1 /*-
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
35 */
36
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39
40 #include "opt_compat.h"
41 #include "opt_kdtrace.h"
42 #include "opt_ktrace.h"
43
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/disk.h>
49 #include <sys/sysent.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/mutex.h>
53 #include <sys/sysproto.h>
54 #include <sys/namei.h>
55 #include <sys/filedesc.h>
56 #include <sys/kernel.h>
57 #include <sys/fcntl.h>
58 #include <sys/file.h>
59 #include <sys/filio.h>
60 #include <sys/limits.h>
61 #include <sys/linker.h>
62 #include <sys/sdt.h>
63 #include <sys/stat.h>
64 #include <sys/sx.h>
65 #include <sys/unistd.h>
66 #include <sys/vnode.h>
67 #include <sys/priv.h>
68 #include <sys/proc.h>
69 #include <sys/dirent.h>
70 #include <sys/jail.h>
71 #include <sys/syscallsubr.h>
72 #include <sys/sysctl.h>
73 #ifdef KTRACE
74 #include <sys/ktrace.h>
75 #endif
76
77 #include <machine/stdarg.h>
78
79 #include <security/audit/audit.h>
80 #include <security/mac/mac_framework.h>
81
82 #include <vm/vm.h>
83 #include <vm/vm_object.h>
84 #include <vm/vm_page.h>
85 #include <vm/uma.h>
86
87 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
88
89 SDT_PROVIDER_DEFINE(vfs);
90 SDT_PROBE_DEFINE(vfs, , stat, mode, mode);
91 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
92 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
93 SDT_PROBE_DEFINE(vfs, , stat, reg, reg);
94 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
95 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
96
97 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
98 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
99 static int setfown(struct thread *td, struct vnode *, uid_t, gid_t);
100 static int setfmode(struct thread *td, struct vnode *, int);
101 static int setfflags(struct thread *td, struct vnode *, int);
102 static int setutimes(struct thread *td, struct vnode *,
103 const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105 struct thread *td);
106
107 /*
108 * The module initialization routine for POSIX asynchronous I/O will
109 * set this to the version of AIO that it implements. (Zero means
110 * that it is not implemented.) This value is used here by pathconf()
111 * and in kern_descrip.c by fpathconf().
112 */
113 int async_io_version;
114
115 #ifdef DEBUG
116 static int syncprt = 0;
117 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
118 #endif
119
120 /*
121 * Sync each mounted filesystem.
122 */
123 #ifndef _SYS_SYSPROTO_H_
124 struct sync_args {
125 int dummy;
126 };
127 #endif
128 /* ARGSUSED */
129 int
130 sync(td, uap)
131 struct thread *td;
132 struct sync_args *uap;
133 {
134 struct mount *mp, *nmp;
135 int save, vfslocked;
136
137 mtx_lock(&mountlist_mtx);
138 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
139 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
140 nmp = TAILQ_NEXT(mp, mnt_list);
141 continue;
142 }
143 vfslocked = VFS_LOCK_GIANT(mp);
144 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
145 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
146 save = curthread_pflags_set(TDP_SYNCIO);
147 vfs_msync(mp, MNT_NOWAIT);
148 VFS_SYNC(mp, MNT_NOWAIT);
149 curthread_pflags_restore(save);
150 vn_finished_write(mp);
151 }
152 VFS_UNLOCK_GIANT(vfslocked);
153 mtx_lock(&mountlist_mtx);
154 nmp = TAILQ_NEXT(mp, mnt_list);
155 vfs_unbusy(mp);
156 }
157 mtx_unlock(&mountlist_mtx);
158 return (0);
159 }
160
161 /*
162 * Change filesystem quotas.
163 */
164 #ifndef _SYS_SYSPROTO_H_
165 struct quotactl_args {
166 char *path;
167 int cmd;
168 int uid;
169 caddr_t arg;
170 };
171 #endif
172 int
173 quotactl(td, uap)
174 struct thread *td;
175 register struct quotactl_args /* {
176 char *path;
177 int cmd;
178 int uid;
179 caddr_t arg;
180 } */ *uap;
181 {
182 struct mount *mp;
183 int vfslocked;
184 int error;
185 struct nameidata nd;
186
187 AUDIT_ARG_CMD(uap->cmd);
188 AUDIT_ARG_UID(uap->uid);
189 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
190 return (EPERM);
191 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
192 UIO_USERSPACE, uap->path, td);
193 if ((error = namei(&nd)) != 0)
194 return (error);
195 vfslocked = NDHASGIANT(&nd);
196 NDFREE(&nd, NDF_ONLY_PNBUF);
197 mp = nd.ni_vp->v_mount;
198 vfs_ref(mp);
199 vput(nd.ni_vp);
200 error = vfs_busy(mp, 0);
201 vfs_rel(mp);
202 if (error) {
203 VFS_UNLOCK_GIANT(vfslocked);
204 return (error);
205 }
206 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
207 vfs_unbusy(mp);
208 VFS_UNLOCK_GIANT(vfslocked);
209 return (error);
210 }
211
212 /*
213 * Used by statfs conversion routines to scale the block size up if
214 * necessary so that all of the block counts are <= 'max_size'. Note
215 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
216 * value of 'n'.
217 */
218 void
219 statfs_scale_blocks(struct statfs *sf, long max_size)
220 {
221 uint64_t count;
222 int shift;
223
224 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
225
226 /*
227 * Attempt to scale the block counts to give a more accurate
228 * overview to userland of the ratio of free space to used
229 * space. To do this, find the largest block count and compute
230 * a divisor that lets it fit into a signed integer <= max_size.
231 */
232 if (sf->f_bavail < 0)
233 count = -sf->f_bavail;
234 else
235 count = sf->f_bavail;
236 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
237 if (count <= max_size)
238 return;
239
240 count >>= flsl(max_size);
241 shift = 0;
242 while (count > 0) {
243 shift++;
244 count >>=1;
245 }
246
247 sf->f_bsize <<= shift;
248 sf->f_blocks >>= shift;
249 sf->f_bfree >>= shift;
250 sf->f_bavail >>= shift;
251 }
252
253 /*
254 * Get filesystem statistics.
255 */
256 #ifndef _SYS_SYSPROTO_H_
257 struct statfs_args {
258 char *path;
259 struct statfs *buf;
260 };
261 #endif
262 int
263 statfs(td, uap)
264 struct thread *td;
265 register struct statfs_args /* {
266 char *path;
267 struct statfs *buf;
268 } */ *uap;
269 {
270 struct statfs sf;
271 int error;
272
273 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
274 if (error == 0)
275 error = copyout(&sf, uap->buf, sizeof(sf));
276 return (error);
277 }
278
279 int
280 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
281 struct statfs *buf)
282 {
283 struct mount *mp;
284 struct statfs *sp, sb;
285 int vfslocked;
286 int error;
287 struct nameidata nd;
288
289 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
290 AUDITVNODE1, pathseg, path, td);
291 error = namei(&nd);
292 if (error)
293 return (error);
294 vfslocked = NDHASGIANT(&nd);
295 mp = nd.ni_vp->v_mount;
296 vfs_ref(mp);
297 NDFREE(&nd, NDF_ONLY_PNBUF);
298 vput(nd.ni_vp);
299 error = vfs_busy(mp, 0);
300 vfs_rel(mp);
301 if (error) {
302 VFS_UNLOCK_GIANT(vfslocked);
303 return (error);
304 }
305 #ifdef MAC
306 error = mac_mount_check_stat(td->td_ucred, mp);
307 if (error)
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)
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 VFS_UNLOCK_GIANT(vfslocked);
330 return (error);
331 }
332
333 /*
334 * Get filesystem statistics.
335 */
336 #ifndef _SYS_SYSPROTO_H_
337 struct fstatfs_args {
338 int fd;
339 struct statfs *buf;
340 };
341 #endif
342 int
343 fstatfs(td, uap)
344 struct thread *td;
345 register struct fstatfs_args /* {
346 int fd;
347 struct statfs *buf;
348 } */ *uap;
349 {
350 struct statfs sf;
351 int error;
352
353 error = kern_fstatfs(td, uap->fd, &sf);
354 if (error == 0)
355 error = copyout(&sf, uap->buf, sizeof(sf));
356 return (error);
357 }
358
359 int
360 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
361 {
362 struct file *fp;
363 struct mount *mp;
364 struct statfs *sp, sb;
365 int vfslocked;
366 struct vnode *vp;
367 int error;
368
369 AUDIT_ARG_FD(fd);
370 error = getvnode(td->td_proc->p_fd, fd, &fp);
371 if (error)
372 return (error);
373 vp = fp->f_vnode;
374 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
375 vn_lock(vp, LK_SHARED | LK_RETRY);
376 #ifdef AUDIT
377 AUDIT_ARG_VNODE1(vp);
378 #endif
379 mp = vp->v_mount;
380 if (mp)
381 vfs_ref(mp);
382 VOP_UNLOCK(vp, 0);
383 fdrop(fp, td);
384 if (mp == NULL) {
385 error = EBADF;
386 goto out;
387 }
388 error = vfs_busy(mp, 0);
389 vfs_rel(mp);
390 if (error) {
391 VFS_UNLOCK_GIANT(vfslocked);
392 return (error);
393 }
394 #ifdef MAC
395 error = mac_mount_check_stat(td->td_ucred, mp);
396 if (error)
397 goto out;
398 #endif
399 /*
400 * Set these in case the underlying filesystem fails to do so.
401 */
402 sp = &mp->mnt_stat;
403 sp->f_version = STATFS_VERSION;
404 sp->f_namemax = NAME_MAX;
405 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
406 error = VFS_STATFS(mp, sp);
407 if (error)
408 goto out;
409 if (priv_check(td, PRIV_VFS_GENERATION)) {
410 bcopy(sp, &sb, sizeof(sb));
411 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
412 prison_enforce_statfs(td->td_ucred, mp, &sb);
413 sp = &sb;
414 }
415 *buf = *sp;
416 out:
417 if (mp)
418 vfs_unbusy(mp);
419 VFS_UNLOCK_GIANT(vfslocked);
420 return (error);
421 }
422
423 /*
424 * Get statistics on all filesystems.
425 */
426 #ifndef _SYS_SYSPROTO_H_
427 struct getfsstat_args {
428 struct statfs *buf;
429 long bufsize;
430 int flags;
431 };
432 #endif
433 int
434 getfsstat(td, uap)
435 struct thread *td;
436 register struct getfsstat_args /* {
437 struct statfs *buf;
438 long bufsize;
439 int flags;
440 } */ *uap;
441 {
442
443 return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
444 uap->flags));
445 }
446
447 /*
448 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
449 * The caller is responsible for freeing memory which will be allocated
450 * in '*buf'.
451 */
452 int
453 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
454 enum uio_seg bufseg, int flags)
455 {
456 struct mount *mp, *nmp;
457 struct statfs *sfsp, *sp, sb;
458 size_t count, maxcount;
459 int vfslocked;
460 int error;
461
462 maxcount = bufsize / sizeof(struct statfs);
463 if (bufsize == 0)
464 sfsp = NULL;
465 else if (bufseg == UIO_USERSPACE)
466 sfsp = *buf;
467 else /* if (bufseg == UIO_SYSSPACE) */ {
468 count = 0;
469 mtx_lock(&mountlist_mtx);
470 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
471 count++;
472 }
473 mtx_unlock(&mountlist_mtx);
474 if (maxcount > count)
475 maxcount = count;
476 sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
477 M_WAITOK);
478 }
479 count = 0;
480 mtx_lock(&mountlist_mtx);
481 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
482 if (prison_canseemount(td->td_ucred, mp) != 0) {
483 nmp = TAILQ_NEXT(mp, mnt_list);
484 continue;
485 }
486 #ifdef MAC
487 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
488 nmp = TAILQ_NEXT(mp, mnt_list);
489 continue;
490 }
491 #endif
492 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
493 nmp = TAILQ_NEXT(mp, mnt_list);
494 continue;
495 }
496 vfslocked = VFS_LOCK_GIANT(mp);
497 if (sfsp && count < maxcount) {
498 sp = &mp->mnt_stat;
499 /*
500 * Set these in case the underlying filesystem
501 * fails to do so.
502 */
503 sp->f_version = STATFS_VERSION;
504 sp->f_namemax = NAME_MAX;
505 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
506 /*
507 * If MNT_NOWAIT or MNT_LAZY is specified, do not
508 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
509 * overrides MNT_WAIT.
510 */
511 if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
512 (flags & MNT_WAIT)) &&
513 (error = VFS_STATFS(mp, sp))) {
514 VFS_UNLOCK_GIANT(vfslocked);
515 mtx_lock(&mountlist_mtx);
516 nmp = TAILQ_NEXT(mp, mnt_list);
517 vfs_unbusy(mp);
518 continue;
519 }
520 if (priv_check(td, PRIV_VFS_GENERATION)) {
521 bcopy(sp, &sb, sizeof(sb));
522 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
523 prison_enforce_statfs(td->td_ucred, mp, &sb);
524 sp = &sb;
525 }
526 if (bufseg == UIO_SYSSPACE)
527 bcopy(sp, sfsp, sizeof(*sp));
528 else /* if (bufseg == UIO_USERSPACE) */ {
529 error = copyout(sp, sfsp, sizeof(*sp));
530 if (error) {
531 vfs_unbusy(mp);
532 VFS_UNLOCK_GIANT(vfslocked);
533 return (error);
534 }
535 }
536 sfsp++;
537 }
538 VFS_UNLOCK_GIANT(vfslocked);
539 count++;
540 mtx_lock(&mountlist_mtx);
541 nmp = TAILQ_NEXT(mp, mnt_list);
542 vfs_unbusy(mp);
543 }
544 mtx_unlock(&mountlist_mtx);
545 if (sfsp && count > maxcount)
546 td->td_retval[0] = maxcount;
547 else
548 td->td_retval[0] = count;
549 return (0);
550 }
551
552 #ifdef COMPAT_FREEBSD4
553 /*
554 * Get old format filesystem statistics.
555 */
556 static void cvtstatfs(struct statfs *, struct ostatfs *);
557
558 #ifndef _SYS_SYSPROTO_H_
559 struct freebsd4_statfs_args {
560 char *path;
561 struct ostatfs *buf;
562 };
563 #endif
564 int
565 freebsd4_statfs(td, uap)
566 struct thread *td;
567 struct freebsd4_statfs_args /* {
568 char *path;
569 struct ostatfs *buf;
570 } */ *uap;
571 {
572 struct ostatfs osb;
573 struct statfs sf;
574 int error;
575
576 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
577 if (error)
578 return (error);
579 cvtstatfs(&sf, &osb);
580 return (copyout(&osb, uap->buf, sizeof(osb)));
581 }
582
583 /*
584 * Get filesystem statistics.
585 */
586 #ifndef _SYS_SYSPROTO_H_
587 struct freebsd4_fstatfs_args {
588 int fd;
589 struct ostatfs *buf;
590 };
591 #endif
592 int
593 freebsd4_fstatfs(td, uap)
594 struct thread *td;
595 struct freebsd4_fstatfs_args /* {
596 int fd;
597 struct ostatfs *buf;
598 } */ *uap;
599 {
600 struct ostatfs osb;
601 struct statfs sf;
602 int error;
603
604 error = kern_fstatfs(td, uap->fd, &sf);
605 if (error)
606 return (error);
607 cvtstatfs(&sf, &osb);
608 return (copyout(&osb, uap->buf, sizeof(osb)));
609 }
610
611 /*
612 * Get statistics on all filesystems.
613 */
614 #ifndef _SYS_SYSPROTO_H_
615 struct freebsd4_getfsstat_args {
616 struct ostatfs *buf;
617 long bufsize;
618 int flags;
619 };
620 #endif
621 int
622 freebsd4_getfsstat(td, uap)
623 struct thread *td;
624 register struct freebsd4_getfsstat_args /* {
625 struct ostatfs *buf;
626 long bufsize;
627 int flags;
628 } */ *uap;
629 {
630 struct statfs *buf, *sp;
631 struct ostatfs osb;
632 size_t count, size;
633 int error;
634
635 count = uap->bufsize / sizeof(struct ostatfs);
636 size = count * sizeof(struct statfs);
637 error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
638 if (size > 0) {
639 count = td->td_retval[0];
640 sp = buf;
641 while (count > 0 && error == 0) {
642 cvtstatfs(sp, &osb);
643 error = copyout(&osb, uap->buf, sizeof(osb));
644 sp++;
645 uap->buf++;
646 count--;
647 }
648 free(buf, M_TEMP);
649 }
650 return (error);
651 }
652
653 /*
654 * Implement fstatfs() for (NFS) file handles.
655 */
656 #ifndef _SYS_SYSPROTO_H_
657 struct freebsd4_fhstatfs_args {
658 struct fhandle *u_fhp;
659 struct ostatfs *buf;
660 };
661 #endif
662 int
663 freebsd4_fhstatfs(td, uap)
664 struct thread *td;
665 struct freebsd4_fhstatfs_args /* {
666 struct fhandle *u_fhp;
667 struct ostatfs *buf;
668 } */ *uap;
669 {
670 struct ostatfs osb;
671 struct statfs sf;
672 fhandle_t fh;
673 int error;
674
675 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
676 if (error)
677 return (error);
678 error = kern_fhstatfs(td, fh, &sf);
679 if (error)
680 return (error);
681 cvtstatfs(&sf, &osb);
682 return (copyout(&osb, uap->buf, sizeof(osb)));
683 }
684
685 /*
686 * Convert a new format statfs structure to an old format statfs structure.
687 */
688 static void
689 cvtstatfs(nsp, osp)
690 struct statfs *nsp;
691 struct ostatfs *osp;
692 {
693
694 statfs_scale_blocks(nsp, LONG_MAX);
695 bzero(osp, sizeof(*osp));
696 osp->f_bsize = nsp->f_bsize;
697 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
698 osp->f_blocks = nsp->f_blocks;
699 osp->f_bfree = nsp->f_bfree;
700 osp->f_bavail = nsp->f_bavail;
701 osp->f_files = MIN(nsp->f_files, LONG_MAX);
702 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
703 osp->f_owner = nsp->f_owner;
704 osp->f_type = nsp->f_type;
705 osp->f_flags = nsp->f_flags;
706 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
707 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
708 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
709 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
710 strlcpy(osp->f_fstypename, nsp->f_fstypename,
711 MIN(MFSNAMELEN, OMFSNAMELEN));
712 strlcpy(osp->f_mntonname, nsp->f_mntonname,
713 MIN(MNAMELEN, OMNAMELEN));
714 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
715 MIN(MNAMELEN, OMNAMELEN));
716 osp->f_fsid = nsp->f_fsid;
717 }
718 #endif /* COMPAT_FREEBSD4 */
719
720 /*
721 * Change current working directory to a given file descriptor.
722 */
723 #ifndef _SYS_SYSPROTO_H_
724 struct fchdir_args {
725 int fd;
726 };
727 #endif
728 int
729 fchdir(td, uap)
730 struct thread *td;
731 struct fchdir_args /* {
732 int fd;
733 } */ *uap;
734 {
735 register struct filedesc *fdp = td->td_proc->p_fd;
736 struct vnode *vp, *tdp, *vpold;
737 struct mount *mp;
738 struct file *fp;
739 int vfslocked;
740 int error;
741
742 AUDIT_ARG_FD(uap->fd);
743 if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
744 return (error);
745 vp = fp->f_vnode;
746 VREF(vp);
747 fdrop(fp, td);
748 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
749 vn_lock(vp, LK_SHARED | LK_RETRY);
750 AUDIT_ARG_VNODE1(vp);
751 error = change_dir(vp, td);
752 while (!error && (mp = vp->v_mountedhere) != NULL) {
753 int tvfslocked;
754 if (vfs_busy(mp, 0))
755 continue;
756 tvfslocked = VFS_LOCK_GIANT(mp);
757 error = VFS_ROOT(mp, LK_SHARED, &tdp);
758 vfs_unbusy(mp);
759 if (error) {
760 VFS_UNLOCK_GIANT(tvfslocked);
761 break;
762 }
763 vput(vp);
764 VFS_UNLOCK_GIANT(vfslocked);
765 vp = tdp;
766 vfslocked = tvfslocked;
767 }
768 if (error) {
769 vput(vp);
770 VFS_UNLOCK_GIANT(vfslocked);
771 return (error);
772 }
773 VOP_UNLOCK(vp, 0);
774 VFS_UNLOCK_GIANT(vfslocked);
775 FILEDESC_XLOCK(fdp);
776 vpold = fdp->fd_cdir;
777 fdp->fd_cdir = vp;
778 FILEDESC_XUNLOCK(fdp);
779 vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
780 vrele(vpold);
781 VFS_UNLOCK_GIANT(vfslocked);
782 return (0);
783 }
784
785 /*
786 * Change current working directory (``.'').
787 */
788 #ifndef _SYS_SYSPROTO_H_
789 struct chdir_args {
790 char *path;
791 };
792 #endif
793 int
794 chdir(td, uap)
795 struct thread *td;
796 struct chdir_args /* {
797 char *path;
798 } */ *uap;
799 {
800
801 return (kern_chdir(td, uap->path, UIO_USERSPACE));
802 }
803
804 int
805 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
806 {
807 register struct filedesc *fdp = td->td_proc->p_fd;
808 int error;
809 struct nameidata nd;
810 struct vnode *vp;
811 int vfslocked;
812
813 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
814 MPSAFE, pathseg, path, td);
815 if ((error = namei(&nd)) != 0)
816 return (error);
817 vfslocked = NDHASGIANT(&nd);
818 if ((error = change_dir(nd.ni_vp, td)) != 0) {
819 vput(nd.ni_vp);
820 VFS_UNLOCK_GIANT(vfslocked);
821 NDFREE(&nd, NDF_ONLY_PNBUF);
822 return (error);
823 }
824 VOP_UNLOCK(nd.ni_vp, 0);
825 VFS_UNLOCK_GIANT(vfslocked);
826 NDFREE(&nd, NDF_ONLY_PNBUF);
827 FILEDESC_XLOCK(fdp);
828 vp = fdp->fd_cdir;
829 fdp->fd_cdir = nd.ni_vp;
830 FILEDESC_XUNLOCK(fdp);
831 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
832 vrele(vp);
833 VFS_UNLOCK_GIANT(vfslocked);
834 return (0);
835 }
836
837 /*
838 * Helper function for raised chroot(2) security function: Refuse if
839 * any filedescriptors are open directories.
840 */
841 static int
842 chroot_refuse_vdir_fds(fdp)
843 struct filedesc *fdp;
844 {
845 struct vnode *vp;
846 struct file *fp;
847 int fd;
848
849 FILEDESC_LOCK_ASSERT(fdp);
850
851 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
852 fp = fget_locked(fdp, fd);
853 if (fp == NULL)
854 continue;
855 if (fp->f_type == DTYPE_VNODE) {
856 vp = fp->f_vnode;
857 if (vp->v_type == VDIR)
858 return (EPERM);
859 }
860 }
861 return (0);
862 }
863
864 /*
865 * This sysctl determines if we will allow a process to chroot(2) if it
866 * has a directory open:
867 * 0: disallowed for all processes.
868 * 1: allowed for processes that were not already chroot(2)'ed.
869 * 2: allowed for all processes.
870 */
871
872 static int chroot_allow_open_directories = 1;
873
874 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
875 &chroot_allow_open_directories, 0, "");
876
877 /*
878 * Change notion of root (``/'') directory.
879 */
880 #ifndef _SYS_SYSPROTO_H_
881 struct chroot_args {
882 char *path;
883 };
884 #endif
885 int
886 chroot(td, uap)
887 struct thread *td;
888 struct chroot_args /* {
889 char *path;
890 } */ *uap;
891 {
892 int error;
893 struct nameidata nd;
894 int vfslocked;
895
896 error = priv_check(td, PRIV_VFS_CHROOT);
897 if (error)
898 return (error);
899 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
900 AUDITVNODE1, UIO_USERSPACE, uap->path, td);
901 error = namei(&nd);
902 if (error)
903 goto error;
904 vfslocked = NDHASGIANT(&nd);
905 if ((error = change_dir(nd.ni_vp, td)) != 0)
906 goto e_vunlock;
907 #ifdef MAC
908 if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
909 goto e_vunlock;
910 #endif
911 VOP_UNLOCK(nd.ni_vp, 0);
912 error = change_root(nd.ni_vp, td);
913 vrele(nd.ni_vp);
914 VFS_UNLOCK_GIANT(vfslocked);
915 NDFREE(&nd, NDF_ONLY_PNBUF);
916 return (error);
917 e_vunlock:
918 vput(nd.ni_vp);
919 VFS_UNLOCK_GIANT(vfslocked);
920 error:
921 NDFREE(&nd, NDF_ONLY_PNBUF);
922 return (error);
923 }
924
925 /*
926 * Common routine for chroot and chdir. Callers must provide a locked vnode
927 * instance.
928 */
929 int
930 change_dir(vp, td)
931 struct vnode *vp;
932 struct thread *td;
933 {
934 int error;
935
936 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
937 if (vp->v_type != VDIR)
938 return (ENOTDIR);
939 #ifdef MAC
940 error = mac_vnode_check_chdir(td->td_ucred, vp);
941 if (error)
942 return (error);
943 #endif
944 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
945 return (error);
946 }
947
948 /*
949 * Common routine for kern_chroot() and jail_attach(). The caller is
950 * responsible for invoking priv_check() and mac_vnode_check_chroot() to
951 * authorize this operation.
952 */
953 int
954 change_root(vp, td)
955 struct vnode *vp;
956 struct thread *td;
957 {
958 struct filedesc *fdp;
959 struct vnode *oldvp;
960 int vfslocked;
961 int error;
962
963 VFS_ASSERT_GIANT(vp->v_mount);
964 fdp = td->td_proc->p_fd;
965 FILEDESC_XLOCK(fdp);
966 if (chroot_allow_open_directories == 0 ||
967 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
968 error = chroot_refuse_vdir_fds(fdp);
969 if (error) {
970 FILEDESC_XUNLOCK(fdp);
971 return (error);
972 }
973 }
974 oldvp = fdp->fd_rdir;
975 fdp->fd_rdir = vp;
976 VREF(fdp->fd_rdir);
977 if (!fdp->fd_jdir) {
978 fdp->fd_jdir = vp;
979 VREF(fdp->fd_jdir);
980 }
981 FILEDESC_XUNLOCK(fdp);
982 vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
983 vrele(oldvp);
984 VFS_UNLOCK_GIANT(vfslocked);
985 return (0);
986 }
987
988 /*
989 * Check permissions, allocate an open file structure, and call the device
990 * open routine if any.
991 */
992 #ifndef _SYS_SYSPROTO_H_
993 struct open_args {
994 char *path;
995 int flags;
996 int mode;
997 };
998 #endif
999 int
1000 open(td, uap)
1001 struct thread *td;
1002 register struct open_args /* {
1003 char *path;
1004 int flags;
1005 int mode;
1006 } */ *uap;
1007 {
1008
1009 return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1010 }
1011
1012 #ifndef _SYS_SYSPROTO_H_
1013 struct openat_args {
1014 int fd;
1015 char *path;
1016 int flag;
1017 int mode;
1018 };
1019 #endif
1020 int
1021 openat(struct thread *td, struct openat_args *uap)
1022 {
1023
1024 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1025 uap->mode));
1026 }
1027
1028 int
1029 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1030 int mode)
1031 {
1032
1033 return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1034 }
1035
1036 int
1037 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1038 int flags, int mode)
1039 {
1040 struct proc *p = td->td_proc;
1041 struct filedesc *fdp = p->p_fd;
1042 struct file *fp;
1043 struct vnode *vp;
1044 int cmode;
1045 struct file *nfp;
1046 int type, indx, error;
1047 struct flock lf;
1048 struct nameidata nd;
1049 int vfslocked;
1050
1051 AUDIT_ARG_FFLAGS(flags);
1052 AUDIT_ARG_MODE(mode);
1053 /* XXX: audit dirfd */
1054 /*
1055 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR may
1056 * be specified.
1057 */
1058 if (flags & O_EXEC) {
1059 if (flags & O_ACCMODE)
1060 return (EINVAL);
1061 } else if ((flags & O_ACCMODE) == O_ACCMODE)
1062 return (EINVAL);
1063 else
1064 flags = FFLAGS(flags);
1065
1066 error = fallocf(td, &nfp, &indx, flags);
1067 if (error)
1068 return (error);
1069 /* An extra reference on `nfp' has been held for us by falloc(). */
1070 fp = nfp;
1071 /* Set the flags early so the finit in devfs can pick them up. */
1072 fp->f_flag = flags & FMASK;
1073 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1074 NDINIT_AT(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg, path, fd,
1075 td);
1076 td->td_dupfd = -1; /* XXX check for fdopen */
1077 error = vn_open(&nd, &flags, cmode, fp);
1078 if (error) {
1079 /*
1080 * If the vn_open replaced the method vector, something
1081 * wonderous happened deep below and we just pass it up
1082 * pretending we know what we do.
1083 */
1084 if (error == ENXIO && fp->f_ops != &badfileops) {
1085 fdrop(fp, td);
1086 td->td_retval[0] = indx;
1087 return (0);
1088 }
1089
1090 /*
1091 * handle special fdopen() case. bleh. dupfdopen() is
1092 * responsible for dropping the old contents of ofiles[indx]
1093 * if it succeeds.
1094 */
1095 if ((error == ENODEV || error == ENXIO) &&
1096 td->td_dupfd >= 0 && /* XXX from fdopen */
1097 (error =
1098 dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
1099 td->td_retval[0] = indx;
1100 fdrop(fp, td);
1101 return (0);
1102 }
1103 /*
1104 * Clean up the descriptor, but only if another thread hadn't
1105 * replaced or closed it.
1106 */
1107 fdclose(fdp, fp, indx, td);
1108 fdrop(fp, td);
1109
1110 return (error);
1111 }
1112 td->td_dupfd = 0;
1113 vfslocked = NDHASGIANT(&nd);
1114 NDFREE(&nd, NDF_ONLY_PNBUF);
1115 vp = nd.ni_vp;
1116
1117 /*
1118 * Store the vnode, for any f_type. Typically, the vnode use
1119 * count is decremented by direct call to vn_closefile() for
1120 * files that switched type in the cdevsw fdopen() method.
1121 */
1122 fp->f_vnode = vp;
1123 /*
1124 * If the file wasn't claimed by devfs bind it to the normal
1125 * vnode operations here.
1126 */
1127 if (fp->f_ops == &badfileops) {
1128 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1129 fp->f_seqcount = 1;
1130 finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1131 }
1132
1133 VOP_UNLOCK(vp, 0);
1134 if (fp->f_type == DTYPE_VNODE && (flags & (O_EXLOCK | O_SHLOCK)) != 0) {
1135 lf.l_whence = SEEK_SET;
1136 lf.l_start = 0;
1137 lf.l_len = 0;
1138 if (flags & O_EXLOCK)
1139 lf.l_type = F_WRLCK;
1140 else
1141 lf.l_type = F_RDLCK;
1142 type = F_FLOCK;
1143 if ((flags & FNONBLOCK) == 0)
1144 type |= F_WAIT;
1145 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1146 type)) != 0)
1147 goto bad;
1148 atomic_set_int(&fp->f_flag, FHASLOCK);
1149 }
1150 if (flags & O_TRUNC) {
1151 error = fo_truncate(fp, 0, td->td_ucred, td);
1152 if (error)
1153 goto bad;
1154 }
1155 VFS_UNLOCK_GIANT(vfslocked);
1156 /*
1157 * Release our private reference, leaving the one associated with
1158 * the descriptor table intact.
1159 */
1160 fdrop(fp, td);
1161 td->td_retval[0] = indx;
1162 return (0);
1163 bad:
1164 VFS_UNLOCK_GIANT(vfslocked);
1165 fdclose(fdp, fp, indx, td);
1166 fdrop(fp, td);
1167 return (error);
1168 }
1169
1170 #ifdef COMPAT_43
1171 /*
1172 * Create a file.
1173 */
1174 #ifndef _SYS_SYSPROTO_H_
1175 struct ocreat_args {
1176 char *path;
1177 int mode;
1178 };
1179 #endif
1180 int
1181 ocreat(td, uap)
1182 struct thread *td;
1183 register struct ocreat_args /* {
1184 char *path;
1185 int mode;
1186 } */ *uap;
1187 {
1188
1189 return (kern_open(td, uap->path, UIO_USERSPACE,
1190 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1191 }
1192 #endif /* COMPAT_43 */
1193
1194 /*
1195 * Create a special file.
1196 */
1197 #ifndef _SYS_SYSPROTO_H_
1198 struct mknod_args {
1199 char *path;
1200 int mode;
1201 int dev;
1202 };
1203 #endif
1204 int
1205 mknod(td, uap)
1206 struct thread *td;
1207 register struct mknod_args /* {
1208 char *path;
1209 int mode;
1210 int dev;
1211 } */ *uap;
1212 {
1213
1214 return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1215 }
1216
1217 #ifndef _SYS_SYSPROTO_H_
1218 struct mknodat_args {
1219 int fd;
1220 char *path;
1221 mode_t mode;
1222 dev_t dev;
1223 };
1224 #endif
1225 int
1226 mknodat(struct thread *td, struct mknodat_args *uap)
1227 {
1228
1229 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1230 uap->dev));
1231 }
1232
1233 int
1234 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1235 int dev)
1236 {
1237
1238 return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1239 }
1240
1241 int
1242 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1243 int mode, int dev)
1244 {
1245 struct vnode *vp;
1246 struct mount *mp;
1247 struct vattr vattr;
1248 int error;
1249 int whiteout = 0;
1250 struct nameidata nd;
1251 int vfslocked;
1252
1253 AUDIT_ARG_MODE(mode);
1254 AUDIT_ARG_DEV(dev);
1255 switch (mode & S_IFMT) {
1256 case S_IFCHR:
1257 case S_IFBLK:
1258 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1259 break;
1260 case S_IFMT:
1261 error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1262 break;
1263 case S_IFWHT:
1264 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1265 break;
1266 case S_IFIFO:
1267 if (dev == 0)
1268 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1269 /* FALLTHROUGH */
1270 default:
1271 error = EINVAL;
1272 break;
1273 }
1274 if (error)
1275 return (error);
1276 restart:
1277 bwillwrite();
1278 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1279 pathseg, path, fd, td);
1280 if ((error = namei(&nd)) != 0)
1281 return (error);
1282 vfslocked = NDHASGIANT(&nd);
1283 vp = nd.ni_vp;
1284 if (vp != NULL) {
1285 NDFREE(&nd, NDF_ONLY_PNBUF);
1286 if (vp == nd.ni_dvp)
1287 vrele(nd.ni_dvp);
1288 else
1289 vput(nd.ni_dvp);
1290 vrele(vp);
1291 VFS_UNLOCK_GIANT(vfslocked);
1292 return (EEXIST);
1293 } else {
1294 VATTR_NULL(&vattr);
1295 vattr.va_mode = (mode & ALLPERMS) &
1296 ~td->td_proc->p_fd->fd_cmask;
1297 vattr.va_rdev = dev;
1298 whiteout = 0;
1299
1300 switch (mode & S_IFMT) {
1301 case S_IFMT: /* used by badsect to flag bad sectors */
1302 vattr.va_type = VBAD;
1303 break;
1304 case S_IFCHR:
1305 vattr.va_type = VCHR;
1306 break;
1307 case S_IFBLK:
1308 vattr.va_type = VBLK;
1309 break;
1310 case S_IFWHT:
1311 whiteout = 1;
1312 break;
1313 default:
1314 panic("kern_mknod: invalid mode");
1315 }
1316 }
1317 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1318 NDFREE(&nd, NDF_ONLY_PNBUF);
1319 vput(nd.ni_dvp);
1320 VFS_UNLOCK_GIANT(vfslocked);
1321 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1322 return (error);
1323 goto restart;
1324 }
1325 #ifdef MAC
1326 if (error == 0 && !whiteout)
1327 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1328 &nd.ni_cnd, &vattr);
1329 #endif
1330 if (!error) {
1331 if (whiteout)
1332 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1333 else {
1334 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1335 &nd.ni_cnd, &vattr);
1336 if (error == 0)
1337 vput(nd.ni_vp);
1338 }
1339 }
1340 NDFREE(&nd, NDF_ONLY_PNBUF);
1341 vput(nd.ni_dvp);
1342 vn_finished_write(mp);
1343 VFS_UNLOCK_GIANT(vfslocked);
1344 return (error);
1345 }
1346
1347 /*
1348 * Create a named pipe.
1349 */
1350 #ifndef _SYS_SYSPROTO_H_
1351 struct mkfifo_args {
1352 char *path;
1353 int mode;
1354 };
1355 #endif
1356 int
1357 mkfifo(td, uap)
1358 struct thread *td;
1359 register struct mkfifo_args /* {
1360 char *path;
1361 int mode;
1362 } */ *uap;
1363 {
1364
1365 return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1366 }
1367
1368 #ifndef _SYS_SYSPROTO_H_
1369 struct mkfifoat_args {
1370 int fd;
1371 char *path;
1372 mode_t mode;
1373 };
1374 #endif
1375 int
1376 mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1377 {
1378
1379 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1380 uap->mode));
1381 }
1382
1383 int
1384 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1385 {
1386
1387 return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1388 }
1389
1390 int
1391 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1392 int mode)
1393 {
1394 struct mount *mp;
1395 struct vattr vattr;
1396 int error;
1397 struct nameidata nd;
1398 int vfslocked;
1399
1400 AUDIT_ARG_MODE(mode);
1401 restart:
1402 bwillwrite();
1403 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1404 pathseg, path, fd, td);
1405 if ((error = namei(&nd)) != 0)
1406 return (error);
1407 vfslocked = NDHASGIANT(&nd);
1408 if (nd.ni_vp != NULL) {
1409 NDFREE(&nd, NDF_ONLY_PNBUF);
1410 if (nd.ni_vp == nd.ni_dvp)
1411 vrele(nd.ni_dvp);
1412 else
1413 vput(nd.ni_dvp);
1414 vrele(nd.ni_vp);
1415 VFS_UNLOCK_GIANT(vfslocked);
1416 return (EEXIST);
1417 }
1418 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1419 NDFREE(&nd, NDF_ONLY_PNBUF);
1420 vput(nd.ni_dvp);
1421 VFS_UNLOCK_GIANT(vfslocked);
1422 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1423 return (error);
1424 goto restart;
1425 }
1426 VATTR_NULL(&vattr);
1427 vattr.va_type = VFIFO;
1428 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1429 #ifdef MAC
1430 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1431 &vattr);
1432 if (error)
1433 goto out;
1434 #endif
1435 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1436 if (error == 0)
1437 vput(nd.ni_vp);
1438 #ifdef MAC
1439 out:
1440 #endif
1441 vput(nd.ni_dvp);
1442 vn_finished_write(mp);
1443 VFS_UNLOCK_GIANT(vfslocked);
1444 NDFREE(&nd, NDF_ONLY_PNBUF);
1445 return (error);
1446 }
1447
1448 /*
1449 * Make a hard file link.
1450 */
1451 #ifndef _SYS_SYSPROTO_H_
1452 struct link_args {
1453 char *path;
1454 char *link;
1455 };
1456 #endif
1457 int
1458 link(td, uap)
1459 struct thread *td;
1460 register struct link_args /* {
1461 char *path;
1462 char *link;
1463 } */ *uap;
1464 {
1465
1466 return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1467 }
1468
1469 #ifndef _SYS_SYSPROTO_H_
1470 struct linkat_args {
1471 int fd1;
1472 char *path1;
1473 int fd2;
1474 char *path2;
1475 int flag;
1476 };
1477 #endif
1478 int
1479 linkat(struct thread *td, struct linkat_args *uap)
1480 {
1481 int flag;
1482
1483 flag = uap->flag;
1484 if (flag & ~AT_SYMLINK_FOLLOW)
1485 return (EINVAL);
1486
1487 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1488 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1489 }
1490
1491 int hardlink_check_uid = 0;
1492 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1493 &hardlink_check_uid, 0,
1494 "Unprivileged processes cannot create hard links to files owned by other "
1495 "users");
1496 static int hardlink_check_gid = 0;
1497 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1498 &hardlink_check_gid, 0,
1499 "Unprivileged processes cannot create hard links to files owned by other "
1500 "groups");
1501
1502 static int
1503 can_hardlink(struct vnode *vp, struct ucred *cred)
1504 {
1505 struct vattr va;
1506 int error;
1507
1508 if (!hardlink_check_uid && !hardlink_check_gid)
1509 return (0);
1510
1511 error = VOP_GETATTR(vp, &va, cred);
1512 if (error != 0)
1513 return (error);
1514
1515 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1516 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1517 if (error)
1518 return (error);
1519 }
1520
1521 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1522 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1523 if (error)
1524 return (error);
1525 }
1526
1527 return (0);
1528 }
1529
1530 int
1531 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1532 {
1533
1534 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1535 }
1536
1537 int
1538 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1539 enum uio_seg segflg, int follow)
1540 {
1541 struct vnode *vp;
1542 struct mount *mp;
1543 struct nameidata nd;
1544 int vfslocked;
1545 int lvfslocked;
1546 int error;
1547
1548 bwillwrite();
1549 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1550 fd1, td);
1551
1552 if ((error = namei(&nd)) != 0)
1553 return (error);
1554 vfslocked = NDHASGIANT(&nd);
1555 NDFREE(&nd, NDF_ONLY_PNBUF);
1556 vp = nd.ni_vp;
1557 if (vp->v_type == VDIR) {
1558 vrele(vp);
1559 VFS_UNLOCK_GIANT(vfslocked);
1560 return (EPERM); /* POSIX */
1561 }
1562 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1563 vrele(vp);
1564 VFS_UNLOCK_GIANT(vfslocked);
1565 return (error);
1566 }
1567 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2,
1568 segflg, path2, fd2, td);
1569 if ((error = namei(&nd)) == 0) {
1570 lvfslocked = NDHASGIANT(&nd);
1571 if (nd.ni_vp != NULL) {
1572 if (nd.ni_dvp == nd.ni_vp)
1573 vrele(nd.ni_dvp);
1574 else
1575 vput(nd.ni_dvp);
1576 vrele(nd.ni_vp);
1577 error = EEXIST;
1578 } else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1579 == 0) {
1580 error = can_hardlink(vp, td->td_ucred);
1581 if (error == 0)
1582 #ifdef MAC
1583 error = mac_vnode_check_link(td->td_ucred,
1584 nd.ni_dvp, vp, &nd.ni_cnd);
1585 if (error == 0)
1586 #endif
1587 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1588 VOP_UNLOCK(vp, 0);
1589 vput(nd.ni_dvp);
1590 }
1591 NDFREE(&nd, NDF_ONLY_PNBUF);
1592 VFS_UNLOCK_GIANT(lvfslocked);
1593 }
1594 vrele(vp);
1595 vn_finished_write(mp);
1596 VFS_UNLOCK_GIANT(vfslocked);
1597 return (error);
1598 }
1599
1600 /*
1601 * Make a symbolic link.
1602 */
1603 #ifndef _SYS_SYSPROTO_H_
1604 struct symlink_args {
1605 char *path;
1606 char *link;
1607 };
1608 #endif
1609 int
1610 symlink(td, uap)
1611 struct thread *td;
1612 register struct symlink_args /* {
1613 char *path;
1614 char *link;
1615 } */ *uap;
1616 {
1617
1618 return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1619 }
1620
1621 #ifndef _SYS_SYSPROTO_H_
1622 struct symlinkat_args {
1623 char *path;
1624 int fd;
1625 char *path2;
1626 };
1627 #endif
1628 int
1629 symlinkat(struct thread *td, struct symlinkat_args *uap)
1630 {
1631
1632 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1633 UIO_USERSPACE));
1634 }
1635
1636 int
1637 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1638 {
1639
1640 return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1641 }
1642
1643 int
1644 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1645 enum uio_seg segflg)
1646 {
1647 struct mount *mp;
1648 struct vattr vattr;
1649 char *syspath;
1650 int error;
1651 struct nameidata nd;
1652 int vfslocked;
1653
1654 if (segflg == UIO_SYSSPACE) {
1655 syspath = path1;
1656 } else {
1657 syspath = uma_zalloc(namei_zone, M_WAITOK);
1658 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1659 goto out;
1660 }
1661 AUDIT_ARG_TEXT(syspath);
1662 restart:
1663 bwillwrite();
1664 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1665 segflg, path2, fd, td);
1666 if ((error = namei(&nd)) != 0)
1667 goto out;
1668 vfslocked = NDHASGIANT(&nd);
1669 if (nd.ni_vp) {
1670 NDFREE(&nd, NDF_ONLY_PNBUF);
1671 if (nd.ni_vp == nd.ni_dvp)
1672 vrele(nd.ni_dvp);
1673 else
1674 vput(nd.ni_dvp);
1675 vrele(nd.ni_vp);
1676 VFS_UNLOCK_GIANT(vfslocked);
1677 error = EEXIST;
1678 goto out;
1679 }
1680 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1681 NDFREE(&nd, NDF_ONLY_PNBUF);
1682 vput(nd.ni_dvp);
1683 VFS_UNLOCK_GIANT(vfslocked);
1684 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1685 goto out;
1686 goto restart;
1687 }
1688 VATTR_NULL(&vattr);
1689 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1690 #ifdef MAC
1691 vattr.va_type = VLNK;
1692 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1693 &vattr);
1694 if (error)
1695 goto out2;
1696 #endif
1697 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1698 if (error == 0)
1699 vput(nd.ni_vp);
1700 #ifdef MAC
1701 out2:
1702 #endif
1703 NDFREE(&nd, NDF_ONLY_PNBUF);
1704 vput(nd.ni_dvp);
1705 vn_finished_write(mp);
1706 VFS_UNLOCK_GIANT(vfslocked);
1707 out:
1708 if (segflg != UIO_SYSSPACE)
1709 uma_zfree(namei_zone, syspath);
1710 return (error);
1711 }
1712
1713 /*
1714 * Delete a whiteout from the filesystem.
1715 */
1716 int
1717 undelete(td, uap)
1718 struct thread *td;
1719 register struct undelete_args /* {
1720 char *path;
1721 } */ *uap;
1722 {
1723 int error;
1724 struct mount *mp;
1725 struct nameidata nd;
1726 int vfslocked;
1727
1728 restart:
1729 bwillwrite();
1730 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1731 UIO_USERSPACE, uap->path, td);
1732 error = namei(&nd);
1733 if (error)
1734 return (error);
1735 vfslocked = NDHASGIANT(&nd);
1736
1737 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1738 NDFREE(&nd, NDF_ONLY_PNBUF);
1739 if (nd.ni_vp == nd.ni_dvp)
1740 vrele(nd.ni_dvp);
1741 else
1742 vput(nd.ni_dvp);
1743 if (nd.ni_vp)
1744 vrele(nd.ni_vp);
1745 VFS_UNLOCK_GIANT(vfslocked);
1746 return (EEXIST);
1747 }
1748 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1749 NDFREE(&nd, NDF_ONLY_PNBUF);
1750 vput(nd.ni_dvp);
1751 VFS_UNLOCK_GIANT(vfslocked);
1752 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1753 return (error);
1754 goto restart;
1755 }
1756 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1757 NDFREE(&nd, NDF_ONLY_PNBUF);
1758 vput(nd.ni_dvp);
1759 vn_finished_write(mp);
1760 VFS_UNLOCK_GIANT(vfslocked);
1761 return (error);
1762 }
1763
1764 /*
1765 * Delete a name from the filesystem.
1766 */
1767 #ifndef _SYS_SYSPROTO_H_
1768 struct unlink_args {
1769 char *path;
1770 };
1771 #endif
1772 int
1773 unlink(td, uap)
1774 struct thread *td;
1775 struct unlink_args /* {
1776 char *path;
1777 } */ *uap;
1778 {
1779
1780 return (kern_unlink(td, uap->path, UIO_USERSPACE));
1781 }
1782
1783 #ifndef _SYS_SYSPROTO_H_
1784 struct unlinkat_args {
1785 int fd;
1786 char *path;
1787 int flag;
1788 };
1789 #endif
1790 int
1791 unlinkat(struct thread *td, struct unlinkat_args *uap)
1792 {
1793 int flag = uap->flag;
1794 int fd = uap->fd;
1795 char *path = uap->path;
1796
1797 if (flag & ~AT_REMOVEDIR)
1798 return (EINVAL);
1799
1800 if (flag & AT_REMOVEDIR)
1801 return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1802 else
1803 return (kern_unlinkat(td, fd, path, UIO_USERSPACE));
1804 }
1805
1806 int
1807 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1808 {
1809
1810 return (kern_unlinkat(td, AT_FDCWD, path, pathseg));
1811 }
1812
1813 int
1814 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
1815 {
1816 struct mount *mp;
1817 struct vnode *vp;
1818 int error;
1819 struct nameidata nd;
1820 int vfslocked;
1821
1822 restart:
1823 bwillwrite();
1824 NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1825 pathseg, path, fd, td);
1826 if ((error = namei(&nd)) != 0)
1827 return (error == EINVAL ? EPERM : error);
1828 vfslocked = NDHASGIANT(&nd);
1829 vp = nd.ni_vp;
1830 if (vp->v_type == VDIR)
1831 error = EPERM; /* POSIX */
1832 else {
1833 /*
1834 * The root of a mounted filesystem cannot be deleted.
1835 *
1836 * XXX: can this only be a VDIR case?
1837 */
1838 if (vp->v_vflag & VV_ROOT)
1839 error = EBUSY;
1840 }
1841 if (error == 0) {
1842 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1843 NDFREE(&nd, NDF_ONLY_PNBUF);
1844 vput(nd.ni_dvp);
1845 if (vp == nd.ni_dvp)
1846 vrele(vp);
1847 else
1848 vput(vp);
1849 VFS_UNLOCK_GIANT(vfslocked);
1850 if ((error = vn_start_write(NULL, &mp,
1851 V_XSLEEP | PCATCH)) != 0)
1852 return (error);
1853 goto restart;
1854 }
1855 #ifdef MAC
1856 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1857 &nd.ni_cnd);
1858 if (error)
1859 goto out;
1860 #endif
1861 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1862 #ifdef MAC
1863 out:
1864 #endif
1865 vn_finished_write(mp);
1866 }
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 VFS_UNLOCK_GIANT(vfslocked);
1874 return (error);
1875 }
1876
1877 /*
1878 * Reposition read/write file offset.
1879 */
1880 #ifndef _SYS_SYSPROTO_H_
1881 struct lseek_args {
1882 int fd;
1883 int pad;
1884 off_t offset;
1885 int whence;
1886 };
1887 #endif
1888 int
1889 lseek(td, uap)
1890 struct thread *td;
1891 register struct lseek_args /* {
1892 int fd;
1893 int pad;
1894 off_t offset;
1895 int whence;
1896 } */ *uap;
1897 {
1898 struct ucred *cred = td->td_ucred;
1899 struct file *fp;
1900 struct vnode *vp;
1901 struct vattr vattr;
1902 off_t offset, size;
1903 int error, noneg;
1904 int vfslocked;
1905
1906 AUDIT_ARG_FD(uap->fd);
1907 if ((error = fget(td, uap->fd, &fp)) != 0)
1908 return (error);
1909 if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1910 fdrop(fp, td);
1911 return (ESPIPE);
1912 }
1913 vp = fp->f_vnode;
1914 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1915 noneg = (vp->v_type != VCHR);
1916 offset = uap->offset;
1917 switch (uap->whence) {
1918 case L_INCR:
1919 if (noneg &&
1920 (fp->f_offset < 0 ||
1921 (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
1922 error = EOVERFLOW;
1923 break;
1924 }
1925 offset += fp->f_offset;
1926 break;
1927 case L_XTND:
1928 vn_lock(vp, LK_SHARED | LK_RETRY);
1929 error = VOP_GETATTR(vp, &vattr, cred);
1930 VOP_UNLOCK(vp, 0);
1931 if (error)
1932 break;
1933
1934 /*
1935 * If the file references a disk device, then fetch
1936 * the media size and use that to determine the ending
1937 * offset.
1938 */
1939 if (vattr.va_size == 0 && vp->v_type == VCHR &&
1940 fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0)
1941 vattr.va_size = size;
1942 if (noneg &&
1943 (vattr.va_size > OFF_MAX ||
1944 (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1945 error = EOVERFLOW;
1946 break;
1947 }
1948 offset += vattr.va_size;
1949 break;
1950 case L_SET:
1951 break;
1952 case SEEK_DATA:
1953 error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1954 break;
1955 case SEEK_HOLE:
1956 error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1957 break;
1958 default:
1959 error = EINVAL;
1960 }
1961 if (error == 0 && noneg && offset < 0)
1962 error = EINVAL;
1963 if (error != 0)
1964 goto drop;
1965 fp->f_offset = offset;
1966 VFS_KNOTE_UNLOCKED(vp, 0);
1967 *(off_t *)(td->td_retval) = fp->f_offset;
1968 drop:
1969 fdrop(fp, td);
1970 VFS_UNLOCK_GIANT(vfslocked);
1971 return (error);
1972 }
1973
1974 #if defined(COMPAT_43)
1975 /*
1976 * Reposition read/write file offset.
1977 */
1978 #ifndef _SYS_SYSPROTO_H_
1979 struct olseek_args {
1980 int fd;
1981 long offset;
1982 int whence;
1983 };
1984 #endif
1985 int
1986 olseek(td, uap)
1987 struct thread *td;
1988 register struct olseek_args /* {
1989 int fd;
1990 long offset;
1991 int whence;
1992 } */ *uap;
1993 {
1994 struct lseek_args /* {
1995 int fd;
1996 int pad;
1997 off_t offset;
1998 int whence;
1999 } */ nuap;
2000
2001 nuap.fd = uap->fd;
2002 nuap.offset = uap->offset;
2003 nuap.whence = uap->whence;
2004 return (lseek(td, &nuap));
2005 }
2006 #endif /* COMPAT_43 */
2007
2008 /* Version with the 'pad' argument */
2009 int
2010 freebsd6_lseek(td, uap)
2011 struct thread *td;
2012 register struct freebsd6_lseek_args *uap;
2013 {
2014 struct lseek_args ouap;
2015
2016 ouap.fd = uap->fd;
2017 ouap.offset = uap->offset;
2018 ouap.whence = uap->whence;
2019 return (lseek(td, &ouap));
2020 }
2021
2022 /*
2023 * Check access permissions using passed credentials.
2024 */
2025 static int
2026 vn_access(vp, user_flags, cred, td)
2027 struct vnode *vp;
2028 int user_flags;
2029 struct ucred *cred;
2030 struct thread *td;
2031 {
2032 int error;
2033 accmode_t accmode;
2034
2035 /* Flags == 0 means only check for existence. */
2036 error = 0;
2037 if (user_flags) {
2038 accmode = 0;
2039 if (user_flags & R_OK)
2040 accmode |= VREAD;
2041 if (user_flags & W_OK)
2042 accmode |= VWRITE;
2043 if (user_flags & X_OK)
2044 accmode |= VEXEC;
2045 #ifdef MAC
2046 error = mac_vnode_check_access(cred, vp, accmode);
2047 if (error)
2048 return (error);
2049 #endif
2050 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2051 error = VOP_ACCESS(vp, accmode, cred, td);
2052 }
2053 return (error);
2054 }
2055
2056 /*
2057 * Check access permissions using "real" credentials.
2058 */
2059 #ifndef _SYS_SYSPROTO_H_
2060 struct access_args {
2061 char *path;
2062 int flags;
2063 };
2064 #endif
2065 int
2066 access(td, uap)
2067 struct thread *td;
2068 register struct access_args /* {
2069 char *path;
2070 int flags;
2071 } */ *uap;
2072 {
2073
2074 return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2075 }
2076
2077 #ifndef _SYS_SYSPROTO_H_
2078 struct faccessat_args {
2079 int dirfd;
2080 char *path;
2081 int mode;
2082 int flag;
2083 }
2084 #endif
2085 int
2086 faccessat(struct thread *td, struct faccessat_args *uap)
2087 {
2088
2089 if (uap->flag & ~AT_EACCESS)
2090 return (EINVAL);
2091 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2092 uap->mode));
2093 }
2094
2095 int
2096 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2097 {
2098
2099 return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2100 }
2101
2102 int
2103 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2104 int flags, int mode)
2105 {
2106 struct ucred *cred, *tmpcred;
2107 struct vnode *vp;
2108 struct nameidata nd;
2109 int vfslocked;
2110 int error;
2111
2112 /*
2113 * Create and modify a temporary credential instead of one that
2114 * is potentially shared.
2115 */
2116 if (!(flags & AT_EACCESS)) {
2117 cred = td->td_ucred;
2118 tmpcred = crdup(cred);
2119 tmpcred->cr_uid = cred->cr_ruid;
2120 tmpcred->cr_groups[0] = cred->cr_rgid;
2121 td->td_ucred = tmpcred;
2122 } else
2123 cred = tmpcred = td->td_ucred;
2124 AUDIT_ARG_VALUE(mode);
2125 NDINIT_AT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2126 AUDITVNODE1, pathseg, path, fd, td);
2127 if ((error = namei(&nd)) != 0)
2128 goto out1;
2129 vfslocked = NDHASGIANT(&nd);
2130 vp = nd.ni_vp;
2131
2132 error = vn_access(vp, mode, tmpcred, td);
2133 NDFREE(&nd, NDF_ONLY_PNBUF);
2134 vput(vp);
2135 VFS_UNLOCK_GIANT(vfslocked);
2136 out1:
2137 if (!(flags & AT_EACCESS)) {
2138 td->td_ucred = cred;
2139 crfree(tmpcred);
2140 }
2141 return (error);
2142 }
2143
2144 /*
2145 * Check access permissions using "effective" credentials.
2146 */
2147 #ifndef _SYS_SYSPROTO_H_
2148 struct eaccess_args {
2149 char *path;
2150 int flags;
2151 };
2152 #endif
2153 int
2154 eaccess(td, uap)
2155 struct thread *td;
2156 register struct eaccess_args /* {
2157 char *path;
2158 int flags;
2159 } */ *uap;
2160 {
2161
2162 return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2163 }
2164
2165 int
2166 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2167 {
2168
2169 return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2170 }
2171
2172 #if defined(COMPAT_43)
2173 /*
2174 * Get file status; this version follows links.
2175 */
2176 #ifndef _SYS_SYSPROTO_H_
2177 struct ostat_args {
2178 char *path;
2179 struct ostat *ub;
2180 };
2181 #endif
2182 int
2183 ostat(td, uap)
2184 struct thread *td;
2185 register struct ostat_args /* {
2186 char *path;
2187 struct ostat *ub;
2188 } */ *uap;
2189 {
2190 struct stat sb;
2191 struct ostat osb;
2192 int error;
2193
2194 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2195 if (error)
2196 return (error);
2197 cvtstat(&sb, &osb);
2198 error = copyout(&osb, uap->ub, sizeof (osb));
2199 return (error);
2200 }
2201
2202 /*
2203 * Get file status; this version does not follow links.
2204 */
2205 #ifndef _SYS_SYSPROTO_H_
2206 struct olstat_args {
2207 char *path;
2208 struct ostat *ub;
2209 };
2210 #endif
2211 int
2212 olstat(td, uap)
2213 struct thread *td;
2214 register struct olstat_args /* {
2215 char *path;
2216 struct ostat *ub;
2217 } */ *uap;
2218 {
2219 struct stat sb;
2220 struct ostat osb;
2221 int error;
2222
2223 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2224 if (error)
2225 return (error);
2226 cvtstat(&sb, &osb);
2227 error = copyout(&osb, uap->ub, sizeof (osb));
2228 return (error);
2229 }
2230
2231 /*
2232 * Convert from an old to a new stat structure.
2233 */
2234 void
2235 cvtstat(st, ost)
2236 struct stat *st;
2237 struct ostat *ost;
2238 {
2239
2240 ost->st_dev = st->st_dev;
2241 ost->st_ino = st->st_ino;
2242 ost->st_mode = st->st_mode;
2243 ost->st_nlink = st->st_nlink;
2244 ost->st_uid = st->st_uid;
2245 ost->st_gid = st->st_gid;
2246 ost->st_rdev = st->st_rdev;
2247 if (st->st_size < (quad_t)1 << 32)
2248 ost->st_size = st->st_size;
2249 else
2250 ost->st_size = -2;
2251 ost->st_atime = st->st_atime;
2252 ost->st_mtime = st->st_mtime;
2253 ost->st_ctime = st->st_ctime;
2254 ost->st_blksize = st->st_blksize;
2255 ost->st_blocks = st->st_blocks;
2256 ost->st_flags = st->st_flags;
2257 ost->st_gen = st->st_gen;
2258 }
2259 #endif /* COMPAT_43 */
2260
2261 /*
2262 * Get file status; this version follows links.
2263 */
2264 #ifndef _SYS_SYSPROTO_H_
2265 struct stat_args {
2266 char *path;
2267 struct stat *ub;
2268 };
2269 #endif
2270 int
2271 stat(td, uap)
2272 struct thread *td;
2273 register struct stat_args /* {
2274 char *path;
2275 struct stat *ub;
2276 } */ *uap;
2277 {
2278 struct stat sb;
2279 int error;
2280
2281 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2282 if (error == 0)
2283 error = copyout(&sb, uap->ub, sizeof (sb));
2284 return (error);
2285 }
2286
2287 #ifndef _SYS_SYSPROTO_H_
2288 struct fstatat_args {
2289 int fd;
2290 char *path;
2291 struct stat *buf;
2292 int flag;
2293 }
2294 #endif
2295 int
2296 fstatat(struct thread *td, struct fstatat_args *uap)
2297 {
2298 struct stat sb;
2299 int error;
2300
2301 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2302 UIO_USERSPACE, &sb);
2303 if (error == 0)
2304 error = copyout(&sb, uap->buf, sizeof (sb));
2305 return (error);
2306 }
2307
2308 int
2309 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2310 {
2311
2312 return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2313 }
2314
2315 int
2316 kern_statat(struct thread *td, int flag, int fd, char *path,
2317 enum uio_seg pathseg, struct stat *sbp)
2318 {
2319
2320 return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2321 }
2322
2323 int
2324 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2325 enum uio_seg pathseg, struct stat *sbp,
2326 void (*hook)(struct vnode *vp, struct stat *sbp))
2327 {
2328 struct nameidata nd;
2329 struct stat sb;
2330 int error, vfslocked;
2331
2332 if (flag & ~AT_SYMLINK_NOFOLLOW)
2333 return (EINVAL);
2334
2335 NDINIT_AT(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2336 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2337 path, fd, td);
2338
2339 if ((error = namei(&nd)) != 0)
2340 return (error);
2341 vfslocked = NDHASGIANT(&nd);
2342 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2343 if (!error) {
2344 SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2345 if (S_ISREG(sb.st_mode))
2346 SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2347 if (__predict_false(hook != NULL))
2348 hook(nd.ni_vp, &sb);
2349 }
2350 NDFREE(&nd, NDF_ONLY_PNBUF);
2351 vput(nd.ni_vp);
2352 VFS_UNLOCK_GIANT(vfslocked);
2353 if (error)
2354 return (error);
2355 *sbp = sb;
2356 #ifdef KTRACE
2357 if (KTRPOINT(td, KTR_STRUCT))
2358 ktrstat(&sb);
2359 #endif
2360 return (0);
2361 }
2362
2363 /*
2364 * Get file status; this version does not follow links.
2365 */
2366 #ifndef _SYS_SYSPROTO_H_
2367 struct lstat_args {
2368 char *path;
2369 struct stat *ub;
2370 };
2371 #endif
2372 int
2373 lstat(td, uap)
2374 struct thread *td;
2375 register struct lstat_args /* {
2376 char *path;
2377 struct stat *ub;
2378 } */ *uap;
2379 {
2380 struct stat sb;
2381 int error;
2382
2383 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2384 if (error == 0)
2385 error = copyout(&sb, uap->ub, sizeof (sb));
2386 return (error);
2387 }
2388
2389 int
2390 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2391 {
2392
2393 return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2394 sbp));
2395 }
2396
2397 /*
2398 * Implementation of the NetBSD [l]stat() functions.
2399 */
2400 void
2401 cvtnstat(sb, nsb)
2402 struct stat *sb;
2403 struct nstat *nsb;
2404 {
2405 bzero(nsb, sizeof *nsb);
2406 nsb->st_dev = sb->st_dev;
2407 nsb->st_ino = sb->st_ino;
2408 nsb->st_mode = sb->st_mode;
2409 nsb->st_nlink = sb->st_nlink;
2410 nsb->st_uid = sb->st_uid;
2411 nsb->st_gid = sb->st_gid;
2412 nsb->st_rdev = sb->st_rdev;
2413 nsb->st_atimespec = sb->st_atimespec;
2414 nsb->st_mtimespec = sb->st_mtimespec;
2415 nsb->st_ctimespec = sb->st_ctimespec;
2416 nsb->st_size = sb->st_size;
2417 nsb->st_blocks = sb->st_blocks;
2418 nsb->st_blksize = sb->st_blksize;
2419 nsb->st_flags = sb->st_flags;
2420 nsb->st_gen = sb->st_gen;
2421 nsb->st_birthtimespec = sb->st_birthtimespec;
2422 }
2423
2424 #ifndef _SYS_SYSPROTO_H_
2425 struct nstat_args {
2426 char *path;
2427 struct nstat *ub;
2428 };
2429 #endif
2430 int
2431 nstat(td, uap)
2432 struct thread *td;
2433 register struct nstat_args /* {
2434 char *path;
2435 struct nstat *ub;
2436 } */ *uap;
2437 {
2438 struct stat sb;
2439 struct nstat nsb;
2440 int error;
2441
2442 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2443 if (error)
2444 return (error);
2445 cvtnstat(&sb, &nsb);
2446 error = copyout(&nsb, uap->ub, sizeof (nsb));
2447 return (error);
2448 }
2449
2450 /*
2451 * NetBSD lstat. Get file status; this version does not follow links.
2452 */
2453 #ifndef _SYS_SYSPROTO_H_
2454 struct lstat_args {
2455 char *path;
2456 struct stat *ub;
2457 };
2458 #endif
2459 int
2460 nlstat(td, uap)
2461 struct thread *td;
2462 register struct nlstat_args /* {
2463 char *path;
2464 struct nstat *ub;
2465 } */ *uap;
2466 {
2467 struct stat sb;
2468 struct nstat nsb;
2469 int error;
2470
2471 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2472 if (error)
2473 return (error);
2474 cvtnstat(&sb, &nsb);
2475 error = copyout(&nsb, uap->ub, sizeof (nsb));
2476 return (error);
2477 }
2478
2479 /*
2480 * Get configurable pathname variables.
2481 */
2482 #ifndef _SYS_SYSPROTO_H_
2483 struct pathconf_args {
2484 char *path;
2485 int name;
2486 };
2487 #endif
2488 int
2489 pathconf(td, uap)
2490 struct thread *td;
2491 register struct pathconf_args /* {
2492 char *path;
2493 int name;
2494 } */ *uap;
2495 {
2496
2497 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2498 }
2499
2500 #ifndef _SYS_SYSPROTO_H_
2501 struct lpathconf_args {
2502 char *path;
2503 int name;
2504 };
2505 #endif
2506 int
2507 lpathconf(td, uap)
2508 struct thread *td;
2509 register struct lpathconf_args /* {
2510 char *path;
2511 int name;
2512 } */ *uap;
2513 {
2514
2515 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW));
2516 }
2517
2518 int
2519 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2520 u_long flags)
2521 {
2522 struct nameidata nd;
2523 int error, vfslocked;
2524
2525 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1 |
2526 flags, pathseg, path, td);
2527 if ((error = namei(&nd)) != 0)
2528 return (error);
2529 vfslocked = NDHASGIANT(&nd);
2530 NDFREE(&nd, NDF_ONLY_PNBUF);
2531
2532 /* If asynchronous I/O is available, it works for all files. */
2533 if (name == _PC_ASYNC_IO)
2534 td->td_retval[0] = async_io_version;
2535 else
2536 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2537 vput(nd.ni_vp);
2538 VFS_UNLOCK_GIANT(vfslocked);
2539 return (error);
2540 }
2541
2542 /*
2543 * Return target name of a symbolic link.
2544 */
2545 #ifndef _SYS_SYSPROTO_H_
2546 struct readlink_args {
2547 char *path;
2548 char *buf;
2549 size_t count;
2550 };
2551 #endif
2552 int
2553 readlink(td, uap)
2554 struct thread *td;
2555 register struct readlink_args /* {
2556 char *path;
2557 char *buf;
2558 size_t count;
2559 } */ *uap;
2560 {
2561
2562 return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2563 UIO_USERSPACE, uap->count));
2564 }
2565 #ifndef _SYS_SYSPROTO_H_
2566 struct readlinkat_args {
2567 int fd;
2568 char *path;
2569 char *buf;
2570 size_t bufsize;
2571 };
2572 #endif
2573 int
2574 readlinkat(struct thread *td, struct readlinkat_args *uap)
2575 {
2576
2577 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2578 uap->buf, UIO_USERSPACE, uap->bufsize));
2579 }
2580
2581 int
2582 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2583 enum uio_seg bufseg, size_t count)
2584 {
2585
2586 return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2587 count));
2588 }
2589
2590 int
2591 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2592 char *buf, enum uio_seg bufseg, size_t count)
2593 {
2594 struct vnode *vp;
2595 struct iovec aiov;
2596 struct uio auio;
2597 int error;
2598 struct nameidata nd;
2599 int vfslocked;
2600
2601 if (count > INT_MAX)
2602 return (EINVAL);
2603
2604 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2605 AUDITVNODE1, pathseg, path, fd, td);
2606
2607 if ((error = namei(&nd)) != 0)
2608 return (error);
2609 NDFREE(&nd, NDF_ONLY_PNBUF);
2610 vfslocked = NDHASGIANT(&nd);
2611 vp = nd.ni_vp;
2612 #ifdef MAC
2613 error = mac_vnode_check_readlink(td->td_ucred, vp);
2614 if (error) {
2615 vput(vp);
2616 VFS_UNLOCK_GIANT(vfslocked);
2617 return (error);
2618 }
2619 #endif
2620 if (vp->v_type != VLNK)
2621 error = EINVAL;
2622 else {
2623 aiov.iov_base = buf;
2624 aiov.iov_len = count;
2625 auio.uio_iov = &aiov;
2626 auio.uio_iovcnt = 1;
2627 auio.uio_offset = 0;
2628 auio.uio_rw = UIO_READ;
2629 auio.uio_segflg = bufseg;
2630 auio.uio_td = td;
2631 auio.uio_resid = count;
2632 error = VOP_READLINK(vp, &auio, td->td_ucred);
2633 }
2634 vput(vp);
2635 VFS_UNLOCK_GIANT(vfslocked);
2636 td->td_retval[0] = count - auio.uio_resid;
2637 return (error);
2638 }
2639
2640 /*
2641 * Common implementation code for chflags() and fchflags().
2642 */
2643 static int
2644 setfflags(td, vp, flags)
2645 struct thread *td;
2646 struct vnode *vp;
2647 int flags;
2648 {
2649 int error;
2650 struct mount *mp;
2651 struct vattr vattr;
2652
2653 /*
2654 * Prevent non-root users from setting flags on devices. When
2655 * a device is reused, users can retain ownership of the device
2656 * if they are allowed to set flags and programs assume that
2657 * chown can't fail when done as root.
2658 */
2659 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2660 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2661 if (error)
2662 return (error);
2663 }
2664
2665 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2666 return (error);
2667 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2668 VATTR_NULL(&vattr);
2669 vattr.va_flags = flags;
2670 #ifdef MAC
2671 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2672 if (error == 0)
2673 #endif
2674 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2675 VOP_UNLOCK(vp, 0);
2676 vn_finished_write(mp);
2677 return (error);
2678 }
2679
2680 /*
2681 * Change flags of a file given a path name.
2682 */
2683 #ifndef _SYS_SYSPROTO_H_
2684 struct chflags_args {
2685 char *path;
2686 int flags;
2687 };
2688 #endif
2689 int
2690 chflags(td, uap)
2691 struct thread *td;
2692 register struct chflags_args /* {
2693 char *path;
2694 int flags;
2695 } */ *uap;
2696 {
2697 int error;
2698 struct nameidata nd;
2699 int vfslocked;
2700
2701 AUDIT_ARG_FFLAGS(uap->flags);
2702 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2703 uap->path, td);
2704 if ((error = namei(&nd)) != 0)
2705 return (error);
2706 NDFREE(&nd, NDF_ONLY_PNBUF);
2707 vfslocked = NDHASGIANT(&nd);
2708 error = setfflags(td, nd.ni_vp, uap->flags);
2709 vrele(nd.ni_vp);
2710 VFS_UNLOCK_GIANT(vfslocked);
2711 return (error);
2712 }
2713
2714 /*
2715 * Same as chflags() but doesn't follow symlinks.
2716 */
2717 int
2718 lchflags(td, uap)
2719 struct thread *td;
2720 register struct lchflags_args /* {
2721 char *path;
2722 int flags;
2723 } */ *uap;
2724 {
2725 int error;
2726 struct nameidata nd;
2727 int vfslocked;
2728
2729 AUDIT_ARG_FFLAGS(uap->flags);
2730 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2731 uap->path, td);
2732 if ((error = namei(&nd)) != 0)
2733 return (error);
2734 vfslocked = NDHASGIANT(&nd);
2735 NDFREE(&nd, NDF_ONLY_PNBUF);
2736 error = setfflags(td, nd.ni_vp, uap->flags);
2737 vrele(nd.ni_vp);
2738 VFS_UNLOCK_GIANT(vfslocked);
2739 return (error);
2740 }
2741
2742 /*
2743 * Change flags of a file given a file descriptor.
2744 */
2745 #ifndef _SYS_SYSPROTO_H_
2746 struct fchflags_args {
2747 int fd;
2748 int flags;
2749 };
2750 #endif
2751 int
2752 fchflags(td, uap)
2753 struct thread *td;
2754 register struct fchflags_args /* {
2755 int fd;
2756 int flags;
2757 } */ *uap;
2758 {
2759 struct file *fp;
2760 int vfslocked;
2761 int error;
2762
2763 AUDIT_ARG_FD(uap->fd);
2764 AUDIT_ARG_FFLAGS(uap->flags);
2765 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2766 return (error);
2767 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2768 #ifdef AUDIT
2769 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2770 AUDIT_ARG_VNODE1(fp->f_vnode);
2771 VOP_UNLOCK(fp->f_vnode, 0);
2772 #endif
2773 error = setfflags(td, fp->f_vnode, uap->flags);
2774 VFS_UNLOCK_GIANT(vfslocked);
2775 fdrop(fp, td);
2776 return (error);
2777 }
2778
2779 /*
2780 * Common implementation code for chmod(), lchmod() and fchmod().
2781 */
2782 static int
2783 setfmode(td, vp, mode)
2784 struct thread *td;
2785 struct vnode *vp;
2786 int mode;
2787 {
2788 int error;
2789 struct mount *mp;
2790 struct vattr vattr;
2791
2792 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2793 return (error);
2794 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2795 VATTR_NULL(&vattr);
2796 vattr.va_mode = mode & ALLPERMS;
2797 #ifdef MAC
2798 error = mac_vnode_check_setmode(td->td_ucred, vp, vattr.va_mode);
2799 if (error == 0)
2800 #endif
2801 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2802 VOP_UNLOCK(vp, 0);
2803 vn_finished_write(mp);
2804 return (error);
2805 }
2806
2807 /*
2808 * Change mode of a file given path name.
2809 */
2810 #ifndef _SYS_SYSPROTO_H_
2811 struct chmod_args {
2812 char *path;
2813 int mode;
2814 };
2815 #endif
2816 int
2817 chmod(td, uap)
2818 struct thread *td;
2819 register struct chmod_args /* {
2820 char *path;
2821 int mode;
2822 } */ *uap;
2823 {
2824
2825 return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2826 }
2827
2828 #ifndef _SYS_SYSPROTO_H_
2829 struct fchmodat_args {
2830 int dirfd;
2831 char *path;
2832 mode_t mode;
2833 int flag;
2834 }
2835 #endif
2836 int
2837 fchmodat(struct thread *td, struct fchmodat_args *uap)
2838 {
2839 int flag = uap->flag;
2840 int fd = uap->fd;
2841 char *path = uap->path;
2842 mode_t mode = uap->mode;
2843
2844 if (flag & ~AT_SYMLINK_NOFOLLOW)
2845 return (EINVAL);
2846
2847 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2848 }
2849
2850 int
2851 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2852 {
2853
2854 return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2855 }
2856
2857 /*
2858 * Change mode of a file given path name (don't follow links.)
2859 */
2860 #ifndef _SYS_SYSPROTO_H_
2861 struct lchmod_args {
2862 char *path;
2863 int mode;
2864 };
2865 #endif
2866 int
2867 lchmod(td, uap)
2868 struct thread *td;
2869 register struct lchmod_args /* {
2870 char *path;
2871 int mode;
2872 } */ *uap;
2873 {
2874
2875 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2876 uap->mode, AT_SYMLINK_NOFOLLOW));
2877 }
2878
2879
2880 int
2881 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2882 mode_t mode, int flag)
2883 {
2884 int error;
2885 struct nameidata nd;
2886 int vfslocked;
2887 int follow;
2888
2889 AUDIT_ARG_MODE(mode);
2890 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2891 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
2892 fd, td);
2893 if ((error = namei(&nd)) != 0)
2894 return (error);
2895 vfslocked = NDHASGIANT(&nd);
2896 NDFREE(&nd, NDF_ONLY_PNBUF);
2897 error = setfmode(td, nd.ni_vp, mode);
2898 vrele(nd.ni_vp);
2899 VFS_UNLOCK_GIANT(vfslocked);
2900 return (error);
2901 }
2902
2903 /*
2904 * Change mode of a file given a file descriptor.
2905 */
2906 #ifndef _SYS_SYSPROTO_H_
2907 struct fchmod_args {
2908 int fd;
2909 int mode;
2910 };
2911 #endif
2912 int
2913 fchmod(td, uap)
2914 struct thread *td;
2915 register struct fchmod_args /* {
2916 int fd;
2917 int mode;
2918 } */ *uap;
2919 {
2920 struct file *fp;
2921 int vfslocked;
2922 int error;
2923
2924 AUDIT_ARG_FD(uap->fd);
2925 AUDIT_ARG_MODE(uap->mode);
2926 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2927 return (error);
2928 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2929 #ifdef AUDIT
2930 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2931 AUDIT_ARG_VNODE1(fp->f_vnode);
2932 VOP_UNLOCK(fp->f_vnode, 0);
2933 #endif
2934 error = setfmode(td, fp->f_vnode, uap->mode);
2935 VFS_UNLOCK_GIANT(vfslocked);
2936 fdrop(fp, td);
2937 return (error);
2938 }
2939
2940 /*
2941 * Common implementation for chown(), lchown(), and fchown()
2942 */
2943 static int
2944 setfown(td, vp, uid, gid)
2945 struct thread *td;
2946 struct vnode *vp;
2947 uid_t uid;
2948 gid_t gid;
2949 {
2950 int error;
2951 struct mount *mp;
2952 struct vattr vattr;
2953
2954 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2955 return (error);
2956 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2957 VATTR_NULL(&vattr);
2958 vattr.va_uid = uid;
2959 vattr.va_gid = gid;
2960 #ifdef MAC
2961 error = mac_vnode_check_setowner(td->td_ucred, vp, vattr.va_uid,
2962 vattr.va_gid);
2963 if (error == 0)
2964 #endif
2965 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2966 VOP_UNLOCK(vp, 0);
2967 vn_finished_write(mp);
2968 return (error);
2969 }
2970
2971 /*
2972 * Set ownership given a path name.
2973 */
2974 #ifndef _SYS_SYSPROTO_H_
2975 struct chown_args {
2976 char *path;
2977 int uid;
2978 int gid;
2979 };
2980 #endif
2981 int
2982 chown(td, uap)
2983 struct thread *td;
2984 register struct chown_args /* {
2985 char *path;
2986 int uid;
2987 int gid;
2988 } */ *uap;
2989 {
2990
2991 return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2992 }
2993
2994 #ifndef _SYS_SYSPROTO_H_
2995 struct fchownat_args {
2996 int fd;
2997 const char * path;
2998 uid_t uid;
2999 gid_t gid;
3000 int flag;
3001 };
3002 #endif
3003 int
3004 fchownat(struct thread *td, struct fchownat_args *uap)
3005 {
3006 int flag;
3007
3008 flag = uap->flag;
3009 if (flag & ~AT_SYMLINK_NOFOLLOW)
3010 return (EINVAL);
3011
3012 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3013 uap->gid, uap->flag));
3014 }
3015
3016 int
3017 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3018 int gid)
3019 {
3020
3021 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3022 }
3023
3024 int
3025 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3026 int uid, int gid, int flag)
3027 {
3028 struct nameidata nd;
3029 int error, vfslocked, follow;
3030
3031 AUDIT_ARG_OWNER(uid, gid);
3032 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3033 NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
3034 fd, td);
3035
3036 if ((error = namei(&nd)) != 0)
3037 return (error);
3038 vfslocked = NDHASGIANT(&nd);
3039 NDFREE(&nd, NDF_ONLY_PNBUF);
3040 error = setfown(td, nd.ni_vp, uid, gid);
3041 vrele(nd.ni_vp);
3042 VFS_UNLOCK_GIANT(vfslocked);
3043 return (error);
3044 }
3045
3046 /*
3047 * Set ownership given a path name, do not cross symlinks.
3048 */
3049 #ifndef _SYS_SYSPROTO_H_
3050 struct lchown_args {
3051 char *path;
3052 int uid;
3053 int gid;
3054 };
3055 #endif
3056 int
3057 lchown(td, uap)
3058 struct thread *td;
3059 register struct lchown_args /* {
3060 char *path;
3061 int uid;
3062 int gid;
3063 } */ *uap;
3064 {
3065
3066 return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3067 }
3068
3069 int
3070 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3071 int gid)
3072 {
3073
3074 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3075 AT_SYMLINK_NOFOLLOW));
3076 }
3077
3078 /*
3079 * Set ownership given a file descriptor.
3080 */
3081 #ifndef _SYS_SYSPROTO_H_
3082 struct fchown_args {
3083 int fd;
3084 int uid;
3085 int gid;
3086 };
3087 #endif
3088 int
3089 fchown(td, uap)
3090 struct thread *td;
3091 register struct fchown_args /* {
3092 int fd;
3093 int uid;
3094 int gid;
3095 } */ *uap;
3096 {
3097 struct file *fp;
3098 int vfslocked;
3099 int error;
3100
3101 AUDIT_ARG_FD(uap->fd);
3102 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3103 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3104 return (error);
3105 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3106 #ifdef AUDIT
3107 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3108 AUDIT_ARG_VNODE1(fp->f_vnode);
3109 VOP_UNLOCK(fp->f_vnode, 0);
3110 #endif
3111 error = setfown(td, fp->f_vnode, uap->uid, uap->gid);
3112 VFS_UNLOCK_GIANT(vfslocked);
3113 fdrop(fp, td);
3114 return (error);
3115 }
3116
3117 /*
3118 * Common implementation code for utimes(), lutimes(), and futimes().
3119 */
3120 static int
3121 getutimes(usrtvp, tvpseg, tsp)
3122 const struct timeval *usrtvp;
3123 enum uio_seg tvpseg;
3124 struct timespec *tsp;
3125 {
3126 struct timeval tv[2];
3127 const struct timeval *tvp;
3128 int error;
3129
3130 if (usrtvp == NULL) {
3131 vfs_timestamp(&tsp[0]);
3132 tsp[1] = tsp[0];
3133 } else {
3134 if (tvpseg == UIO_SYSSPACE) {
3135 tvp = usrtvp;
3136 } else {
3137 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3138 return (error);
3139 tvp = tv;
3140 }
3141
3142 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3143 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3144 return (EINVAL);
3145 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3146 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3147 }
3148 return (0);
3149 }
3150
3151 /*
3152 * Common implementation code for utimes(), lutimes(), and futimes().
3153 */
3154 static int
3155 setutimes(td, vp, ts, numtimes, nullflag)
3156 struct thread *td;
3157 struct vnode *vp;
3158 const struct timespec *ts;
3159 int numtimes;
3160 int nullflag;
3161 {
3162 int error, setbirthtime;
3163 struct mount *mp;
3164 struct vattr vattr;
3165
3166 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3167 return (error);
3168 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3169 setbirthtime = 0;
3170 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3171 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3172 setbirthtime = 1;
3173 VATTR_NULL(&vattr);
3174 vattr.va_atime = ts[0];
3175 vattr.va_mtime = ts[1];
3176 if (setbirthtime)
3177 vattr.va_birthtime = ts[1];
3178 if (numtimes > 2)
3179 vattr.va_birthtime = ts[2];
3180 if (nullflag)
3181 vattr.va_vaflags |= VA_UTIMES_NULL;
3182 #ifdef MAC
3183 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3184 vattr.va_mtime);
3185 #endif
3186 if (error == 0)
3187 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3188 VOP_UNLOCK(vp, 0);
3189 vn_finished_write(mp);
3190 return (error);
3191 }
3192
3193 /*
3194 * Set the access and modification times of a file.
3195 */
3196 #ifndef _SYS_SYSPROTO_H_
3197 struct utimes_args {
3198 char *path;
3199 struct timeval *tptr;
3200 };
3201 #endif
3202 int
3203 utimes(td, uap)
3204 struct thread *td;
3205 register struct utimes_args /* {
3206 char *path;
3207 struct timeval *tptr;
3208 } */ *uap;
3209 {
3210
3211 return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3212 UIO_USERSPACE));
3213 }
3214
3215 #ifndef _SYS_SYSPROTO_H_
3216 struct futimesat_args {
3217 int fd;
3218 const char * path;
3219 const struct timeval * times;
3220 };
3221 #endif
3222 int
3223 futimesat(struct thread *td, struct futimesat_args *uap)
3224 {
3225
3226 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3227 uap->times, UIO_USERSPACE));
3228 }
3229
3230 int
3231 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3232 struct timeval *tptr, enum uio_seg tptrseg)
3233 {
3234
3235 return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3236 }
3237
3238 int
3239 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3240 struct timeval *tptr, enum uio_seg tptrseg)
3241 {
3242 struct nameidata nd;
3243 struct timespec ts[2];
3244 int error, vfslocked;
3245
3246 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3247 return (error);
3248 NDINIT_AT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path,
3249 fd, td);
3250
3251 if ((error = namei(&nd)) != 0)
3252 return (error);
3253 vfslocked = NDHASGIANT(&nd);
3254 NDFREE(&nd, NDF_ONLY_PNBUF);
3255 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3256 vrele(nd.ni_vp);
3257 VFS_UNLOCK_GIANT(vfslocked);
3258 return (error);
3259 }
3260
3261 /*
3262 * Set the access and modification times of a file.
3263 */
3264 #ifndef _SYS_SYSPROTO_H_
3265 struct lutimes_args {
3266 char *path;
3267 struct timeval *tptr;
3268 };
3269 #endif
3270 int
3271 lutimes(td, uap)
3272 struct thread *td;
3273 register struct lutimes_args /* {
3274 char *path;
3275 struct timeval *tptr;
3276 } */ *uap;
3277 {
3278
3279 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3280 UIO_USERSPACE));
3281 }
3282
3283 int
3284 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3285 struct timeval *tptr, enum uio_seg tptrseg)
3286 {
3287 struct timespec ts[2];
3288 int error;
3289 struct nameidata nd;
3290 int vfslocked;
3291
3292 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3293 return (error);
3294 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3295 if ((error = namei(&nd)) != 0)
3296 return (error);
3297 vfslocked = NDHASGIANT(&nd);
3298 NDFREE(&nd, NDF_ONLY_PNBUF);
3299 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3300 vrele(nd.ni_vp);
3301 VFS_UNLOCK_GIANT(vfslocked);
3302 return (error);
3303 }
3304
3305 /*
3306 * Set the access and modification times of a file.
3307 */
3308 #ifndef _SYS_SYSPROTO_H_
3309 struct futimes_args {
3310 int fd;
3311 struct timeval *tptr;
3312 };
3313 #endif
3314 int
3315 futimes(td, uap)
3316 struct thread *td;
3317 register struct futimes_args /* {
3318 int fd;
3319 struct timeval *tptr;
3320 } */ *uap;
3321 {
3322
3323 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3324 }
3325
3326 int
3327 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3328 enum uio_seg tptrseg)
3329 {
3330 struct timespec ts[2];
3331 struct file *fp;
3332 int vfslocked;
3333 int error;
3334
3335 AUDIT_ARG_FD(fd);
3336 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3337 return (error);
3338 if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
3339 return (error);
3340 vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3341 #ifdef AUDIT
3342 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3343 AUDIT_ARG_VNODE1(fp->f_vnode);
3344 VOP_UNLOCK(fp->f_vnode, 0);
3345 #endif
3346 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3347 VFS_UNLOCK_GIANT(vfslocked);
3348 fdrop(fp, td);
3349 return (error);
3350 }
3351
3352 /*
3353 * Truncate a file given its path name.
3354 */
3355 #ifndef _SYS_SYSPROTO_H_
3356 struct truncate_args {
3357 char *path;
3358 int pad;
3359 off_t length;
3360 };
3361 #endif
3362 int
3363 truncate(td, uap)
3364 struct thread *td;
3365 register struct truncate_args /* {
3366 char *path;
3367 int pad;
3368 off_t length;
3369 } */ *uap;
3370 {
3371
3372 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3373 }
3374
3375 int
3376 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3377 {
3378 struct mount *mp;
3379 struct vnode *vp;
3380 struct vattr vattr;
3381 int error;
3382 struct nameidata nd;
3383 int vfslocked;
3384
3385 if (length < 0)
3386 return(EINVAL);
3387 NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3388 if ((error = namei(&nd)) != 0)
3389 return (error);
3390 vfslocked = NDHASGIANT(&nd);
3391 vp = nd.ni_vp;
3392 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3393 vrele(vp);
3394 VFS_UNLOCK_GIANT(vfslocked);
3395 return (error);
3396 }
3397 NDFREE(&nd, NDF_ONLY_PNBUF);
3398 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3399 if (vp->v_type == VDIR)
3400 error = EISDIR;
3401 #ifdef MAC
3402 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3403 }
3404 #endif
3405 else if ((error = vn_writechk(vp)) == 0 &&
3406 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3407 VATTR_NULL(&vattr);
3408 vattr.va_size = length;
3409 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3410 }
3411 vput(vp);
3412 vn_finished_write(mp);
3413 VFS_UNLOCK_GIANT(vfslocked);
3414 return (error);
3415 }
3416
3417 #if defined(COMPAT_43)
3418 /*
3419 * Truncate a file given its path name.
3420 */
3421 #ifndef _SYS_SYSPROTO_H_
3422 struct otruncate_args {
3423 char *path;
3424 long length;
3425 };
3426 #endif
3427 int
3428 otruncate(td, uap)
3429 struct thread *td;
3430 register struct otruncate_args /* {
3431 char *path;
3432 long length;
3433 } */ *uap;
3434 {
3435 struct truncate_args /* {
3436 char *path;
3437 int pad;
3438 off_t length;
3439 } */ nuap;
3440
3441 nuap.path = uap->path;
3442 nuap.length = uap->length;
3443 return (truncate(td, &nuap));
3444 }
3445 #endif /* COMPAT_43 */
3446
3447 /* Versions with the pad argument */
3448 int
3449 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3450 {
3451 struct truncate_args ouap;
3452
3453 ouap.path = uap->path;
3454 ouap.length = uap->length;
3455 return (truncate(td, &ouap));
3456 }
3457
3458 int
3459 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3460 {
3461 struct ftruncate_args ouap;
3462
3463 ouap.fd = uap->fd;
3464 ouap.length = uap->length;
3465 return (ftruncate(td, &ouap));
3466 }
3467
3468 /*
3469 * Sync an open file.
3470 */
3471 #ifndef _SYS_SYSPROTO_H_
3472 struct fsync_args {
3473 int fd;
3474 };
3475 #endif
3476 int
3477 fsync(td, uap)
3478 struct thread *td;
3479 struct fsync_args /* {
3480 int fd;
3481 } */ *uap;
3482 {
3483 struct vnode *vp;
3484 struct mount *mp;
3485 struct file *fp;
3486 int vfslocked;
3487 int error, lock_flags;
3488
3489 AUDIT_ARG_FD(uap->fd);
3490 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3491 return (error);
3492 vp = fp->f_vnode;
3493 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3494 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3495 goto drop;
3496 if (MNT_SHARED_WRITES(mp) ||
3497 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3498 lock_flags = LK_SHARED;
3499 } else {
3500 lock_flags = LK_EXCLUSIVE;
3501 }
3502 vn_lock(vp, lock_flags | LK_RETRY);
3503 AUDIT_ARG_VNODE1(vp);
3504 if (vp->v_object != NULL) {
3505 VM_OBJECT_LOCK(vp->v_object);
3506 vm_object_page_clean(vp->v_object, 0, 0, 0);
3507 VM_OBJECT_UNLOCK(vp->v_object);
3508 }
3509 error = VOP_FSYNC(vp, MNT_WAIT, td);
3510
3511 VOP_UNLOCK(vp, 0);
3512 vn_finished_write(mp);
3513 drop:
3514 VFS_UNLOCK_GIANT(vfslocked);
3515 fdrop(fp, td);
3516 return (error);
3517 }
3518
3519 /*
3520 * Rename files. Source and destination must either both be directories, or
3521 * both not be directories. If target is a directory, it must be empty.
3522 */
3523 #ifndef _SYS_SYSPROTO_H_
3524 struct rename_args {
3525 char *from;
3526 char *to;
3527 };
3528 #endif
3529 int
3530 rename(td, uap)
3531 struct thread *td;
3532 register struct rename_args /* {
3533 char *from;
3534 char *to;
3535 } */ *uap;
3536 {
3537
3538 return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3539 }
3540
3541 #ifndef _SYS_SYSPROTO_H_
3542 struct renameat_args {
3543 int oldfd;
3544 char *old;
3545 int newfd;
3546 char *new;
3547 };
3548 #endif
3549 int
3550 renameat(struct thread *td, struct renameat_args *uap)
3551 {
3552
3553 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3554 UIO_USERSPACE));
3555 }
3556
3557 int
3558 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3559 {
3560
3561 return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3562 }
3563
3564 int
3565 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3566 enum uio_seg pathseg)
3567 {
3568 struct mount *mp = NULL;
3569 struct vnode *tvp, *fvp, *tdvp;
3570 struct nameidata fromnd, tond;
3571 int tvfslocked;
3572 int fvfslocked;
3573 int error;
3574
3575 bwillwrite();
3576 #ifdef MAC
3577 NDINIT_AT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | MPSAFE |
3578 AUDITVNODE1, pathseg, old, oldfd, td);
3579 #else
3580 NDINIT_AT(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3581 AUDITVNODE1, pathseg, old, oldfd, td);
3582 #endif
3583
3584 if ((error = namei(&fromnd)) != 0)
3585 return (error);
3586 fvfslocked = NDHASGIANT(&fromnd);
3587 tvfslocked = 0;
3588 #ifdef MAC
3589 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3590 fromnd.ni_vp, &fromnd.ni_cnd);
3591 VOP_UNLOCK(fromnd.ni_dvp, 0);
3592 if (fromnd.ni_dvp != fromnd.ni_vp)
3593 VOP_UNLOCK(fromnd.ni_vp, 0);
3594 #endif
3595 fvp = fromnd.ni_vp;
3596 if (error == 0)
3597 error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3598 if (error != 0) {
3599 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3600 vrele(fromnd.ni_dvp);
3601 vrele(fvp);
3602 goto out1;
3603 }
3604 NDINIT_AT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3605 MPSAFE | AUDITVNODE2, pathseg, new, newfd, td);
3606 if (fromnd.ni_vp->v_type == VDIR)
3607 tond.ni_cnd.cn_flags |= WILLBEDIR;
3608 if ((error = namei(&tond)) != 0) {
3609 /* Translate error code for rename("dir1", "dir2/."). */
3610 if (error == EISDIR && fvp->v_type == VDIR)
3611 error = EINVAL;
3612 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3613 vrele(fromnd.ni_dvp);
3614 vrele(fvp);
3615 vn_finished_write(mp);
3616 goto out1;
3617 }
3618 tvfslocked = NDHASGIANT(&tond);
3619 tdvp = tond.ni_dvp;
3620 tvp = tond.ni_vp;
3621 if (tvp != NULL) {
3622 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3623 error = ENOTDIR;
3624 goto out;
3625 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3626 error = EISDIR;
3627 goto out;
3628 }
3629 }
3630 if (fvp == tdvp) {
3631 error = EINVAL;
3632 goto out;
3633 }
3634 /*
3635 * If the source is the same as the destination (that is, if they
3636 * are links to the same vnode), then there is nothing to do.
3637 */
3638 if (fvp == tvp)
3639 error = -1;
3640 #ifdef MAC
3641 else
3642 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3643 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3644 #endif
3645 out:
3646 if (!error) {
3647 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3648 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3649 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3650 NDFREE(&tond, NDF_ONLY_PNBUF);
3651 } else {
3652 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3653 NDFREE(&tond, NDF_ONLY_PNBUF);
3654 if (tvp)
3655 vput(tvp);
3656 if (tdvp == tvp)
3657 vrele(tdvp);
3658 else
3659 vput(tdvp);
3660 vrele(fromnd.ni_dvp);
3661 vrele(fvp);
3662 }
3663 vrele(tond.ni_startdir);
3664 vn_finished_write(mp);
3665 out1:
3666 if (fromnd.ni_startdir)
3667 vrele(fromnd.ni_startdir);
3668 VFS_UNLOCK_GIANT(fvfslocked);
3669 VFS_UNLOCK_GIANT(tvfslocked);
3670 if (error == -1)
3671 return (0);
3672 return (error);
3673 }
3674
3675 /*
3676 * Make a directory file.
3677 */
3678 #ifndef _SYS_SYSPROTO_H_
3679 struct mkdir_args {
3680 char *path;
3681 int mode;
3682 };
3683 #endif
3684 int
3685 mkdir(td, uap)
3686 struct thread *td;
3687 register struct mkdir_args /* {
3688 char *path;
3689 int mode;
3690 } */ *uap;
3691 {
3692
3693 return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3694 }
3695
3696 #ifndef _SYS_SYSPROTO_H_
3697 struct mkdirat_args {
3698 int fd;
3699 char *path;
3700 mode_t mode;
3701 };
3702 #endif
3703 int
3704 mkdirat(struct thread *td, struct mkdirat_args *uap)
3705 {
3706
3707 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3708 }
3709
3710 int
3711 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3712 {
3713
3714 return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3715 }
3716
3717 int
3718 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3719 int mode)
3720 {
3721 struct mount *mp;
3722 struct vnode *vp;
3723 struct vattr vattr;
3724 int error;
3725 struct nameidata nd;
3726 int vfslocked;
3727
3728 AUDIT_ARG_MODE(mode);
3729 restart:
3730 bwillwrite();
3731 NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
3732 segflg, path, fd, td);
3733 nd.ni_cnd.cn_flags |= WILLBEDIR;
3734 if ((error = namei(&nd)) != 0)
3735 return (error);
3736 vfslocked = NDHASGIANT(&nd);
3737 vp = nd.ni_vp;
3738 if (vp != NULL) {
3739 NDFREE(&nd, NDF_ONLY_PNBUF);
3740 /*
3741 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3742 * the strange behaviour of leaving the vnode unlocked
3743 * if the target is the same vnode as the parent.
3744 */
3745 if (vp == nd.ni_dvp)
3746 vrele(nd.ni_dvp);
3747 else
3748 vput(nd.ni_dvp);
3749 vrele(vp);
3750 VFS_UNLOCK_GIANT(vfslocked);
3751 return (EEXIST);
3752 }
3753 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3754 NDFREE(&nd, NDF_ONLY_PNBUF);
3755 vput(nd.ni_dvp);
3756 VFS_UNLOCK_GIANT(vfslocked);
3757 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3758 return (error);
3759 goto restart;
3760 }
3761 VATTR_NULL(&vattr);
3762 vattr.va_type = VDIR;
3763 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3764 #ifdef MAC
3765 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3766 &vattr);
3767 if (error)
3768 goto out;
3769 #endif
3770 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3771 #ifdef MAC
3772 out:
3773 #endif
3774 NDFREE(&nd, NDF_ONLY_PNBUF);
3775 vput(nd.ni_dvp);
3776 if (!error)
3777 vput(nd.ni_vp);
3778 vn_finished_write(mp);
3779 VFS_UNLOCK_GIANT(vfslocked);
3780 return (error);
3781 }
3782
3783 /*
3784 * Remove a directory file.
3785 */
3786 #ifndef _SYS_SYSPROTO_H_
3787 struct rmdir_args {
3788 char *path;
3789 };
3790 #endif
3791 int
3792 rmdir(td, uap)
3793 struct thread *td;
3794 struct rmdir_args /* {
3795 char *path;
3796 } */ *uap;
3797 {
3798
3799 return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3800 }
3801
3802 int
3803 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3804 {
3805
3806 return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3807 }
3808
3809 int
3810 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3811 {
3812 struct mount *mp;
3813 struct vnode *vp;
3814 int error;
3815 struct nameidata nd;
3816 int vfslocked;
3817
3818 restart:
3819 bwillwrite();
3820 NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
3821 pathseg, path, fd, td);
3822 if ((error = namei(&nd)) != 0)
3823 return (error);
3824 vfslocked = NDHASGIANT(&nd);
3825 vp = nd.ni_vp;
3826 if (vp->v_type != VDIR) {
3827 error = ENOTDIR;
3828 goto out;
3829 }
3830 /*
3831 * No rmdir "." please.
3832 */
3833 if (nd.ni_dvp == vp) {
3834 error = EINVAL;
3835 goto out;
3836 }
3837 /*
3838 * The root of a mounted filesystem cannot be deleted.
3839 */
3840 if (vp->v_vflag & VV_ROOT) {
3841 error = EBUSY;
3842 goto out;
3843 }
3844 #ifdef MAC
3845 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3846 &nd.ni_cnd);
3847 if (error)
3848 goto out;
3849 #endif
3850 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3851 NDFREE(&nd, NDF_ONLY_PNBUF);
3852 vput(vp);
3853 if (nd.ni_dvp == vp)
3854 vrele(nd.ni_dvp);
3855 else
3856 vput(nd.ni_dvp);
3857 VFS_UNLOCK_GIANT(vfslocked);
3858 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3859 return (error);
3860 goto restart;
3861 }
3862 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3863 vn_finished_write(mp);
3864 out:
3865 NDFREE(&nd, NDF_ONLY_PNBUF);
3866 vput(vp);
3867 if (nd.ni_dvp == vp)
3868 vrele(nd.ni_dvp);
3869 else
3870 vput(nd.ni_dvp);
3871 VFS_UNLOCK_GIANT(vfslocked);
3872 return (error);
3873 }
3874
3875 #ifdef COMPAT_43
3876 /*
3877 * Read a block of directory entries in a filesystem independent format.
3878 */
3879 #ifndef _SYS_SYSPROTO_H_
3880 struct ogetdirentries_args {
3881 int fd;
3882 char *buf;
3883 u_int count;
3884 long *basep;
3885 };
3886 #endif
3887 int
3888 ogetdirentries(td, uap)
3889 struct thread *td;
3890 register struct ogetdirentries_args /* {
3891 int fd;
3892 char *buf;
3893 u_int count;
3894 long *basep;
3895 } */ *uap;
3896 {
3897 struct vnode *vp;
3898 struct file *fp;
3899 struct uio auio, kuio;
3900 struct iovec aiov, kiov;
3901 struct dirent *dp, *edp;
3902 caddr_t dirbuf;
3903 int error, eofflag, readcnt, vfslocked;
3904 long loff;
3905
3906 /* XXX arbitrary sanity limit on `count'. */
3907 if (uap->count > 64 * 1024)
3908 return (EINVAL);
3909 if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3910 return (error);
3911 if ((fp->f_flag & FREAD) == 0) {
3912 fdrop(fp, td);
3913 return (EBADF);
3914 }
3915 vp = fp->f_vnode;
3916 unionread:
3917 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3918 if (vp->v_type != VDIR) {
3919 VFS_UNLOCK_GIANT(vfslocked);
3920 fdrop(fp, td);
3921 return (EINVAL);
3922 }
3923 aiov.iov_base = uap->buf;
3924 aiov.iov_len = uap->count;
3925 auio.uio_iov = &aiov;
3926 auio.uio_iovcnt = 1;
3927 auio.uio_rw = UIO_READ;
3928 auio.uio_segflg = UIO_USERSPACE;
3929 auio.uio_td = td;
3930 auio.uio_resid = uap->count;
3931 vn_lock(vp, LK_SHARED | LK_RETRY);
3932 loff = auio.uio_offset = fp->f_offset;
3933 #ifdef MAC
3934 error = mac_vnode_check_readdir(td->td_ucred, vp);
3935 if (error) {
3936 VOP_UNLOCK(vp, 0);
3937 VFS_UNLOCK_GIANT(vfslocked);
3938 fdrop(fp, td);
3939 return (error);
3940 }
3941 #endif
3942 # if (BYTE_ORDER != LITTLE_ENDIAN)
3943 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3944 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3945 NULL, NULL);
3946 fp->f_offset = auio.uio_offset;
3947 } else
3948 # endif
3949 {
3950 kuio = auio;
3951 kuio.uio_iov = &kiov;
3952 kuio.uio_segflg = UIO_SYSSPACE;
3953 kiov.iov_len = uap->count;
3954 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3955 kiov.iov_base = dirbuf;
3956 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3957 NULL, NULL);
3958 fp->f_offset = kuio.uio_offset;
3959 if (error == 0) {
3960 readcnt = uap->count - kuio.uio_resid;
3961 edp = (struct dirent *)&dirbuf[readcnt];
3962 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3963 # if (BYTE_ORDER == LITTLE_ENDIAN)
3964 /*
3965 * The expected low byte of
3966 * dp->d_namlen is our dp->d_type.
3967 * The high MBZ byte of dp->d_namlen
3968 * is our dp->d_namlen.
3969 */
3970 dp->d_type = dp->d_namlen;
3971 dp->d_namlen = 0;
3972 # else
3973 /*
3974 * The dp->d_type is the high byte
3975 * of the expected dp->d_namlen,
3976 * so must be zero'ed.
3977 */
3978 dp->d_type = 0;
3979 # endif
3980 if (dp->d_reclen > 0) {
3981 dp = (struct dirent *)
3982 ((char *)dp + dp->d_reclen);
3983 } else {
3984 error = EIO;
3985 break;
3986 }
3987 }
3988 if (dp >= edp)
3989 error = uiomove(dirbuf, readcnt, &auio);
3990 }
3991 free(dirbuf, M_TEMP);
3992 }
3993 if (error) {
3994 VOP_UNLOCK(vp, 0);
3995 VFS_UNLOCK_GIANT(vfslocked);
3996 fdrop(fp, td);
3997 return (error);
3998 }
3999 if (uap->count == auio.uio_resid &&
4000 (vp->v_vflag & VV_ROOT) &&
4001 (vp->v_mount->mnt_flag & MNT_UNION)) {
4002 struct vnode *tvp = vp;
4003 vp = vp->v_mount->mnt_vnodecovered;
4004 VREF(vp);
4005 fp->f_vnode = vp;
4006 fp->f_data = vp;
4007 fp->f_offset = 0;
4008 vput(tvp);
4009 VFS_UNLOCK_GIANT(vfslocked);
4010 goto unionread;
4011 }
4012 VOP_UNLOCK(vp, 0);
4013 VFS_UNLOCK_GIANT(vfslocked);
4014 error = copyout(&loff, uap->basep, sizeof(long));
4015 fdrop(fp, td);
4016 td->td_retval[0] = uap->count - auio.uio_resid;
4017 return (error);
4018 }
4019 #endif /* COMPAT_43 */
4020
4021 /*
4022 * Read a block of directory entries in a filesystem independent format.
4023 */
4024 #ifndef _SYS_SYSPROTO_H_
4025 struct getdirentries_args {
4026 int fd;
4027 char *buf;
4028 u_int count;
4029 long *basep;
4030 };
4031 #endif
4032 int
4033 getdirentries(td, uap)
4034 struct thread *td;
4035 register struct getdirentries_args /* {
4036 int fd;
4037 char *buf;
4038 u_int count;
4039 long *basep;
4040 } */ *uap;
4041 {
4042 long base;
4043 int error;
4044
4045 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4046 if (error)
4047 return (error);
4048 if (uap->basep != NULL)
4049 error = copyout(&base, uap->basep, sizeof(long));
4050 return (error);
4051 }
4052
4053 int
4054 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4055 long *basep)
4056 {
4057 struct vnode *vp;
4058 struct file *fp;
4059 struct uio auio;
4060 struct iovec aiov;
4061 int vfslocked;
4062 long loff;
4063 int error, eofflag;
4064
4065 AUDIT_ARG_FD(fd);
4066 if (count > INT_MAX)
4067 return (EINVAL);
4068 if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
4069 return (error);
4070 if ((fp->f_flag & FREAD) == 0) {
4071 fdrop(fp, td);
4072 return (EBADF);
4073 }
4074 vp = fp->f_vnode;
4075 unionread:
4076 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4077 if (vp->v_type != VDIR) {
4078 VFS_UNLOCK_GIANT(vfslocked);
4079 error = EINVAL;
4080 goto fail;
4081 }
4082 aiov.iov_base = buf;
4083 aiov.iov_len = count;
4084 auio.uio_iov = &aiov;
4085 auio.uio_iovcnt = 1;
4086 auio.uio_rw = UIO_READ;
4087 auio.uio_segflg = UIO_USERSPACE;
4088 auio.uio_td = td;
4089 auio.uio_resid = count;
4090 vn_lock(vp, LK_SHARED | LK_RETRY);
4091 AUDIT_ARG_VNODE1(vp);
4092 loff = auio.uio_offset = fp->f_offset;
4093 #ifdef MAC
4094 error = mac_vnode_check_readdir(td->td_ucred, vp);
4095 if (error == 0)
4096 #endif
4097 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4098 NULL);
4099 fp->f_offset = auio.uio_offset;
4100 if (error) {
4101 VOP_UNLOCK(vp, 0);
4102 VFS_UNLOCK_GIANT(vfslocked);
4103 goto fail;
4104 }
4105 if (count == auio.uio_resid &&
4106 (vp->v_vflag & VV_ROOT) &&
4107 (vp->v_mount->mnt_flag & MNT_UNION)) {
4108 struct vnode *tvp = vp;
4109 vp = vp->v_mount->mnt_vnodecovered;
4110 VREF(vp);
4111 fp->f_vnode = vp;
4112 fp->f_data = vp;
4113 fp->f_offset = 0;
4114 vput(tvp);
4115 VFS_UNLOCK_GIANT(vfslocked);
4116 goto unionread;
4117 }
4118 VOP_UNLOCK(vp, 0);
4119 VFS_UNLOCK_GIANT(vfslocked);
4120 *basep = loff;
4121 td->td_retval[0] = count - auio.uio_resid;
4122 fail:
4123 fdrop(fp, td);
4124 return (error);
4125 }
4126
4127 #ifndef _SYS_SYSPROTO_H_
4128 struct getdents_args {
4129 int fd;
4130 char *buf;
4131 size_t count;
4132 };
4133 #endif
4134 int
4135 getdents(td, uap)
4136 struct thread *td;
4137 register struct getdents_args /* {
4138 int fd;
4139 char *buf;
4140 u_int count;
4141 } */ *uap;
4142 {
4143 struct getdirentries_args ap;
4144 ap.fd = uap->fd;
4145 ap.buf = uap->buf;
4146 ap.count = uap->count;
4147 ap.basep = NULL;
4148 return (getdirentries(td, &ap));
4149 }
4150
4151 /*
4152 * Set the mode mask for creation of filesystem nodes.
4153 */
4154 #ifndef _SYS_SYSPROTO_H_
4155 struct umask_args {
4156 int newmask;
4157 };
4158 #endif
4159 int
4160 umask(td, uap)
4161 struct thread *td;
4162 struct umask_args /* {
4163 int newmask;
4164 } */ *uap;
4165 {
4166 register struct filedesc *fdp;
4167
4168 FILEDESC_XLOCK(td->td_proc->p_fd);
4169 fdp = td->td_proc->p_fd;
4170 td->td_retval[0] = fdp->fd_cmask;
4171 fdp->fd_cmask = uap->newmask & ALLPERMS;
4172 FILEDESC_XUNLOCK(td->td_proc->p_fd);
4173 return (0);
4174 }
4175
4176 /*
4177 * Void all references to file by ripping underlying filesystem away from
4178 * vnode.
4179 */
4180 #ifndef _SYS_SYSPROTO_H_
4181 struct revoke_args {
4182 char *path;
4183 };
4184 #endif
4185 int
4186 revoke(td, uap)
4187 struct thread *td;
4188 register struct revoke_args /* {
4189 char *path;
4190 } */ *uap;
4191 {
4192 struct vnode *vp;
4193 struct vattr vattr;
4194 int error;
4195 struct nameidata nd;
4196 int vfslocked;
4197
4198 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4199 UIO_USERSPACE, uap->path, td);
4200 if ((error = namei(&nd)) != 0)
4201 return (error);
4202 vfslocked = NDHASGIANT(&nd);
4203 vp = nd.ni_vp;
4204 NDFREE(&nd, NDF_ONLY_PNBUF);
4205 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4206 error = EINVAL;
4207 goto out;
4208 }
4209 #ifdef MAC
4210 error = mac_vnode_check_revoke(td->td_ucred, vp);
4211 if (error)
4212 goto out;
4213 #endif
4214 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4215 if (error)
4216 goto out;
4217 if (td->td_ucred->cr_uid != vattr.va_uid) {
4218 error = priv_check(td, PRIV_VFS_ADMIN);
4219 if (error)
4220 goto out;
4221 }
4222 if (vcount(vp) > 1)
4223 VOP_REVOKE(vp, REVOKEALL);
4224 out:
4225 vput(vp);
4226 VFS_UNLOCK_GIANT(vfslocked);
4227 return (error);
4228 }
4229
4230 /*
4231 * Convert a user file descriptor to a kernel file entry.
4232 * A reference on the file entry is held upon returning.
4233 */
4234 int
4235 getvnode(fdp, fd, fpp)
4236 struct filedesc *fdp;
4237 int fd;
4238 struct file **fpp;
4239 {
4240 int error;
4241 struct file *fp;
4242
4243 error = 0;
4244 fp = NULL;
4245 if (fdp == NULL || (fp = fget_unlocked(fdp, fd)) == NULL)
4246 error = EBADF;
4247 else if (fp->f_vnode == NULL) {
4248 error = EINVAL;
4249 fdrop(fp, curthread);
4250 }
4251 *fpp = fp;
4252 return (error);
4253 }
4254
4255 /*
4256 * Get an (NFS) file handle.
4257 */
4258 #ifndef _SYS_SYSPROTO_H_
4259 struct lgetfh_args {
4260 char *fname;
4261 fhandle_t *fhp;
4262 };
4263 #endif
4264 int
4265 lgetfh(td, uap)
4266 struct thread *td;
4267 register struct lgetfh_args *uap;
4268 {
4269 struct nameidata nd;
4270 fhandle_t fh;
4271 register struct vnode *vp;
4272 int vfslocked;
4273 int error;
4274
4275 error = priv_check(td, PRIV_VFS_GETFH);
4276 if (error)
4277 return (error);
4278 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4279 UIO_USERSPACE, uap->fname, td);
4280 error = namei(&nd);
4281 if (error)
4282 return (error);
4283 vfslocked = NDHASGIANT(&nd);
4284 NDFREE(&nd, NDF_ONLY_PNBUF);
4285 vp = nd.ni_vp;
4286 bzero(&fh, sizeof(fh));
4287 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4288 error = VOP_VPTOFH(vp, &fh.fh_fid);
4289 vput(vp);
4290 VFS_UNLOCK_GIANT(vfslocked);
4291 if (error)
4292 return (error);
4293 error = copyout(&fh, uap->fhp, sizeof (fh));
4294 return (error);
4295 }
4296
4297 #ifndef _SYS_SYSPROTO_H_
4298 struct getfh_args {
4299 char *fname;
4300 fhandle_t *fhp;
4301 };
4302 #endif
4303 int
4304 getfh(td, uap)
4305 struct thread *td;
4306 register struct getfh_args *uap;
4307 {
4308 struct nameidata nd;
4309 fhandle_t fh;
4310 register struct vnode *vp;
4311 int vfslocked;
4312 int error;
4313
4314 error = priv_check(td, PRIV_VFS_GETFH);
4315 if (error)
4316 return (error);
4317 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4318 UIO_USERSPACE, uap->fname, td);
4319 error = namei(&nd);
4320 if (error)
4321 return (error);
4322 vfslocked = NDHASGIANT(&nd);
4323 NDFREE(&nd, NDF_ONLY_PNBUF);
4324 vp = nd.ni_vp;
4325 bzero(&fh, sizeof(fh));
4326 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4327 error = VOP_VPTOFH(vp, &fh.fh_fid);
4328 vput(vp);
4329 VFS_UNLOCK_GIANT(vfslocked);
4330 if (error)
4331 return (error);
4332 error = copyout(&fh, uap->fhp, sizeof (fh));
4333 return (error);
4334 }
4335
4336 /*
4337 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4338 * open descriptor.
4339 *
4340 * warning: do not remove the priv_check() call or this becomes one giant
4341 * security hole.
4342 */
4343 #ifndef _SYS_SYSPROTO_H_
4344 struct fhopen_args {
4345 const struct fhandle *u_fhp;
4346 int flags;
4347 };
4348 #endif
4349 int
4350 fhopen(td, uap)
4351 struct thread *td;
4352 struct fhopen_args /* {
4353 const struct fhandle *u_fhp;
4354 int flags;
4355 } */ *uap;
4356 {
4357 struct proc *p = td->td_proc;
4358 struct mount *mp;
4359 struct vnode *vp;
4360 struct fhandle fhp;
4361 struct vattr vat;
4362 struct vattr *vap = &vat;
4363 struct flock lf;
4364 struct file *fp;
4365 register struct filedesc *fdp = p->p_fd;
4366 int fmode, error, type;
4367 accmode_t accmode;
4368 struct file *nfp;
4369 int vfslocked;
4370 int indx;
4371
4372 error = priv_check(td, PRIV_VFS_FHOPEN);
4373 if (error)
4374 return (error);
4375 fmode = FFLAGS(uap->flags);
4376 /* why not allow a non-read/write open for our lockd? */
4377 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4378 return (EINVAL);
4379 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4380 if (error)
4381 return(error);
4382 /* find the mount point */
4383 mp = vfs_busyfs(&fhp.fh_fsid);
4384 if (mp == NULL)
4385 return (ESTALE);
4386 vfslocked = VFS_LOCK_GIANT(mp);
4387 /* now give me my vnode, it gets returned to me locked */
4388 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
4389 vfs_unbusy(mp);
4390 if (error)
4391 goto out;
4392 /*
4393 * from now on we have to make sure not
4394 * to forget about the vnode
4395 * any error that causes an abort must vput(vp)
4396 * just set error = err and 'goto bad;'.
4397 */
4398
4399 /*
4400 * from vn_open
4401 */
4402 if (vp->v_type == VLNK) {
4403 error = EMLINK;
4404 goto bad;
4405 }
4406 if (vp->v_type == VSOCK) {
4407 error = EOPNOTSUPP;
4408 goto bad;
4409 }
4410 accmode = 0;
4411 if (fmode & (FWRITE | O_TRUNC)) {
4412 if (vp->v_type == VDIR) {
4413 error = EISDIR;
4414 goto bad;
4415 }
4416 error = vn_writechk(vp);
4417 if (error)
4418 goto bad;
4419 accmode |= VWRITE;
4420 }
4421 if (fmode & FREAD)
4422 accmode |= VREAD;
4423 if ((fmode & O_APPEND) && (fmode & FWRITE))
4424 accmode |= VAPPEND;
4425 #ifdef MAC
4426 error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4427 if (error)
4428 goto bad;
4429 #endif
4430 if (accmode) {
4431 error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4432 if (error)
4433 goto bad;
4434 }
4435 if (fmode & O_TRUNC) {
4436 vfs_ref(mp);
4437 VOP_UNLOCK(vp, 0); /* XXX */
4438 if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4439 vrele(vp);
4440 vfs_rel(mp);
4441 goto out;
4442 }
4443 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
4444 vfs_rel(mp);
4445 #ifdef MAC
4446 /*
4447 * We don't yet have fp->f_cred, so use td->td_ucred, which
4448 * should be right.
4449 */
4450 error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4451 if (error == 0) {
4452 #endif
4453 VATTR_NULL(vap);
4454 vap->va_size = 0;
4455 error = VOP_SETATTR(vp, vap, td->td_ucred);
4456 #ifdef MAC
4457 }
4458 #endif
4459 vn_finished_write(mp);
4460 if (error)
4461 goto bad;
4462 }
4463 error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4464 if (error)
4465 goto bad;
4466
4467 if (fmode & FWRITE)
4468 vp->v_writecount++;
4469
4470 /*
4471 * end of vn_open code
4472 */
4473
4474 if ((error = fallocf(td, &nfp, &indx, fmode)) != 0) {
4475 if (fmode & FWRITE)
4476 vp->v_writecount--;
4477 goto bad;
4478 }
4479 /* An extra reference on `nfp' has been held for us by falloc(). */
4480 fp = nfp;
4481 nfp->f_vnode = vp;
4482 finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4483 if (fmode & (O_EXLOCK | O_SHLOCK)) {
4484 lf.l_whence = SEEK_SET;
4485 lf.l_start = 0;
4486 lf.l_len = 0;
4487 if (fmode & O_EXLOCK)
4488 lf.l_type = F_WRLCK;
4489 else
4490 lf.l_type = F_RDLCK;
4491 type = F_FLOCK;
4492 if ((fmode & FNONBLOCK) == 0)
4493 type |= F_WAIT;
4494 VOP_UNLOCK(vp, 0);
4495 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4496 type)) != 0) {
4497 /*
4498 * The lock request failed. Normally close the
4499 * descriptor but handle the case where someone might
4500 * have dup()d or close()d it when we weren't looking.
4501 */
4502 fdclose(fdp, fp, indx, td);
4503
4504 /*
4505 * release our private reference
4506 */
4507 fdrop(fp, td);
4508 goto out;
4509 }
4510 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4511 atomic_set_int(&fp->f_flag, FHASLOCK);
4512 }
4513
4514 VOP_UNLOCK(vp, 0);
4515 fdrop(fp, td);
4516 VFS_UNLOCK_GIANT(vfslocked);
4517 td->td_retval[0] = indx;
4518 return (0);
4519
4520 bad:
4521 vput(vp);
4522 out:
4523 VFS_UNLOCK_GIANT(vfslocked);
4524 return (error);
4525 }
4526
4527 /*
4528 * Stat an (NFS) file handle.
4529 */
4530 #ifndef _SYS_SYSPROTO_H_
4531 struct fhstat_args {
4532 struct fhandle *u_fhp;
4533 struct stat *sb;
4534 };
4535 #endif
4536 int
4537 fhstat(td, uap)
4538 struct thread *td;
4539 register struct fhstat_args /* {
4540 struct fhandle *u_fhp;
4541 struct stat *sb;
4542 } */ *uap;
4543 {
4544 struct stat sb;
4545 fhandle_t fh;
4546 struct mount *mp;
4547 struct vnode *vp;
4548 int vfslocked;
4549 int error;
4550
4551 error = priv_check(td, PRIV_VFS_FHSTAT);
4552 if (error)
4553 return (error);
4554 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4555 if (error)
4556 return (error);
4557 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4558 return (ESTALE);
4559 vfslocked = VFS_LOCK_GIANT(mp);
4560 error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4561 vfs_unbusy(mp);
4562 if (error) {
4563 VFS_UNLOCK_GIANT(vfslocked);
4564 return (error);
4565 }
4566 error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4567 vput(vp);
4568 VFS_UNLOCK_GIANT(vfslocked);
4569 if (error)
4570 return (error);
4571 error = copyout(&sb, uap->sb, sizeof(sb));
4572 return (error);
4573 }
4574
4575 /*
4576 * Implement fstatfs() for (NFS) file handles.
4577 */
4578 #ifndef _SYS_SYSPROTO_H_
4579 struct fhstatfs_args {
4580 struct fhandle *u_fhp;
4581 struct statfs *buf;
4582 };
4583 #endif
4584 int
4585 fhstatfs(td, uap)
4586 struct thread *td;
4587 struct fhstatfs_args /* {
4588 struct fhandle *u_fhp;
4589 struct statfs *buf;
4590 } */ *uap;
4591 {
4592 struct statfs sf;
4593 fhandle_t fh;
4594 int error;
4595
4596 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4597 if (error)
4598 return (error);
4599 error = kern_fhstatfs(td, fh, &sf);
4600 if (error)
4601 return (error);
4602 return (copyout(&sf, uap->buf, sizeof(sf)));
4603 }
4604
4605 int
4606 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4607 {
4608 struct statfs *sp;
4609 struct mount *mp;
4610 struct vnode *vp;
4611 int vfslocked;
4612 int error;
4613
4614 error = priv_check(td, PRIV_VFS_FHSTATFS);
4615 if (error)
4616 return (error);
4617 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4618 return (ESTALE);
4619 vfslocked = VFS_LOCK_GIANT(mp);
4620 error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4621 if (error) {
4622 vfs_unbusy(mp);
4623 VFS_UNLOCK_GIANT(vfslocked);
4624 return (error);
4625 }
4626 vput(vp);
4627 error = prison_canseemount(td->td_ucred, mp);
4628 if (error)
4629 goto out;
4630 #ifdef MAC
4631 error = mac_mount_check_stat(td->td_ucred, mp);
4632 if (error)
4633 goto out;
4634 #endif
4635 /*
4636 * Set these in case the underlying filesystem fails to do so.
4637 */
4638 sp = &mp->mnt_stat;
4639 sp->f_version = STATFS_VERSION;
4640 sp->f_namemax = NAME_MAX;
4641 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4642 error = VFS_STATFS(mp, sp);
4643 if (error == 0)
4644 *buf = *sp;
4645 out:
4646 vfs_unbusy(mp);
4647 VFS_UNLOCK_GIANT(vfslocked);
4648 return (error);
4649 }
4650
4651 int
4652 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4653 {
4654 struct file *fp;
4655 struct mount *mp;
4656 struct vnode *vp;
4657 off_t olen, ooffset;
4658 int error, vfslocked;
4659
4660 fp = NULL;
4661 vfslocked = 0;
4662 error = fget(td, fd, &fp);
4663 if (error != 0)
4664 goto out;
4665
4666 switch (fp->f_type) {
4667 case DTYPE_VNODE:
4668 break;
4669 case DTYPE_PIPE:
4670 case DTYPE_FIFO:
4671 error = ESPIPE;
4672 goto out;
4673 default:
4674 error = ENODEV;
4675 goto out;
4676 }
4677 if ((fp->f_flag & FWRITE) == 0) {
4678 error = EBADF;
4679 goto out;
4680 }
4681 vp = fp->f_vnode;
4682 if (vp->v_type != VREG) {
4683 error = ENODEV;
4684 goto out;
4685 }
4686 if (offset < 0 || len <= 0) {
4687 error = EINVAL;
4688 goto out;
4689 }
4690 /* Check for wrap. */
4691 if (offset > OFF_MAX - len) {
4692 error = EFBIG;
4693 goto out;
4694 }
4695
4696 /* Allocating blocks may take a long time, so iterate. */
4697 for (;;) {
4698 olen = len;
4699 ooffset = offset;
4700
4701 bwillwrite();
4702 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4703 mp = NULL;
4704 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4705 if (error != 0) {
4706 VFS_UNLOCK_GIANT(vfslocked);
4707 break;
4708 }
4709 error = vn_lock(vp, LK_EXCLUSIVE);
4710 if (error != 0) {
4711 vn_finished_write(mp);
4712 VFS_UNLOCK_GIANT(vfslocked);
4713 break;
4714 }
4715 #ifdef MAC
4716 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4717 if (error == 0)
4718 #endif
4719 error = VOP_ALLOCATE(vp, &offset, &len);
4720 VOP_UNLOCK(vp, 0);
4721 vn_finished_write(mp);
4722 VFS_UNLOCK_GIANT(vfslocked);
4723
4724 if (olen + ooffset != offset + len) {
4725 panic("offset + len changed from %jx/%jx to %jx/%jx",
4726 ooffset, olen, offset, len);
4727 }
4728 if (error != 0 || len == 0)
4729 break;
4730 KASSERT(olen > len, ("Iteration did not make progress?"));
4731 maybe_yield();
4732 }
4733 out:
4734 if (fp != NULL)
4735 fdrop(fp, td);
4736 return (error);
4737 }
4738
4739 int
4740 posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4741 {
4742
4743 return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4744 }
4745
4746 /*
4747 * Unlike madvise(2), we do not make a best effort to remember every
4748 * possible caching hint. Instead, we remember the last setting with
4749 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4750 * region of any current setting.
4751 */
4752 int
4753 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4754 int advice)
4755 {
4756 struct fadvise_info *fa, *new;
4757 struct file *fp;
4758 struct vnode *vp;
4759 off_t end;
4760 int error;
4761
4762 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4763 return (EINVAL);
4764 switch (advice) {
4765 case POSIX_FADV_SEQUENTIAL:
4766 case POSIX_FADV_RANDOM:
4767 case POSIX_FADV_NOREUSE:
4768 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4769 break;
4770 case POSIX_FADV_NORMAL:
4771 case POSIX_FADV_WILLNEED:
4772 case POSIX_FADV_DONTNEED:
4773 new = NULL;
4774 break;
4775 default:
4776 return (EINVAL);
4777 }
4778 error = fget(td, fd, &fp);
4779 if (error != 0)
4780 goto out;
4781
4782 switch (fp->f_type) {
4783 case DTYPE_VNODE:
4784 break;
4785 case DTYPE_PIPE:
4786 case DTYPE_FIFO:
4787 error = ESPIPE;
4788 goto out;
4789 default:
4790 error = ENODEV;
4791 goto out;
4792 }
4793 vp = fp->f_vnode;
4794 if (vp->v_type != VREG) {
4795 error = ENODEV;
4796 goto out;
4797 }
4798 if (len == 0)
4799 end = OFF_MAX;
4800 else
4801 end = offset + len - 1;
4802 switch (advice) {
4803 case POSIX_FADV_SEQUENTIAL:
4804 case POSIX_FADV_RANDOM:
4805 case POSIX_FADV_NOREUSE:
4806 /*
4807 * Try to merge any existing non-standard region with
4808 * this new region if possible, otherwise create a new
4809 * non-standard region for this request.
4810 */
4811 mtx_pool_lock(mtxpool_sleep, fp);
4812 fa = fp->f_advice;
4813 if (fa != NULL && fa->fa_advice == advice &&
4814 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4815 (end != OFF_MAX && fa->fa_start == end + 1) ||
4816 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4817 if (offset < fa->fa_start)
4818 fa->fa_start = offset;
4819 if (end > fa->fa_end)
4820 fa->fa_end = end;
4821 } else {
4822 new->fa_advice = advice;
4823 new->fa_start = offset;
4824 new->fa_end = end;
4825 new->fa_prevstart = 0;
4826 new->fa_prevend = 0;
4827 fp->f_advice = new;
4828 new = fa;
4829 }
4830 mtx_pool_unlock(mtxpool_sleep, fp);
4831 break;
4832 case POSIX_FADV_NORMAL:
4833 /*
4834 * If a the "normal" region overlaps with an existing
4835 * non-standard region, trim or remove the
4836 * non-standard region.
4837 */
4838 mtx_pool_lock(mtxpool_sleep, fp);
4839 fa = fp->f_advice;
4840 if (fa != NULL) {
4841 if (offset <= fa->fa_start && end >= fa->fa_end) {
4842 new = fa;
4843 fp->f_advice = NULL;
4844 } else if (offset <= fa->fa_start &&
4845 end >= fa->fa_start)
4846 fa->fa_start = end + 1;
4847 else if (offset <= fa->fa_end && end >= fa->fa_end)
4848 fa->fa_end = offset - 1;
4849 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4850 /*
4851 * If the "normal" region is a middle
4852 * portion of the existing
4853 * non-standard region, just remove
4854 * the whole thing rather than picking
4855 * one side or the other to
4856 * preserve.
4857 */
4858 new = fa;
4859 fp->f_advice = NULL;
4860 }
4861 }
4862 mtx_pool_unlock(mtxpool_sleep, fp);
4863 break;
4864 case POSIX_FADV_WILLNEED:
4865 case POSIX_FADV_DONTNEED:
4866 error = VOP_ADVISE(vp, offset, end, advice);
4867 break;
4868 }
4869 out:
4870 if (fp != NULL)
4871 fdrop(fp, td);
4872 free(new, M_FADVISE);
4873 return (error);
4874 }
4875
4876 int
4877 posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4878 {
4879
4880 return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4881 uap->advice));
4882 }
Cache object: 273f6511e7be0ab91dd908a4959268a9
|