);
1596 				vput(nd.ni_dvp);
1597 				NDFREE(&nd, NDF_ONLY_PNBUF);
1598 				return (error);
1599 			}
1600 			error = vn_start_write(vp, &mp, V_NOWAIT);
1601 			if (error != 0) {
1602 				vput(vp);
1603 				vput(nd.ni_dvp);
1604 				NDFREE(&nd, NDF_ONLY_PNBUF);
1605 				error = vn_start_write(NULL, &mp,
1606 				    V_XSLEEP | PCATCH);
1607 				if (error != 0)
1608 					return (error);
1609 				goto again;
1610 			}
1611 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1612 			VOP_UNLOCK(vp, 0);
1613 			vput(nd.ni_dvp);
1614 			vn_finished_write(mp);
1615 			NDFREE(&nd, NDF_ONLY_PNBUF);
1616 		} else {
1617 			vput(nd.ni_dvp);
1618 			NDFREE(&nd, NDF_ONLY_PNBUF);
1619 			vrele(vp);
1620 			goto again;
1621 		}
1622 	}
1623 	vrele(vp);
1624 	return (error);
1625 }
1626 
1627 /*
1628  * Make a symbolic link.
1629  */
1630 #ifndef _SYS_SYSPROTO_H_
1631 struct symlink_args {
1632 	char	*path;
1633 	char	*link;
1634 };
1635 #endif
1636 int
sys_symlink(td,uap)1637 sys_symlink(td, uap)
1638 	struct thread *td;
1639 	register struct symlink_args /* {
1640 		char *path;
1641 		char *link;
1642 	} */ *uap;
1643 {
1644 
1645 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1646 }
1647 
1648 #ifndef _SYS_SYSPROTO_H_
1649 struct symlinkat_args {
1650 	char	*path;
1651 	int	fd;
1652 	char	*path2;
1653 };
1654 #endif
1655 int
sys_symlinkat(struct thread * td,struct symlinkat_args * uap)1656 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1657 {
1658 
1659 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1660 	    UIO_USERSPACE));
1661 }
1662 
1663 int
kern_symlink(struct thread * td,char * path,char * link,enum uio_seg segflg)1664 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1665 {
1666 
1667 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1668 }
1669 
1670 int
kern_symlinkat(struct thread * td,char * path1,int fd,char * path2,enum uio_seg segflg)1671 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1672     enum uio_seg segflg)
1673 {
1674 	struct mount *mp;
1675 	struct vattr vattr;
1676 	char *syspath;
1677 	struct nameidata nd;
1678 	int error;
1679 	cap_rights_t rights;
1680 
1681 	if (segflg == UIO_SYSSPACE) {
1682 		syspath = path1;
1683 	} else {
1684 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1685 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1686 			goto out;
1687 	}
1688 	AUDIT_ARG_TEXT(syspath);
1689 restart:
1690 	bwillwrite();
1691 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1692 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1693 	    td);
1694 	if ((error = namei(&nd)) != 0)
1695 		goto out;
1696 	if (nd.ni_vp) {
1697 		NDFREE(&nd, NDF_ONLY_PNBUF);
1698 		if (nd.ni_vp == nd.ni_dvp)
1699 			vrele(nd.ni_dvp);
1700 		else
1701 			vput(nd.ni_dvp);
1702 		vrele(nd.ni_vp);
1703 		error = EEXIST;
1704 		goto out;
1705 	}
1706 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1707 		NDFREE(&nd, NDF_ONLY_PNBUF);
1708 		vput(nd.ni_dvp);
1709 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1710 			goto out;
1711 		goto restart;
1712 	}
1713 	VATTR_NULL(&vattr);
1714 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1715 #ifdef MAC
1716 	vattr.va_type = VLNK;
1717 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1718 	    &vattr);
1719 	if (error != 0)
1720 		goto out2;
1721 #endif
1722 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1723 	if (error == 0)
1724 		vput(nd.ni_vp);
1725 #ifdef MAC
1726 out2:
1727 #endif
1728 	NDFREE(&nd, NDF_ONLY_PNBUF);
1729 	vput(nd.ni_dvp);
1730 	vn_finished_write(mp);
1731 out:
1732 	if (segflg != UIO_SYSSPACE)
1733 		uma_zfree(namei_zone, syspath);
1734 	return (error);
1735 }
1736 
1737 /*
1738  * Delete a whiteout from the filesystem.
1739  */
1740 int
sys_undelete(td,uap)1741 sys_undelete(td, uap)
1742 	struct thread *td;
1743 	register struct undelete_args /* {
1744 		char *path;
1745 	} */ *uap;
1746 {
1747 	struct mount *mp;
1748 	struct nameidata nd;
1749 	int error;
1750 
1751 restart:
1752 	bwillwrite();
1753 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1754 	    UIO_USERSPACE, uap->path, td);
1755 	error = namei(&nd);
1756 	if (error != 0)
1757 		return (error);
1758 
1759 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1760 		NDFREE(&nd, NDF_ONLY_PNBUF);
1761 		if (nd.ni_vp == nd.ni_dvp)
1762 			vrele(nd.ni_dvp);
1763 		else
1764 			vput(nd.ni_dvp);
1765 		if (nd.ni_vp)
1766 			vrele(nd.ni_vp);
1767 		return (EEXIST);
1768 	}
1769 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1770 		NDFREE(&nd, NDF_ONLY_PNBUF);
1771 		vput(nd.ni_dvp);
1772 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1773 			return (error);
1774 		goto restart;
1775 	}
1776 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1777 	NDFREE(&nd, NDF_ONLY_PNBUF);
1778 	vput(nd.ni_dvp);
1779 	vn_finished_write(mp);
1780 	return (error);
1781 }
1782 
1783 /*
1784  * Delete a name from the filesystem.
1785  */
1786 #ifndef _SYS_SYSPROTO_H_
1787 struct unlink_args {
1788 	char	*path;
1789 };
1790 #endif
1791 int
sys_unlink(td,uap)1792 sys_unlink(td, uap)
1793 	struct thread *td;
1794 	struct unlink_args /* {
1795 		char *path;
1796 	} */ *uap;
1797 {
1798 
1799 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1800 }
1801 
1802 #ifndef _SYS_SYSPROTO_H_
1803 struct unlinkat_args {
1804 	int	fd;
1805 	char	*path;
1806 	int	flag;
1807 };
1808 #endif
1809 int
sys_unlinkat(struct thread * td,struct unlinkat_args * uap)1810 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1811 {
1812 	int flag = uap->flag;
1813 	int fd = uap->fd;
1814 	char *path = uap->path;
1815 
1816 	if (flag & ~AT_REMOVEDIR)
1817 		return (EINVAL);
1818 
1819 	if (flag & AT_REMOVEDIR)
1820 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1821 	else
1822 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1823 }
1824 
1825 int
kern_unlink(struct thread * td,char * path,enum uio_seg pathseg)1826 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1827 {
1828 
1829 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1830 }
1831 
1832 int
kern_unlinkat(struct thread * td,int fd,char * path,enum uio_seg pathseg,ino_t oldinum)1833 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1834     ino_t oldinum)
1835 {
1836 	struct mount *mp;
1837 	struct vnode *vp;
1838 	struct nameidata nd;
1839 	struct stat sb;
1840 	cap_rights_t rights;
1841 	int error;
1842 
1843 restart:
1844 	bwillwrite();
1845 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1846 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1847 	if ((error = namei(&nd)) != 0)
1848 		return (error == EINVAL ? EPERM : error);
1849 	vp = nd.ni_vp;
1850 	if (vp->v_type == VDIR && oldinum == 0) {
1851 		error = EPERM;		/* POSIX */
1852 	} else if (oldinum != 0 &&
1853 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1854 		  sb.st_ino != oldinum) {
1855 			error = EIDRM;	/* Identifier removed */
1856 	} else {
1857 		/*
1858 		 * The root of a mounted filesystem cannot be deleted.
1859 		 *
1860 		 * XXX: can this only be a VDIR case?
1861 		 */
1862 		if (vp->v_vflag & VV_ROOT)
1863 			error = EBUSY;
1864 	}
1865 	if (error == 0) {
1866 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1867 			NDFREE(&nd, NDF_ONLY_PNBUF);
1868 			vput(nd.ni_dvp);
1869 			if (vp == nd.ni_dvp)
1870 				vrele(vp);
1871 			else
1872 				vput(vp);
1873 			if ((error = vn_start_write(NULL, &mp,
1874 			    V_XSLEEP | PCATCH)) != 0)
1875 				return (error);
1876 			goto restart;
1877 		}
1878 #ifdef MAC
1879 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1880 		    &nd.ni_cnd);
1881 		if (error != 0)
1882 			goto out;
1883 #endif
1884 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1885 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1886 #ifdef MAC
1887 out:
1888 #endif
1889 		vn_finished_write(mp);
1890 	}
1891 	NDFREE(&nd, NDF_ONLY_PNBUF);
1892 	vput(nd.ni_dvp);
1893 	if (vp == nd.ni_dvp)
1894 		vrele(vp);
1895 	else
1896 		vput(vp);
1897 	return (error);
1898 }
1899 
1900 /*
1901  * Reposition read/write file offset.
1902  */
1903 #ifndef _SYS_SYSPROTO_H_
1904 struct lseek_args {
1905 	int	fd;
1906 	int	pad;
1907 	off_t	offset;
1908 	int	whence;
1909 };
1910 #endif
1911 int
sys_lseek(td,uap)1912 sys_lseek(td, uap)
1913 	struct thread *td;
1914 	register struct lseek_args /* {
1915 		int fd;
1916 		int pad;
1917 		off_t offset;
1918 		int whence;
1919 	} */ *uap;
1920 {
1921 	struct file *fp;
1922 	cap_rights_t rights;
1923 	int error;
1924 
1925 	AUDIT_ARG_FD(uap->fd);
1926 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1927 	if (error != 0)
1928 		return (error);
1929 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1930 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1931 	fdrop(fp, td);
1932 	return (error);
1933 }
1934 
1935 #if defined(COMPAT_43)
1936 /*
1937  * Reposition read/write file offset.
1938  */
1939 #ifndef _SYS_SYSPROTO_H_
1940 struct olseek_args {
1941 	int	fd;
1942 	long	offset;
1943 	int	whence;
1944 };
1945 #endif
1946 int
olseek(td,uap)1947 olseek(td, uap)
1948 	struct thread *td;
1949 	register struct olseek_args /* {
1950 		int fd;
1951 		long offset;
1952 		int whence;
1953 	} */ *uap;
1954 {
1955 	struct lseek_args /* {
1956 		int fd;
1957 		int pad;
1958 		off_t offset;
1959 		int whence;
1960 	} */ nuap;
1961 
1962 	nuap.fd = uap->fd;
1963 	nuap.offset = uap->offset;
1964 	nuap.whence = uap->whence;
1965 	return (sys_lseek(td, &nuap));
1966 }
1967 #endif /* COMPAT_43 */
1968 
1969 /* Version with the 'pad' argument */
1970 int
freebsd6_lseek(td,uap)1971 freebsd6_lseek(td, uap)
1972 	struct thread *td;
1973 	register struct freebsd6_lseek_args *uap;
1974 {
1975 	struct lseek_args ouap;
1976 
1977 	ouap.fd = uap->fd;
1978 	ouap.offset = uap->offset;
1979 	ouap.whence = uap->whence;
1980 	return (sys_lseek(td, &ouap));
1981 }
1982 
1983 /*
1984  * Check access permissions using passed credentials.
1985  */
1986 static int
vn_access(vp,user_flags,cred,td)1987 vn_access(vp, user_flags, cred, td)
1988 	struct vnode	*vp;
1989 	int		user_flags;
1990 	struct ucred	*cred;
1991 	struct thread	*td;
1992 {
1993 	accmode_t accmode;
1994 	int error;
1995 
1996 	/* Flags == 0 means only check for existence. */
1997 	error = 0;
1998 	if (user_flags) {
1999 		accmode = 0;
2000 		if (user_flags & R_OK)
2001 			accmode |= VREAD;
2002 		if (user_flags & W_OK)
2003 			accmode |= VWRITE;
2004 		if (user_flags & X_OK)
2005 			accmode |= VEXEC;
2006 #ifdef MAC
2007 		error = mac_vnode_check_access(cred, vp, accmode);
2008 		if (error != 0)
2009 			return (error);
2010 #endif
2011 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2012 			error = VOP_ACCESS(vp, accmode, cred, td);
2013 	}
2014 	return (error);
2015 }
2016 
2017 /*
2018  * Check access permissions using "real" credentials.
2019  */
2020 #ifndef _SYS_SYSPROTO_H_
2021 struct access_args {
2022 	char	*path;
2023 	int	amode;
2024 };
2025 #endif
2026 int
sys_access(td,uap)2027 sys_access(td, uap)
2028 	struct thread *td;
2029 	register struct access_args /* {
2030 		char *path;
2031 		int amode;
2032 	} */ *uap;
2033 {
2034 
2035 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2036 }
2037 
2038 #ifndef _SYS_SYSPROTO_H_
2039 struct faccessat_args {
2040 	int	dirfd;
2041 	char	*path;
2042 	int	amode;
2043 	int	flag;
2044 }
2045 #endif
2046 int
sys_faccessat(struct thread * td,struct faccessat_args * uap)2047 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2048 {
2049 
2050 	if (uap->flag & ~AT_EACCESS)
2051 		return (EINVAL);
2052 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2053 	    uap->amode));
2054 }
2055 
2056 int
kern_access(struct thread * td,char * path,enum uio_seg pathseg,int amode)2057 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2058 {
2059 
2060 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2061 }
2062 
2063 int
kern_accessat(struct thread * td,int fd,char * path,enum uio_seg pathseg,int flag,int amode)2064 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2065     int flag, int amode)
2066 {
2067 	struct ucred *cred, *tmpcred;
2068 	struct vnode *vp;
2069 	struct nameidata nd;
2070 	cap_rights_t rights;
2071 	int error;
2072 
2073 	/*
2074 	 * Create and modify a temporary credential instead of one that
2075 	 * is potentially shared.
2076 	 */
2077 	if (!(flag & AT_EACCESS)) {
2078 		cred = td->td_ucred;
2079 		tmpcred = crdup(cred);
2080 		tmpcred->cr_uid = cred->cr_ruid;
2081 		tmpcred->cr_groups[0] = cred->cr_rgid;
2082 		td->td_ucred = tmpcred;
2083 	} else
2084 		cred = tmpcred = td->td_ucred;
2085 	AUDIT_ARG_VALUE(amode);
2086 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2087 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
2088 	    td);
2089 	if ((error = namei(&nd)) != 0)
2090 		goto out1;
2091 	vp = nd.ni_vp;
2092 
2093 	error = vn_access(vp, amode, tmpcred, td);
2094 	NDFREE(&nd, NDF_ONLY_PNBUF);
2095 	vput(vp);
2096 out1:
2097 	if (!(flag & AT_EACCESS)) {
2098 		td->td_ucred = cred;
2099 		crfree(tmpcred);
2100 	}
2101 	return (error);
2102 }
2103 
2104 /*
2105  * Check access permissions using "effective" credentials.
2106  */
2107 #ifndef _SYS_SYSPROTO_H_
2108 struct eaccess_args {
2109 	char	*path;
2110 	int	amode;
2111 };
2112 #endif
2113 int
sys_eaccess(td,uap)2114 sys_eaccess(td, uap)
2115 	struct thread *td;
2116 	register struct eaccess_args /* {
2117 		char *path;
2118 		int amode;
2119 	} */ *uap;
2120 {
2121 
2122 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2123 }
2124 
2125 int
kern_eaccess(struct thread * td,char * path,enum uio_seg pathseg,int amode)2126 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2127 {
2128 
2129 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2130 }
2131 
2132 #if defined(COMPAT_43)
2133 /*
2134  * Get file status; this version follows links.
2135  */
2136 #ifndef _SYS_SYSPROTO_H_
2137 struct ostat_args {
2138 	char	*path;
2139 	struct ostat *ub;
2140 };
2141 #endif
2142 int
ostat(td,uap)2143 ostat(td, uap)
2144 	struct thread *td;
2145 	register struct ostat_args /* {
2146 		char *path;
2147 		struct ostat *ub;
2148 	} */ *uap;
2149 {
2150 	struct stat sb;
2151 	struct ostat osb;
2152 	int error;
2153 
2154 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2155 	if (error != 0)
2156 		return (error);
2157 	cvtstat(&sb, &osb);
2158 	return (copyout(&osb, uap->ub, sizeof (osb)));
2159 }
2160 
2161 /*
2162  * Get file status; this version does not follow links.
2163  */
2164 #ifndef _SYS_SYSPROTO_H_
2165 struct olstat_args {
2166 	char	*path;
2167 	struct ostat *ub;
2168 };
2169 #endif
2170 int
olstat(td,uap)2171 olstat(td, uap)
2172 	struct thread *td;
2173 	register struct olstat_args /* {
2174 		char *path;
2175 		struct ostat *ub;
2176 	} */ *uap;
2177 {
2178 	struct stat sb;
2179 	struct ostat osb;
2180 	int error;
2181 
2182 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2183 	if (error != 0)
2184 		return (error);
2185 	cvtstat(&sb, &osb);
2186 	return (copyout(&osb, uap->ub, sizeof (osb)));
2187 }
2188 
2189 /*
2190  * Convert from an old to a new stat structure.
2191  */
2192 void
cvtstat(st,ost)2193 cvtstat(st, ost)
2194 	struct stat *st;
2195 	struct ostat *ost;
2196 {
2197 
2198 	ost->st_dev = st->st_dev;
2199 	ost->st_ino = st->st_ino;
2200 	ost->st_mode = st->st_mode;
2201 	ost->st_nlink = st->st_nlink;
2202 	ost->st_uid = st->st_uid;
2203 	ost->st_gid = st->st_gid;
2204 	ost->st_rdev = st->st_rdev;
2205 	if (st->st_size < (quad_t)1 << 32)
2206 		ost->st_size = st->st_size;
2207 	else
2208 		ost->st_size = -2;
2209 	ost->st_atim = st->st_atim;
2210 	ost->st_mtim = st->st_mtim;
2211 	ost->st_ctim = st->st_ctim;
2212 	ost->st_blksize = st->st_blksize;
2213 	ost->st_blocks = st->st_blocks;
2214 	ost->st_flags = st->st_flags;
2215 	ost->st_gen = st->st_gen;
2216 }
2217 #endif /* COMPAT_43 */
2218 
2219 /*
2220  * Get file status; this version follows links.
2221  */
2222 #ifndef _SYS_SYSPROTO_H_
2223 struct stat_args {
2224 	char	*path;
2225 	struct stat *ub;
2226 };
2227 #endif
2228 int
sys_stat(td,uap)2229 sys_stat(td, uap)
2230 	struct thread *td;
2231 	register struct stat_args /* {
2232 		char *path;
2233 		struct stat *ub;
2234 	} */ *uap;
2235 {
2236 	struct stat sb;
2237 	int error;
2238 
2239 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2240 	if (error == 0)
2241 		error = copyout(&sb, uap->ub, sizeof (sb));
2242 	return (error);
2243 }
2244 
2245 #ifndef _SYS_SYSPROTO_H_
2246 struct fstatat_args {
2247 	int	fd;
2248 	char	*path;
2249 	struct stat	*buf;
2250 	int	flag;
2251 }
2252 #endif
2253 int
sys_fstatat(struct thread * td,struct fstatat_args * uap)2254 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2255 {
2256 	struct stat sb;
2257 	int error;
2258 
2259 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2260 	    UIO_USERSPACE, &sb);
2261 	if (error == 0)
2262 		error = copyout(&sb, uap->buf, sizeof (sb));
2263 	return (error);
2264 }
2265 
2266 int
kern_stat(struct thread * td,char * path,enum uio_seg pathseg,struct stat * sbp)2267 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2268 {
2269 
2270 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2271 }
2272 
2273 int
kern_statat(struct thread * td,int flag,int fd,char * path,enum uio_seg pathseg,struct stat * sbp)2274 kern_statat(struct thread *td, int flag, int fd, char *path,
2275     enum uio_seg pathseg, struct stat *sbp)
2276 {
2277 
2278 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2279 }
2280 
2281 int
kern_statat_vnhook(struct thread * td,int flag,int fd,char * path,enum uio_seg pathseg,struct stat * sbp,void (* hook)(struct vnode * vp,struct stat * sbp))2282 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2283     enum uio_seg pathseg, struct stat *sbp,
2284     void (*hook)(struct vnode *vp, struct stat *sbp))
2285 {
2286 	struct nameidata nd;
2287 	struct stat sb;
2288 	cap_rights_t rights;
2289 	int error;
2290 
2291 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2292 		return (EINVAL);
2293 
2294 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2295 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2296 	    cap_rights_init(&rights, CAP_FSTAT), td);
2297 
2298 	if ((error = namei(&nd)) != 0)
2299 		return (error);
2300 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2301 	if (error == 0) {
2302 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2303 		if (S_ISREG(sb.st_mode))
2304 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2305 		if (__predict_false(hook != NULL))
2306 			hook(nd.ni_vp, &sb);
2307 	}
2308 	NDFREE(&nd, NDF_ONLY_PNBUF);
2309 	vput(nd.ni_vp);
2310 	if (error != 0)
2311 		return (error);
2312 	*sbp = sb;
2313 #ifdef KTRACE
2314 	if (KTRPOINT(td, KTR_STRUCT))
2315 		ktrstat(&sb);
2316 #endif
2317 	return (0);
2318 }
2319 
2320 /*
2321  * Get file status; this version does not follow links.
2322  */
2323 #ifndef _SYS_SYSPROTO_H_
2324 struct lstat_args {
2325 	char	*path;
2326 	struct stat *ub;
2327 };
2328 #endif
2329 int
sys_lstat(td,uap)2330 sys_lstat(td, uap)
2331 	struct thread *td;
2332 	register struct lstat_args /* {
2333 		char *path;
2334 		struct stat *ub;
2335 	} */ *uap;
2336 {
2337 	struct stat sb;
2338 	int error;
2339 
2340 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2341 	if (error == 0)
2342 		error = copyout(&sb, uap->ub, sizeof (sb));
2343 	return (error);
2344 }
2345 
2346 int
kern_lstat(struct thread * td,char * path,enum uio_seg pathseg,struct stat * sbp)2347 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2348 {
2349 
2350 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2351 	    sbp));
2352 }
2353 
2354 /*
2355  * Implementation of the NetBSD [l]stat() functions.
2356  */
2357 void
cvtnstat(sb,nsb)2358 cvtnstat(sb, nsb)
2359 	struct stat *sb;
2360 	struct nstat *nsb;
2361 {
2362 
2363 	bzero(nsb, sizeof *nsb);
2364 	nsb->st_dev = sb->st_dev;
2365 	nsb->st_ino = sb->st_ino;
2366 	nsb->st_mode = sb->st_mode;
2367 	nsb->st_nlink = sb->st_nlink;
2368 	nsb->st_uid = sb->st_uid;
2369 	nsb->st_gid = sb->st_gid;
2370 	nsb->st_rdev = sb->st_rdev;
2371 	nsb->st_atim = sb->st_atim;
2372 	nsb->st_mtim = sb->st_mtim;
2373 	nsb->st_ctim = sb->st_ctim;
2374 	nsb->st_size = sb->st_size;
2375 	nsb->st_blocks = sb->st_blocks;
2376 	nsb->st_blksize = sb->st_blksize;
2377 	nsb->st_flags = sb->st_flags;
2378 	nsb->st_gen = sb->st_gen;
2379 	nsb->st_birthtim = sb->st_birthtim;
2380 }
2381 
2382 #ifndef _SYS_SYSPROTO_H_
2383 struct nstat_args {
2384 	char	*path;
2385 	struct nstat *ub;
2386 };
2387 #endif
2388 int
sys_nstat(td,uap)2389 sys_nstat(td, uap)
2390 	struct thread *td;
2391 	register struct nstat_args /* {
2392 		char *path;
2393 		struct nstat *ub;
2394 	} */ *uap;
2395 {
2396 	struct stat sb;
2397 	struct nstat nsb;
2398 	int error;
2399 
2400 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2401 	if (error != 0)
2402 		return (error);
2403 	cvtnstat(&sb, &nsb);
2404 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2405 }
2406 
2407 /*
2408  * NetBSD lstat.  Get file status; this version does not follow links.
2409  */
2410 #ifndef _SYS_SYSPROTO_H_
2411 struct lstat_args {
2412 	char	*path;
2413 	struct stat *ub;
2414 };
2415 #endif
2416 int
sys_nlstat(td,uap)2417 sys_nlstat(td, uap)
2418 	struct thread *td;
2419 	register struct nlstat_args /* {
2420 		char *path;
2421 		struct nstat *ub;
2422 	} */ *uap;
2423 {
2424 	struct stat sb;
2425 	struct nstat nsb;
2426 	int error;
2427 
2428 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2429 	if (error != 0)
2430 		return (error);
2431 	cvtnstat(&sb, &nsb);
2432 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2433 }
2434 
2435 /*
2436  * Get configurable pathname variables.
2437  */
2438 #ifndef _SYS_SYSPROTO_H_
2439 struct pathconf_args {
2440 	char	*path;
2441 	int	name;
2442 };
2443 #endif
2444 int
sys_pathconf(td,uap)2445 sys_pathconf(td, uap)
2446 	struct thread *td;
2447 	register struct pathconf_args /* {
2448 		char *path;
2449 		int name;
2450 	} */ *uap;
2451 {
2452 
2453 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2454 }
2455 
2456 #ifndef _SYS_SYSPROTO_H_
2457 struct lpathconf_args {
2458 	char	*path;
2459 	int	name;
2460 };
2461 #endif
2462 int
sys_lpathconf(td,uap)2463 sys_lpathconf(td, uap)
2464 	struct thread *td;
2465 	register struct lpathconf_args /* {
2466 		char *path;
2467 		int name;
2468 	} */ *uap;
2469 {
2470 
2471 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2472 	    NOFOLLOW));
2473 }
2474 
2475 int
kern_pathconf(struct thread * td,char * path,enum uio_seg pathseg,int name,u_long flags)2476 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2477     u_long flags)
2478 {
2479 	struct nameidata nd;
2480 	int error;
2481 
2482 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2483 	    pathseg, path, td);
2484 	if ((error = namei(&nd)) != 0)
2485 		return (error);
2486 	NDFREE(&nd, NDF_ONLY_PNBUF);
2487 
2488 	/* If asynchronous I/O is available, it works for all files. */
2489 	if (name == _PC_ASYNC_IO)
2490 		td->td_retval[0] = async_io_version;
2491 	else
2492 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2493 	vput(nd.ni_vp);
2494 	return (error);
2495 }
2496 
2497 /*
2498  * Return target name of a symbolic link.
2499  */
2500 #ifndef _SYS_SYSPROTO_H_
2501 struct readlink_args {
2502 	char	*path;
2503 	char	*buf;
2504 	size_t	count;
2505 };
2506 #endif
2507 int
sys_readlink(td,uap)2508 sys_readlink(td, uap)
2509 	struct thread *td;
2510 	register struct readlink_args /* {
2511 		char *path;
2512 		char *buf;
2513 		size_t count;
2514 	} */ *uap;
2515 {
2516 
2517 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2518 	    UIO_USERSPACE, uap->count));
2519 }
2520 #ifndef _SYS_SYSPROTO_H_
2521 struct readlinkat_args {
2522 	int	fd;
2523 	char	*path;
2524 	char	*buf;
2525 	size_t	bufsize;
2526 };
2527 #endif
2528 int
sys_readlinkat(struct thread * td,struct readlinkat_args * uap)2529 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2530 {
2531 
2532 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2533 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2534 }
2535 
2536 int
kern_readlink(struct thread * td,char * path,enum uio_seg pathseg,char * buf,enum uio_seg bufseg,size_t count)2537 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2538     enum uio_seg bufseg, size_t count)
2539 {
2540 
2541 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2542 	    count));
2543 }
2544 
2545 int
kern_readlinkat(struct thread * td,int fd,char * path,enum uio_seg pathseg,char * buf,enum uio_seg bufseg,size_t count)2546 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2547     char *buf, enum uio_seg bufseg, size_t count)
2548 {
2549 	struct vnode *vp;
2550 	struct iovec aiov;
2551 	struct uio auio;
2552 	struct nameidata nd;
2553 	int error;
2554 
2555 	if (count > IOSIZE_MAX)
2556 		return (EINVAL);
2557 
2558 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2559 	    pathseg, path, fd, td);
2560 
2561 	if ((error = namei(&nd)) != 0)
2562 		return (error);
2563 	NDFREE(&nd, NDF_ONLY_PNBUF);
2564 	vp = nd.ni_vp;
2565 #ifdef MAC
2566 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2567 	if (error != 0) {
2568 		vput(vp);
2569 		return (error);
2570 	}
2571 #endif
2572 	if (vp->v_type != VLNK)
2573 		error = EINVAL;
2574 	else {
2575 		aiov.iov_base = buf;
2576 		aiov.iov_len = count;
2577 		auio.uio_iov = &aiov;
2578 		auio.uio_iovcnt = 1;
2579 		auio.uio_offset = 0;
2580 		auio.uio_rw = UIO_READ;
2581 		auio.uio_segflg = bufseg;
2582 		auio.uio_td = td;
2583 		auio.uio_resid = count;
2584 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2585 		td->td_retval[0] = count - auio.uio_resid;
2586 	}
2587 	vput(vp);
2588 	return (error);
2589 }
2590 
2591 /*
2592  * Common implementation code for chflags() and fchflags().
2593  */
2594 static int
setfflags(td,vp,flags)2595 setfflags(td, vp, flags)
2596 	struct thread *td;
2597 	struct vnode *vp;
2598 	u_long flags;
2599 {
2600 	struct mount *mp;
2601 	struct vattr vattr;
2602 	int error;
2603 
2604 	/* We can't support the value matching VNOVAL. */
2605 	if (flags == VNOVAL)
2606 		return (EOPNOTSUPP);
2607 
2608 	/*
2609 	 * Prevent non-root users from setting flags on devices.  When
2610 	 * a device is reused, users can retain ownership of the device
2611 	 * if they are allowed to set flags and programs assume that
2612 	 * chown can't fail when done as root.
2613 	 */
2614 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2615 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2616 		if (error != 0)
2617 			return (error);
2618 	}
2619 
2620 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2621 		return (error);
2622 	VATTR_NULL(&vattr);
2623 	vattr.va_flags = flags;
2624 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2625 #ifdef MAC
2626 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2627 	if (error == 0)
2628 #endif
2629 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2630 	VOP_UNLOCK(vp, 0);
2631 	vn_finished_write(mp);
2632 	return (error);
2633 }
2634 
2635 /*
2636  * Change flags of a file given a path name.
2637  */
2638 #ifndef _SYS_SYSPROTO_H_
2639 struct chflags_args {
2640 	const char *path;
2641 	u_long	flags;
2642 };
2643 #endif
2644 int
sys_chflags(td,uap)2645 sys_chflags(td, uap)
2646 	struct thread *td;
2647 	register struct chflags_args /* {
2648 		const char *path;
2649 		u_long flags;
2650 	} */ *uap;
2651 {
2652 
2653 	return (kern_chflags(td, uap->path, UIO_USERSPACE, uap->flags));
2654 }
2655 
2656 #ifndef _SYS_SYSPROTO_H_
2657 struct chflagsat_args {
2658 	int	fd;
2659 	const char *path;
2660 	u_long	flags;
2661 	int	atflag;
2662 }
2663 #endif
2664 int
sys_chflagsat(struct thread * td,struct chflagsat_args * uap)2665 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2666 {
2667 	int fd = uap->fd;
2668 	const char *path = uap->path;
2669 	u_long flags = uap->flags;
2670 	int atflag = uap->atflag;
2671 
2672 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2673 		return (EINVAL);
2674 
2675 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2676 }
2677 
2678 static int
kern_chflags(struct thread * td,const char * path,enum uio_seg pathseg,u_long flags)2679 kern_chflags(struct thread *td, const char *path, enum uio_seg pathseg,
2680     u_long flags)
2681 {
2682 
2683 	return (kern_chflagsat(td, AT_FDCWD, path, pathseg, flags, 0));
2684 }
2685 
2686 /*
2687  * Same as chflags() but doesn't follow symlinks.
2688  */
2689 int
sys_lchflags(td,uap)2690 sys_lchflags(td, uap)
2691 	struct thread *td;
2692 	register struct lchflags_args /* {
2693 		const char *path;
2694 		u_long flags;
2695 	} */ *uap;
2696 {
2697 
2698 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2699 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2700 }
2701 
2702 static int
kern_chflagsat(struct thread * td,int fd,const char * path,enum uio_seg pathseg,u_long flags,int atflag)2703 kern_chflagsat(struct thread *td, int fd, const char *path,
2704     enum uio_seg pathseg, u_long flags, int atflag)
2705 {
2706 	struct nameidata nd;
2707 	cap_rights_t rights;
2708 	int error, follow;
2709 
2710 	AUDIT_ARG_FFLAGS(flags);
2711 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2712 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2713 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2714 	if ((error = namei(&nd)) != 0)
2715 		return (error);
2716 	NDFREE(&nd, NDF_ONLY_PNBUF);
2717 	error = setfflags(td, nd.ni_vp, flags);
2718 	vrele(nd.ni_vp);
2719 	return (error);
2720 }
2721 
2722 /*
2723  * Change flags of a file given a file descriptor.
2724  */
2725 #ifndef _SYS_SYSPROTO_H_
2726 struct fchflags_args {
2727 	int	fd;
2728 	u_long	flags;
2729 };
2730 #endif
2731 int
sys_fchflags(td,uap)2732 sys_fchflags(td, uap)
2733 	struct thread *td;
2734 	register struct fchflags_args /* {
2735 		int fd;
2736 		u_long flags;
2737 	} */ *uap;
2738 {
2739 	struct file *fp;
2740 	cap_rights_t rights;
2741 	int error;
2742 
2743 	AUDIT_ARG_FD(uap->fd);
2744 	AUDIT_ARG_FFLAGS(uap->flags);
2745 	error = getvnode(td->td_proc->p_fd, uap->fd,
2746 	    cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
2747 	if (error != 0)
2748 		return (error);
2749 #ifdef AUDIT
2750 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2751 	AUDIT_ARG_VNODE1(fp->f_vnode);
2752 	VOP_UNLOCK(fp->f_vnode, 0);
2753 #endif
2754 	error = setfflags(td, fp->f_vnode, uap->flags);
2755 	fdrop(fp, td);
2756 	return (error);
2757 }
2758 
2759 /*
2760  * Common implementation code for chmod(), lchmod() and fchmod().
2761  */
2762 int
setfmode(td,cred,vp,mode)2763 setfmode(td, cred, vp, mode)
2764 	struct thread *td;
2765 	struct ucred *cred;
2766 	struct vnode *vp;
2767 	int mode;
2768 {
2769 	struct mount *mp;
2770 	struct vattr vattr;
2771 	int error;
2772 
2773 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2774 		return (error);
2775 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2776 	VATTR_NULL(&vattr);
2777 	vattr.va_mode = mode & ALLPERMS;
2778 #ifdef MAC
2779 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2780 	if (error == 0)
2781 #endif
2782 		error = VOP_SETATTR(vp, &vattr, cred);
2783 	VOP_UNLOCK(vp, 0);
2784 	vn_finished_write(mp);
2785 	return (error);
2786 }
2787 
2788 /*
2789  * Change mode of a file given path name.
2790  */
2791 #ifndef _SYS_SYSPROTO_H_
2792 struct chmod_args {
2793 	char	*path;
2794 	int	mode;
2795 };
2796 #endif
2797 int
sys_chmod(td,uap)2798 sys_chmod(td, uap)
2799 	struct thread *td;
2800 	register struct chmod_args /* {
2801 		char *path;
2802 		int mode;
2803 	} */ *uap;
2804 {
2805 
2806 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2807 }
2808 
2809 #ifndef _SYS_SYSPROTO_H_
2810 struct fchmodat_args {
2811 	int	dirfd;
2812 	char	*path;
2813 	mode_t	mode;
2814 	int	flag;
2815 }
2816 #endif
2817 int
sys_fchmodat(struct thread * td,struct fchmodat_args * uap)2818 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2819 {
2820 	int flag = uap->flag;
2821 	int fd = uap->fd;
2822 	char *path = uap->path;
2823 	mode_t mode = uap->mode;
2824 
2825 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2826 		return (EINVAL);
2827 
2828 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2829 }
2830 
2831 int
kern_chmod(struct thread * td,char * path,enum uio_seg pathseg,int mode)2832 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2833 {
2834 
2835 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2836 }
2837 
2838 /*
2839  * Change mode of a file given path name (don't follow links.)
2840  */
2841 #ifndef _SYS_SYSPROTO_H_
2842 struct lchmod_args {
2843 	char	*path;
2844 	int	mode;
2845 };
2846 #endif
2847 int
sys_lchmod(td,uap)2848 sys_lchmod(td, uap)
2849 	struct thread *td;
2850 	register struct lchmod_args /* {
2851 		char *path;
2852 		int mode;
2853 	} */ *uap;
2854 {
2855 
2856 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2857 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2858 }
2859 
2860 int
kern_fchmodat(struct thread * td,int fd,char * path,enum uio_seg pathseg,mode_t mode,int flag)2861 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2862     mode_t mode, int flag)
2863 {
2864 	struct nameidata nd;
2865 	cap_rights_t rights;
2866 	int error, follow;
2867 
2868 	AUDIT_ARG_MODE(mode);
2869 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2870 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2871 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2872 	if ((error = namei(&nd)) != 0)
2873 		return (error);
2874 	NDFREE(&nd, NDF_ONLY_PNBUF);
2875 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2876 	vrele(nd.ni_vp);
2877 	return (error);
2878 }
2879 
2880 /*
2881  * Change mode of a file given a file descriptor.
2882  */
2883 #ifndef _SYS_SYSPROTO_H_
2884 struct fchmod_args {
2885 	int	fd;
2886 	int	mode;
2887 };
2888 #endif
2889 int
sys_fchmod(struct thread * td,struct fchmod_args * uap)2890 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2891 {
2892 	struct file *fp;
2893 	cap_rights_t rights;
2894 	int error;
2895 
2896 	AUDIT_ARG_FD(uap->fd);
2897 	AUDIT_ARG_MODE(uap->mode);
2898 
2899 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2900 	if (error != 0)
2901 		return (error);
2902 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2903 	fdrop(fp, td);
2904 	return (error);
2905 }
2906 
2907 /*
2908  * Common implementation for chown(), lchown(), and fchown()
2909  */
2910 int
setfown(td,cred,vp,uid,gid)2911 setfown(td, cred, vp, uid, gid)
2912 	struct thread *td;
2913 	struct ucred *cred;
2914 	struct vnode *vp;
2915 	uid_t uid;
2916 	gid_t gid;
2917 {
2918 	struct mount *mp;
2919 	struct vattr vattr;
2920 	int error;
2921 
2922 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2923 		return (error);
2924 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2925 	VATTR_NULL(&vattr);
2926 	vattr.va_uid = uid;
2927 	vattr.va_gid = gid;
2928 #ifdef MAC
2929 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2930 	    vattr.va_gid);
2931 	if (error == 0)
2932 #endif
2933 		error = VOP_SETATTR(vp, &vattr, cred);
2934 	VOP_UNLOCK(vp, 0);
2935 	vn_finished_write(mp);
2936 	return (error);
2937 }
2938 
2939 /*
2940  * Set ownership given a path name.
2941  */
2942 #ifndef _SYS_SYSPROTO_H_
2943 struct chown_args {
2944 	char	*path;
2945 	int	uid;
2946 	int	gid;
2947 };
2948 #endif
2949 int
sys_chown(td,uap)2950 sys_chown(td, uap)
2951 	struct thread *td;
2952 	register struct chown_args /* {
2953 		char *path;
2954 		int uid;
2955 		int gid;
2956 	} */ *uap;
2957 {
2958 
2959 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2960 }
2961 
2962 #ifndef _SYS_SYSPROTO_H_
2963 struct fchownat_args {
2964 	int fd;
2965 	const char * path;
2966 	uid_t uid;
2967 	gid_t gid;
2968 	int flag;
2969 };
2970 #endif
2971 int
sys_fchownat(struct thread * td,struct fchownat_args * uap)2972 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2973 {
2974 	int flag;
2975 
2976 	flag = uap->flag;
2977 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2978 		return (EINVAL);
2979 
2980 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2981 	    uap->gid, uap->flag));
2982 }
2983 
2984 int
kern_chown(struct thread * td,char * path,enum uio_seg pathseg,int uid,int gid)2985 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2986     int gid)
2987 {
2988 
2989 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2990 }
2991 
2992 int
kern_fchownat(struct thread * td,int fd,char * path,enum uio_seg pathseg,int uid,int gid,int flag)2993 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2994     int uid, int gid, int flag)
2995 {
2996 	struct nameidata nd;
2997 	cap_rights_t rights;
2998 	int error, follow;
2999 
3000 	AUDIT_ARG_OWNER(uid, gid);
3001 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3002 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
3003 	    cap_rights_init(&rights, CAP_FCHOWN), td);
3004 
3005 	if ((error = namei(&nd)) != 0)
3006 		return (error);
3007 	NDFREE(&nd, NDF_ONLY_PNBUF);
3008 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3009 	vrele(nd.ni_vp);
3010 	return (error);
3011 }
3012 
3013 /*
3014  * Set ownership given a path name, do not cross symlinks.
3015  */
3016 #ifndef _SYS_SYSPROTO_H_
3017 struct lchown_args {
3018 	char	*path;
3019 	int	uid;
3020 	int	gid;
3021 };
3022 #endif
3023 int
sys_lchown(td,uap)3024 sys_lchown(td, uap)
3025 	struct thread *td;
3026 	register struct lchown_args /* {
3027 		char *path;
3028 		int uid;
3029 		int gid;
3030 	} */ *uap;
3031 {
3032 
3033 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3034 }
3035 
3036 int
kern_lchown(struct thread * td,char * path,enum uio_seg pathseg,int uid,int gid)3037 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3038     int gid)
3039 {
3040 
3041 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3042 	    AT_SYMLINK_NOFOLLOW));
3043 }
3044 
3045 /*
3046  * Set ownership given a file descriptor.
3047  */
3048 #ifndef _SYS_SYSPROTO_H_
3049 struct fchown_args {
3050 	int	fd;
3051 	int	uid;
3052 	int	gid;
3053 };
3054 #endif
3055 int
sys_fchown(td,uap)3056 sys_fchown(td, uap)
3057 	struct thread *td;
3058 	register struct fchown_args /* {
3059 		int fd;
3060 		int uid;
3061 		int gid;
3062 	} */ *uap;
3063 {
3064 	struct file *fp;
3065 	cap_rights_t rights;
3066 	int error;
3067 
3068 	AUDIT_ARG_FD(uap->fd);
3069 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3070 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
3071 	if (error != 0)
3072 		return (error);
3073 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3074 	fdrop(fp, td);
3075 	return (error);
3076 }
3077 
3078 /*
3079  * Common implementation code for utimes(), lutimes(), and futimes().
3080  */
3081 static int
getutimes(usrtvp,tvpseg,tsp)3082 getutimes(usrtvp, tvpseg, tsp)
3083 	const struct timeval *usrtvp;
3084 	enum uio_seg tvpseg;
3085 	struct timespec *tsp;
3086 {
3087 	struct timeval tv[2];
3088 	const struct timeval *tvp;
3089 	int error;
3090 
3091 	if (usrtvp == NULL) {
3092 		vfs_timestamp(&tsp[0]);
3093 		tsp[1] = tsp[0];
3094 	} else {
3095 		if (tvpseg == UIO_SYSSPACE) {
3096 			tvp = usrtvp;
3097 		} else {
3098 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3099 				return (error);
3100 			tvp = tv;
3101 		}
3102 
3103 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3104 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3105 			return (EINVAL);
3106 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3107 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3108 	}
3109 	return (0);
3110 }
3111 
3112 /*
3113  * Common implementation code for utimes(), lutimes(), and futimes().
3114  */
3115 static int
setutimes(td,vp,ts,numtimes,nullflag)3116 setutimes(td, vp, ts, numtimes, nullflag)
3117 	struct thread *td;
3118 	struct vnode *vp;
3119 	const struct timespec *ts;
3120 	int numtimes;
3121 	int nullflag;
3122 {
3123 	struct mount *mp;
3124 	struct vattr vattr;
3125 	int error, setbirthtime;
3126 
3127 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3128 		return (error);
3129 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3130 	setbirthtime = 0;
3131 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3132 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3133 		setbirthtime = 1;
3134 	VATTR_NULL(&vattr);
3135 	vattr.va_atime = ts[0];
3136 	vattr.va_mtime = ts[1];
3137 	if (setbirthtime)
3138 		vattr.va_birthtime = ts[1];
3139 	if (numtimes > 2)
3140 		vattr.va_birthtime = ts[2];
3141 	if (nullflag)
3142 		vattr.va_vaflags |= VA_UTIMES_NULL;
3143 #ifdef MAC
3144 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3145 	    vattr.va_mtime);
3146 #endif
3147 	if (error == 0)
3148 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3149 	VOP_UNLOCK(vp, 0);
3150 	vn_finished_write(mp);
3151 	return (error);
3152 }
3153 
3154 /*
3155  * Set the access and modification times of a file.
3156  */
3157 #ifndef _SYS_SYSPROTO_H_
3158 struct utimes_args {
3159 	char	*path;
3160 	struct	timeval *tptr;
3161 };
3162 #endif
3163 int
sys_utimes(td,uap)3164 sys_utimes(td, uap)
3165 	struct thread *td;
3166 	register struct utimes_args /* {
3167 		char *path;
3168 		struct timeval *tptr;
3169 	} */ *uap;
3170 {
3171 
3172 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3173 	    UIO_USERSPACE));
3174 }
3175 
3176 #ifndef _SYS_SYSPROTO_H_
3177 struct futimesat_args {
3178 	int fd;
3179 	const char * path;
3180 	const struct timeval * times;
3181 };
3182 #endif
3183 int
sys_futimesat(struct thread * td,struct futimesat_args * uap)3184 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3185 {
3186 
3187 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3188 	    uap->times, UIO_USERSPACE));
3189 }
3190 
3191 int
kern_utimes(struct thread * td,char * path,enum uio_seg pathseg,struct timeval * tptr,enum uio_seg tptrseg)3192 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3193     struct timeval *tptr, enum uio_seg tptrseg)
3194 {
3195 
3196 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3197 }
3198 
3199 int
kern_utimesat(struct thread * td,int fd,char * path,enum uio_seg pathseg,struct timeval * tptr,enum uio_seg tptrseg)3200 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3201     struct timeval *tptr, enum uio_seg tptrseg)
3202 {
3203 	struct nameidata nd;
3204 	struct timespec ts[2];
3205 	cap_rights_t rights;
3206 	int error;
3207 
3208 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3209 		return (error);
3210 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3211 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3212 
3213 	if ((error = namei(&nd)) != 0)
3214 		return (error);
3215 	NDFREE(&nd, NDF_ONLY_PNBUF);
3216 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3217 	vrele(nd.ni_vp);
3218 	return (error);
3219 }
3220 
3221 /*
3222  * Set the access and modification times of a file.
3223  */
3224 #ifndef _SYS_SYSPROTO_H_
3225 struct lutimes_args {
3226 	char	*path;
3227 	struct	timeval *tptr;
3228 };
3229 #endif
3230 int
sys_lutimes(td,uap)3231 sys_lutimes(td, uap)
3232 	struct thread *td;
3233 	register struct lutimes_args /* {
3234 		char *path;
3235 		struct timeval *tptr;
3236 	} */ *uap;
3237 {
3238 
3239 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3240 	    UIO_USERSPACE));
3241 }
3242 
3243 int
kern_lutimes(struct thread * td,char * path,enum uio_seg pathseg,struct timeval * tptr,enum uio_seg tptrseg)3244 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3245     struct timeval *tptr, enum uio_seg tptrseg)
3246 {
3247 	struct timespec ts[2];
3248 	struct nameidata nd;
3249 	int error;
3250 
3251 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3252 		return (error);
3253 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3254 	if ((error = namei(&nd)) != 0)
3255 		return (error);
3256 	NDFREE(&nd, NDF_ONLY_PNBUF);
3257 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3258 	vrele(nd.ni_vp);
3259 	return (error);
3260 }
3261 
3262 /*
3263  * Set the access and modification times of a file.
3264  */
3265 #ifndef _SYS_SYSPROTO_H_
3266 struct futimes_args {
3267 	int	fd;
3268 	struct	timeval *tptr;
3269 };
3270 #endif
3271 int
sys_futimes(td,uap)3272 sys_futimes(td, uap)
3273 	struct thread *td;
3274 	register struct futimes_args /* {
3275 		int  fd;
3276 		struct timeval *tptr;
3277 	} */ *uap;
3278 {
3279 
3280 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3281 }
3282 
3283 int
kern_futimes(struct thread * td,int fd,struct timeval * tptr,enum uio_seg tptrseg)3284 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3285     enum uio_seg tptrseg)
3286 {
3287 	struct timespec ts[2];
3288 	struct file *fp;
3289 	cap_rights_t rights;
3290 	int error;
3291 
3292 	AUDIT_ARG_FD(fd);
3293 	error = getutimes(tptr, tptrseg, ts);
3294 	if (error != 0)
3295 		return (error);
3296 	error = getvnode(td->td_proc->p_fd, fd,
3297 	    cap_rights_init(&rights, CAP_FUTIMES), &fp);
3298 	if (error != 0)
3299 		return (error);
3300 #ifdef AUDIT
3301 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3302 	AUDIT_ARG_VNODE1(fp->f_vnode);
3303 	VOP_UNLOCK(fp->f_vnode, 0);
3304 #endif
3305 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3306 	fdrop(fp, td);
3307 	return (error);
3308 }
3309 
3310 /*
3311  * Truncate a file given its path name.
3312  */
3313 #ifndef _SYS_SYSPROTO_H_
3314 struct truncate_args {
3315 	char	*path;
3316 	int	pad;
3317 	off_t	length;
3318 };
3319 #endif
3320 int
sys_truncate(td,uap)3321 sys_truncate(td, uap)
3322 	struct thread *td;
3323 	register struct truncate_args /* {
3324 		char *path;
3325 		int pad;
3326 		off_t length;
3327 	} */ *uap;
3328 {
3329 
3330 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3331 }
3332 
3333 int
kern_truncate(struct thread * td,char * path,enum uio_seg pathseg,off_t length)3334 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3335 {
3336 	struct mount *mp;
3337 	struct vnode *vp;
3338 	void *rl_cookie;
3339 	struct vattr vattr;
3340 	struct nameidata nd;
3341 	int error;
3342 
3343 	if (length < 0)
3344 		return(EINVAL);
3345 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3346 	if ((error = namei(&nd)) != 0)
3347 		return (error);
3348 	vp = nd.ni_vp;
3349 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3350 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3351 		vn_rangelock_unlock(vp, rl_cookie);
3352 		vrele(vp);
3353 		return (error);
3354 	}
3355 	NDFREE(&nd, NDF_ONLY_PNBUF);
3356 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3357 	if (vp->v_type == VDIR)
3358 		error = EISDIR;
3359 #ifdef MAC
3360 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3361 	}
3362 #endif
3363 	else if ((error = vn_writechk(vp)) == 0 &&
3364 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3365 		VATTR_NULL(&vattr);
3366 		vattr.va_size = length;
3367 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3368 	}
3369 	VOP_UNLOCK(vp, 0);
3370 	vn_finished_write(mp);
3371 	vn_rangelock_unlock(vp, rl_cookie);
3372 	vrele(vp);
3373 	return (error);
3374 }
3375 
3376 #if defined(COMPAT_43)
3377 /*
3378  * Truncate a file given its path name.
3379  */
3380 #ifndef _SYS_SYSPROTO_H_
3381 struct otruncate_args {
3382 	char	*path;
3383 	long	length;
3384 };
3385 #endif
3386 int
otruncate(td,uap)3387 otruncate(td, uap)
3388 	struct thread *td;
3389 	register struct otruncate_args /* {
3390 		char *path;
3391 		long length;
3392 	} */ *uap;
3393 {
3394 	struct truncate_args /* {
3395 		char *path;
3396 		int pad;
3397 		off_t length;
3398 	} */ nuap;
3399 
3400 	nuap.path = uap->path;
3401 	nuap.length = uap->length;
3402 	return (sys_truncate(td, &nuap));
3403 }
3404 #endif /* COMPAT_43 */
3405 
3406 /* Versions with the pad argument */
3407 int
freebsd6_truncate(struct thread * td,struct freebsd6_truncate_args * uap)3408 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3409 {
3410 	struct truncate_args ouap;
3411 
3412 	ouap.path = uap->path;
3413 	ouap.length = uap->length;
3414 	return (sys_truncate(td, &ouap));
3415 }
3416 
3417 int
freebsd6_ftruncate(struct thread * td,struct freebsd6_ftruncate_args * uap)3418 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3419 {
3420 	struct ftruncate_args ouap;
3421 
3422 	ouap.fd = uap->fd;
3423 	ouap.length = uap->length;
3424 	return (sys_ftruncate(td, &ouap));
3425 }
3426 
3427 /*
3428  * Sync an open file.
3429  */
3430 #ifndef _SYS_SYSPROTO_H_
3431 struct fsync_args {
3432 	int	fd;
3433 };
3434 #endif
3435 int
sys_fsync(td,uap)3436 sys_fsync(td, uap)
3437 	struct thread *td;
3438 	struct fsync_args /* {
3439 		int fd;
3440 	} */ *uap;
3441 {
3442 	struct vnode *vp;
3443 	struct mount *mp;
3444 	struct file *fp;
3445 	cap_rights_t rights;
3446 	int error, lock_flags;
3447 
3448 	AUDIT_ARG_FD(uap->fd);
3449 	error = getvnode(td->td_proc->p_fd, uap->fd,
3450 	    cap_rights_init(&rights, CAP_FSYNC), &fp);
3451 	if (error != 0)
3452 		return (error);
3453 	vp = fp->f_vnode;
3454 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3455 	if (error != 0)
3456 		goto drop;
3457 	if (MNT_SHARED_WRITES(mp) ||
3458 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3459 		lock_flags = LK_SHARED;
3460 	} else {
3461 		lock_flags = LK_EXCLUSIVE;
3462 	}
3463 	vn_lock(vp, lock_flags | LK_RETRY);
3464 	AUDIT_ARG_VNODE1(vp);
3465 	if (vp->v_object != NULL) {
3466 		VM_OBJECT_WLOCK(vp->v_object);
3467 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3468 		VM_OBJECT_WUNLOCK(vp->v_object);
3469 	}
3470 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3471 
3472 	VOP_UNLOCK(vp, 0);
3473 	vn_finished_write(mp);
3474 drop:
3475 	fdrop(fp, td);
3476 	return (error);
3477 }
3478 
3479 /*
3480  * Rename files.  Source and destination must either both be directories, or
3481  * both not be directories.  If target is a directory, it must be empty.
3482  */
3483 #ifndef _SYS_SYSPROTO_H_
3484 struct rename_args {
3485 	char	*from;
3486 	char	*to;
3487 };
3488 #endif
3489 int
sys_rename(td,uap)3490 sys_rename(td, uap)
3491 	struct thread *td;
3492 	register struct rename_args /* {
3493 		char *from;
3494 		char *to;
3495 	} */ *uap;
3496 {
3497 
3498 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3499 }
3500 
3501 #ifndef _SYS_SYSPROTO_H_
3502 struct renameat_args {
3503 	int	oldfd;
3504 	char	*old;
3505 	int	newfd;
3506 	char	*new;
3507 };
3508 #endif
3509 int
sys_renameat(struct thread * td,struct renameat_args * uap)3510 sys_renameat(struct thread *td, struct renameat_args *uap)
3511 {
3512 
3513 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3514 	    UIO_USERSPACE));
3515 }
3516 
3517 int
kern_rename(struct thread * td,char * from,char * to,enum uio_seg pathseg)3518 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3519 {
3520 
3521 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3522 }
3523 
3524 int
kern_renameat(struct thread * td,int oldfd,char * old,int newfd,char * new,enum uio_seg pathseg)3525 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3526     enum uio_seg pathseg)
3527 {
3528 	struct mount *mp = NULL;
3529 	struct vnode *tvp, *fvp, *tdvp;
3530 	struct nameidata fromnd, tond;
3531 	cap_rights_t rights;
3532 	int error;
3533 
3534 again:
3535 	bwillwrite();
3536 #ifdef MAC
3537 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3538 	    AUDITVNODE1, pathseg, old, oldfd,
3539 	    cap_rights_init(&rights, CAP_RENAMEAT), td);
3540 #else
3541 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3542 	    pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3543 #endif
3544 
3545 	if ((error = namei(&fromnd)) != 0)
3546 		return (error);
3547 #ifdef MAC
3548 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3549 	    fromnd.ni_vp, &fromnd.ni_cnd);
3550 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3551 	if (fromnd.ni_dvp != fromnd.ni_vp)
3552 		VOP_UNLOCK(fromnd.ni_vp, 0);
3553 #endif
3554 	fvp = fromnd.ni_vp;
3555 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3556 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3557 	    cap_rights_init(&rights, CAP_LINKAT), td);
3558 	if (fromnd.ni_vp->v_type == VDIR)
3559 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3560 	if ((error = namei(&tond)) != 0) {
3561 		/* Translate error code for rename("dir1", "dir2/."). */
3562 		if (error == EISDIR && fvp->v_type == VDIR)
3563 			error = EINVAL;
3564 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3565 		vrele(fromnd.ni_dvp);
3566 		vrele(fvp);
3567 		goto out1;
3568 	}
3569 	tdvp = tond.ni_dvp;
3570 	tvp = tond.ni_vp;
3571 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3572 	if (error != 0) {
3573 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3574 		NDFREE(&tond, NDF_ONLY_PNBUF);
3575 		if (tvp != NULL)
3576 			vput(tvp);
3577 		if (tdvp == tvp)
3578 			vrele(tdvp);
3579 		else
3580 			vput(tdvp);
3581 		vrele(fromnd.ni_dvp);
3582 		vrele(fvp);
3583 		vrele(tond.ni_startdir);
3584 		if (fromnd.ni_startdir != NULL)
3585 			vrele(fromnd.ni_startdir);
3586 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3587 		if (error != 0)
3588 			return (error);
3589 		goto again;
3590 	}
3591 	if (tvp != NULL) {
3592 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3593 			error = ENOTDIR;
3594 			goto out;
3595 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3596 			error = EISDIR;
3597 			goto out;
3598 		}
3599 #ifdef CAPABILITIES
3600 		if (newfd != AT_FDCWD) {
3601 			/*
3602 			 * If the target already exists we require CAP_UNLINKAT
3603 			 * from 'newfd'.
3604 			 */
3605 			error = cap_check(&tond.ni_filecaps.fc_rights,
3606 			    cap_rights_init(&rights, CAP_UNLINKAT));
3607 			if (error != 0)
3608 				goto out;
3609 		}
3610 #endif
3611 	}
3612 	if (fvp == tdvp) {
3613 		error = EINVAL;
3614 		goto out;
3615 	}
3616 	/*
3617 	 * If the source is the same as the destination (that is, if they
3618 	 * are links to the same vnode), then there is nothing to do.
3619 	 */
3620 	if (fvp == tvp)
3621 		error = -1;
3622 #ifdef MAC
3623 	else
3624 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3625 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3626 #endif
3627 out:
3628 	if (error == 0) {
3629 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3630 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3631 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3632 		NDFREE(&tond, NDF_ONLY_PNBUF);
3633 	} else {
3634 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3635 		NDFREE(&tond, NDF_ONLY_PNBUF);
3636 		if (tvp != NULL)
3637 			vput(tvp);
3638 		if (tdvp == tvp)
3639 			vrele(tdvp);
3640 		else
3641 			vput(tdvp);
3642 		vrele(fromnd.ni_dvp);
3643 		vrele(fvp);
3644 	}
3645 	vrele(tond.ni_startdir);
3646 	vn_finished_write(mp);
3647 out1:
3648 	if (fromnd.ni_startdir)
3649 		vrele(fromnd.ni_startdir);
3650 	if (error == -1)
3651 		return (0);
3652 	return (error);
3653 }
3654 
3655 /*
3656  * Make a directory file.
3657  */
3658 #ifndef _SYS_SYSPROTO_H_
3659 struct mkdir_args {
3660 	char	*path;
3661 	int	mode;
3662 };
3663 #endif
3664 int
sys_mkdir(td,uap)3665 sys_mkdir(td, uap)
3666 	struct thread *td;
3667 	register struct mkdir_args /* {
3668 		char *path;
3669 		int mode;
3670 	} */ *uap;
3671 {
3672 
3673 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3674 }
3675 
3676 #ifndef _SYS_SYSPROTO_H_
3677 struct mkdirat_args {
3678 	int	fd;
3679 	char	*path;
3680 	mode_t	mode;
3681 };
3682 #endif
3683 int
sys_mkdirat(struct thread * td,struct mkdirat_args * uap)3684 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3685 {
3686 
3687 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3688 }
3689 
3690 int
kern_mkdir(struct thread * td,char * path,enum uio_seg segflg,int mode)3691 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3692 {
3693 
3694 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3695 }
3696 
3697 int
kern_mkdirat(struct thread * td,int fd,char * path,enum uio_seg segflg,int mode)3698 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3699     int mode)
3700 {
3701 	struct mount *mp;
3702 	struct vnode *vp;
3703 	struct vattr vattr;
3704 	struct nameidata nd;
3705 	cap_rights_t rights;
3706 	int error;
3707 
3708 	AUDIT_ARG_MODE(mode);
3709 restart:
3710 	bwillwrite();
3711 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3712 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3713 	    td);
3714 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3715 	if ((error = namei(&nd)) != 0)
3716 		return (error);
3717 	vp = nd.ni_vp;
3718 	if (vp != NULL) {
3719 		NDFREE(&nd, NDF_ONLY_PNBUF);
3720 		/*
3721 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3722 		 * the strange behaviour of leaving the vnode unlocked
3723 		 * if the target is the same vnode as the parent.
3724 		 */
3725 		if (vp == nd.ni_dvp)
3726 			vrele(nd.ni_dvp);
3727 		else
3728 			vput(nd.ni_dvp);
3729 		vrele(vp);
3730 		return (EEXIST);
3731 	}
3732 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3733 		NDFREE(&nd, NDF_ONLY_PNBUF);
3734 		vput(nd.ni_dvp);
3735 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3736 			return (error);
3737 		goto restart;
3738 	}
3739 	VATTR_NULL(&vattr);
3740 	vattr.va_type = VDIR;
3741 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3742 #ifdef MAC
3743 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3744 	    &vattr);
3745 	if (error != 0)
3746 		goto out;
3747 #endif
3748 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3749 #ifdef MAC
3750 out:
3751 #endif
3752 	NDFREE(&nd, NDF_ONLY_PNBUF);
3753 	vput(nd.ni_dvp);
3754 	if (error == 0)
3755 		vput(nd.ni_vp);
3756 	vn_finished_write(mp);
3757 	return (error);
3758 }
3759 
3760 /*
3761  * Remove a directory file.
3762  */
3763 #ifndef _SYS_SYSPROTO_H_
3764 struct rmdir_args {
3765 	char	*path;
3766 };
3767 #endif
3768 int
sys_rmdir(td,uap)3769 sys_rmdir(td, uap)
3770 	struct thread *td;
3771 	struct rmdir_args /* {
3772 		char *path;
3773 	} */ *uap;
3774 {
3775 
3776 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3777 }
3778 
3779 int
kern_rmdir(struct thread * td,char * path,enum uio_seg pathseg)3780 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3781 {
3782 
3783 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3784 }
3785 
3786 int
kern_rmdirat(struct thread * td,int fd,char * path,enum uio_seg pathseg)3787 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3788 {
3789 	struct mount *mp;
3790 	struct vnode *vp;
3791 	struct nameidata nd;
3792 	cap_rights_t rights;
3793 	int error;
3794 
3795 restart:
3796 	bwillwrite();
3797 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3798 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3799 	if ((error = namei(&nd)) != 0)
3800 		return (error);
3801 	vp = nd.ni_vp;
3802 	if (vp->v_type != VDIR) {
3803 		error = ENOTDIR;
3804 		goto out;
3805 	}
3806 	/*
3807 	 * No rmdir "." please.
3808 	 */
3809 	if (nd.ni_dvp == vp) {
3810 		error = EINVAL;
3811 		goto out;
3812 	}
3813 	/*
3814 	 * The root of a mounted filesystem cannot be deleted.
3815 	 */
3816 	if (vp->v_vflag & VV_ROOT) {
3817 		error = EBUSY;
3818 		goto out;
3819 	}
3820 #ifdef MAC
3821 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3822 	    &nd.ni_cnd);
3823 	if (error != 0)
3824 		goto out;
3825 #endif
3826 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3827 		NDFREE(&nd, NDF_ONLY_PNBUF);
3828 		vput(vp);
3829 		if (nd.ni_dvp == vp)
3830 			vrele(nd.ni_dvp);
3831 		else
3832 			vput(nd.ni_dvp);
3833 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3834 			return (error);
3835 		goto restart;
3836 	}
3837 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3838 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3839 	vn_finished_write(mp);
3840 out:
3841 	NDFREE(&nd, NDF_ONLY_PNBUF);
3842 	vput(vp);
3843 	if (nd.ni_dvp == vp)
3844 		vrele(nd.ni_dvp);
3845 	else
3846 		vput(nd.ni_dvp);
3847 	return (error);
3848 }
3849 
3850 #ifdef COMPAT_43
3851 /*
3852  * Read a block of directory entries in a filesystem independent format.
3853  */
3854 #ifndef _SYS_SYSPROTO_H_
3855 struct ogetdirentries_args {
3856 	int	fd;
3857 	char	*buf;
3858 	u_int	count;
3859 	long	*basep;
3860 };
3861 #endif
3862 int
ogetdirentries(struct thread * td,struct ogetdirentries_args * uap)3863 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3864 {
3865 	long loff;
3866 	int error;
3867 
3868 	error = kern_ogetdirentries(td, uap, &loff);
3869 	if (error == 0)
3870 		error = copyout(&loff, uap->basep, sizeof(long));
3871 	return (error);
3872 }
3873 
3874 int
kern_ogetdirentries(struct thread * td,struct ogetdirentries_args * uap,long * ploff)3875 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3876     long *ploff)
3877 {
3878 	struct vnode *vp;
3879 	struct file *fp;
3880 	struct uio auio, kuio;
3881 	struct iovec aiov, kiov;
3882 	struct dirent *dp, *edp;
3883 	cap_rights_t rights;
3884 	caddr_t dirbuf;
3885 	int error, eofflag, readcnt;
3886 	long loff;
3887 	off_t foffset;
3888 
3889 	/* XXX arbitrary sanity limit on `count'. */
3890 	if (uap->count > 64 * 1024)
3891 		return (EINVAL);
3892 	error = getvnode(td->td_proc->p_fd, uap->fd,
3893 	    cap_rights_init(&rights, CAP_READ), &fp);
3894 	if (error != 0)
3895 		return (error);
3896 	if ((fp->f_flag & FREAD) == 0) {
3897 		fdrop(fp, td);
3898 		return (EBADF);
3899 	}
3900 	vp = fp->f_vnode;
3901 	foffset = foffset_lock(fp, 0);
3902 unionread:
3903 	if (vp->v_type != VDIR) {
3904 		foffset_unlock(fp, foffset, 0);
3905 		fdrop(fp, td);
3906 		return (EINVAL);
3907 	}
3908 	aiov.iov_base = uap->buf;
3909 	aiov.iov_len = uap->count;
3910 	auio.uio_iov = &aiov;
3911 	auio.uio_iovcnt = 1;
3912 	auio.uio_rw = UIO_READ;
3913 	auio.uio_segflg = UIO_USERSPACE;
3914 	auio.uio_td = td;
3915 	auio.uio_resid = uap->count;
3916 	vn_lock(vp, LK_SHARED | LK_RETRY);
3917 	loff = auio.uio_offset = foffset;
3918 #ifdef MAC
3919 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3920 	if (error != 0) {
3921 		VOP_UNLOCK(vp, 0);
3922 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3923 		fdrop(fp, td);
3924 		return (error);
3925 	}
3926 #endif
3927 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3928 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3929 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3930 			    NULL, NULL);
3931 			foffset = auio.uio_offset;
3932 		} else
3933 #	endif
3934 	{
3935 		kuio = auio;
3936 		kuio.uio_iov = &kiov;
3937 		kuio.uio_segflg = UIO_SYSSPACE;
3938 		kiov.iov_len = uap->count;
3939 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3940 		kiov.iov_base = dirbuf;
3941 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3942 			    NULL, NULL);
3943 		foffset = kuio.uio_offset;
3944 		if (error == 0) {
3945 			readcnt = uap->count - kuio.uio_resid;
3946 			edp = (struct dirent *)&dirbuf[readcnt];
3947 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3948 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3949 					/*
3950 					 * The expected low byte of
3951 					 * dp->d_namlen is our dp->d_type.
3952 					 * The high MBZ byte of dp->d_namlen
3953 					 * is our dp->d_namlen.
3954 					 */
3955 					dp->d_type = dp->d_namlen;
3956 					dp->d_namlen = 0;
3957 #				else
3958 					/*
3959 					 * The dp->d_type is the high byte
3960 					 * of the expected dp->d_namlen,
3961 					 * so must be zero'ed.
3962 					 */
3963 					dp->d_type = 0;
3964 #				endif
3965 				if (dp->d_reclen > 0) {
3966 					dp = (struct dirent *)
3967 					    ((char *)dp + dp->d_reclen);
3968 				} else {
3969 					error = EIO;
3970 					break;
3971 				}
3972 			}
3973 			if (dp >= edp)
3974 				error = uiomove(dirbuf, readcnt, &auio);
3975 		}
3976 		free(dirbuf, M_TEMP);
3977 	}
3978 	if (error != 0) {
3979 		VOP_UNLOCK(vp, 0);
3980 		foffset_unlock(fp, foffset, 0);
3981 		fdrop(fp, td);
3982 		return (error);
3983 	}
3984 	if (uap->count == auio.uio_resid &&
3985 	    (vp->v_vflag & VV_ROOT) &&
3986 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3987 		struct vnode *tvp = vp;
3988 		vp = vp->v_mount->mnt_vnodecovered;
3989 		VREF(vp);
3990 		fp->f_vnode = vp;
3991 		fp->f_data = vp;
3992 		foffset = 0;
3993 		vput(tvp);
3994 		goto unionread;
3995 	}
3996 	VOP_UNLOCK(vp, 0);
3997 	foffset_unlock(fp, foffset, 0);
3998 	fdrop(fp, td);
3999 	td->td_retval[0] = uap->count - auio.uio_resid;
4000 	if (error == 0)
4001 		*ploff = loff;
4002 	return (error);
4003 }
4004 #endif /* COMPAT_43 */
4005 
4006 /*
4007  * Read a block of directory entries in a filesystem independent format.
4008  */
4009 #ifndef _SYS_SYSPROTO_H_
4010 struct getdirentries_args {
4011 	int	fd;
4012 	char	*buf;
4013 	u_int	count;
4014 	long	*basep;
4015 };
4016 #endif
4017 int
sys_getdirentries(td,uap)4018 sys_getdirentries(td, uap)
4019 	struct thread *td;
4020 	register struct getdirentries_args /* {
4021 		int fd;
4022 		char *buf;
4023 		u_int count;
4024 		long *basep;
4025 	} */ *uap;
4026 {
4027 	long base;
4028 	int error;
4029 
4030 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4031 	    NULL, UIO_USERSPACE);
4032 	if (error != 0)
4033 		return (error);
4034 	if (uap->basep != NULL)
4035 		error = copyout(&base, uap->basep, sizeof(long));
4036 	return (error);
4037 }
4038 
4039 int
kern_getdirentries(struct thread * td,int fd,char * buf,u_int count,long * basep,ssize_t * residp,enum uio_seg bufseg)4040 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4041     long *basep, ssize_t *residp, enum uio_seg bufseg)
4042 {
4043 	struct vnode *vp;
4044 	struct file *fp;
4045 	struct uio auio;
4046 	struct iovec aiov;
4047 	cap_rights_t rights;
4048 	long loff;
4049 	int error, eofflag;
4050 	off_t foffset;
4051 
4052 	AUDIT_ARG_FD(fd);
4053 	if (count > IOSIZE_MAX)
4054 		return (EINVAL);
4055 	auio.uio_resid = count;
4056 	error = getvnode(td->td_proc->p_fd, fd,
4057 	    cap_rights_init(&rights, CAP_READ), &fp);
4058 	if (error != 0)
4059 		return (error);
4060 	if ((fp->f_flag & FREAD) == 0) {
4061 		fdrop(fp, td);
4062 		return (EBADF);
4063 	}
4064 	vp = fp->f_vnode;
4065 	foffset = foffset_lock(fp, 0);
4066 unionread:
4067 	if (vp->v_type != VDIR) {
4068 		error = EINVAL;
4069 		goto fail;
4070 	}
4071 	aiov.iov_base = buf;
4072 	aiov.iov_len = count;
4073 	auio.uio_iov = &aiov;
4074 	auio.uio_iovcnt = 1;
4075 	auio.uio_rw = UIO_READ;
4076 	auio.uio_segflg = bufseg;
4077 	auio.uio_td = td;
4078 	vn_lock(vp, LK_SHARED | LK_RETRY);
4079 	AUDIT_ARG_VNODE1(vp);
4080 	loff = auio.uio_offset = foffset;
4081 #ifdef MAC
4082 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4083 	if (error == 0)
4084 #endif
4085 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4086 		    NULL);
4087 	foffset = auio.uio_offset;
4088 	if (error != 0) {
4089 		VOP_UNLOCK(vp, 0);
4090 		goto fail;
4091 	}
4092 	if (count == auio.uio_resid &&
4093 	    (vp->v_vflag & VV_ROOT) &&
4094 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4095 		struct vnode *tvp = vp;
4096 
4097 		vp = vp->v_mount->mnt_vnodecovered;
4098 		VREF(vp);
4099 		fp->f_vnode = vp;
4100 		fp->f_data = vp;
4101 		foffset = 0;
4102 		vput(tvp);
4103 		goto unionread;
4104 	}
4105 	VOP_UNLOCK(vp, 0);
4106 	*basep = loff;
4107 	if (residp != NULL)
4108 		*residp = auio.uio_resid;
4109 	td->td_retval[0] = count - auio.uio_resid;
4110 fail:
4111 	foffset_unlock(fp, foffset, 0);
4112 	fdrop(fp, td);
4113 	return (error);
4114 }
4115 
4116 #ifndef _SYS_SYSPROTO_H_
4117 struct getdents_args {
4118 	int fd;
4119 	char *buf;
4120 	size_t count;
4121 };
4122 #endif
4123 int
sys_getdents(td,uap)4124 sys_getdents(td, uap)
4125 	struct thread *td;
4126 	register struct getdents_args /* {
4127 		int fd;
4128 		char *buf;
4129 		u_int count;
4130 	} */ *uap;
4131 {
4132 	struct getdirentries_args ap;
4133 
4134 	ap.fd = uap->fd;
4135 	ap.buf = uap->buf;
4136 	ap.count = uap->count;
4137 	ap.basep = NULL;
4138 	return (sys_getdirentries(td, &ap));
4139 }
4140 
4141 /*
4142  * Set the mode mask for creation of filesystem nodes.
4143  */
4144 #ifndef _SYS_SYSPROTO_H_
4145 struct umask_args {
4146 	int	newmask;
4147 };
4148 #endif
4149 int
sys_umask(td,uap)4150 sys_umask(td, uap)
4151 	struct thread *td;
4152 	struct umask_args /* {
4153 		int newmask;
4154 	} */ *uap;
4155 {
4156 	register struct filedesc *fdp;
4157 
4158 	FILEDESC_XLOCK(td->td_proc->p_fd);
4159 	fdp = td->td_proc->p_fd;
4160 	td->td_retval[0] = fdp->fd_cmask;
4161 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4162 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4163 	return (0);
4164 }
4165 
4166 /*
4167  * Void all references to file by ripping underlying filesystem away from
4168  * vnode.
4169  */
4170 #ifndef _SYS_SYSPROTO_H_
4171 struct revoke_args {
4172 	char	*path;
4173 };
4174 #endif
4175 int
sys_revoke(td,uap)4176 sys_revoke(td, uap)
4177 	struct thread *td;
4178 	register struct revoke_args /* {
4179 		char *path;
4180 	} */ *uap;
4181 {
4182 	struct vnode *vp;
4183 	struct vattr vattr;
4184 	struct nameidata nd;
4185 	int error;
4186 
4187 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4188 	    uap->path, td);
4189 	if ((error = namei(&nd)) != 0)
4190 		return (error);
4191 	vp = nd.ni_vp;
4192 	NDFREE(&nd, NDF_ONLY_PNBUF);
4193 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4194 		error = EINVAL;
4195 		goto out;
4196 	}
4197 #ifdef MAC
4198 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4199 	if (error != 0)
4200 		goto out;
4201 #endif
4202 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4203 	if (error != 0)
4204 		goto out;
4205 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4206 		error = priv_check(td, PRIV_VFS_ADMIN);
4207 		if (error != 0)
4208 			goto out;
4209 	}
4210 	if (vcount(vp) > 1)
4211 		VOP_REVOKE(vp, REVOKEALL);
4212 out:
4213 	vput(vp);
4214 	return (error);
4215 }
4216 
4217 /*
4218  * Convert a user file descriptor to a kernel file entry and check that, if it
4219  * is a capability, the correct rights are present. A reference on the file
4220  * entry is held upon returning.
4221  */
4222 int
getvnode(struct filedesc * fdp,int fd,cap_rights_t * rightsp,struct file ** fpp)4223 getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
4224 {
4225 	struct file *fp;
4226 	int error;
4227 
4228 	error = fget_unlocked(fdp, fd, rightsp, 0, &fp, NULL);
4229 	if (error != 0)
4230 		return (error);
4231 
4232 	/*
4233 	 * The file could be not of the vnode type, or it may be not
4234 	 * yet fully initialized, in which case the f_vnode pointer
4235 	 * may be set, but f_ops is still badfileops.  E.g.,
4236 	 * devfs_open() transiently create such situation to
4237 	 * facilitate csw d_fdopen().
4238 	 *
4239 	 * Dupfdopen() handling in kern_openat() installs the
4240 	 * half-baked file into the process descriptor table, allowing
4241 	 * other thread to dereference it. Guard against the race by
4242 	 * checking f_ops.
4243 	 */
4244 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4245 		fdrop(fp, curthread);
4246 		return (EINVAL);
4247 	}
4248 	*fpp = fp;
4249 	return (0);
4250 }
4251 
4252 
4253 /*
4254  * Get an (NFS) file handle.
4255  */
4256 #ifndef _SYS_SYSPROTO_H_
4257 struct lgetfh_args {
4258 	char	*fname;
4259 	fhandle_t *fhp;
4260 };
4261 #endif
4262 int
sys_lgetfh(td,uap)4263 sys_lgetfh(td, uap)
4264 	struct thread *td;
4265 	register struct lgetfh_args *uap;
4266 {
4267 	struct nameidata nd;
4268 	fhandle_t fh;
4269 	register struct vnode *vp;
4270 	int error;
4271 
4272 	error = priv_check(td, PRIV_VFS_GETFH);
4273 	if (error != 0)
4274 		return (error);
4275 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4276 	    uap->fname, td);
4277 	error = namei(&nd);
4278 	if (error != 0)
4279 		return (error);
4280 	NDFREE(&nd, NDF_ONLY_PNBUF);
4281 	vp = nd.ni_vp;
4282 	bzero(&fh, sizeof(fh));
4283 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4284 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4285 	vput(vp);
4286 	if (error == 0)
4287 		error = copyout(&fh, uap->fhp, sizeof (fh));
4288 	return (error);
4289 }
4290 
4291 #ifndef _SYS_SYSPROTO_H_
4292 struct getfh_args {
4293 	char	*fname;
4294 	fhandle_t *fhp;
4295 };
4296 #endif
4297 int
sys_getfh(td,uap)4298 sys_getfh(td, uap)
4299 	struct thread *td;
4300 	register struct getfh_args *uap;
4301 {
4302 	struct nameidata nd;
4303 	fhandle_t fh;
4304 	register struct vnode *vp;
4305 	int error;
4306 
4307 	error = priv_check(td, PRIV_VFS_GETFH);
4308 	if (error != 0)
4309 		return (error);
4310 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4311 	    uap->fname, td);
4312 	error = namei(&nd);
4313 	if (error != 0)
4314 		return (error);
4315 	NDFREE(&nd, NDF_ONLY_PNBUF);
4316 	vp = nd.ni_vp;
4317 	bzero(&fh, sizeof(fh));
4318 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4319 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4320 	vput(vp);
4321 	if (error == 0)
4322 		error = copyout(&fh, uap->fhp, sizeof (fh));
4323 	return (error);
4324 }
4325 
4326 /*
4327  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4328  * open descriptor.
4329  *
4330  * warning: do not remove the priv_check() call or this becomes one giant
4331  * security hole.
4332  */
4333 #ifndef _SYS_SYSPROTO_H_
4334 struct fhopen_args {
4335 	const struct fhandle *u_fhp;
4336 	int flags;
4337 };
4338 #endif
4339 int
sys_fhopen(td,uap)4340 sys_fhopen(td, uap)
4341 	struct thread *td;
4342 	struct fhopen_args /* {
4343 		const struct fhandle *u_fhp;
4344 		int flags;
4345 	} */ *uap;
4346 {
4347 	struct mount *mp;
4348 	struct vnode *vp;
4349 	struct fhandle fhp;
4350 	struct file *fp;
4351 	int fmode, error;
4352 	int indx;
4353 
4354 	error = priv_check(td, PRIV_VFS_FHOPEN);
4355 	if (error != 0)
4356 		return (error);
4357 	indx = -1;
4358 	fmode = FFLAGS(uap->flags);
4359 	/* why not allow a non-read/write open for our lockd? */
4360 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4361 		return (EINVAL);
4362 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4363 	if (error != 0)
4364 		return(error);
4365 	/* find the mount point */
4366 	mp = vfs_busyfs(&fhp.fh_fsid);
4367 	if (mp == NULL)
4368 		return (ESTALE);
4369 	/* now give me my vnode, it gets returned to me locked */
4370 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4371 	vfs_unbusy(mp);
4372 	if (error != 0)
4373 		return (error);
4374 
4375 	error = falloc_noinstall(td, &fp);
4376 	if (error != 0) {
4377 		vput(vp);
4378 		return (error);
4379 	}
4380 	/*
4381 	 * An extra reference on `fp' has been held for us by
4382 	 * falloc_noinstall().
4383 	 */
4384 
4385 #ifdef INVARIANTS
4386 	td->td_dupfd = -1;
4387 #endif
4388 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4389 	if (error != 0) {
4390 		KASSERT(fp->f_ops == &badfileops,
4391 		    ("VOP_OPEN in fhopen() set f_ops"));
4392 		KASSERT(td->td_dupfd < 0,
4393 		    ("fhopen() encountered fdopen()"));
4394 
4395 		vput(vp);
4396 		goto bad;
4397 	}
4398 #ifdef INVARIANTS
4399 	td->td_dupfd = 0;
4400 #endif
4401 	fp->f_vnode = vp;
4402 	fp->f_seqcount = 1;
4403 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4404 	    &vnops);
4405 	VOP_UNLOCK(vp, 0);
4406 	if ((fmode & O_TRUNC) != 0) {
4407 		error = fo_truncate(fp, 0, td->td_ucred, td);
4408 		if (error != 0)
4409 			goto bad;
4410 	}
4411 
4412 	error = finstall(td, fp, &indx, fmode, NULL);
4413 bad:
4414 	fdrop(fp, td);
4415 	td->td_retval[0] = indx;
4416 	return (error);
4417 }
4418 
4419 /*
4420  * Stat an (NFS) file handle.
4421  */
4422 #ifndef _SYS_SYSPROTO_H_
4423 struct fhstat_args {
4424 	struct fhandle *u_fhp;
4425 	struct stat *sb;
4426 };
4427 #endif
4428 int
sys_fhstat(td,uap)4429 sys_fhstat(td, uap)
4430 	struct thread *td;
4431 	register struct fhstat_args /* {
4432 		struct fhandle *u_fhp;
4433 		struct stat *sb;
4434 	} */ *uap;
4435 {
4436 	struct stat sb;
4437 	struct fhandle fh;
4438 	int error;
4439 
4440 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4441 	if (error != 0)
4442 		return (error);
4443 	error = kern_fhstat(td, fh, &sb);
4444 	if (error == 0)
4445 		error = copyout(&sb, uap->sb, sizeof(sb));
4446 	return (error);
4447 }
4448 
4449 int
kern_fhstat(struct thread * td,struct fhandle fh,struct stat * sb)4450 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4451 {
4452 	struct mount *mp;
4453 	struct vnode *vp;
4454 	int error;
4455 
4456 	error = priv_check(td, PRIV_VFS_FHSTAT);
4457 	if (error != 0)
4458 		return (error);
4459 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4460 		return (ESTALE);
4461 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4462 	vfs_unbusy(mp);
4463 	if (error != 0)
4464 		return (error);
4465 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4466 	vput(vp);
4467 	return (error);
4468 }
4469 
4470 /*
4471  * Implement fstatfs() for (NFS) file handles.
4472  */
4473 #ifndef _SYS_SYSPROTO_H_
4474 struct fhstatfs_args {
4475 	struct fhandle *u_fhp;
4476 	struct statfs *buf;
4477 };
4478 #endif
4479 int
sys_fhstatfs(td,uap)4480 sys_fhstatfs(td, uap)
4481 	struct thread *td;
4482 	struct fhstatfs_args /* {
4483 		struct fhandle *u_fhp;
4484 		struct statfs *buf;
4485 	} */ *uap;
4486 {
4487 	struct statfs sf;
4488 	fhandle_t fh;
4489 	int error;
4490 
4491 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4492 	if (error != 0)
4493 		return (error);
4494 	error = kern_fhstatfs(td, fh, &sf);
4495 	if (error != 0)
4496 		return (error);
4497 	return (copyout(&sf, uap->buf, sizeof(sf)));
4498 }
4499 
4500 int
kern_fhstatfs(struct thread * td,fhandle_t fh,struct statfs * buf)4501 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4502 {
4503 	struct statfs *sp;
4504 	struct mount *mp;
4505 	struct vnode *vp;
4506 	int error;
4507 
4508 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4509 	if (error != 0)
4510 		return (error);
4511 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4512 		return (ESTALE);
4513 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4514 	if (error != 0) {
4515 		vfs_unbusy(mp);
4516 		return (error);
4517 	}
4518 	vput(vp);
4519 	error = prison_canseemount(td->td_ucred, mp);
4520 	if (error != 0)
4521 		goto out;
4522 #ifdef MAC
4523 	error = mac_mount_check_stat(td->td_ucred, mp);
4524 	if (error != 0)
4525 		goto out;
4526 #endif
4527 	/*
4528 	 * Set these in case the underlying filesystem fails to do so.
4529 	 */
4530 	sp = &mp->mnt_stat;
4531 	sp->f_version = STATFS_VERSION;
4532 	sp->f_namemax = NAME_MAX;
4533 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4534 	error = VFS_STATFS(mp, sp);
4535 	if (error == 0)
4536 		*buf = *sp;
4537 out:
4538 	vfs_unbusy(mp);
4539 	return (error);
4540 }
4541 
4542 int
kern_posix_fallocate(struct thread * td,int fd,off_t offset,off_t len)4543 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4544 {
4545 	struct file *fp;
4546 	struct mount *mp;
4547 	struct vnode *vp;
4548 	cap_rights_t rights;
4549 	off_t olen, ooffset;
4550 	int error;
4551 
4552 	if (offset < 0 || len <= 0)
4553 		return (EINVAL);
4554 	/* Check for wrap. */
4555 	if (offset > OFF_MAX - len)
4556 		return (EFBIG);
4557 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4558 	if (error != 0)
4559 		return (error);
4560 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4561 		error = ESPIPE;
4562 		goto out;
4563 	}
4564 	if ((fp->f_flag & FWRITE) == 0) {
4565 		error = EBADF;
4566 		goto out;
4567 	}
4568 	if (fp->f_type != DTYPE_VNODE) {
4569 		error = ENODEV;
4570 		goto out;
4571 	}
4572 	vp = fp->f_vnode;
4573 	if (vp->v_type != VREG) {
4574 		error = ENODEV;
4575 		goto out;
4576 	}
4577 
4578 	/* Allocating blocks may take a long time, so iterate. */
4579 	for (;;) {
4580 		olen = len;
4581 		ooffset = offset;
4582 
4583 		bwillwrite();
4584 		mp = NULL;
4585 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4586 		if (error != 0)
4587 			break;
4588 		error = vn_lock(vp, LK_EXCLUSIVE);
4589 		if (error != 0) {
4590 			vn_finished_write(mp);
4591 			break;
4592 		}
4593 #ifdef MAC
4594 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4595 		if (error == 0)
4596 #endif
4597 			error = VOP_ALLOCATE(vp, &offset, &len);
4598 		VOP_UNLOCK(vp, 0);
4599 		vn_finished_write(mp);
4600 
4601 		if (olen + ooffset != offset + len) {
4602 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4603 			    ooffset, olen, offset, len);
4604 		}
4605 		if (error != 0 || len == 0)
4606 			break;
4607 		KASSERT(olen > len, ("Iteration did not make progress?"));
4608 		maybe_yield();
4609 	}
4610  out:
4611 	fdrop(fp, td);
4612 	return (error);
4613 }
4614 
4615 int
sys_posix_fallocate(struct thread * td,struct posix_fallocate_args * uap)4616 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4617 {
4618 
4619 	td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4620 	    uap->len);
4621 	return (0);
4622 }
4623 
4624 /*
4625  * Unlike madvise(2), we do not make a best effort to remember every
4626  * possible caching hint.  Instead, we remember the last setting with
4627  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4628  * region of any current setting.
4629  */
4630 int
kern_posix_fadvise(struct thread * td,int fd,off_t offset,off_t len,int advice)4631 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4632     int advice)
4633 {
4634 	struct fadvise_info *fa, *new;
4635 	struct file *fp;
4636 	struct vnode *vp;
4637 	cap_rights_t rights;
4638 	off_t end;
4639 	int error;
4640 
4641 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4642 		return (EINVAL);
4643 	switch (advice) {
4644 	case POSIX_FADV_SEQUENTIAL:
4645 	case POSIX_FADV_RANDOM:
4646 	case POSIX_FADV_NOREUSE:
4647 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4648 		break;
4649 	case POSIX_FADV_NORMAL:
4650 	case POSIX_FADV_WILLNEED:
4651 	case POSIX_FADV_DONTNEED:
4652 		new = NULL;
4653 		break;
4654 	default:
4655 		return (EINVAL);
4656 	}
4657 	/* XXX: CAP_POSIX_FADVISE? */
4658 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4659 	if (error != 0)
4660 		goto out;
4661 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4662 		error = ESPIPE;
4663 		goto out;
4664 	}
4665 	if (fp->f_type != DTYPE_VNODE) {
4666 		error = ENODEV;
4667 		goto out;
4668 	}
4669 	vp = fp->f_vnode;
4670 	if (vp->v_type != VREG) {
4671 		error = ENODEV;
4672 		goto out;
4673 	}
4674 	if (len == 0)
4675 		end = OFF_MAX;
4676 	else
4677 		end = offset + len - 1;
4678 	switch (advice) {
4679 	case POSIX_FADV_SEQUENTIAL:
4680 	case POSIX_FADV_RANDOM:
4681 	case POSIX_FADV_NOREUSE:
4682 		/*
4683 		 * Try to merge any existing non-standard region with
4684 		 * this new region if possible, otherwise create a new
4685 		 * non-standard region for this request.
4686 		 */
4687 		mtx_pool_lock(mtxpool_sleep, fp);
4688 		fa = fp->f_advice;
4689 		if (fa != NULL && fa->fa_advice == advice &&
4690 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4691 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4692 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4693 			if (offset < fa->fa_start)
4694 				fa->fa_start = offset;
4695 			if (end > fa->fa_end)
4696 				fa->fa_end = end;
4697 		} else {
4698 			new->fa_advice = advice;
4699 			new->fa_start = offset;
4700 			new->fa_end = end;
4701 			new->fa_prevstart = 0;
4702 			new->fa_prevend = 0;
4703 			fp->f_advice = new;
4704 			new = fa;
4705 		}
4706 		mtx_pool_unlock(mtxpool_sleep, fp);
4707 		break;
4708 	case POSIX_FADV_NORMAL:
4709 		/*
4710 		 * If a the "normal" region overlaps with an existing
4711 		 * non-standard region, trim or remove the
4712 		 * non-standard region.
4713 		 */
4714 		mtx_pool_lock(mtxpool_sleep, fp);
4715 		fa = fp->f_advice;
4716 		if (fa != NULL) {
4717 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4718 				new = fa;
4719 				fp->f_advice = NULL;
4720 			} else if (offset <= fa->fa_start &&
4721 			    end >= fa->fa_start)
4722 				fa->fa_start = end + 1;
4723 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4724 				fa->fa_end = offset - 1;
4725 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4726 				/*
4727 				 * If the "normal" region is a middle
4728 				 * portion of the existing
4729 				 * non-standard region, just remove
4730 				 * the whole thing rather than picking
4731 				 * one side or the other to
4732 				 * preserve.
4733 				 */
4734 				new = fa;
4735 				fp->f_advice = NULL;
4736 			}
4737 		}
4738 		mtx_pool_unlock(mtxpool_sleep, fp);
4739 		break;
4740 	case POSIX_FADV_WILLNEED:
4741 	case POSIX_FADV_DONTNEED:
4742 		error = VOP_ADVISE(vp, offset, end, advice);
4743 		break;
4744 	}
4745 out:
4746 	if (fp != NULL)
4747 		fdrop(fp, td);
4748 	free(new, M_FADVISE);
4749 	return (error);
4750 }
4751 
4752 int
sys_posix_fadvise(struct thread * td,struct posix_fadvise_args * uap)4753 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4754 {
4755 
4756 	td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4757 	    uap->len, uap->advice);
4758 	return (0);
4759 }
4760