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: releng/10.2/sys/kern/vfs_syscalls.c 301051 2016-05-31 16:55:45Z glebius $");
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
2193 cvtstat(st, ost)
2194 struct stat *st;
2195 struct ostat *ost;
2196 {
2197
2198 bzero(ost, sizeof(*ost));
2199 ost->st_dev = st->st_dev;
2200 ost->st_ino = st->st_ino;
2201 ost->st_mode = st->st_mode;
2202 ost->st_nlink = st->st_nlink;
2203 ost->st_uid = st->st_uid;
2204 ost->st_gid = st->st_gid;
2205 ost->st_rdev = st->st_rdev;
2206 if (st->st_size < (quad_t)1 << 32)
2207 ost->st_size = st->st_size;
2208 else
2209 ost->st_size = -2;
2210 ost->st_atim = st->st_atim;
2211 ost->st_mtim = st->st_mtim;
2212 ost->st_ctim = st->st_ctim;
2213 ost->st_blksize = st->st_blksize;
2214 ost->st_blocks = st->st_blocks;
2215 ost->st_flags = st->st_flags;
2216 ost->st_gen = st->st_gen;
2217 }
2218 #endif /* COMPAT_43 */
2219
2220 /*
2221 * Get file status; this version follows links.
2222 */
2223 #ifndef _SYS_SYSPROTO_H_
2224 struct stat_args {
2225 char *path;
2226 struct stat *ub;
2227 };
2228 #endif
2229 int
2230 sys_stat(td, uap)
2231 struct thread *td;
2232 register struct stat_args /* {
2233 char *path;
2234 struct stat *ub;
2235 } */ *uap;
2236 {
2237 struct stat sb;
2238 int error;
2239
2240 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2241 if (error == 0)
2242 error = copyout(&sb, uap->ub, sizeof (sb));
2243 return (error);
2244 }
2245
2246 #ifndef _SYS_SYSPROTO_H_
2247 struct fstatat_args {
2248 int fd;
2249 char *path;
2250 struct stat *buf;
2251 int flag;
2252 }
2253 #endif
2254 int
2255 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2256 {
2257 struct stat sb;
2258 int error;
2259
2260 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2261 UIO_USERSPACE, &sb);
2262 if (error == 0)
2263 error = copyout(&sb, uap->buf, sizeof (sb));
2264 return (error);
2265 }
2266
2267 int
2268 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2269 {
2270
2271 return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2272 }
2273
2274 int
2275 kern_statat(struct thread *td, int flag, int fd, char *path,
2276 enum uio_seg pathseg, struct stat *sbp)
2277 {
2278
2279 return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2280 }
2281
2282 int
2283 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2284 enum uio_seg pathseg, struct stat *sbp,
2285 void (*hook)(struct vnode *vp, struct stat *sbp))
2286 {
2287 struct nameidata nd;
2288 struct stat sb;
2289 cap_rights_t rights;
2290 int error;
2291
2292 if (flag & ~AT_SYMLINK_NOFOLLOW)
2293 return (EINVAL);
2294
2295 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2296 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2297 cap_rights_init(&rights, CAP_FSTAT), td);
2298
2299 if ((error = namei(&nd)) != 0)
2300 return (error);
2301 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2302 if (error == 0) {
2303 SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2304 if (S_ISREG(sb.st_mode))
2305 SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2306 if (__predict_false(hook != NULL))
2307 hook(nd.ni_vp, &sb);
2308 }
2309 NDFREE(&nd, NDF_ONLY_PNBUF);
2310 vput(nd.ni_vp);
2311 if (error != 0)
2312 return (error);
2313 *sbp = sb;
2314 #ifdef KTRACE
2315 if (KTRPOINT(td, KTR_STRUCT))
2316 ktrstat(&sb);
2317 #endif
2318 return (0);
2319 }
2320
2321 /*
2322 * Get file status; this version does not follow links.
2323 */
2324 #ifndef _SYS_SYSPROTO_H_
2325 struct lstat_args {
2326 char *path;
2327 struct stat *ub;
2328 };
2329 #endif
2330 int
2331 sys_lstat(td, uap)
2332 struct thread *td;
2333 register struct lstat_args /* {
2334 char *path;
2335 struct stat *ub;
2336 } */ *uap;
2337 {
2338 struct stat sb;
2339 int error;
2340
2341 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2342 if (error == 0)
2343 error = copyout(&sb, uap->ub, sizeof (sb));
2344 return (error);
2345 }
2346
2347 int
2348 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2349 {
2350
2351 return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2352 sbp));
2353 }
2354
2355 /*
2356 * Implementation of the NetBSD [l]stat() functions.
2357 */
2358 void
2359 cvtnstat(sb, nsb)
2360 struct stat *sb;
2361 struct nstat *nsb;
2362 {
2363
2364 bzero(nsb, sizeof *nsb);
2365 nsb->st_dev = sb->st_dev;
2366 nsb->st_ino = sb->st_ino;
2367 nsb->st_mode = sb->st_mode;
2368 nsb->st_nlink = sb->st_nlink;
2369 nsb->st_uid = sb->st_uid;
2370 nsb->st_gid = sb->st_gid;
2371 nsb->st_rdev = sb->st_rdev;
2372 nsb->st_atim = sb->st_atim;
2373 nsb->st_mtim = sb->st_mtim;
2374 nsb->st_ctim = sb->st_ctim;
2375 nsb->st_size = sb->st_size;
2376 nsb->st_blocks = sb->st_blocks;
2377 nsb->st_blksize = sb->st_blksize;
2378 nsb->st_flags = sb->st_flags;
2379 nsb->st_gen = sb->st_gen;
2380 nsb->st_birthtim = sb->st_birthtim;
2381 }
2382
2383 #ifndef _SYS_SYSPROTO_H_
2384 struct nstat_args {
2385 char *path;
2386 struct nstat *ub;
2387 };
2388 #endif
2389 int
2390 sys_nstat(td, uap)
2391 struct thread *td;
2392 register struct nstat_args /* {
2393 char *path;
2394 struct nstat *ub;
2395 } */ *uap;
2396 {
2397 struct stat sb;
2398 struct nstat nsb;
2399 int error;
2400
2401 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2402 if (error != 0)
2403 return (error);
2404 cvtnstat(&sb, &nsb);
2405 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2406 }
2407
2408 /*
2409 * NetBSD lstat. Get file status; this version does not follow links.
2410 */
2411 #ifndef _SYS_SYSPROTO_H_
2412 struct lstat_args {
2413 char *path;
2414 struct stat *ub;
2415 };
2416 #endif
2417 int
2418 sys_nlstat(td, uap)
2419 struct thread *td;
2420 register struct nlstat_args /* {
2421 char *path;
2422 struct nstat *ub;
2423 } */ *uap;
2424 {
2425 struct stat sb;
2426 struct nstat nsb;
2427 int error;
2428
2429 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2430 if (error != 0)
2431 return (error);
2432 cvtnstat(&sb, &nsb);
2433 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2434 }
2435
2436 /*
2437 * Get configurable pathname variables.
2438 */
2439 #ifndef _SYS_SYSPROTO_H_
2440 struct pathconf_args {
2441 char *path;
2442 int name;
2443 };
2444 #endif
2445 int
2446 sys_pathconf(td, uap)
2447 struct thread *td;
2448 register struct pathconf_args /* {
2449 char *path;
2450 int name;
2451 } */ *uap;
2452 {
2453
2454 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2455 }
2456
2457 #ifndef _SYS_SYSPROTO_H_
2458 struct lpathconf_args {
2459 char *path;
2460 int name;
2461 };
2462 #endif
2463 int
2464 sys_lpathconf(td, uap)
2465 struct thread *td;
2466 register struct lpathconf_args /* {
2467 char *path;
2468 int name;
2469 } */ *uap;
2470 {
2471
2472 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2473 NOFOLLOW));
2474 }
2475
2476 int
2477 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2478 u_long flags)
2479 {
2480 struct nameidata nd;
2481 int error;
2482
2483 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2484 pathseg, path, td);
2485 if ((error = namei(&nd)) != 0)
2486 return (error);
2487 NDFREE(&nd, NDF_ONLY_PNBUF);
2488
2489 /* If asynchronous I/O is available, it works for all files. */
2490 if (name == _PC_ASYNC_IO)
2491 td->td_retval[0] = async_io_version;
2492 else
2493 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2494 vput(nd.ni_vp);
2495 return (error);
2496 }
2497
2498 /*
2499 * Return target name of a symbolic link.
2500 */
2501 #ifndef _SYS_SYSPROTO_H_
2502 struct readlink_args {
2503 char *path;
2504 char *buf;
2505 size_t count;
2506 };
2507 #endif
2508 int
2509 sys_readlink(td, uap)
2510 struct thread *td;
2511 register struct readlink_args /* {
2512 char *path;
2513 char *buf;
2514 size_t count;
2515 } */ *uap;
2516 {
2517
2518 return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2519 UIO_USERSPACE, uap->count));
2520 }
2521 #ifndef _SYS_SYSPROTO_H_
2522 struct readlinkat_args {
2523 int fd;
2524 char *path;
2525 char *buf;
2526 size_t bufsize;
2527 };
2528 #endif
2529 int
2530 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2531 {
2532
2533 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2534 uap->buf, UIO_USERSPACE, uap->bufsize));
2535 }
2536
2537 int
2538 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2539 enum uio_seg bufseg, size_t count)
2540 {
2541
2542 return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2543 count));
2544 }
2545
2546 int
2547 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2548 char *buf, enum uio_seg bufseg, size_t count)
2549 {
2550 struct vnode *vp;
2551 struct iovec aiov;
2552 struct uio auio;
2553 struct nameidata nd;
2554 int error;
2555
2556 if (count > IOSIZE_MAX)
2557 return (EINVAL);
2558
2559 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2560 pathseg, path, fd, td);
2561
2562 if ((error = namei(&nd)) != 0)
2563 return (error);
2564 NDFREE(&nd, NDF_ONLY_PNBUF);
2565 vp = nd.ni_vp;
2566 #ifdef MAC
2567 error = mac_vnode_check_readlink(td->td_ucred, vp);
2568 if (error != 0) {
2569 vput(vp);
2570 return (error);
2571 }
2572 #endif
2573 if (vp->v_type != VLNK)
2574 error = EINVAL;
2575 else {
2576 aiov.iov_base = buf;
2577 aiov.iov_len = count;
2578 auio.uio_iov = &aiov;
2579 auio.uio_iovcnt = 1;
2580 auio.uio_offset = 0;
2581 auio.uio_rw = UIO_READ;
2582 auio.uio_segflg = bufseg;
2583 auio.uio_td = td;
2584 auio.uio_resid = count;
2585 error = VOP_READLINK(vp, &auio, td->td_ucred);
2586 td->td_retval[0] = count - auio.uio_resid;
2587 }
2588 vput(vp);
2589 return (error);
2590 }
2591
2592 /*
2593 * Common implementation code for chflags() and fchflags().
2594 */
2595 static int
2596 setfflags(td, vp, flags)
2597 struct thread *td;
2598 struct vnode *vp;
2599 u_long flags;
2600 {
2601 struct mount *mp;
2602 struct vattr vattr;
2603 int error;
2604
2605 /* We can't support the value matching VNOVAL. */
2606 if (flags == VNOVAL)
2607 return (EOPNOTSUPP);
2608
2609 /*
2610 * Prevent non-root users from setting flags on devices. When
2611 * a device is reused, users can retain ownership of the device
2612 * if they are allowed to set flags and programs assume that
2613 * chown can't fail when done as root.
2614 */
2615 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2616 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2617 if (error != 0)
2618 return (error);
2619 }
2620
2621 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2622 return (error);
2623 VATTR_NULL(&vattr);
2624 vattr.va_flags = flags;
2625 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2626 #ifdef MAC
2627 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2628 if (error == 0)
2629 #endif
2630 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2631 VOP_UNLOCK(vp, 0);
2632 vn_finished_write(mp);
2633 return (error);
2634 }
2635
2636 /*
2637 * Change flags of a file given a path name.
2638 */
2639 #ifndef _SYS_SYSPROTO_H_
2640 struct chflags_args {
2641 const char *path;
2642 u_long flags;
2643 };
2644 #endif
2645 int
2646 sys_chflags(td, uap)
2647 struct thread *td;
2648 register struct chflags_args /* {
2649 const char *path;
2650 u_long flags;
2651 } */ *uap;
2652 {
2653
2654 return (kern_chflags(td, uap->path, UIO_USERSPACE, uap->flags));
2655 }
2656
2657 #ifndef _SYS_SYSPROTO_H_
2658 struct chflagsat_args {
2659 int fd;
2660 const char *path;
2661 u_long flags;
2662 int atflag;
2663 }
2664 #endif
2665 int
2666 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2667 {
2668 int fd = uap->fd;
2669 const char *path = uap->path;
2670 u_long flags = uap->flags;
2671 int atflag = uap->atflag;
2672
2673 if (atflag & ~AT_SYMLINK_NOFOLLOW)
2674 return (EINVAL);
2675
2676 return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2677 }
2678
2679 static int
2680 kern_chflags(struct thread *td, const char *path, enum uio_seg pathseg,
2681 u_long flags)
2682 {
2683
2684 return (kern_chflagsat(td, AT_FDCWD, path, pathseg, flags, 0));
2685 }
2686
2687 /*
2688 * Same as chflags() but doesn't follow symlinks.
2689 */
2690 int
2691 sys_lchflags(td, uap)
2692 struct thread *td;
2693 register struct lchflags_args /* {
2694 const char *path;
2695 u_long flags;
2696 } */ *uap;
2697 {
2698
2699 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2700 uap->flags, AT_SYMLINK_NOFOLLOW));
2701 }
2702
2703 static int
2704 kern_chflagsat(struct thread *td, int fd, const char *path,
2705 enum uio_seg pathseg, u_long flags, int atflag)
2706 {
2707 struct nameidata nd;
2708 cap_rights_t rights;
2709 int error, follow;
2710
2711 AUDIT_ARG_FFLAGS(flags);
2712 follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2713 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2714 cap_rights_init(&rights, CAP_FCHFLAGS), td);
2715 if ((error = namei(&nd)) != 0)
2716 return (error);
2717 NDFREE(&nd, NDF_ONLY_PNBUF);
2718 error = setfflags(td, nd.ni_vp, flags);
2719 vrele(nd.ni_vp);
2720 return (error);
2721 }
2722
2723 /*
2724 * Change flags of a file given a file descriptor.
2725 */
2726 #ifndef _SYS_SYSPROTO_H_
2727 struct fchflags_args {
2728 int fd;
2729 u_long flags;
2730 };
2731 #endif
2732 int
2733 sys_fchflags(td, uap)
2734 struct thread *td;
2735 register struct fchflags_args /* {
2736 int fd;
2737 u_long flags;
2738 } */ *uap;
2739 {
2740 struct file *fp;
2741 cap_rights_t rights;
2742 int error;
2743
2744 AUDIT_ARG_FD(uap->fd);
2745 AUDIT_ARG_FFLAGS(uap->flags);
2746 error = getvnode(td->td_proc->p_fd, uap->fd,
2747 cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
2748 if (error != 0)
2749 return (error);
2750 #ifdef AUDIT
2751 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2752 AUDIT_ARG_VNODE1(fp->f_vnode);
2753 VOP_UNLOCK(fp->f_vnode, 0);
2754 #endif
2755 error = setfflags(td, fp->f_vnode, uap->flags);
2756 fdrop(fp, td);
2757 return (error);
2758 }
2759
2760 /*
2761 * Common implementation code for chmod(), lchmod() and fchmod().
2762 */
2763 int
2764 setfmode(td, cred, vp, mode)
2765 struct thread *td;
2766 struct ucred *cred;
2767 struct vnode *vp;
2768 int mode;
2769 {
2770 struct mount *mp;
2771 struct vattr vattr;
2772 int error;
2773
2774 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2775 return (error);
2776 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2777 VATTR_NULL(&vattr);
2778 vattr.va_mode = mode & ALLPERMS;
2779 #ifdef MAC
2780 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2781 if (error == 0)
2782 #endif
2783 error = VOP_SETATTR(vp, &vattr, cred);
2784 VOP_UNLOCK(vp, 0);
2785 vn_finished_write(mp);
2786 return (error);
2787 }
2788
2789 /*
2790 * Change mode of a file given path name.
2791 */
2792 #ifndef _SYS_SYSPROTO_H_
2793 struct chmod_args {
2794 char *path;
2795 int mode;
2796 };
2797 #endif
2798 int
2799 sys_chmod(td, uap)
2800 struct thread *td;
2801 register struct chmod_args /* {
2802 char *path;
2803 int mode;
2804 } */ *uap;
2805 {
2806
2807 return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2808 }
2809
2810 #ifndef _SYS_SYSPROTO_H_
2811 struct fchmodat_args {
2812 int dirfd;
2813 char *path;
2814 mode_t mode;
2815 int flag;
2816 }
2817 #endif
2818 int
2819 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2820 {
2821 int flag = uap->flag;
2822 int fd = uap->fd;
2823 char *path = uap->path;
2824 mode_t mode = uap->mode;
2825
2826 if (flag & ~AT_SYMLINK_NOFOLLOW)
2827 return (EINVAL);
2828
2829 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2830 }
2831
2832 int
2833 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2834 {
2835
2836 return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2837 }
2838
2839 /*
2840 * Change mode of a file given path name (don't follow links.)
2841 */
2842 #ifndef _SYS_SYSPROTO_H_
2843 struct lchmod_args {
2844 char *path;
2845 int mode;
2846 };
2847 #endif
2848 int
2849 sys_lchmod(td, uap)
2850 struct thread *td;
2851 register struct lchmod_args /* {
2852 char *path;
2853 int mode;
2854 } */ *uap;
2855 {
2856
2857 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2858 uap->mode, AT_SYMLINK_NOFOLLOW));
2859 }
2860
2861 int
2862 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2863 mode_t mode, int flag)
2864 {
2865 struct nameidata nd;
2866 cap_rights_t rights;
2867 int error, follow;
2868
2869 AUDIT_ARG_MODE(mode);
2870 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2871 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2872 cap_rights_init(&rights, CAP_FCHMOD), td);
2873 if ((error = namei(&nd)) != 0)
2874 return (error);
2875 NDFREE(&nd, NDF_ONLY_PNBUF);
2876 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2877 vrele(nd.ni_vp);
2878 return (error);
2879 }
2880
2881 /*
2882 * Change mode of a file given a file descriptor.
2883 */
2884 #ifndef _SYS_SYSPROTO_H_
2885 struct fchmod_args {
2886 int fd;
2887 int mode;
2888 };
2889 #endif
2890 int
2891 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2892 {
2893 struct file *fp;
2894 cap_rights_t rights;
2895 int error;
2896
2897 AUDIT_ARG_FD(uap->fd);
2898 AUDIT_ARG_MODE(uap->mode);
2899
2900 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2901 if (error != 0)
2902 return (error);
2903 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2904 fdrop(fp, td);
2905 return (error);
2906 }
2907
2908 /*
2909 * Common implementation for chown(), lchown(), and fchown()
2910 */
2911 int
2912 setfown(td, cred, vp, uid, gid)
2913 struct thread *td;
2914 struct ucred *cred;
2915 struct vnode *vp;
2916 uid_t uid;
2917 gid_t gid;
2918 {
2919 struct mount *mp;
2920 struct vattr vattr;
2921 int error;
2922
2923 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2924 return (error);
2925 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2926 VATTR_NULL(&vattr);
2927 vattr.va_uid = uid;
2928 vattr.va_gid = gid;
2929 #ifdef MAC
2930 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2931 vattr.va_gid);
2932 if (error == 0)
2933 #endif
2934 error = VOP_SETATTR(vp, &vattr, cred);
2935 VOP_UNLOCK(vp, 0);
2936 vn_finished_write(mp);
2937 return (error);
2938 }
2939
2940 /*
2941 * Set ownership given a path name.
2942 */
2943 #ifndef _SYS_SYSPROTO_H_
2944 struct chown_args {
2945 char *path;
2946 int uid;
2947 int gid;
2948 };
2949 #endif
2950 int
2951 sys_chown(td, uap)
2952 struct thread *td;
2953 register struct chown_args /* {
2954 char *path;
2955 int uid;
2956 int gid;
2957 } */ *uap;
2958 {
2959
2960 return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2961 }
2962
2963 #ifndef _SYS_SYSPROTO_H_
2964 struct fchownat_args {
2965 int fd;
2966 const char * path;
2967 uid_t uid;
2968 gid_t gid;
2969 int flag;
2970 };
2971 #endif
2972 int
2973 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2974 {
2975 int flag;
2976
2977 flag = uap->flag;
2978 if (flag & ~AT_SYMLINK_NOFOLLOW)
2979 return (EINVAL);
2980
2981 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2982 uap->gid, uap->flag));
2983 }
2984
2985 int
2986 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2987 int gid)
2988 {
2989
2990 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2991 }
2992
2993 int
2994 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2995 int uid, int gid, int flag)
2996 {
2997 struct nameidata nd;
2998 cap_rights_t rights;
2999 int error, follow;
3000
3001 AUDIT_ARG_OWNER(uid, gid);
3002 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3003 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
3004 cap_rights_init(&rights, CAP_FCHOWN), td);
3005
3006 if ((error = namei(&nd)) != 0)
3007 return (error);
3008 NDFREE(&nd, NDF_ONLY_PNBUF);
3009 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3010 vrele(nd.ni_vp);
3011 return (error);
3012 }
3013
3014 /*
3015 * Set ownership given a path name, do not cross symlinks.
3016 */
3017 #ifndef _SYS_SYSPROTO_H_
3018 struct lchown_args {
3019 char *path;
3020 int uid;
3021 int gid;
3022 };
3023 #endif
3024 int
3025 sys_lchown(td, uap)
3026 struct thread *td;
3027 register struct lchown_args /* {
3028 char *path;
3029 int uid;
3030 int gid;
3031 } */ *uap;
3032 {
3033
3034 return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3035 }
3036
3037 int
3038 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3039 int gid)
3040 {
3041
3042 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3043 AT_SYMLINK_NOFOLLOW));
3044 }
3045
3046 /*
3047 * Set ownership given a file descriptor.
3048 */
3049 #ifndef _SYS_SYSPROTO_H_
3050 struct fchown_args {
3051 int fd;
3052 int uid;
3053 int gid;
3054 };
3055 #endif
3056 int
3057 sys_fchown(td, uap)
3058 struct thread *td;
3059 register struct fchown_args /* {
3060 int fd;
3061 int uid;
3062 int gid;
3063 } */ *uap;
3064 {
3065 struct file *fp;
3066 cap_rights_t rights;
3067 int error;
3068
3069 AUDIT_ARG_FD(uap->fd);
3070 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3071 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
3072 if (error != 0)
3073 return (error);
3074 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3075 fdrop(fp, td);
3076 return (error);
3077 }
3078
3079 /*
3080 * Common implementation code for utimes(), lutimes(), and futimes().
3081 */
3082 static int
3083 getutimes(usrtvp, tvpseg, tsp)
3084 const struct timeval *usrtvp;
3085 enum uio_seg tvpseg;
3086 struct timespec *tsp;
3087 {
3088 struct timeval tv[2];
3089 const struct timeval *tvp;
3090 int error;
3091
3092 if (usrtvp == NULL) {
3093 vfs_timestamp(&tsp[0]);
3094 tsp[1] = tsp[0];
3095 } else {
3096 if (tvpseg == UIO_SYSSPACE) {
3097 tvp = usrtvp;
3098 } else {
3099 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3100 return (error);
3101 tvp = tv;
3102 }
3103
3104 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3105 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3106 return (EINVAL);
3107 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3108 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3109 }
3110 return (0);
3111 }
3112
3113 /*
3114 * Common implementation code for utimes(), lutimes(), and futimes().
3115 */
3116 static int
3117 setutimes(td, vp, ts, numtimes, nullflag)
3118 struct thread *td;
3119 struct vnode *vp;
3120 const struct timespec *ts;
3121 int numtimes;
3122 int nullflag;
3123 {
3124 struct mount *mp;
3125 struct vattr vattr;
3126 int error, setbirthtime;
3127
3128 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3129 return (error);
3130 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3131 setbirthtime = 0;
3132 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3133 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3134 setbirthtime = 1;
3135 VATTR_NULL(&vattr);
3136 vattr.va_atime = ts[0];
3137 vattr.va_mtime = ts[1];
3138 if (setbirthtime)
3139 vattr.va_birthtime = ts[1];
3140 if (numtimes > 2)
3141 vattr.va_birthtime = ts[2];
3142 if (nullflag)
3143 vattr.va_vaflags |= VA_UTIMES_NULL;
3144 #ifdef MAC
3145 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3146 vattr.va_mtime);
3147 #endif
3148 if (error == 0)
3149 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3150 VOP_UNLOCK(vp, 0);
3151 vn_finished_write(mp);
3152 return (error);
3153 }
3154
3155 /*
3156 * Set the access and modification times of a file.
3157 */
3158 #ifndef _SYS_SYSPROTO_H_
3159 struct utimes_args {
3160 char *path;
3161 struct timeval *tptr;
3162 };
3163 #endif
3164 int
3165 sys_utimes(td, uap)
3166 struct thread *td;
3167 register struct utimes_args /* {
3168 char *path;
3169 struct timeval *tptr;
3170 } */ *uap;
3171 {
3172
3173 return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3174 UIO_USERSPACE));
3175 }
3176
3177 #ifndef _SYS_SYSPROTO_H_
3178 struct futimesat_args {
3179 int fd;
3180 const char * path;
3181 const struct timeval * times;
3182 };
3183 #endif
3184 int
3185 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3186 {
3187
3188 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3189 uap->times, UIO_USERSPACE));
3190 }
3191
3192 int
3193 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3194 struct timeval *tptr, enum uio_seg tptrseg)
3195 {
3196
3197 return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3198 }
3199
3200 int
3201 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3202 struct timeval *tptr, enum uio_seg tptrseg)
3203 {
3204 struct nameidata nd;
3205 struct timespec ts[2];
3206 cap_rights_t rights;
3207 int error;
3208
3209 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3210 return (error);
3211 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3212 cap_rights_init(&rights, CAP_FUTIMES), td);
3213
3214 if ((error = namei(&nd)) != 0)
3215 return (error);
3216 NDFREE(&nd, NDF_ONLY_PNBUF);
3217 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3218 vrele(nd.ni_vp);
3219 return (error);
3220 }
3221
3222 /*
3223 * Set the access and modification times of a file.
3224 */
3225 #ifndef _SYS_SYSPROTO_H_
3226 struct lutimes_args {
3227 char *path;
3228 struct timeval *tptr;
3229 };
3230 #endif
3231 int
3232 sys_lutimes(td, uap)
3233 struct thread *td;
3234 register struct lutimes_args /* {
3235 char *path;
3236 struct timeval *tptr;
3237 } */ *uap;
3238 {
3239
3240 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3241 UIO_USERSPACE));
3242 }
3243
3244 int
3245 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3246 struct timeval *tptr, enum uio_seg tptrseg)
3247 {
3248 struct timespec ts[2];
3249 struct nameidata nd;
3250 int error;
3251
3252 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3253 return (error);
3254 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3255 if ((error = namei(&nd)) != 0)
3256 return (error);
3257 NDFREE(&nd, NDF_ONLY_PNBUF);
3258 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3259 vrele(nd.ni_vp);
3260 return (error);
3261 }
3262
3263 /*
3264 * Set the access and modification times of a file.
3265 */
3266 #ifndef _SYS_SYSPROTO_H_
3267 struct futimes_args {
3268 int fd;
3269 struct timeval *tptr;
3270 };
3271 #endif
3272 int
3273 sys_futimes(td, uap)
3274 struct thread *td;
3275 register struct futimes_args /* {
3276 int fd;
3277 struct timeval *tptr;
3278 } */ *uap;
3279 {
3280
3281 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3282 }
3283
3284 int
3285 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3286 enum uio_seg tptrseg)
3287 {
3288 struct timespec ts[2];
3289 struct file *fp;
3290 cap_rights_t rights;
3291 int error;
3292
3293 AUDIT_ARG_FD(fd);
3294 error = getutimes(tptr, tptrseg, ts);
3295 if (error != 0)
3296 return (error);
3297 error = getvnode(td->td_proc->p_fd, fd,
3298 cap_rights_init(&rights, CAP_FUTIMES), &fp);
3299 if (error != 0)
3300 return (error);
3301 #ifdef AUDIT
3302 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3303 AUDIT_ARG_VNODE1(fp->f_vnode);
3304 VOP_UNLOCK(fp->f_vnode, 0);
3305 #endif
3306 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3307 fdrop(fp, td);
3308 return (error);
3309 }
3310
3311 /*
3312 * Truncate a file given its path name.
3313 */
3314 #ifndef _SYS_SYSPROTO_H_
3315 struct truncate_args {
3316 char *path;
3317 int pad;
3318 off_t length;
3319 };
3320 #endif
3321 int
3322 sys_truncate(td, uap)
3323 struct thread *td;
3324 register struct truncate_args /* {
3325 char *path;
3326 int pad;
3327 off_t length;
3328 } */ *uap;
3329 {
3330
3331 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3332 }
3333
3334 int
3335 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3336 {
3337 struct mount *mp;
3338 struct vnode *vp;
3339 void *rl_cookie;
3340 struct vattr vattr;
3341 struct nameidata nd;
3342 int error;
3343
3344 if (length < 0)
3345 return(EINVAL);
3346 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3347 if ((error = namei(&nd)) != 0)
3348 return (error);
3349 vp = nd.ni_vp;
3350 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3351 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3352 vn_rangelock_unlock(vp, rl_cookie);
3353 vrele(vp);
3354 return (error);
3355 }
3356 NDFREE(&nd, NDF_ONLY_PNBUF);
3357 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3358 if (vp->v_type == VDIR)
3359 error = EISDIR;
3360 #ifdef MAC
3361 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3362 }
3363 #endif
3364 else if ((error = vn_writechk(vp)) == 0 &&
3365 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3366 VATTR_NULL(&vattr);
3367 vattr.va_size = length;
3368 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3369 }
3370 VOP_UNLOCK(vp, 0);
3371 vn_finished_write(mp);
3372 vn_rangelock_unlock(vp, rl_cookie);
3373 vrele(vp);
3374 return (error);
3375 }
3376
3377 #if defined(COMPAT_43)
3378 /*
3379 * Truncate a file given its path name.
3380 */
3381 #ifndef _SYS_SYSPROTO_H_
3382 struct otruncate_args {
3383 char *path;
3384 long length;
3385 };
3386 #endif
3387 int
3388 otruncate(td, uap)
3389 struct thread *td;
3390 register struct otruncate_args /* {
3391 char *path;
3392 long length;
3393 } */ *uap;
3394 {
3395 struct truncate_args /* {
3396 char *path;
3397 int pad;
3398 off_t length;
3399 } */ nuap;
3400
3401 nuap.path = uap->path;
3402 nuap.length = uap->length;
3403 return (sys_truncate(td, &nuap));
3404 }
3405 #endif /* COMPAT_43 */
3406
3407 /* Versions with the pad argument */
3408 int
3409 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3410 {
3411 struct truncate_args ouap;
3412
3413 ouap.path = uap->path;
3414 ouap.length = uap->length;
3415 return (sys_truncate(td, &ouap));
3416 }
3417
3418 int
3419 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3420 {
3421 struct ftruncate_args ouap;
3422
3423 ouap.fd = uap->fd;
3424 ouap.length = uap->length;
3425 return (sys_ftruncate(td, &ouap));
3426 }
3427
3428 /*
3429 * Sync an open file.
3430 */
3431 #ifndef _SYS_SYSPROTO_H_
3432 struct fsync_args {
3433 int fd;
3434 };
3435 #endif
3436 int
3437 sys_fsync(td, uap)
3438 struct thread *td;
3439 struct fsync_args /* {
3440 int fd;
3441 } */ *uap;
3442 {
3443 struct vnode *vp;
3444 struct mount *mp;
3445 struct file *fp;
3446 cap_rights_t rights;
3447 int error, lock_flags;
3448
3449 AUDIT_ARG_FD(uap->fd);
3450 error = getvnode(td->td_proc->p_fd, uap->fd,
3451 cap_rights_init(&rights, CAP_FSYNC), &fp);
3452 if (error != 0)
3453 return (error);
3454 vp = fp->f_vnode;
3455 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3456 if (error != 0)
3457 goto drop;
3458 if (MNT_SHARED_WRITES(mp) ||
3459 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3460 lock_flags = LK_SHARED;
3461 } else {
3462 lock_flags = LK_EXCLUSIVE;
3463 }
3464 vn_lock(vp, lock_flags | LK_RETRY);
3465 AUDIT_ARG_VNODE1(vp);
3466 if (vp->v_object != NULL) {
3467 VM_OBJECT_WLOCK(vp->v_object);
3468 vm_object_page_clean(vp->v_object, 0, 0, 0);
3469 VM_OBJECT_WUNLOCK(vp->v_object);
3470 }
3471 error = VOP_FSYNC(vp, MNT_WAIT, td);
3472
3473 VOP_UNLOCK(vp, 0);
3474 vn_finished_write(mp);
3475 drop:
3476 fdrop(fp, td);
3477 return (error);
3478 }
3479
3480 /*
3481 * Rename files. Source and destination must either both be directories, or
3482 * both not be directories. If target is a directory, it must be empty.
3483 */
3484 #ifndef _SYS_SYSPROTO_H_
3485 struct rename_args {
3486 char *from;
3487 char *to;
3488 };
3489 #endif
3490 int
3491 sys_rename(td, uap)
3492 struct thread *td;
3493 register struct rename_args /* {
3494 char *from;
3495 char *to;
3496 } */ *uap;
3497 {
3498
3499 return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3500 }
3501
3502 #ifndef _SYS_SYSPROTO_H_
3503 struct renameat_args {
3504 int oldfd;
3505 char *old;
3506 int newfd;
3507 char *new;
3508 };
3509 #endif
3510 int
3511 sys_renameat(struct thread *td, struct renameat_args *uap)
3512 {
3513
3514 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3515 UIO_USERSPACE));
3516 }
3517
3518 int
3519 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3520 {
3521
3522 return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3523 }
3524
3525 int
3526 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3527 enum uio_seg pathseg)
3528 {
3529 struct mount *mp = NULL;
3530 struct vnode *tvp, *fvp, *tdvp;
3531 struct nameidata fromnd, tond;
3532 cap_rights_t rights;
3533 int error;
3534
3535 again:
3536 bwillwrite();
3537 #ifdef MAC
3538 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3539 AUDITVNODE1, pathseg, old, oldfd,
3540 cap_rights_init(&rights, CAP_RENAMEAT), td);
3541 #else
3542 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3543 pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3544 #endif
3545
3546 if ((error = namei(&fromnd)) != 0)
3547 return (error);
3548 #ifdef MAC
3549 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3550 fromnd.ni_vp, &fromnd.ni_cnd);
3551 VOP_UNLOCK(fromnd.ni_dvp, 0);
3552 if (fromnd.ni_dvp != fromnd.ni_vp)
3553 VOP_UNLOCK(fromnd.ni_vp, 0);
3554 #endif
3555 fvp = fromnd.ni_vp;
3556 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3557 SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3558 cap_rights_init(&rights, CAP_LINKAT), td);
3559 if (fromnd.ni_vp->v_type == VDIR)
3560 tond.ni_cnd.cn_flags |= WILLBEDIR;
3561 if ((error = namei(&tond)) != 0) {
3562 /* Translate error code for rename("dir1", "dir2/."). */
3563 if (error == EISDIR && fvp->v_type == VDIR)
3564 error = EINVAL;
3565 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3566 vrele(fromnd.ni_dvp);
3567 vrele(fvp);
3568 goto out1;
3569 }
3570 tdvp = tond.ni_dvp;
3571 tvp = tond.ni_vp;
3572 error = vn_start_write(fvp, &mp, V_NOWAIT);
3573 if (error != 0) {
3574 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3575 NDFREE(&tond, NDF_ONLY_PNBUF);
3576 if (tvp != NULL)
3577 vput(tvp);
3578 if (tdvp == tvp)
3579 vrele(tdvp);
3580 else
3581 vput(tdvp);
3582 vrele(fromnd.ni_dvp);
3583 vrele(fvp);
3584 vrele(tond.ni_startdir);
3585 if (fromnd.ni_startdir != NULL)
3586 vrele(fromnd.ni_startdir);
3587 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3588 if (error != 0)
3589 return (error);
3590 goto again;
3591 }
3592 if (tvp != NULL) {
3593 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3594 error = ENOTDIR;
3595 goto out;
3596 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3597 error = EISDIR;
3598 goto out;
3599 }
3600 #ifdef CAPABILITIES
3601 if (newfd != AT_FDCWD) {
3602 /*
3603 * If the target already exists we require CAP_UNLINKAT
3604 * from 'newfd'.
3605 */
3606 error = cap_check(&tond.ni_filecaps.fc_rights,
3607 cap_rights_init(&rights, CAP_UNLINKAT));
3608 if (error != 0)
3609 goto out;
3610 }
3611 #endif
3612 }
3613 if (fvp == tdvp) {
3614 error = EINVAL;
3615 goto out;
3616 }
3617 /*
3618 * If the source is the same as the destination (that is, if they
3619 * are links to the same vnode), then there is nothing to do.
3620 */
3621 if (fvp == tvp)
3622 error = -1;
3623 #ifdef MAC
3624 else
3625 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3626 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3627 #endif
3628 out:
3629 if (error == 0) {
3630 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3631 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3632 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3633 NDFREE(&tond, NDF_ONLY_PNBUF);
3634 } else {
3635 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3636 NDFREE(&tond, NDF_ONLY_PNBUF);
3637 if (tvp != NULL)
3638 vput(tvp);
3639 if (tdvp == tvp)
3640 vrele(tdvp);
3641 else
3642 vput(tdvp);
3643 vrele(fromnd.ni_dvp);
3644 vrele(fvp);
3645 }
3646 vrele(tond.ni_startdir);
3647 vn_finished_write(mp);
3648 out1:
3649 if (fromnd.ni_startdir)
3650 vrele(fromnd.ni_startdir);
3651 if (error == -1)
3652 return (0);
3653 return (error);
3654 }
3655
3656 /*
3657 * Make a directory file.
3658 */
3659 #ifndef _SYS_SYSPROTO_H_
3660 struct mkdir_args {
3661 char *path;
3662 int mode;
3663 };
3664 #endif
3665 int
3666 sys_mkdir(td, uap)
3667 struct thread *td;
3668 register struct mkdir_args /* {
3669 char *path;
3670 int mode;
3671 } */ *uap;
3672 {
3673
3674 return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3675 }
3676
3677 #ifndef _SYS_SYSPROTO_H_
3678 struct mkdirat_args {
3679 int fd;
3680 char *path;
3681 mode_t mode;
3682 };
3683 #endif
3684 int
3685 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3686 {
3687
3688 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3689 }
3690
3691 int
3692 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3693 {
3694
3695 return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3696 }
3697
3698 int
3699 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3700 int mode)
3701 {
3702 struct mount *mp;
3703 struct vnode *vp;
3704 struct vattr vattr;
3705 struct nameidata nd;
3706 cap_rights_t rights;
3707 int error;
3708
3709 AUDIT_ARG_MODE(mode);
3710 restart:
3711 bwillwrite();
3712 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3713 NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3714 td);
3715 nd.ni_cnd.cn_flags |= WILLBEDIR;
3716 if ((error = namei(&nd)) != 0)
3717 return (error);
3718 vp = nd.ni_vp;
3719 if (vp != NULL) {
3720 NDFREE(&nd, NDF_ONLY_PNBUF);
3721 /*
3722 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3723 * the strange behaviour of leaving the vnode unlocked
3724 * if the target is the same vnode as the parent.
3725 */
3726 if (vp == nd.ni_dvp)
3727 vrele(nd.ni_dvp);
3728 else
3729 vput(nd.ni_dvp);
3730 vrele(vp);
3731 return (EEXIST);
3732 }
3733 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3734 NDFREE(&nd, NDF_ONLY_PNBUF);
3735 vput(nd.ni_dvp);
3736 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3737 return (error);
3738 goto restart;
3739 }
3740 VATTR_NULL(&vattr);
3741 vattr.va_type = VDIR;
3742 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3743 #ifdef MAC
3744 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3745 &vattr);
3746 if (error != 0)
3747 goto out;
3748 #endif
3749 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3750 #ifdef MAC
3751 out:
3752 #endif
3753 NDFREE(&nd, NDF_ONLY_PNBUF);
3754 vput(nd.ni_dvp);
3755 if (error == 0)
3756 vput(nd.ni_vp);
3757 vn_finished_write(mp);
3758 return (error);
3759 }
3760
3761 /*
3762 * Remove a directory file.
3763 */
3764 #ifndef _SYS_SYSPROTO_H_
3765 struct rmdir_args {
3766 char *path;
3767 };
3768 #endif
3769 int
3770 sys_rmdir(td, uap)
3771 struct thread *td;
3772 struct rmdir_args /* {
3773 char *path;
3774 } */ *uap;
3775 {
3776
3777 return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3778 }
3779
3780 int
3781 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3782 {
3783
3784 return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3785 }
3786
3787 int
3788 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3789 {
3790 struct mount *mp;
3791 struct vnode *vp;
3792 struct nameidata nd;
3793 cap_rights_t rights;
3794 int error;
3795
3796 restart:
3797 bwillwrite();
3798 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3799 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3800 if ((error = namei(&nd)) != 0)
3801 return (error);
3802 vp = nd.ni_vp;
3803 if (vp->v_type != VDIR) {
3804 error = ENOTDIR;
3805 goto out;
3806 }
3807 /*
3808 * No rmdir "." please.
3809 */
3810 if (nd.ni_dvp == vp) {
3811 error = EINVAL;
3812 goto out;
3813 }
3814 /*
3815 * The root of a mounted filesystem cannot be deleted.
3816 */
3817 if (vp->v_vflag & VV_ROOT) {
3818 error = EBUSY;
3819 goto out;
3820 }
3821 #ifdef MAC
3822 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3823 &nd.ni_cnd);
3824 if (error != 0)
3825 goto out;
3826 #endif
3827 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3828 NDFREE(&nd, NDF_ONLY_PNBUF);
3829 vput(vp);
3830 if (nd.ni_dvp == vp)
3831 vrele(nd.ni_dvp);
3832 else
3833 vput(nd.ni_dvp);
3834 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3835 return (error);
3836 goto restart;
3837 }
3838 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3839 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3840 vn_finished_write(mp);
3841 out:
3842 NDFREE(&nd, NDF_ONLY_PNBUF);
3843 vput(vp);
3844 if (nd.ni_dvp == vp)
3845 vrele(nd.ni_dvp);
3846 else
3847 vput(nd.ni_dvp);
3848 return (error);
3849 }
3850
3851 #ifdef COMPAT_43
3852 /*
3853 * Read a block of directory entries in a filesystem independent format.
3854 */
3855 #ifndef _SYS_SYSPROTO_H_
3856 struct ogetdirentries_args {
3857 int fd;
3858 char *buf;
3859 u_int count;
3860 long *basep;
3861 };
3862 #endif
3863 int
3864 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3865 {
3866 long loff;
3867 int error;
3868
3869 error = kern_ogetdirentries(td, uap, &loff);
3870 if (error == 0)
3871 error = copyout(&loff, uap->basep, sizeof(long));
3872 return (error);
3873 }
3874
3875 int
3876 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3877 long *ploff)
3878 {
3879 struct vnode *vp;
3880 struct file *fp;
3881 struct uio auio, kuio;
3882 struct iovec aiov, kiov;
3883 struct dirent *dp, *edp;
3884 cap_rights_t rights;
3885 caddr_t dirbuf;
3886 int error, eofflag, readcnt;
3887 long loff;
3888 off_t foffset;
3889
3890 /* XXX arbitrary sanity limit on `count'. */
3891 if (uap->count > 64 * 1024)
3892 return (EINVAL);
3893 error = getvnode(td->td_proc->p_fd, uap->fd,
3894 cap_rights_init(&rights, CAP_READ), &fp);
3895 if (error != 0)
3896 return (error);
3897 if ((fp->f_flag & FREAD) == 0) {
3898 fdrop(fp, td);
3899 return (EBADF);
3900 }
3901 vp = fp->f_vnode;
3902 foffset = foffset_lock(fp, 0);
3903 unionread:
3904 if (vp->v_type != VDIR) {
3905 foffset_unlock(fp, foffset, 0);
3906 fdrop(fp, td);
3907 return (EINVAL);
3908 }
3909 aiov.iov_base = uap->buf;
3910 aiov.iov_len = uap->count;
3911 auio.uio_iov = &aiov;
3912 auio.uio_iovcnt = 1;
3913 auio.uio_rw = UIO_READ;
3914 auio.uio_segflg = UIO_USERSPACE;
3915 auio.uio_td = td;
3916 auio.uio_resid = uap->count;
3917 vn_lock(vp, LK_SHARED | LK_RETRY);
3918 loff = auio.uio_offset = foffset;
3919 #ifdef MAC
3920 error = mac_vnode_check_readdir(td->td_ucred, vp);
3921 if (error != 0) {
3922 VOP_UNLOCK(vp, 0);
3923 foffset_unlock(fp, foffset, FOF_NOUPDATE);
3924 fdrop(fp, td);
3925 return (error);
3926 }
3927 #endif
3928 # if (BYTE_ORDER != LITTLE_ENDIAN)
3929 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3930 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3931 NULL, NULL);
3932 foffset = auio.uio_offset;
3933 } else
3934 # endif
3935 {
3936 kuio = auio;
3937 kuio.uio_iov = &kiov;
3938 kuio.uio_segflg = UIO_SYSSPACE;
3939 kiov.iov_len = uap->count;
3940 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3941 kiov.iov_base = dirbuf;
3942 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3943 NULL, NULL);
3944 foffset = kuio.uio_offset;
3945 if (error == 0) {
3946 readcnt = uap->count - kuio.uio_resid;
3947 edp = (struct dirent *)&dirbuf[readcnt];
3948 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3949 # if (BYTE_ORDER == LITTLE_ENDIAN)
3950 /*
3951 * The expected low byte of
3952 * dp->d_namlen is our dp->d_type.
3953 * The high MBZ byte of dp->d_namlen
3954 * is our dp->d_namlen.
3955 */
3956 dp->d_type = dp->d_namlen;
3957 dp->d_namlen = 0;
3958 # else
3959 /*
3960 * The dp->d_type is the high byte
3961 * of the expected dp->d_namlen,
3962 * so must be zero'ed.
3963 */
3964 dp->d_type = 0;
3965 # endif
3966 if (dp->d_reclen > 0) {
3967 dp = (struct dirent *)
3968 ((char *)dp + dp->d_reclen);
3969 } else {
3970 error = EIO;
3971 break;
3972 }
3973 }
3974 if (dp >= edp)
3975 error = uiomove(dirbuf, readcnt, &auio);
3976 }
3977 free(dirbuf, M_TEMP);
3978 }
3979 if (error != 0) {
3980 VOP_UNLOCK(vp, 0);
3981 foffset_unlock(fp, foffset, 0);
3982 fdrop(fp, td);
3983 return (error);
3984 }
3985 if (uap->count == auio.uio_resid &&
3986 (vp->v_vflag & VV_ROOT) &&
3987 (vp->v_mount->mnt_flag & MNT_UNION)) {
3988 struct vnode *tvp = vp;
3989 vp = vp->v_mount->mnt_vnodecovered;
3990 VREF(vp);
3991 fp->f_vnode = vp;
3992 fp->f_data = vp;
3993 foffset = 0;
3994 vput(tvp);
3995 goto unionread;
3996 }
3997 VOP_UNLOCK(vp, 0);
3998 foffset_unlock(fp, foffset, 0);
3999 fdrop(fp, td);
4000 td->td_retval[0] = uap->count - auio.uio_resid;
4001 if (error == 0)
4002 *ploff = loff;
4003 return (error);
4004 }
4005 #endif /* COMPAT_43 */
4006
4007 /*
4008 * Read a block of directory entries in a filesystem independent format.
4009 */
4010 #ifndef _SYS_SYSPROTO_H_
4011 struct getdirentries_args {
4012 int fd;
4013 char *buf;
4014 u_int count;
4015 long *basep;
4016 };
4017 #endif
4018 int
4019 sys_getdirentries(td, uap)
4020 struct thread *td;
4021 register struct getdirentries_args /* {
4022 int fd;
4023 char *buf;
4024 u_int count;
4025 long *basep;
4026 } */ *uap;
4027 {
4028 long base;
4029 int error;
4030
4031 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4032 NULL, UIO_USERSPACE);
4033 if (error != 0)
4034 return (error);
4035 if (uap->basep != NULL)
4036 error = copyout(&base, uap->basep, sizeof(long));
4037 return (error);
4038 }
4039
4040 int
4041 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4042 long *basep, ssize_t *residp, enum uio_seg bufseg)
4043 {
4044 struct vnode *vp;
4045 struct file *fp;
4046 struct uio auio;
4047 struct iovec aiov;
4048 cap_rights_t rights;
4049 long loff;
4050 int error, eofflag;
4051 off_t foffset;
4052
4053 AUDIT_ARG_FD(fd);
4054 if (count > IOSIZE_MAX)
4055 return (EINVAL);
4056 auio.uio_resid = count;
4057 error = getvnode(td->td_proc->p_fd, fd,
4058 cap_rights_init(&rights, CAP_READ), &fp);
4059 if (error != 0)
4060 return (error);
4061 if ((fp->f_flag & FREAD) == 0) {
4062 fdrop(fp, td);
4063 return (EBADF);
4064 }
4065 vp = fp->f_vnode;
4066 foffset = foffset_lock(fp, 0);
4067 unionread:
4068 if (vp->v_type != VDIR) {
4069 error = EINVAL;
4070 goto fail;
4071 }
4072 aiov.iov_base = buf;
4073 aiov.iov_len = count;
4074 auio.uio_iov = &aiov;
4075 auio.uio_iovcnt = 1;
4076 auio.uio_rw = UIO_READ;
4077 auio.uio_segflg = bufseg;
4078 auio.uio_td = td;
4079 vn_lock(vp, LK_SHARED | LK_RETRY);
4080 AUDIT_ARG_VNODE1(vp);
4081 loff = auio.uio_offset = foffset;
4082 #ifdef MAC
4083 error = mac_vnode_check_readdir(td->td_ucred, vp);
4084 if (error == 0)
4085 #endif
4086 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4087 NULL);
4088 foffset = auio.uio_offset;
4089 if (error != 0) {
4090 VOP_UNLOCK(vp, 0);
4091 goto fail;
4092 }
4093 if (count == auio.uio_resid &&
4094 (vp->v_vflag & VV_ROOT) &&
4095 (vp->v_mount->mnt_flag & MNT_UNION)) {
4096 struct vnode *tvp = vp;
4097
4098 vp = vp->v_mount->mnt_vnodecovered;
4099 VREF(vp);
4100 fp->f_vnode = vp;
4101 fp->f_data = vp;
4102 foffset = 0;
4103 vput(tvp);
4104 goto unionread;
4105 }
4106 VOP_UNLOCK(vp, 0);
4107 *basep = loff;
4108 if (residp != NULL)
4109 *residp = auio.uio_resid;
4110 td->td_retval[0] = count - auio.uio_resid;
4111 fail:
4112 foffset_unlock(fp, foffset, 0);
4113 fdrop(fp, td);
4114 return (error);
4115 }
4116
4117 #ifndef _SYS_SYSPROTO_H_
4118 struct getdents_args {
4119 int fd;
4120 char *buf;
4121 size_t count;
4122 };
4123 #endif
4124 int
4125 sys_getdents(td, uap)
4126 struct thread *td;
4127 register struct getdents_args /* {
4128 int fd;
4129 char *buf;
4130 u_int count;
4131 } */ *uap;
4132 {
4133 struct getdirentries_args ap;
4134
4135 ap.fd = uap->fd;
4136 ap.buf = uap->buf;
4137 ap.count = uap->count;
4138 ap.basep = NULL;
4139 return (sys_getdirentries(td, &ap));
4140 }
4141
4142 /*
4143 * Set the mode mask for creation of filesystem nodes.
4144 */
4145 #ifndef _SYS_SYSPROTO_H_
4146 struct umask_args {
4147 int newmask;
4148 };
4149 #endif
4150 int
4151 sys_umask(td, uap)
4152 struct thread *td;
4153 struct umask_args /* {
4154 int newmask;
4155 } */ *uap;
4156 {
4157 register struct filedesc *fdp;
4158
4159 FILEDESC_XLOCK(td->td_proc->p_fd);
4160 fdp = td->td_proc->p_fd;
4161 td->td_retval[0] = fdp->fd_cmask;
4162 fdp->fd_cmask = uap->newmask & ALLPERMS;
4163 FILEDESC_XUNLOCK(td->td_proc->p_fd);
4164 return (0);
4165 }
4166
4167 /*
4168 * Void all references to file by ripping underlying filesystem away from
4169 * vnode.
4170 */
4171 #ifndef _SYS_SYSPROTO_H_
4172 struct revoke_args {
4173 char *path;
4174 };
4175 #endif
4176 int
4177 sys_revoke(td, uap)
4178 struct thread *td;
4179 register struct revoke_args /* {
4180 char *path;
4181 } */ *uap;
4182 {
4183 struct vnode *vp;
4184 struct vattr vattr;
4185 struct nameidata nd;
4186 int error;
4187
4188 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4189 uap->path, td);
4190 if ((error = namei(&nd)) != 0)
4191 return (error);
4192 vp = nd.ni_vp;
4193 NDFREE(&nd, NDF_ONLY_PNBUF);
4194 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4195 error = EINVAL;
4196 goto out;
4197 }
4198 #ifdef MAC
4199 error = mac_vnode_check_revoke(td->td_ucred, vp);
4200 if (error != 0)
4201 goto out;
4202 #endif
4203 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4204 if (error != 0)
4205 goto out;
4206 if (td->td_ucred->cr_uid != vattr.va_uid) {
4207 error = priv_check(td, PRIV_VFS_ADMIN);
4208 if (error != 0)
4209 goto out;
4210 }
4211 if (vcount(vp) > 1)
4212 VOP_REVOKE(vp, REVOKEALL);
4213 out:
4214 vput(vp);
4215 return (error);
4216 }
4217
4218 /*
4219 * Convert a user file descriptor to a kernel file entry and check that, if it
4220 * is a capability, the correct rights are present. A reference on the file
4221 * entry is held upon returning.
4222 */
4223 int
4224 getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
4225 {
4226 struct file *fp;
4227 int error;
4228
4229 error = fget_unlocked(fdp, fd, rightsp, 0, &fp, NULL);
4230 if (error != 0)
4231 return (error);
4232
4233 /*
4234 * The file could be not of the vnode type, or it may be not
4235 * yet fully initialized, in which case the f_vnode pointer
4236 * may be set, but f_ops is still badfileops. E.g.,
4237 * devfs_open() transiently create such situation to
4238 * facilitate csw d_fdopen().
4239 *
4240 * Dupfdopen() handling in kern_openat() installs the
4241 * half-baked file into the process descriptor table, allowing
4242 * other thread to dereference it. Guard against the race by
4243 * checking f_ops.
4244 */
4245 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4246 fdrop(fp, curthread);
4247 return (EINVAL);
4248 }
4249 *fpp = fp;
4250 return (0);
4251 }
4252
4253
4254 /*
4255 * Get an (NFS) file handle.
4256 */
4257 #ifndef _SYS_SYSPROTO_H_
4258 struct lgetfh_args {
4259 char *fname;
4260 fhandle_t *fhp;
4261 };
4262 #endif
4263 int
4264 sys_lgetfh(td, uap)
4265 struct thread *td;
4266 register struct lgetfh_args *uap;
4267 {
4268 struct nameidata nd;
4269 fhandle_t fh;
4270 register struct vnode *vp;
4271 int error;
4272
4273 error = priv_check(td, PRIV_VFS_GETFH);
4274 if (error != 0)
4275 return (error);
4276 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4277 uap->fname, td);
4278 error = namei(&nd);
4279 if (error != 0)
4280 return (error);
4281 NDFREE(&nd, NDF_ONLY_PNBUF);
4282 vp = nd.ni_vp;
4283 bzero(&fh, sizeof(fh));
4284 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4285 error = VOP_VPTOFH(vp, &fh.fh_fid);
4286 vput(vp);
4287 if (error == 0)
4288 error = copyout(&fh, uap->fhp, sizeof (fh));
4289 return (error);
4290 }
4291
4292 #ifndef _SYS_SYSPROTO_H_
4293 struct getfh_args {
4294 char *fname;
4295 fhandle_t *fhp;
4296 };
4297 #endif
4298 int
4299 sys_getfh(td, uap)
4300 struct thread *td;
4301 register struct getfh_args *uap;
4302 {
4303 struct nameidata nd;
4304 fhandle_t fh;
4305 register struct vnode *vp;
4306 int error;
4307
4308 error = priv_check(td, PRIV_VFS_GETFH);
4309 if (error != 0)
4310 return (error);
4311 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4312 uap->fname, td);
4313 error = namei(&nd);
4314 if (error != 0)
4315 return (error);
4316 NDFREE(&nd, NDF_ONLY_PNBUF);
4317 vp = nd.ni_vp;
4318 bzero(&fh, sizeof(fh));
4319 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4320 error = VOP_VPTOFH(vp, &fh.fh_fid);
4321 vput(vp);
4322 if (error == 0)
4323 error = copyout(&fh, uap->fhp, sizeof (fh));
4324 return (error);
4325 }
4326
4327 /*
4328 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4329 * open descriptor.
4330 *
4331 * warning: do not remove the priv_check() call or this becomes one giant
4332 * security hole.
4333 */
4334 #ifndef _SYS_SYSPROTO_H_
4335 struct fhopen_args {
4336 const struct fhandle *u_fhp;
4337 int flags;
4338 };
4339 #endif
4340 int
4341 sys_fhopen(td, uap)
4342 struct thread *td;
4343 struct fhopen_args /* {
4344 const struct fhandle *u_fhp;
4345 int flags;
4346 } */ *uap;
4347 {
4348 struct mount *mp;
4349 struct vnode *vp;
4350 struct fhandle fhp;
4351 struct file *fp;
4352 int fmode, error;
4353 int indx;
4354
4355 error = priv_check(td, PRIV_VFS_FHOPEN);
4356 if (error != 0)
4357 return (error);
4358 indx = -1;
4359 fmode = FFLAGS(uap->flags);
4360 /* why not allow a non-read/write open for our lockd? */
4361 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4362 return (EINVAL);
4363 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4364 if (error != 0)
4365 return(error);
4366 /* find the mount point */
4367 mp = vfs_busyfs(&fhp.fh_fsid);
4368 if (mp == NULL)
4369 return (ESTALE);
4370 /* now give me my vnode, it gets returned to me locked */
4371 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4372 vfs_unbusy(mp);
4373 if (error != 0)
4374 return (error);
4375
4376 error = falloc_noinstall(td, &fp);
4377 if (error != 0) {
4378 vput(vp);
4379 return (error);
4380 }
4381 /*
4382 * An extra reference on `fp' has been held for us by
4383 * falloc_noinstall().
4384 */
4385
4386 #ifdef INVARIANTS
4387 td->td_dupfd = -1;
4388 #endif
4389 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4390 if (error != 0) {
4391 KASSERT(fp->f_ops == &badfileops,
4392 ("VOP_OPEN in fhopen() set f_ops"));
4393 KASSERT(td->td_dupfd < 0,
4394 ("fhopen() encountered fdopen()"));
4395
4396 vput(vp);
4397 goto bad;
4398 }
4399 #ifdef INVARIANTS
4400 td->td_dupfd = 0;
4401 #endif
4402 fp->f_vnode = vp;
4403 fp->f_seqcount = 1;
4404 finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4405 &vnops);
4406 VOP_UNLOCK(vp, 0);
4407 if ((fmode & O_TRUNC) != 0) {
4408 error = fo_truncate(fp, 0, td->td_ucred, td);
4409 if (error != 0)
4410 goto bad;
4411 }
4412
4413 error = finstall(td, fp, &indx, fmode, NULL);
4414 bad:
4415 fdrop(fp, td);
4416 td->td_retval[0] = indx;
4417 return (error);
4418 }
4419
4420 /*
4421 * Stat an (NFS) file handle.
4422 */
4423 #ifndef _SYS_SYSPROTO_H_
4424 struct fhstat_args {
4425 struct fhandle *u_fhp;
4426 struct stat *sb;
4427 };
4428 #endif
4429 int
4430 sys_fhstat(td, uap)
4431 struct thread *td;
4432 register struct fhstat_args /* {
4433 struct fhandle *u_fhp;
4434 struct stat *sb;
4435 } */ *uap;
4436 {
4437 struct stat sb;
4438 struct fhandle fh;
4439 int error;
4440
4441 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4442 if (error != 0)
4443 return (error);
4444 error = kern_fhstat(td, fh, &sb);
4445 if (error == 0)
4446 error = copyout(&sb, uap->sb, sizeof(sb));
4447 return (error);
4448 }
4449
4450 int
4451 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4452 {
4453 struct mount *mp;
4454 struct vnode *vp;
4455 int error;
4456
4457 error = priv_check(td, PRIV_VFS_FHSTAT);
4458 if (error != 0)
4459 return (error);
4460 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4461 return (ESTALE);
4462 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4463 vfs_unbusy(mp);
4464 if (error != 0)
4465 return (error);
4466 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4467 vput(vp);
4468 return (error);
4469 }
4470
4471 /*
4472 * Implement fstatfs() for (NFS) file handles.
4473 */
4474 #ifndef _SYS_SYSPROTO_H_
4475 struct fhstatfs_args {
4476 struct fhandle *u_fhp;
4477 struct statfs *buf;
4478 };
4479 #endif
4480 int
4481 sys_fhstatfs(td, uap)
4482 struct thread *td;
4483 struct fhstatfs_args /* {
4484 struct fhandle *u_fhp;
4485 struct statfs *buf;
4486 } */ *uap;
4487 {
4488 struct statfs sf;
4489 fhandle_t fh;
4490 int error;
4491
4492 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4493 if (error != 0)
4494 return (error);
4495 error = kern_fhstatfs(td, fh, &sf);
4496 if (error != 0)
4497 return (error);
4498 return (copyout(&sf, uap->buf, sizeof(sf)));
4499 }
4500
4501 int
4502 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4503 {
4504 struct statfs *sp;
4505 struct mount *mp;
4506 struct vnode *vp;
4507 int error;
4508
4509 error = priv_check(td, PRIV_VFS_FHSTATFS);
4510 if (error != 0)
4511 return (error);
4512 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4513 return (ESTALE);
4514 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4515 if (error != 0) {
4516 vfs_unbusy(mp);
4517 return (error);
4518 }
4519 vput(vp);
4520 error = prison_canseemount(td->td_ucred, mp);
4521 if (error != 0)
4522 goto out;
4523 #ifdef MAC
4524 error = mac_mount_check_stat(td->td_ucred, mp);
4525 if (error != 0)
4526 goto out;
4527 #endif
4528 /*
4529 * Set these in case the underlying filesystem fails to do so.
4530 */
4531 sp = &mp->mnt_stat;
4532 sp->f_version = STATFS_VERSION;
4533 sp->f_namemax = NAME_MAX;
4534 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4535 error = VFS_STATFS(mp, sp);
4536 if (error == 0)
4537 *buf = *sp;
4538 out:
4539 vfs_unbusy(mp);
4540 return (error);
4541 }
4542
4543 int
4544 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4545 {
4546 struct file *fp;
4547 struct mount *mp;
4548 struct vnode *vp;
4549 cap_rights_t rights;
4550 off_t olen, ooffset;
4551 int error;
4552
4553 if (offset < 0 || len <= 0)
4554 return (EINVAL);
4555 /* Check for wrap. */
4556 if (offset > OFF_MAX - len)
4557 return (EFBIG);
4558 error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4559 if (error != 0)
4560 return (error);
4561 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4562 error = ESPIPE;
4563 goto out;
4564 }
4565 if ((fp->f_flag & FWRITE) == 0) {
4566 error = EBADF;
4567 goto out;
4568 }
4569 if (fp->f_type != DTYPE_VNODE) {
4570 error = ENODEV;
4571 goto out;
4572 }
4573 vp = fp->f_vnode;
4574 if (vp->v_type != VREG) {
4575 error = ENODEV;
4576 goto out;
4577 }
4578
4579 /* Allocating blocks may take a long time, so iterate. */
4580 for (;;) {
4581 olen = len;
4582 ooffset = offset;
4583
4584 bwillwrite();
4585 mp = NULL;
4586 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4587 if (error != 0)
4588 break;
4589 error = vn_lock(vp, LK_EXCLUSIVE);
4590 if (error != 0) {
4591 vn_finished_write(mp);
4592 break;
4593 }
4594 #ifdef MAC
4595 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4596 if (error == 0)
4597 #endif
4598 error = VOP_ALLOCATE(vp, &offset, &len);
4599 VOP_UNLOCK(vp, 0);
4600 vn_finished_write(mp);
4601
4602 if (olen + ooffset != offset + len) {
4603 panic("offset + len changed from %jx/%jx to %jx/%jx",
4604 ooffset, olen, offset, len);
4605 }
4606 if (error != 0 || len == 0)
4607 break;
4608 KASSERT(olen > len, ("Iteration did not make progress?"));
4609 maybe_yield();
4610 }
4611 out:
4612 fdrop(fp, td);
4613 return (error);
4614 }
4615
4616 int
4617 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4618 {
4619
4620 td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4621 uap->len);
4622 return (0);
4623 }
4624
4625 /*
4626 * Unlike madvise(2), we do not make a best effort to remember every
4627 * possible caching hint. Instead, we remember the last setting with
4628 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4629 * region of any current setting.
4630 */
4631 int
4632 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4633 int advice)
4634 {
4635 struct fadvise_info *fa, *new;
4636 struct file *fp;
4637 struct vnode *vp;
4638 cap_rights_t rights;
4639 off_t end;
4640 int error;
4641
4642 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4643 return (EINVAL);
4644 switch (advice) {
4645 case POSIX_FADV_SEQUENTIAL:
4646 case POSIX_FADV_RANDOM:
4647 case POSIX_FADV_NOREUSE:
4648 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4649 break;
4650 case POSIX_FADV_NORMAL:
4651 case POSIX_FADV_WILLNEED:
4652 case POSIX_FADV_DONTNEED:
4653 new = NULL;
4654 break;
4655 default:
4656 return (EINVAL);
4657 }
4658 /* XXX: CAP_POSIX_FADVISE? */
4659 error = fget(td, fd, cap_rights_init(&rights), &fp);
4660 if (error != 0)
4661 goto out;
4662 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4663 error = ESPIPE;
4664 goto out;
4665 }
4666 if (fp->f_type != DTYPE_VNODE) {
4667 error = ENODEV;
4668 goto out;
4669 }
4670 vp = fp->f_vnode;
4671 if (vp->v_type != VREG) {
4672 error = ENODEV;
4673 goto out;
4674 }
4675 if (len == 0)
4676 end = OFF_MAX;
4677 else
4678 end = offset + len - 1;
4679 switch (advice) {
4680 case POSIX_FADV_SEQUENTIAL:
4681 case POSIX_FADV_RANDOM:
4682 case POSIX_FADV_NOREUSE:
4683 /*
4684 * Try to merge any existing non-standard region with
4685 * this new region if possible, otherwise create a new
4686 * non-standard region for this request.
4687 */
4688 mtx_pool_lock(mtxpool_sleep, fp);
4689 fa = fp->f_advice;
4690 if (fa != NULL && fa->fa_advice == advice &&
4691 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4692 (end != OFF_MAX && fa->fa_start == end + 1) ||
4693 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4694 if (offset < fa->fa_start)
4695 fa->fa_start = offset;
4696 if (end > fa->fa_end)
4697 fa->fa_end = end;
4698 } else {
4699 new->fa_advice = advice;
4700 new->fa_start = offset;
4701 new->fa_end = end;
4702 new->fa_prevstart = 0;
4703 new->fa_prevend = 0;
4704 fp->f_advice = new;
4705 new = fa;
4706 }
4707 mtx_pool_unlock(mtxpool_sleep, fp);
4708 break;
4709 case POSIX_FADV_NORMAL:
4710 /*
4711 * If a the "normal" region overlaps with an existing
4712 * non-standard region, trim or remove the
4713 * non-standard region.
4714 */
4715 mtx_pool_lock(mtxpool_sleep, fp);
4716 fa = fp->f_advice;
4717 if (fa != NULL) {
4718 if (offset <= fa->fa_start && end >= fa->fa_end) {
4719 new = fa;
4720 fp->f_advice = NULL;
4721 } else if (offset <= fa->fa_start &&
4722 end >= fa->fa_start)
4723 fa->fa_start = end + 1;
4724 else if (offset <= fa->fa_end && end >= fa->fa_end)
4725 fa->fa_end = offset - 1;
4726 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4727 /*
4728 * If the "normal" region is a middle
4729 * portion of the existing
4730 * non-standard region, just remove
4731 * the whole thing rather than picking
4732 * one side or the other to
4733 * preserve.
4734 */
4735 new = fa;
4736 fp->f_advice = NULL;
4737 }
4738 }
4739 mtx_pool_unlock(mtxpool_sleep, fp);
4740 break;
4741 case POSIX_FADV_WILLNEED:
4742 case POSIX_FADV_DONTNEED:
4743 error = VOP_ADVISE(vp, offset, end, advice);
4744 break;
4745 }
4746 out:
4747 if (fp != NULL)
4748 fdrop(fp, td);
4749 free(new, M_FADVISE);
4750 return (error);
4751 }
4752
4753 int
4754 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4755 {
4756
4757 td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4758 uap->len, uap->advice);
4759 return (0);
4760 }
Cache object: 9f2bf84de4cf7e0bcee97bec294601c8
|