1 /*-
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
35 */
36
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_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/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78
79 #include <machine/stdarg.h>
80
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88
89 #include <ufs/ufs/quota.h>
90
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
96
97 static int kern_chflagsat(struct thread *td, int fd, const char *path,
98 enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102 struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104 const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106 struct thread *td);
107
108 /*
109 * Sync each mounted filesystem.
110 */
111 #ifndef _SYS_SYSPROTO_H_
112 struct sync_args {
113 int dummy;
114 };
115 #endif
116 /* ARGSUSED */
117 int
118 sys_sync(struct thread *td, struct sync_args *uap)
119 {
120 struct mount *mp, *nmp;
121 int save;
122
123 mtx_lock(&mountlist_mtx);
124 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
125 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
126 nmp = TAILQ_NEXT(mp, mnt_list);
127 continue;
128 }
129 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
130 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
131 save = curthread_pflags_set(TDP_SYNCIO);
132 vfs_msync(mp, MNT_NOWAIT);
133 VFS_SYNC(mp, MNT_NOWAIT);
134 curthread_pflags_restore(save);
135 vn_finished_write(mp);
136 }
137 mtx_lock(&mountlist_mtx);
138 nmp = TAILQ_NEXT(mp, mnt_list);
139 vfs_unbusy(mp);
140 }
141 mtx_unlock(&mountlist_mtx);
142 return (0);
143 }
144
145 /*
146 * Change filesystem quotas.
147 */
148 #ifndef _SYS_SYSPROTO_H_
149 struct quotactl_args {
150 char *path;
151 int cmd;
152 int uid;
153 caddr_t arg;
154 };
155 #endif
156 int
157 sys_quotactl(struct thread *td, struct quotactl_args *uap)
158 {
159 struct mount *mp;
160 struct nameidata nd;
161 int error;
162
163 AUDIT_ARG_CMD(uap->cmd);
164 AUDIT_ARG_UID(uap->uid);
165 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
166 return (EPERM);
167 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
168 uap->path, td);
169 if ((error = namei(&nd)) != 0)
170 return (error);
171 NDFREE(&nd, NDF_ONLY_PNBUF);
172 mp = nd.ni_vp->v_mount;
173 vfs_ref(mp);
174 vput(nd.ni_vp);
175 error = vfs_busy(mp, 0);
176 vfs_rel(mp);
177 if (error != 0)
178 return (error);
179 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
180
181 /*
182 * Since quota on operation typically needs to open quota
183 * file, the Q_QUOTAON handler needs to unbusy the mount point
184 * before calling into namei. Otherwise, unmount might be
185 * started between two vfs_busy() invocations (first is our,
186 * second is from mount point cross-walk code in lookup()),
187 * causing deadlock.
188 *
189 * Require that Q_QUOTAON handles the vfs_busy() reference on
190 * its own, always returning with ubusied mount point.
191 */
192 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
193 (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
194 vfs_unbusy(mp);
195 return (error);
196 }
197
198 /*
199 * Used by statfs conversion routines to scale the block size up if
200 * necessary so that all of the block counts are <= 'max_size'. Note
201 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
202 * value of 'n'.
203 */
204 void
205 statfs_scale_blocks(struct statfs *sf, long max_size)
206 {
207 uint64_t count;
208 int shift;
209
210 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
211
212 /*
213 * Attempt to scale the block counts to give a more accurate
214 * overview to userland of the ratio of free space to used
215 * space. To do this, find the largest block count and compute
216 * a divisor that lets it fit into a signed integer <= max_size.
217 */
218 if (sf->f_bavail < 0)
219 count = -sf->f_bavail;
220 else
221 count = sf->f_bavail;
222 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
223 if (count <= max_size)
224 return;
225
226 count >>= flsl(max_size);
227 shift = 0;
228 while (count > 0) {
229 shift++;
230 count >>=1;
231 }
232
233 sf->f_bsize <<= shift;
234 sf->f_blocks >>= shift;
235 sf->f_bfree >>= shift;
236 sf->f_bavail >>= shift;
237 }
238
239 static int
240 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
241 {
242 struct statfs *sp;
243 int error;
244
245 if (mp == NULL)
246 return (EBADF);
247 error = vfs_busy(mp, 0);
248 vfs_rel(mp);
249 if (error != 0)
250 return (error);
251 #ifdef MAC
252 error = mac_mount_check_stat(td->td_ucred, mp);
253 if (error != 0)
254 goto out;
255 #endif
256 /*
257 * Set these in case the underlying filesystem fails to do so.
258 */
259 sp = &mp->mnt_stat;
260 sp->f_version = STATFS_VERSION;
261 sp->f_namemax = NAME_MAX;
262 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
263 error = VFS_STATFS(mp, sp);
264 if (error != 0)
265 goto out;
266 *buf = *sp;
267 if (priv_check(td, PRIV_VFS_GENERATION)) {
268 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
269 prison_enforce_statfs(td->td_ucred, mp, buf);
270 }
271 out:
272 vfs_unbusy(mp);
273 return (error);
274 }
275
276 /*
277 * Get filesystem statistics.
278 */
279 #ifndef _SYS_SYSPROTO_H_
280 struct statfs_args {
281 char *path;
282 struct statfs *buf;
283 };
284 #endif
285 int
286 sys_statfs(struct thread *td, struct statfs_args *uap)
287 {
288 struct statfs *sfp;
289 int error;
290
291 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
292 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
293 if (error == 0)
294 error = copyout(sfp, uap->buf, sizeof(struct statfs));
295 free(sfp, M_STATFS);
296 return (error);
297 }
298
299 int
300 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
301 struct statfs *buf)
302 {
303 struct mount *mp;
304 struct nameidata nd;
305 int error;
306
307 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
308 pathseg, path, td);
309 error = namei(&nd);
310 if (error != 0)
311 return (error);
312 mp = nd.ni_vp->v_mount;
313 vfs_ref(mp);
314 NDFREE(&nd, NDF_ONLY_PNBUF);
315 vput(nd.ni_vp);
316 return (kern_do_statfs(td, mp, buf));
317 }
318
319 /*
320 * Get filesystem statistics.
321 */
322 #ifndef _SYS_SYSPROTO_H_
323 struct fstatfs_args {
324 int fd;
325 struct statfs *buf;
326 };
327 #endif
328 int
329 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
330 {
331 struct statfs *sfp;
332 int error;
333
334 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
335 error = kern_fstatfs(td, uap->fd, sfp);
336 if (error == 0)
337 error = copyout(sfp, uap->buf, sizeof(struct statfs));
338 free(sfp, M_STATFS);
339 return (error);
340 }
341
342 int
343 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
344 {
345 struct file *fp;
346 struct mount *mp;
347 struct vnode *vp;
348 cap_rights_t rights;
349 int error;
350
351 AUDIT_ARG_FD(fd);
352 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
353 if (error != 0)
354 return (error);
355 vp = fp->f_vnode;
356 vn_lock(vp, LK_SHARED | LK_RETRY);
357 #ifdef AUDIT
358 AUDIT_ARG_VNODE1(vp);
359 #endif
360 mp = vp->v_mount;
361 if (mp != NULL)
362 vfs_ref(mp);
363 VOP_UNLOCK(vp, 0);
364 fdrop(fp, td);
365 return (kern_do_statfs(td, mp, buf));
366 }
367
368 /*
369 * Get statistics on all filesystems.
370 */
371 #ifndef _SYS_SYSPROTO_H_
372 struct getfsstat_args {
373 struct statfs *buf;
374 long bufsize;
375 int mode;
376 };
377 #endif
378 int
379 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
380 {
381 size_t count;
382 int error;
383
384 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
385 return (EINVAL);
386 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
387 UIO_USERSPACE, uap->mode);
388 if (error == 0)
389 td->td_retval[0] = count;
390 return (error);
391 }
392
393 /*
394 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
395 * The caller is responsible for freeing memory which will be allocated
396 * in '*buf'.
397 */
398 int
399 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
400 size_t *countp, enum uio_seg bufseg, int mode)
401 {
402 struct mount *mp, *nmp;
403 struct statfs *sfsp, *sp, *sptmp, *tofree;
404 size_t count, maxcount;
405 int error;
406
407 switch (mode) {
408 case MNT_WAIT:
409 case MNT_NOWAIT:
410 break;
411 default:
412 if (bufseg == UIO_SYSSPACE)
413 *buf = NULL;
414 return (EINVAL);
415 }
416 restart:
417 maxcount = bufsize / sizeof(struct statfs);
418 if (bufsize == 0) {
419 sfsp = NULL;
420 tofree = NULL;
421 } else if (bufseg == UIO_USERSPACE) {
422 sfsp = *buf;
423 tofree = NULL;
424 } else /* if (bufseg == UIO_SYSSPACE) */ {
425 count = 0;
426 mtx_lock(&mountlist_mtx);
427 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
428 count++;
429 }
430 mtx_unlock(&mountlist_mtx);
431 if (maxcount > count)
432 maxcount = count;
433 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
434 M_STATFS, M_WAITOK);
435 }
436 count = 0;
437 mtx_lock(&mountlist_mtx);
438 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
439 if (prison_canseemount(td->td_ucred, mp) != 0) {
440 nmp = TAILQ_NEXT(mp, mnt_list);
441 continue;
442 }
443 #ifdef MAC
444 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
445 nmp = TAILQ_NEXT(mp, mnt_list);
446 continue;
447 }
448 #endif
449 if (mode == MNT_WAIT) {
450 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
451 /*
452 * If vfs_busy() failed, and MBF_NOWAIT
453 * wasn't passed, then the mp is gone.
454 * Furthermore, because of MBF_MNTLSTLOCK,
455 * the mountlist_mtx was dropped. We have
456 * no other choice than to start over.
457 */
458 mtx_unlock(&mountlist_mtx);
459 free(tofree, M_STATFS);
460 goto restart;
461 }
462 } else {
463 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
464 nmp = TAILQ_NEXT(mp, mnt_list);
465 continue;
466 }
467 }
468 if (sfsp != NULL && count < maxcount) {
469 sp = &mp->mnt_stat;
470 /*
471 * Set these in case the underlying filesystem
472 * fails to do so.
473 */
474 sp->f_version = STATFS_VERSION;
475 sp->f_namemax = NAME_MAX;
476 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
477 /*
478 * If MNT_NOWAIT is specified, do not refresh
479 * the fsstat cache.
480 */
481 if (mode != MNT_NOWAIT) {
482 error = VFS_STATFS(mp, sp);
483 if (error != 0) {
484 mtx_lock(&mountlist_mtx);
485 nmp = TAILQ_NEXT(mp, mnt_list);
486 vfs_unbusy(mp);
487 continue;
488 }
489 }
490 if (priv_check(td, PRIV_VFS_GENERATION)) {
491 sptmp = malloc(sizeof(struct statfs), M_STATFS,
492 M_WAITOK);
493 *sptmp = *sp;
494 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
495 prison_enforce_statfs(td->td_ucred, mp, sptmp);
496 sp = sptmp;
497 } else
498 sptmp = NULL;
499 if (bufseg == UIO_SYSSPACE) {
500 bcopy(sp, sfsp, sizeof(*sp));
501 free(sptmp, M_STATFS);
502 } else /* if (bufseg == UIO_USERSPACE) */ {
503 error = copyout(sp, sfsp, sizeof(*sp));
504 free(sptmp, M_STATFS);
505 if (error != 0) {
506 vfs_unbusy(mp);
507 return (error);
508 }
509 }
510 sfsp++;
511 }
512 count++;
513 mtx_lock(&mountlist_mtx);
514 nmp = TAILQ_NEXT(mp, mnt_list);
515 vfs_unbusy(mp);
516 }
517 mtx_unlock(&mountlist_mtx);
518 if (sfsp != NULL && count > maxcount)
519 *countp = maxcount;
520 else
521 *countp = count;
522 return (0);
523 }
524
525 #ifdef COMPAT_FREEBSD4
526 /*
527 * Get old format filesystem statistics.
528 */
529 static void cvtstatfs(struct statfs *, struct ostatfs *);
530
531 #ifndef _SYS_SYSPROTO_H_
532 struct freebsd4_statfs_args {
533 char *path;
534 struct ostatfs *buf;
535 };
536 #endif
537 int
538 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
539 {
540 struct ostatfs osb;
541 struct statfs *sfp;
542 int error;
543
544 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
545 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
546 if (error == 0) {
547 cvtstatfs(sfp, &osb);
548 error = copyout(&osb, uap->buf, sizeof(osb));
549 }
550 free(sfp, M_STATFS);
551 return (error);
552 }
553
554 /*
555 * Get filesystem statistics.
556 */
557 #ifndef _SYS_SYSPROTO_H_
558 struct freebsd4_fstatfs_args {
559 int fd;
560 struct ostatfs *buf;
561 };
562 #endif
563 int
564 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
565 {
566 struct ostatfs osb;
567 struct statfs *sfp;
568 int error;
569
570 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
571 error = kern_fstatfs(td, uap->fd, sfp);
572 if (error == 0) {
573 cvtstatfs(sfp, &osb);
574 error = copyout(&osb, uap->buf, sizeof(osb));
575 }
576 free(sfp, M_STATFS);
577 return (error);
578 }
579
580 /*
581 * Get statistics on all filesystems.
582 */
583 #ifndef _SYS_SYSPROTO_H_
584 struct freebsd4_getfsstat_args {
585 struct ostatfs *buf;
586 long bufsize;
587 int mode;
588 };
589 #endif
590 int
591 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
592 {
593 struct statfs *buf, *sp;
594 struct ostatfs osb;
595 size_t count, size;
596 int error;
597
598 if (uap->bufsize < 0)
599 return (EINVAL);
600 count = uap->bufsize / sizeof(struct ostatfs);
601 if (count > SIZE_MAX / sizeof(struct statfs))
602 return (EINVAL);
603 size = count * sizeof(struct statfs);
604 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
605 uap->mode);
606 if (buf == NULL)
607 return (EINVAL);
608 td->td_retval[0] = count;
609 if (size != 0) {
610 sp = buf;
611 while (count != 0 && error == 0) {
612 cvtstatfs(sp, &osb);
613 error = copyout(&osb, uap->buf, sizeof(osb));
614 sp++;
615 uap->buf++;
616 count--;
617 }
618 free(buf, M_STATFS);
619 }
620 return (error);
621 }
622
623 /*
624 * Implement fstatfs() for (NFS) file handles.
625 */
626 #ifndef _SYS_SYSPROTO_H_
627 struct freebsd4_fhstatfs_args {
628 struct fhandle *u_fhp;
629 struct ostatfs *buf;
630 };
631 #endif
632 int
633 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
634 {
635 struct ostatfs osb;
636 struct statfs *sfp;
637 fhandle_t fh;
638 int error;
639
640 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
641 if (error != 0)
642 return (error);
643 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
644 error = kern_fhstatfs(td, fh, sfp);
645 if (error == 0) {
646 cvtstatfs(sfp, &osb);
647 error = copyout(&osb, uap->buf, sizeof(osb));
648 }
649 free(sfp, M_STATFS);
650 return (error);
651 }
652
653 /*
654 * Convert a new format statfs structure to an old format statfs structure.
655 */
656 static void
657 cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
658 {
659
660 statfs_scale_blocks(nsp, LONG_MAX);
661 bzero(osp, sizeof(*osp));
662 osp->f_bsize = nsp->f_bsize;
663 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
664 osp->f_blocks = nsp->f_blocks;
665 osp->f_bfree = nsp->f_bfree;
666 osp->f_bavail = nsp->f_bavail;
667 osp->f_files = MIN(nsp->f_files, LONG_MAX);
668 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
669 osp->f_owner = nsp->f_owner;
670 osp->f_type = nsp->f_type;
671 osp->f_flags = nsp->f_flags;
672 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
673 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
674 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
675 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
676 strlcpy(osp->f_fstypename, nsp->f_fstypename,
677 MIN(MFSNAMELEN, OMFSNAMELEN));
678 strlcpy(osp->f_mntonname, nsp->f_mntonname,
679 MIN(MNAMELEN, OMNAMELEN));
680 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
681 MIN(MNAMELEN, OMNAMELEN));
682 osp->f_fsid = nsp->f_fsid;
683 }
684 #endif /* COMPAT_FREEBSD4 */
685
686 /*
687 * Change current working directory to a given file descriptor.
688 */
689 #ifndef _SYS_SYSPROTO_H_
690 struct fchdir_args {
691 int fd;
692 };
693 #endif
694 int
695 sys_fchdir(struct thread *td, struct fchdir_args *uap)
696 {
697 struct vnode *vp, *tdp;
698 struct mount *mp;
699 struct file *fp;
700 cap_rights_t rights;
701 int error;
702
703 AUDIT_ARG_FD(uap->fd);
704 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
705 &fp);
706 if (error != 0)
707 return (error);
708 vp = fp->f_vnode;
709 vrefact(vp);
710 fdrop(fp, td);
711 vn_lock(vp, LK_SHARED | LK_RETRY);
712 AUDIT_ARG_VNODE1(vp);
713 error = change_dir(vp, td);
714 while (!error && (mp = vp->v_mountedhere) != NULL) {
715 if (vfs_busy(mp, 0))
716 continue;
717 error = VFS_ROOT(mp, LK_SHARED, &tdp);
718 vfs_unbusy(mp);
719 if (error != 0)
720 break;
721 vput(vp);
722 vp = tdp;
723 }
724 if (error != 0) {
725 vput(vp);
726 return (error);
727 }
728 VOP_UNLOCK(vp, 0);
729 pwd_chdir(td, vp);
730 return (0);
731 }
732
733 /*
734 * Change current working directory (``.'').
735 */
736 #ifndef _SYS_SYSPROTO_H_
737 struct chdir_args {
738 char *path;
739 };
740 #endif
741 int
742 sys_chdir(struct thread *td, struct chdir_args *uap)
743 {
744
745 return (kern_chdir(td, uap->path, UIO_USERSPACE));
746 }
747
748 int
749 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
750 {
751 struct nameidata nd;
752 int error;
753
754 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
755 pathseg, path, td);
756 if ((error = namei(&nd)) != 0)
757 return (error);
758 if ((error = change_dir(nd.ni_vp, td)) != 0) {
759 vput(nd.ni_vp);
760 NDFREE(&nd, NDF_ONLY_PNBUF);
761 return (error);
762 }
763 VOP_UNLOCK(nd.ni_vp, 0);
764 NDFREE(&nd, NDF_ONLY_PNBUF);
765 pwd_chdir(td, nd.ni_vp);
766 return (0);
767 }
768
769 /*
770 * Change notion of root (``/'') directory.
771 */
772 #ifndef _SYS_SYSPROTO_H_
773 struct chroot_args {
774 char *path;
775 };
776 #endif
777 int
778 sys_chroot(struct thread *td, struct chroot_args *uap)
779 {
780 struct nameidata nd;
781 int error;
782
783 error = priv_check(td, PRIV_VFS_CHROOT);
784 if (error != 0)
785 return (error);
786 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
787 UIO_USERSPACE, uap->path, td);
788 error = namei(&nd);
789 if (error != 0)
790 goto error;
791 error = change_dir(nd.ni_vp, td);
792 if (error != 0)
793 goto e_vunlock;
794 #ifdef MAC
795 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
796 if (error != 0)
797 goto e_vunlock;
798 #endif
799 VOP_UNLOCK(nd.ni_vp, 0);
800 error = pwd_chroot(td, nd.ni_vp);
801 vrele(nd.ni_vp);
802 NDFREE(&nd, NDF_ONLY_PNBUF);
803 return (error);
804 e_vunlock:
805 vput(nd.ni_vp);
806 error:
807 NDFREE(&nd, NDF_ONLY_PNBUF);
808 return (error);
809 }
810
811 /*
812 * Common routine for chroot and chdir. Callers must provide a locked vnode
813 * instance.
814 */
815 int
816 change_dir(struct vnode *vp, struct thread *td)
817 {
818 #ifdef MAC
819 int error;
820 #endif
821
822 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
823 if (vp->v_type != VDIR)
824 return (ENOTDIR);
825 #ifdef MAC
826 error = mac_vnode_check_chdir(td->td_ucred, vp);
827 if (error != 0)
828 return (error);
829 #endif
830 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
831 }
832
833 static __inline void
834 flags_to_rights(int flags, cap_rights_t *rightsp)
835 {
836
837 if (flags & O_EXEC) {
838 cap_rights_set(rightsp, CAP_FEXECVE);
839 } else {
840 switch ((flags & O_ACCMODE)) {
841 case O_RDONLY:
842 cap_rights_set(rightsp, CAP_READ);
843 break;
844 case O_RDWR:
845 cap_rights_set(rightsp, CAP_READ);
846 /* FALLTHROUGH */
847 case O_WRONLY:
848 cap_rights_set(rightsp, CAP_WRITE);
849 if (!(flags & (O_APPEND | O_TRUNC)))
850 cap_rights_set(rightsp, CAP_SEEK);
851 break;
852 }
853 }
854
855 if (flags & O_CREAT)
856 cap_rights_set(rightsp, CAP_CREATE);
857
858 if (flags & O_TRUNC)
859 cap_rights_set(rightsp, CAP_FTRUNCATE);
860
861 if (flags & (O_SYNC | O_FSYNC))
862 cap_rights_set(rightsp, CAP_FSYNC);
863
864 if (flags & (O_EXLOCK | O_SHLOCK))
865 cap_rights_set(rightsp, CAP_FLOCK);
866 }
867
868 /*
869 * Check permissions, allocate an open file structure, and call the device
870 * open routine if any.
871 */
872 #ifndef _SYS_SYSPROTO_H_
873 struct open_args {
874 char *path;
875 int flags;
876 int mode;
877 };
878 #endif
879 int
880 sys_open(struct thread *td, struct open_args *uap)
881 {
882
883 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
884 uap->flags, uap->mode));
885 }
886
887 #ifndef _SYS_SYSPROTO_H_
888 struct openat_args {
889 int fd;
890 char *path;
891 int flag;
892 int mode;
893 };
894 #endif
895 int
896 sys_openat(struct thread *td, struct openat_args *uap)
897 {
898
899 AUDIT_ARG_FD(uap->fd);
900 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
901 uap->mode));
902 }
903
904 int
905 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
906 int flags, int mode)
907 {
908 struct proc *p = td->td_proc;
909 struct filedesc *fdp = p->p_fd;
910 struct file *fp;
911 struct vnode *vp;
912 struct nameidata nd;
913 cap_rights_t rights;
914 int cmode, error, indx;
915
916 indx = -1;
917
918 AUDIT_ARG_FFLAGS(flags);
919 AUDIT_ARG_MODE(mode);
920 cap_rights_init(&rights, CAP_LOOKUP);
921 flags_to_rights(flags, &rights);
922 /*
923 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
924 * may be specified.
925 */
926 if (flags & O_EXEC) {
927 if (flags & O_ACCMODE)
928 return (EINVAL);
929 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
930 return (EINVAL);
931 } else {
932 flags = FFLAGS(flags);
933 }
934
935 /*
936 * Allocate a file structure. The descriptor to reference it
937 * is allocated and set by finstall() below.
938 */
939 error = falloc_noinstall(td, &fp);
940 if (error != 0)
941 return (error);
942 /*
943 * An extra reference on `fp' has been held for us by
944 * falloc_noinstall().
945 */
946 /* Set the flags early so the finit in devfs can pick them up. */
947 fp->f_flag = flags & FMASK;
948 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
949 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
950 &rights, td);
951 td->td_dupfd = -1; /* XXX check for fdopen */
952 error = vn_open(&nd, &flags, cmode, fp);
953 if (error != 0) {
954 /*
955 * If the vn_open replaced the method vector, something
956 * wonderous happened deep below and we just pass it up
957 * pretending we know what we do.
958 */
959 if (error == ENXIO && fp->f_ops != &badfileops)
960 goto success;
961
962 /*
963 * Handle special fdopen() case. bleh.
964 *
965 * Don't do this for relative (capability) lookups; we don't
966 * understand exactly what would happen, and we don't think
967 * that it ever should.
968 */
969 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
970 (error == ENODEV || error == ENXIO) &&
971 td->td_dupfd >= 0) {
972 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
973 &indx);
974 if (error == 0)
975 goto success;
976 }
977
978 goto bad;
979 }
980 td->td_dupfd = 0;
981 NDFREE(&nd, NDF_ONLY_PNBUF);
982 vp = nd.ni_vp;
983
984 /*
985 * Store the vnode, for any f_type. Typically, the vnode use
986 * count is decremented by direct call to vn_closefile() for
987 * files that switched type in the cdevsw fdopen() method.
988 */
989 fp->f_vnode = vp;
990 /*
991 * If the file wasn't claimed by devfs bind it to the normal
992 * vnode operations here.
993 */
994 if (fp->f_ops == &badfileops) {
995 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
996 fp->f_seqcount = 1;
997 finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
998 DTYPE_VNODE, vp, &vnops);
999 }
1000
1001 VOP_UNLOCK(vp, 0);
1002 if (flags & O_TRUNC) {
1003 error = fo_truncate(fp, 0, td->td_ucred, td);
1004 if (error != 0)
1005 goto bad;
1006 }
1007 success:
1008 /*
1009 * If we haven't already installed the FD (for dupfdopen), do so now.
1010 */
1011 if (indx == -1) {
1012 struct filecaps *fcaps;
1013
1014 #ifdef CAPABILITIES
1015 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1016 fcaps = &nd.ni_filecaps;
1017 else
1018 #endif
1019 fcaps = NULL;
1020 error = finstall(td, fp, &indx, flags, fcaps);
1021 /* On success finstall() consumes fcaps. */
1022 if (error != 0) {
1023 filecaps_free(&nd.ni_filecaps);
1024 goto bad;
1025 }
1026 } else {
1027 filecaps_free(&nd.ni_filecaps);
1028 }
1029
1030 /*
1031 * Release our private reference, leaving the one associated with
1032 * the descriptor table intact.
1033 */
1034 fdrop(fp, td);
1035 td->td_retval[0] = indx;
1036 return (0);
1037 bad:
1038 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1039 fdrop(fp, td);
1040 return (error);
1041 }
1042
1043 #ifdef COMPAT_43
1044 /*
1045 * Create a file.
1046 */
1047 #ifndef _SYS_SYSPROTO_H_
1048 struct ocreat_args {
1049 char *path;
1050 int mode;
1051 };
1052 #endif
1053 int
1054 ocreat(struct thread *td, struct ocreat_args *uap)
1055 {
1056
1057 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1058 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1059 }
1060 #endif /* COMPAT_43 */
1061
1062 /*
1063 * Create a special file.
1064 */
1065 #ifndef _SYS_SYSPROTO_H_
1066 struct mknod_args {
1067 char *path;
1068 int mode;
1069 int dev;
1070 };
1071 #endif
1072 int
1073 sys_mknod(struct thread *td, struct mknod_args *uap)
1074 {
1075
1076 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1077 uap->mode, uap->dev));
1078 }
1079
1080 #ifndef _SYS_SYSPROTO_H_
1081 struct mknodat_args {
1082 int fd;
1083 char *path;
1084 mode_t mode;
1085 dev_t dev;
1086 };
1087 #endif
1088 int
1089 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1090 {
1091
1092 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1093 uap->dev));
1094 }
1095
1096 int
1097 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1098 int mode, int dev)
1099 {
1100 struct vnode *vp;
1101 struct mount *mp;
1102 struct vattr vattr;
1103 struct nameidata nd;
1104 cap_rights_t rights;
1105 int error, whiteout = 0;
1106
1107 AUDIT_ARG_MODE(mode);
1108 AUDIT_ARG_DEV(dev);
1109 switch (mode & S_IFMT) {
1110 case S_IFCHR:
1111 case S_IFBLK:
1112 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1113 if (error == 0 && dev == VNOVAL)
1114 error = EINVAL;
1115 break;
1116 case S_IFWHT:
1117 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1118 break;
1119 case S_IFIFO:
1120 if (dev == 0)
1121 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1122 /* FALLTHROUGH */
1123 default:
1124 error = EINVAL;
1125 break;
1126 }
1127 if (error != 0)
1128 return (error);
1129 restart:
1130 bwillwrite();
1131 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1132 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1133 td);
1134 if ((error = namei(&nd)) != 0)
1135 return (error);
1136 vp = nd.ni_vp;
1137 if (vp != NULL) {
1138 NDFREE(&nd, NDF_ONLY_PNBUF);
1139 if (vp == nd.ni_dvp)
1140 vrele(nd.ni_dvp);
1141 else
1142 vput(nd.ni_dvp);
1143 vrele(vp);
1144 return (EEXIST);
1145 } else {
1146 VATTR_NULL(&vattr);
1147 vattr.va_mode = (mode & ALLPERMS) &
1148 ~td->td_proc->p_fd->fd_cmask;
1149 vattr.va_rdev = dev;
1150 whiteout = 0;
1151
1152 switch (mode & S_IFMT) {
1153 case S_IFCHR:
1154 vattr.va_type = VCHR;
1155 break;
1156 case S_IFBLK:
1157 vattr.va_type = VBLK;
1158 break;
1159 case S_IFWHT:
1160 whiteout = 1;
1161 break;
1162 default:
1163 panic("kern_mknod: invalid mode");
1164 }
1165 }
1166 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1167 NDFREE(&nd, NDF_ONLY_PNBUF);
1168 vput(nd.ni_dvp);
1169 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1170 return (error);
1171 goto restart;
1172 }
1173 #ifdef MAC
1174 if (error == 0 && !whiteout)
1175 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1176 &nd.ni_cnd, &vattr);
1177 #endif
1178 if (error == 0) {
1179 if (whiteout)
1180 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1181 else {
1182 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1183 &nd.ni_cnd, &vattr);
1184 if (error == 0)
1185 vput(nd.ni_vp);
1186 }
1187 }
1188 NDFREE(&nd, NDF_ONLY_PNBUF);
1189 vput(nd.ni_dvp);
1190 vn_finished_write(mp);
1191 return (error);
1192 }
1193
1194 /*
1195 * Create a named pipe.
1196 */
1197 #ifndef _SYS_SYSPROTO_H_
1198 struct mkfifo_args {
1199 char *path;
1200 int mode;
1201 };
1202 #endif
1203 int
1204 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1205 {
1206
1207 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1208 uap->mode));
1209 }
1210
1211 #ifndef _SYS_SYSPROTO_H_
1212 struct mkfifoat_args {
1213 int fd;
1214 char *path;
1215 mode_t mode;
1216 };
1217 #endif
1218 int
1219 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1220 {
1221
1222 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1223 uap->mode));
1224 }
1225
1226 int
1227 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1228 int mode)
1229 {
1230 struct mount *mp;
1231 struct vattr vattr;
1232 struct nameidata nd;
1233 cap_rights_t rights;
1234 int error;
1235
1236 AUDIT_ARG_MODE(mode);
1237 restart:
1238 bwillwrite();
1239 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1240 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1241 td);
1242 if ((error = namei(&nd)) != 0)
1243 return (error);
1244 if (nd.ni_vp != NULL) {
1245 NDFREE(&nd, NDF_ONLY_PNBUF);
1246 if (nd.ni_vp == nd.ni_dvp)
1247 vrele(nd.ni_dvp);
1248 else
1249 vput(nd.ni_dvp);
1250 vrele(nd.ni_vp);
1251 return (EEXIST);
1252 }
1253 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1254 NDFREE(&nd, NDF_ONLY_PNBUF);
1255 vput(nd.ni_dvp);
1256 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1257 return (error);
1258 goto restart;
1259 }
1260 VATTR_NULL(&vattr);
1261 vattr.va_type = VFIFO;
1262 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1263 #ifdef MAC
1264 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1265 &vattr);
1266 if (error != 0)
1267 goto out;
1268 #endif
1269 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1270 if (error == 0)
1271 vput(nd.ni_vp);
1272 #ifdef MAC
1273 out:
1274 #endif
1275 vput(nd.ni_dvp);
1276 vn_finished_write(mp);
1277 NDFREE(&nd, NDF_ONLY_PNBUF);
1278 return (error);
1279 }
1280
1281 /*
1282 * Make a hard file link.
1283 */
1284 #ifndef _SYS_SYSPROTO_H_
1285 struct link_args {
1286 char *path;
1287 char *link;
1288 };
1289 #endif
1290 int
1291 sys_link(struct thread *td, struct link_args *uap)
1292 {
1293
1294 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1295 UIO_USERSPACE, FOLLOW));
1296 }
1297
1298 #ifndef _SYS_SYSPROTO_H_
1299 struct linkat_args {
1300 int fd1;
1301 char *path1;
1302 int fd2;
1303 char *path2;
1304 int flag;
1305 };
1306 #endif
1307 int
1308 sys_linkat(struct thread *td, struct linkat_args *uap)
1309 {
1310 int flag;
1311
1312 flag = uap->flag;
1313 if (flag & ~AT_SYMLINK_FOLLOW)
1314 return (EINVAL);
1315
1316 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1317 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1318 }
1319
1320 int hardlink_check_uid = 0;
1321 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1322 &hardlink_check_uid, 0,
1323 "Unprivileged processes cannot create hard links to files owned by other "
1324 "users");
1325 static int hardlink_check_gid = 0;
1326 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1327 &hardlink_check_gid, 0,
1328 "Unprivileged processes cannot create hard links to files owned by other "
1329 "groups");
1330
1331 static int
1332 can_hardlink(struct vnode *vp, struct ucred *cred)
1333 {
1334 struct vattr va;
1335 int error;
1336
1337 if (!hardlink_check_uid && !hardlink_check_gid)
1338 return (0);
1339
1340 error = VOP_GETATTR(vp, &va, cred);
1341 if (error != 0)
1342 return (error);
1343
1344 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1345 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1346 if (error != 0)
1347 return (error);
1348 }
1349
1350 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1351 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1352 if (error != 0)
1353 return (error);
1354 }
1355
1356 return (0);
1357 }
1358
1359 int
1360 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1361 enum uio_seg segflg, int follow)
1362 {
1363 struct vnode *vp;
1364 struct mount *mp;
1365 struct nameidata nd;
1366 cap_rights_t rights;
1367 int error;
1368
1369 again:
1370 bwillwrite();
1371 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1372 cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1373
1374 if ((error = namei(&nd)) != 0)
1375 return (error);
1376 NDFREE(&nd, NDF_ONLY_PNBUF);
1377 vp = nd.ni_vp;
1378 if (vp->v_type == VDIR) {
1379 vrele(vp);
1380 return (EPERM); /* POSIX */
1381 }
1382 NDINIT_ATRIGHTS(&nd, CREATE,
1383 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1384 cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1385 if ((error = namei(&nd)) == 0) {
1386 if (nd.ni_vp != NULL) {
1387 NDFREE(&nd, NDF_ONLY_PNBUF);
1388 if (nd.ni_dvp == nd.ni_vp)
1389 vrele(nd.ni_dvp);
1390 else
1391 vput(nd.ni_dvp);
1392 vrele(nd.ni_vp);
1393 vrele(vp);
1394 return (EEXIST);
1395 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1396 /*
1397 * Cross-device link. No need to recheck
1398 * vp->v_type, since it cannot change, except
1399 * to VBAD.
1400 */
1401 NDFREE(&nd, NDF_ONLY_PNBUF);
1402 vput(nd.ni_dvp);
1403 vrele(vp);
1404 return (EXDEV);
1405 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1406 error = can_hardlink(vp, td->td_ucred);
1407 #ifdef MAC
1408 if (error == 0)
1409 error = mac_vnode_check_link(td->td_ucred,
1410 nd.ni_dvp, vp, &nd.ni_cnd);
1411 #endif
1412 if (error != 0) {
1413 vput(vp);
1414 vput(nd.ni_dvp);
1415 NDFREE(&nd, NDF_ONLY_PNBUF);
1416 return (error);
1417 }
1418 error = vn_start_write(vp, &mp, V_NOWAIT);
1419 if (error != 0) {
1420 vput(vp);
1421 vput(nd.ni_dvp);
1422 NDFREE(&nd, NDF_ONLY_PNBUF);
1423 error = vn_start_write(NULL, &mp,
1424 V_XSLEEP | PCATCH);
1425 if (error != 0)
1426 return (error);
1427 goto again;
1428 }
1429 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1430 VOP_UNLOCK(vp, 0);
1431 vput(nd.ni_dvp);
1432 vn_finished_write(mp);
1433 NDFREE(&nd, NDF_ONLY_PNBUF);
1434 } else {
1435 vput(nd.ni_dvp);
1436 NDFREE(&nd, NDF_ONLY_PNBUF);
1437 vrele(vp);
1438 goto again;
1439 }
1440 }
1441 vrele(vp);
1442 return (error);
1443 }
1444
1445 /*
1446 * Make a symbolic link.
1447 */
1448 #ifndef _SYS_SYSPROTO_H_
1449 struct symlink_args {
1450 char *path;
1451 char *link;
1452 };
1453 #endif
1454 int
1455 sys_symlink(struct thread *td, struct symlink_args *uap)
1456 {
1457
1458 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1459 UIO_USERSPACE));
1460 }
1461
1462 #ifndef _SYS_SYSPROTO_H_
1463 struct symlinkat_args {
1464 char *path;
1465 int fd;
1466 char *path2;
1467 };
1468 #endif
1469 int
1470 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1471 {
1472
1473 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1474 UIO_USERSPACE));
1475 }
1476
1477 int
1478 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1479 enum uio_seg segflg)
1480 {
1481 struct mount *mp;
1482 struct vattr vattr;
1483 char *syspath;
1484 struct nameidata nd;
1485 int error;
1486 cap_rights_t rights;
1487
1488 if (segflg == UIO_SYSSPACE) {
1489 syspath = path1;
1490 } else {
1491 syspath = uma_zalloc(namei_zone, M_WAITOK);
1492 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1493 goto out;
1494 }
1495 AUDIT_ARG_TEXT(syspath);
1496 restart:
1497 bwillwrite();
1498 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1499 NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1500 td);
1501 if ((error = namei(&nd)) != 0)
1502 goto out;
1503 if (nd.ni_vp) {
1504 NDFREE(&nd, NDF_ONLY_PNBUF);
1505 if (nd.ni_vp == nd.ni_dvp)
1506 vrele(nd.ni_dvp);
1507 else
1508 vput(nd.ni_dvp);
1509 vrele(nd.ni_vp);
1510 error = EEXIST;
1511 goto out;
1512 }
1513 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1514 NDFREE(&nd, NDF_ONLY_PNBUF);
1515 vput(nd.ni_dvp);
1516 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1517 goto out;
1518 goto restart;
1519 }
1520 VATTR_NULL(&vattr);
1521 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1522 #ifdef MAC
1523 vattr.va_type = VLNK;
1524 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1525 &vattr);
1526 if (error != 0)
1527 goto out2;
1528 #endif
1529 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1530 if (error == 0)
1531 vput(nd.ni_vp);
1532 #ifdef MAC
1533 out2:
1534 #endif
1535 NDFREE(&nd, NDF_ONLY_PNBUF);
1536 vput(nd.ni_dvp);
1537 vn_finished_write(mp);
1538 out:
1539 if (segflg != UIO_SYSSPACE)
1540 uma_zfree(namei_zone, syspath);
1541 return (error);
1542 }
1543
1544 /*
1545 * Delete a whiteout from the filesystem.
1546 */
1547 #ifndef _SYS_SYSPROTO_H_
1548 struct undelete_args {
1549 char *path;
1550 };
1551 #endif
1552 int
1553 sys_undelete(struct thread *td, struct undelete_args *uap)
1554 {
1555 struct mount *mp;
1556 struct nameidata nd;
1557 int error;
1558
1559 restart:
1560 bwillwrite();
1561 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1562 UIO_USERSPACE, uap->path, td);
1563 error = namei(&nd);
1564 if (error != 0)
1565 return (error);
1566
1567 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1568 NDFREE(&nd, NDF_ONLY_PNBUF);
1569 if (nd.ni_vp == nd.ni_dvp)
1570 vrele(nd.ni_dvp);
1571 else
1572 vput(nd.ni_dvp);
1573 if (nd.ni_vp)
1574 vrele(nd.ni_vp);
1575 return (EEXIST);
1576 }
1577 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1578 NDFREE(&nd, NDF_ONLY_PNBUF);
1579 vput(nd.ni_dvp);
1580 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1581 return (error);
1582 goto restart;
1583 }
1584 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1585 NDFREE(&nd, NDF_ONLY_PNBUF);
1586 vput(nd.ni_dvp);
1587 vn_finished_write(mp);
1588 return (error);
1589 }
1590
1591 /*
1592 * Delete a name from the filesystem.
1593 */
1594 #ifndef _SYS_SYSPROTO_H_
1595 struct unlink_args {
1596 char *path;
1597 };
1598 #endif
1599 int
1600 sys_unlink(struct thread *td, struct unlink_args *uap)
1601 {
1602
1603 return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1604 }
1605
1606 #ifndef _SYS_SYSPROTO_H_
1607 struct unlinkat_args {
1608 int fd;
1609 char *path;
1610 int flag;
1611 };
1612 #endif
1613 int
1614 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1615 {
1616 int flag = uap->flag;
1617 int fd = uap->fd;
1618 char *path = uap->path;
1619
1620 if (flag & ~AT_REMOVEDIR)
1621 return (EINVAL);
1622
1623 if (flag & AT_REMOVEDIR)
1624 return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1625 else
1626 return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1627 }
1628
1629 int
1630 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1631 ino_t oldinum)
1632 {
1633 struct mount *mp;
1634 struct vnode *vp;
1635 struct nameidata nd;
1636 struct stat sb;
1637 cap_rights_t rights;
1638 int error;
1639
1640 restart:
1641 bwillwrite();
1642 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1643 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1644 if ((error = namei(&nd)) != 0)
1645 return (error == EINVAL ? EPERM : error);
1646 vp = nd.ni_vp;
1647 if (vp->v_type == VDIR && oldinum == 0) {
1648 error = EPERM; /* POSIX */
1649 } else if (oldinum != 0 &&
1650 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1651 sb.st_ino != oldinum) {
1652 error = EIDRM; /* Identifier removed */
1653 } else {
1654 /*
1655 * The root of a mounted filesystem cannot be deleted.
1656 *
1657 * XXX: can this only be a VDIR case?
1658 */
1659 if (vp->v_vflag & VV_ROOT)
1660 error = EBUSY;
1661 }
1662 if (error == 0) {
1663 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1664 NDFREE(&nd, NDF_ONLY_PNBUF);
1665 vput(nd.ni_dvp);
1666 if (vp == nd.ni_dvp)
1667 vrele(vp);
1668 else
1669 vput(vp);
1670 if ((error = vn_start_write(NULL, &mp,
1671 V_XSLEEP | PCATCH)) != 0)
1672 return (error);
1673 goto restart;
1674 }
1675 #ifdef MAC
1676 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1677 &nd.ni_cnd);
1678 if (error != 0)
1679 goto out;
1680 #endif
1681 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1682 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1683 #ifdef MAC
1684 out:
1685 #endif
1686 vn_finished_write(mp);
1687 }
1688 NDFREE(&nd, NDF_ONLY_PNBUF);
1689 vput(nd.ni_dvp);
1690 if (vp == nd.ni_dvp)
1691 vrele(vp);
1692 else
1693 vput(vp);
1694 return (error);
1695 }
1696
1697 /*
1698 * Reposition read/write file offset.
1699 */
1700 #ifndef _SYS_SYSPROTO_H_
1701 struct lseek_args {
1702 int fd;
1703 int pad;
1704 off_t offset;
1705 int whence;
1706 };
1707 #endif
1708 int
1709 sys_lseek(struct thread *td, struct lseek_args *uap)
1710 {
1711
1712 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1713 }
1714
1715 int
1716 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1717 {
1718 struct file *fp;
1719 cap_rights_t rights;
1720 int error;
1721
1722 AUDIT_ARG_FD(fd);
1723 error = fget(td, fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1724 if (error != 0)
1725 return (error);
1726 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1727 fo_seek(fp, offset, whence, td) : ESPIPE;
1728 fdrop(fp, td);
1729 return (error);
1730 }
1731
1732 #if defined(COMPAT_43)
1733 /*
1734 * Reposition read/write file offset.
1735 */
1736 #ifndef _SYS_SYSPROTO_H_
1737 struct olseek_args {
1738 int fd;
1739 long offset;
1740 int whence;
1741 };
1742 #endif
1743 int
1744 olseek(struct thread *td, struct olseek_args *uap)
1745 {
1746
1747 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1748 }
1749 #endif /* COMPAT_43 */
1750
1751 #if defined(COMPAT_FREEBSD6)
1752 /* Version with the 'pad' argument */
1753 int
1754 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1755 {
1756
1757 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1758 }
1759 #endif
1760
1761 /*
1762 * Check access permissions using passed credentials.
1763 */
1764 static int
1765 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
1766 struct thread *td)
1767 {
1768 accmode_t accmode;
1769 int error;
1770
1771 /* Flags == 0 means only check for existence. */
1772 if (user_flags == 0)
1773 return (0);
1774
1775 accmode = 0;
1776 if (user_flags & R_OK)
1777 accmode |= VREAD;
1778 if (user_flags & W_OK)
1779 accmode |= VWRITE;
1780 if (user_flags & X_OK)
1781 accmode |= VEXEC;
1782 #ifdef MAC
1783 error = mac_vnode_check_access(cred, vp, accmode);
1784 if (error != 0)
1785 return (error);
1786 #endif
1787 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1788 error = VOP_ACCESS(vp, accmode, cred, td);
1789 return (error);
1790 }
1791
1792 /*
1793 * Check access permissions using "real" credentials.
1794 */
1795 #ifndef _SYS_SYSPROTO_H_
1796 struct access_args {
1797 char *path;
1798 int amode;
1799 };
1800 #endif
1801 int
1802 sys_access(struct thread *td, struct access_args *uap)
1803 {
1804
1805 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1806 0, uap->amode));
1807 }
1808
1809 #ifndef _SYS_SYSPROTO_H_
1810 struct faccessat_args {
1811 int dirfd;
1812 char *path;
1813 int amode;
1814 int flag;
1815 }
1816 #endif
1817 int
1818 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1819 {
1820
1821 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1822 uap->amode));
1823 }
1824
1825 int
1826 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1827 int flag, int amode)
1828 {
1829 struct ucred *cred, *usecred;
1830 struct vnode *vp;
1831 struct nameidata nd;
1832 cap_rights_t rights;
1833 int error;
1834
1835 if (flag & ~AT_EACCESS)
1836 return (EINVAL);
1837 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1838 return (EINVAL);
1839
1840 /*
1841 * Create and modify a temporary credential instead of one that
1842 * is potentially shared (if we need one).
1843 */
1844 cred = td->td_ucred;
1845 if ((flag & AT_EACCESS) == 0 &&
1846 ((cred->cr_uid != cred->cr_ruid ||
1847 cred->cr_rgid != cred->cr_groups[0]))) {
1848 usecred = crdup(cred);
1849 usecred->cr_uid = cred->cr_ruid;
1850 usecred->cr_groups[0] = cred->cr_rgid;
1851 td->td_ucred = usecred;
1852 } else
1853 usecred = cred;
1854 AUDIT_ARG_VALUE(amode);
1855 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1856 AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1857 td);
1858 if ((error = namei(&nd)) != 0)
1859 goto out;
1860 vp = nd.ni_vp;
1861
1862 error = vn_access(vp, amode, usecred, td);
1863 NDFREE(&nd, NDF_ONLY_PNBUF);
1864 vput(vp);
1865 out:
1866 if (usecred != cred) {
1867 td->td_ucred = cred;
1868 crfree(usecred);
1869 }
1870 return (error);
1871 }
1872
1873 /*
1874 * Check access permissions using "effective" credentials.
1875 */
1876 #ifndef _SYS_SYSPROTO_H_
1877 struct eaccess_args {
1878 char *path;
1879 int amode;
1880 };
1881 #endif
1882 int
1883 sys_eaccess(struct thread *td, struct eaccess_args *uap)
1884 {
1885
1886 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1887 AT_EACCESS, uap->amode));
1888 }
1889
1890 #if defined(COMPAT_43)
1891 /*
1892 * Get file status; this version follows links.
1893 */
1894 #ifndef _SYS_SYSPROTO_H_
1895 struct ostat_args {
1896 char *path;
1897 struct ostat *ub;
1898 };
1899 #endif
1900 int
1901 ostat(struct thread *td, struct ostat_args *uap)
1902 {
1903 struct stat sb;
1904 struct ostat osb;
1905 int error;
1906
1907 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
1908 &sb, NULL);
1909 if (error != 0)
1910 return (error);
1911 cvtstat(&sb, &osb);
1912 return (copyout(&osb, uap->ub, sizeof (osb)));
1913 }
1914
1915 /*
1916 * Get file status; this version does not follow links.
1917 */
1918 #ifndef _SYS_SYSPROTO_H_
1919 struct olstat_args {
1920 char *path;
1921 struct ostat *ub;
1922 };
1923 #endif
1924 int
1925 olstat(struct thread *td, struct olstat_args *uap)
1926 {
1927 struct stat sb;
1928 struct ostat osb;
1929 int error;
1930
1931 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
1932 UIO_USERSPACE, &sb, NULL);
1933 if (error != 0)
1934 return (error);
1935 cvtstat(&sb, &osb);
1936 return (copyout(&osb, uap->ub, sizeof (osb)));
1937 }
1938
1939 /*
1940 * Convert from an old to a new stat structure.
1941 */
1942 void
1943 cvtstat(struct stat *st, struct ostat *ost)
1944 {
1945
1946 bzero(ost, sizeof(*ost));
1947 ost->st_dev = st->st_dev;
1948 ost->st_ino = st->st_ino;
1949 ost->st_mode = st->st_mode;
1950 ost->st_nlink = st->st_nlink;
1951 ost->st_uid = st->st_uid;
1952 ost->st_gid = st->st_gid;
1953 ost->st_rdev = st->st_rdev;
1954 if (st->st_size < (quad_t)1 << 32)
1955 ost->st_size = st->st_size;
1956 else
1957 ost->st_size = -2;
1958 ost->st_atim = st->st_atim;
1959 ost->st_mtim = st->st_mtim;
1960 ost->st_ctim = st->st_ctim;
1961 ost->st_blksize = st->st_blksize;
1962 ost->st_blocks = st->st_blocks;
1963 ost->st_flags = st->st_flags;
1964 ost->st_gen = st->st_gen;
1965 }
1966 #endif /* COMPAT_43 */
1967
1968 /*
1969 * Get file status; this version follows links.
1970 */
1971 #ifndef _SYS_SYSPROTO_H_
1972 struct stat_args {
1973 char *path;
1974 struct stat *ub;
1975 };
1976 #endif
1977 int
1978 sys_stat(struct thread *td, struct stat_args *uap)
1979 {
1980 struct stat sb;
1981 int error;
1982
1983 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
1984 &sb, NULL);
1985 if (error == 0)
1986 error = copyout(&sb, uap->ub, sizeof (sb));
1987 return (error);
1988 }
1989
1990 #ifndef _SYS_SYSPROTO_H_
1991 struct fstatat_args {
1992 int fd;
1993 char *path;
1994 struct stat *buf;
1995 int flag;
1996 }
1997 #endif
1998 int
1999 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2000 {
2001 struct stat sb;
2002 int error;
2003
2004 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2005 UIO_USERSPACE, &sb, NULL);
2006 if (error == 0)
2007 error = copyout(&sb, uap->buf, sizeof (sb));
2008 return (error);
2009 }
2010
2011 int
2012 kern_statat(struct thread *td, int flag, int fd, char *path,
2013 enum uio_seg pathseg, struct stat *sbp,
2014 void (*hook)(struct vnode *vp, struct stat *sbp))
2015 {
2016 struct nameidata nd;
2017 struct stat sb;
2018 cap_rights_t rights;
2019 int error;
2020
2021 if (flag & ~AT_SYMLINK_NOFOLLOW)
2022 return (EINVAL);
2023
2024 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2025 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2026 cap_rights_init(&rights, CAP_FSTAT), td);
2027
2028 if ((error = namei(&nd)) != 0)
2029 return (error);
2030 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2031 if (error == 0) {
2032 SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2033 if (S_ISREG(sb.st_mode))
2034 SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2035 if (__predict_false(hook != NULL))
2036 hook(nd.ni_vp, &sb);
2037 }
2038 NDFREE(&nd, NDF_ONLY_PNBUF);
2039 vput(nd.ni_vp);
2040 if (error != 0)
2041 return (error);
2042 *sbp = sb;
2043 #ifdef KTRACE
2044 if (KTRPOINT(td, KTR_STRUCT))
2045 ktrstat(&sb);
2046 #endif
2047 return (0);
2048 }
2049
2050 /*
2051 * Get file status; this version does not follow links.
2052 */
2053 #ifndef _SYS_SYSPROTO_H_
2054 struct lstat_args {
2055 char *path;
2056 struct stat *ub;
2057 };
2058 #endif
2059 int
2060 sys_lstat(struct thread *td, struct lstat_args *uap)
2061 {
2062 struct stat sb;
2063 int error;
2064
2065 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2066 UIO_USERSPACE, &sb, NULL);
2067 if (error == 0)
2068 error = copyout(&sb, uap->ub, sizeof (sb));
2069 return (error);
2070 }
2071
2072 /*
2073 * Implementation of the NetBSD [l]stat() functions.
2074 */
2075 void
2076 cvtnstat( struct stat *sb, struct nstat *nsb)
2077 {
2078
2079 bzero(nsb, sizeof *nsb);
2080 nsb->st_dev = sb->st_dev;
2081 nsb->st_ino = sb->st_ino;
2082 nsb->st_mode = sb->st_mode;
2083 nsb->st_nlink = sb->st_nlink;
2084 nsb->st_uid = sb->st_uid;
2085 nsb->st_gid = sb->st_gid;
2086 nsb->st_rdev = sb->st_rdev;
2087 nsb->st_atim = sb->st_atim;
2088 nsb->st_mtim = sb->st_mtim;
2089 nsb->st_ctim = sb->st_ctim;
2090 nsb->st_size = sb->st_size;
2091 nsb->st_blocks = sb->st_blocks;
2092 nsb->st_blksize = sb->st_blksize;
2093 nsb->st_flags = sb->st_flags;
2094 nsb->st_gen = sb->st_gen;
2095 nsb->st_birthtim = sb->st_birthtim;
2096 }
2097
2098 #ifndef _SYS_SYSPROTO_H_
2099 struct nstat_args {
2100 char *path;
2101 struct nstat *ub;
2102 };
2103 #endif
2104 int
2105 sys_nstat(struct thread *td, struct nstat_args *uap)
2106 {
2107 struct stat sb;
2108 struct nstat nsb;
2109 int error;
2110
2111 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2112 &sb, NULL);
2113 if (error != 0)
2114 return (error);
2115 cvtnstat(&sb, &nsb);
2116 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2117 }
2118
2119 /*
2120 * NetBSD lstat. Get file status; this version does not follow links.
2121 */
2122 #ifndef _SYS_SYSPROTO_H_
2123 struct lstat_args {
2124 char *path;
2125 struct stat *ub;
2126 };
2127 #endif
2128 int
2129 sys_nlstat(struct thread *td, struct nlstat_args *uap)
2130 {
2131 struct stat sb;
2132 struct nstat nsb;
2133 int error;
2134
2135 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2136 UIO_USERSPACE, &sb, NULL);
2137 if (error != 0)
2138 return (error);
2139 cvtnstat(&sb, &nsb);
2140 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2141 }
2142
2143 /*
2144 * Get configurable pathname variables.
2145 */
2146 #ifndef _SYS_SYSPROTO_H_
2147 struct pathconf_args {
2148 char *path;
2149 int name;
2150 };
2151 #endif
2152 int
2153 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2154 {
2155
2156 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2157 }
2158
2159 #ifndef _SYS_SYSPROTO_H_
2160 struct lpathconf_args {
2161 char *path;
2162 int name;
2163 };
2164 #endif
2165 int
2166 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2167 {
2168
2169 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2170 NOFOLLOW));
2171 }
2172
2173 int
2174 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2175 u_long flags)
2176 {
2177 struct nameidata nd;
2178 int error;
2179
2180 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2181 pathseg, path, td);
2182 if ((error = namei(&nd)) != 0)
2183 return (error);
2184 NDFREE(&nd, NDF_ONLY_PNBUF);
2185
2186 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2187 vput(nd.ni_vp);
2188 return (error);
2189 }
2190
2191 /*
2192 * Return target name of a symbolic link.
2193 */
2194 #ifndef _SYS_SYSPROTO_H_
2195 struct readlink_args {
2196 char *path;
2197 char *buf;
2198 size_t count;
2199 };
2200 #endif
2201 int
2202 sys_readlink(struct thread *td, struct readlink_args *uap)
2203 {
2204
2205 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2206 uap->buf, UIO_USERSPACE, uap->count));
2207 }
2208 #ifndef _SYS_SYSPROTO_H_
2209 struct readlinkat_args {
2210 int fd;
2211 char *path;
2212 char *buf;
2213 size_t bufsize;
2214 };
2215 #endif
2216 int
2217 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2218 {
2219
2220 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2221 uap->buf, UIO_USERSPACE, uap->bufsize));
2222 }
2223
2224 int
2225 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2226 char *buf, enum uio_seg bufseg, size_t count)
2227 {
2228 struct vnode *vp;
2229 struct iovec aiov;
2230 struct uio auio;
2231 struct nameidata nd;
2232 int error;
2233
2234 if (count > IOSIZE_MAX)
2235 return (EINVAL);
2236
2237 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2238 pathseg, path, fd, td);
2239
2240 if ((error = namei(&nd)) != 0)
2241 return (error);
2242 NDFREE(&nd, NDF_ONLY_PNBUF);
2243 vp = nd.ni_vp;
2244 #ifdef MAC
2245 error = mac_vnode_check_readlink(td->td_ucred, vp);
2246 if (error != 0) {
2247 vput(vp);
2248 return (error);
2249 }
2250 #endif
2251 if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2252 error = EINVAL;
2253 else {
2254 aiov.iov_base = buf;
2255 aiov.iov_len = count;
2256 auio.uio_iov = &aiov;
2257 auio.uio_iovcnt = 1;
2258 auio.uio_offset = 0;
2259 auio.uio_rw = UIO_READ;
2260 auio.uio_segflg = bufseg;
2261 auio.uio_td = td;
2262 auio.uio_resid = count;
2263 error = VOP_READLINK(vp, &auio, td->td_ucred);
2264 td->td_retval[0] = count - auio.uio_resid;
2265 }
2266 vput(vp);
2267 return (error);
2268 }
2269
2270 /*
2271 * Common implementation code for chflags() and fchflags().
2272 */
2273 static int
2274 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2275 {
2276 struct mount *mp;
2277 struct vattr vattr;
2278 int error;
2279
2280 /* We can't support the value matching VNOVAL. */
2281 if (flags == VNOVAL)
2282 return (EOPNOTSUPP);
2283
2284 /*
2285 * Prevent non-root users from setting flags on devices. When
2286 * a device is reused, users can retain ownership of the device
2287 * if they are allowed to set flags and programs assume that
2288 * chown can't fail when done as root.
2289 */
2290 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2291 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2292 if (error != 0)
2293 return (error);
2294 }
2295
2296 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2297 return (error);
2298 VATTR_NULL(&vattr);
2299 vattr.va_flags = flags;
2300 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2301 #ifdef MAC
2302 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2303 if (error == 0)
2304 #endif
2305 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2306 VOP_UNLOCK(vp, 0);
2307 vn_finished_write(mp);
2308 return (error);
2309 }
2310
2311 /*
2312 * Change flags of a file given a path name.
2313 */
2314 #ifndef _SYS_SYSPROTO_H_
2315 struct chflags_args {
2316 const char *path;
2317 u_long flags;
2318 };
2319 #endif
2320 int
2321 sys_chflags(struct thread *td, struct chflags_args *uap)
2322 {
2323
2324 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2325 uap->flags, 0));
2326 }
2327
2328 #ifndef _SYS_SYSPROTO_H_
2329 struct chflagsat_args {
2330 int fd;
2331 const char *path;
2332 u_long flags;
2333 int atflag;
2334 }
2335 #endif
2336 int
2337 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2338 {
2339 int fd = uap->fd;
2340 const char *path = uap->path;
2341 u_long flags = uap->flags;
2342 int atflag = uap->atflag;
2343
2344 if (atflag & ~AT_SYMLINK_NOFOLLOW)
2345 return (EINVAL);
2346
2347 return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2348 }
2349
2350 /*
2351 * Same as chflags() but doesn't follow symlinks.
2352 */
2353 #ifndef _SYS_SYSPROTO_H_
2354 struct lchflags_args {
2355 const char *path;
2356 u_long flags;
2357 };
2358 #endif
2359 int
2360 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2361 {
2362
2363 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2364 uap->flags, AT_SYMLINK_NOFOLLOW));
2365 }
2366
2367 static int
2368 kern_chflagsat(struct thread *td, int fd, const char *path,
2369 enum uio_seg pathseg, u_long flags, int atflag)
2370 {
2371 struct nameidata nd;
2372 cap_rights_t rights;
2373 int error, follow;
2374
2375 AUDIT_ARG_FFLAGS(flags);
2376 follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2377 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2378 cap_rights_init(&rights, CAP_FCHFLAGS), td);
2379 if ((error = namei(&nd)) != 0)
2380 return (error);
2381 NDFREE(&nd, NDF_ONLY_PNBUF);
2382 error = setfflags(td, nd.ni_vp, flags);
2383 vrele(nd.ni_vp);
2384 return (error);
2385 }
2386
2387 /*
2388 * Change flags of a file given a file descriptor.
2389 */
2390 #ifndef _SYS_SYSPROTO_H_
2391 struct fchflags_args {
2392 int fd;
2393 u_long flags;
2394 };
2395 #endif
2396 int
2397 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2398 {
2399 struct file *fp;
2400 cap_rights_t rights;
2401 int error;
2402
2403 AUDIT_ARG_FD(uap->fd);
2404 AUDIT_ARG_FFLAGS(uap->flags);
2405 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2406 &fp);
2407 if (error != 0)
2408 return (error);
2409 #ifdef AUDIT
2410 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2411 AUDIT_ARG_VNODE1(fp->f_vnode);
2412 VOP_UNLOCK(fp->f_vnode, 0);
2413 #endif
2414 error = setfflags(td, fp->f_vnode, uap->flags);
2415 fdrop(fp, td);
2416 return (error);
2417 }
2418
2419 /*
2420 * Common implementation code for chmod(), lchmod() and fchmod().
2421 */
2422 int
2423 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2424 {
2425 struct mount *mp;
2426 struct vattr vattr;
2427 int error;
2428
2429 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2430 return (error);
2431 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2432 VATTR_NULL(&vattr);
2433 vattr.va_mode = mode & ALLPERMS;
2434 #ifdef MAC
2435 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2436 if (error == 0)
2437 #endif
2438 error = VOP_SETATTR(vp, &vattr, cred);
2439 VOP_UNLOCK(vp, 0);
2440 vn_finished_write(mp);
2441 return (error);
2442 }
2443
2444 /*
2445 * Change mode of a file given path name.
2446 */
2447 #ifndef _SYS_SYSPROTO_H_
2448 struct chmod_args {
2449 char *path;
2450 int mode;
2451 };
2452 #endif
2453 int
2454 sys_chmod(struct thread *td, struct chmod_args *uap)
2455 {
2456
2457 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2458 uap->mode, 0));
2459 }
2460
2461 #ifndef _SYS_SYSPROTO_H_
2462 struct fchmodat_args {
2463 int dirfd;
2464 char *path;
2465 mode_t mode;
2466 int flag;
2467 }
2468 #endif
2469 int
2470 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2471 {
2472 int flag = uap->flag;
2473 int fd = uap->fd;
2474 char *path = uap->path;
2475 mode_t mode = uap->mode;
2476
2477 if (flag & ~AT_SYMLINK_NOFOLLOW)
2478 return (EINVAL);
2479
2480 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2481 }
2482
2483 /*
2484 * Change mode of a file given path name (don't follow links.)
2485 */
2486 #ifndef _SYS_SYSPROTO_H_
2487 struct lchmod_args {
2488 char *path;
2489 int mode;
2490 };
2491 #endif
2492 int
2493 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2494 {
2495
2496 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2497 uap->mode, AT_SYMLINK_NOFOLLOW));
2498 }
2499
2500 int
2501 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2502 mode_t mode, int flag)
2503 {
2504 struct nameidata nd;
2505 cap_rights_t rights;
2506 int error, follow;
2507
2508 AUDIT_ARG_MODE(mode);
2509 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2510 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2511 cap_rights_init(&rights, CAP_FCHMOD), td);
2512 if ((error = namei(&nd)) != 0)
2513 return (error);
2514 NDFREE(&nd, NDF_ONLY_PNBUF);
2515 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2516 vrele(nd.ni_vp);
2517 return (error);
2518 }
2519
2520 /*
2521 * Change mode of a file given a file descriptor.
2522 */
2523 #ifndef _SYS_SYSPROTO_H_
2524 struct fchmod_args {
2525 int fd;
2526 int mode;
2527 };
2528 #endif
2529 int
2530 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2531 {
2532 struct file *fp;
2533 cap_rights_t rights;
2534 int error;
2535
2536 AUDIT_ARG_FD(uap->fd);
2537 AUDIT_ARG_MODE(uap->mode);
2538
2539 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2540 if (error != 0)
2541 return (error);
2542 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2543 fdrop(fp, td);
2544 return (error);
2545 }
2546
2547 /*
2548 * Common implementation for chown(), lchown(), and fchown()
2549 */
2550 int
2551 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2552 gid_t gid)
2553 {
2554 struct mount *mp;
2555 struct vattr vattr;
2556 int error;
2557
2558 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2559 return (error);
2560 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2561 VATTR_NULL(&vattr);
2562 vattr.va_uid = uid;
2563 vattr.va_gid = gid;
2564 #ifdef MAC
2565 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2566 vattr.va_gid);
2567 if (error == 0)
2568 #endif
2569 error = VOP_SETATTR(vp, &vattr, cred);
2570 VOP_UNLOCK(vp, 0);
2571 vn_finished_write(mp);
2572 return (error);
2573 }
2574
2575 /*
2576 * Set ownership given a path name.
2577 */
2578 #ifndef _SYS_SYSPROTO_H_
2579 struct chown_args {
2580 char *path;
2581 int uid;
2582 int gid;
2583 };
2584 #endif
2585 int
2586 sys_chown(struct thread *td, struct chown_args *uap)
2587 {
2588
2589 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2590 uap->gid, 0));
2591 }
2592
2593 #ifndef _SYS_SYSPROTO_H_
2594 struct fchownat_args {
2595 int fd;
2596 const char * path;
2597 uid_t uid;
2598 gid_t gid;
2599 int flag;
2600 };
2601 #endif
2602 int
2603 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2604 {
2605 int flag;
2606
2607 flag = uap->flag;
2608 if (flag & ~AT_SYMLINK_NOFOLLOW)
2609 return (EINVAL);
2610
2611 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2612 uap->gid, uap->flag));
2613 }
2614
2615 int
2616 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2617 int uid, int gid, int flag)
2618 {
2619 struct nameidata nd;
2620 cap_rights_t rights;
2621 int error, follow;
2622
2623 AUDIT_ARG_OWNER(uid, gid);
2624 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2625 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2626 cap_rights_init(&rights, CAP_FCHOWN), td);
2627
2628 if ((error = namei(&nd)) != 0)
2629 return (error);
2630 NDFREE(&nd, NDF_ONLY_PNBUF);
2631 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2632 vrele(nd.ni_vp);
2633 return (error);
2634 }
2635
2636 /*
2637 * Set ownership given a path name, do not cross symlinks.
2638 */
2639 #ifndef _SYS_SYSPROTO_H_
2640 struct lchown_args {
2641 char *path;
2642 int uid;
2643 int gid;
2644 };
2645 #endif
2646 int
2647 sys_lchown(struct thread *td, struct lchown_args *uap)
2648 {
2649
2650 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2651 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2652 }
2653
2654 /*
2655 * Set ownership given a file descriptor.
2656 */
2657 #ifndef _SYS_SYSPROTO_H_
2658 struct fchown_args {
2659 int fd;
2660 int uid;
2661 int gid;
2662 };
2663 #endif
2664 int
2665 sys_fchown(struct thread *td, struct fchown_args *uap)
2666 {
2667 struct file *fp;
2668 cap_rights_t rights;
2669 int error;
2670
2671 AUDIT_ARG_FD(uap->fd);
2672 AUDIT_ARG_OWNER(uap->uid, uap->gid);
2673 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2674 if (error != 0)
2675 return (error);
2676 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2677 fdrop(fp, td);
2678 return (error);
2679 }
2680
2681 /*
2682 * Common implementation code for utimes(), lutimes(), and futimes().
2683 */
2684 static int
2685 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
2686 struct timespec *tsp)
2687 {
2688 struct timeval tv[2];
2689 const struct timeval *tvp;
2690 int error;
2691
2692 if (usrtvp == NULL) {
2693 vfs_timestamp(&tsp[0]);
2694 tsp[1] = tsp[0];
2695 } else {
2696 if (tvpseg == UIO_SYSSPACE) {
2697 tvp = usrtvp;
2698 } else {
2699 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2700 return (error);
2701 tvp = tv;
2702 }
2703
2704 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2705 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2706 return (EINVAL);
2707 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2708 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2709 }
2710 return (0);
2711 }
2712
2713 /*
2714 * Common implementation code for futimens(), utimensat().
2715 */
2716 #define UTIMENS_NULL 0x1
2717 #define UTIMENS_EXIT 0x2
2718 static int
2719 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2720 struct timespec *tsp, int *retflags)
2721 {
2722 struct timespec tsnow;
2723 int error;
2724
2725 vfs_timestamp(&tsnow);
2726 *retflags = 0;
2727 if (usrtsp == NULL) {
2728 tsp[0] = tsnow;
2729 tsp[1] = tsnow;
2730 *retflags |= UTIMENS_NULL;
2731 return (0);
2732 }
2733 if (tspseg == UIO_SYSSPACE) {
2734 tsp[0] = usrtsp[0];
2735 tsp[1] = usrtsp[1];
2736 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2737 return (error);
2738 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2739 *retflags |= UTIMENS_EXIT;
2740 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2741 *retflags |= UTIMENS_NULL;
2742 if (tsp[0].tv_nsec == UTIME_OMIT)
2743 tsp[0].tv_sec = VNOVAL;
2744 else if (tsp[0].tv_nsec == UTIME_NOW)
2745 tsp[0] = tsnow;
2746 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2747 return (EINVAL);
2748 if (tsp[1].tv_nsec == UTIME_OMIT)
2749 tsp[1].tv_sec = VNOVAL;
2750 else if (tsp[1].tv_nsec == UTIME_NOW)
2751 tsp[1] = tsnow;
2752 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2753 return (EINVAL);
2754
2755 return (0);
2756 }
2757
2758 /*
2759 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2760 * and utimensat().
2761 */
2762 static int
2763 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
2764 int numtimes, int nullflag)
2765 {
2766 struct mount *mp;
2767 struct vattr vattr;
2768 int error, setbirthtime;
2769
2770 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2771 return (error);
2772 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2773 setbirthtime = 0;
2774 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
2775 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
2776 setbirthtime = 1;
2777 VATTR_NULL(&vattr);
2778 vattr.va_atime = ts[0];
2779 vattr.va_mtime = ts[1];
2780 if (setbirthtime)
2781 vattr.va_birthtime = ts[1];
2782 if (numtimes > 2)
2783 vattr.va_birthtime = ts[2];
2784 if (nullflag)
2785 vattr.va_vaflags |= VA_UTIMES_NULL;
2786 #ifdef MAC
2787 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
2788 vattr.va_mtime);
2789 #endif
2790 if (error == 0)
2791 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2792 VOP_UNLOCK(vp, 0);
2793 vn_finished_write(mp);
2794 return (error);
2795 }
2796
2797 /*
2798 * Set the access and modification times of a file.
2799 */
2800 #ifndef _SYS_SYSPROTO_H_
2801 struct utimes_args {
2802 char *path;
2803 struct timeval *tptr;
2804 };
2805 #endif
2806 int
2807 sys_utimes(struct thread *td, struct utimes_args *uap)
2808 {
2809
2810 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2811 uap->tptr, UIO_USERSPACE));
2812 }
2813
2814 #ifndef _SYS_SYSPROTO_H_
2815 struct futimesat_args {
2816 int fd;
2817 const char * path;
2818 const struct timeval * times;
2819 };
2820 #endif
2821 int
2822 sys_futimesat(struct thread *td, struct futimesat_args *uap)
2823 {
2824
2825 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
2826 uap->times, UIO_USERSPACE));
2827 }
2828
2829 int
2830 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2831 struct timeval *tptr, enum uio_seg tptrseg)
2832 {
2833 struct nameidata nd;
2834 struct timespec ts[2];
2835 cap_rights_t rights;
2836 int error;
2837
2838 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
2839 return (error);
2840 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
2841 cap_rights_init(&rights, CAP_FUTIMES), td);
2842
2843 if ((error = namei(&nd)) != 0)
2844 return (error);
2845 NDFREE(&nd, NDF_ONLY_PNBUF);
2846 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
2847 vrele(nd.ni_vp);
2848 return (error);
2849 }
2850
2851 /*
2852 * Set the access and modification times of a file.
2853 */
2854 #ifndef _SYS_SYSPROTO_H_
2855 struct lutimes_args {
2856 char *path;
2857 struct timeval *tptr;
2858 };
2859 #endif
2860 int
2861 sys_lutimes(struct thread *td, struct lutimes_args *uap)
2862 {
2863
2864 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
2865 UIO_USERSPACE));
2866 }
2867
2868 int
2869 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
2870 struct timeval *tptr, enum uio_seg tptrseg)
2871 {
2872 struct timespec ts[2];
2873 struct nameidata nd;
2874 int error;
2875
2876 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
2877 return (error);
2878 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
2879 if ((error = namei(&nd)) != 0)
2880 return (error);
2881 NDFREE(&nd, NDF_ONLY_PNBUF);
2882 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
2883 vrele(nd.ni_vp);
2884 return (error);
2885 }
2886
2887 /*
2888 * Set the access and modification times of a file.
2889 */
2890 #ifndef _SYS_SYSPROTO_H_
2891 struct futimes_args {
2892 int fd;
2893 struct timeval *tptr;
2894 };
2895 #endif
2896 int
2897 sys_futimes(struct thread *td, struct futimes_args *uap)
2898 {
2899
2900 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
2901 }
2902
2903 int
2904 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
2905 enum uio_seg tptrseg)
2906 {
2907 struct timespec ts[2];
2908 struct file *fp;
2909 cap_rights_t rights;
2910 int error;
2911
2912 AUDIT_ARG_FD(fd);
2913 error = getutimes(tptr, tptrseg, ts);
2914 if (error != 0)
2915 return (error);
2916 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
2917 if (error != 0)
2918 return (error);
2919 #ifdef AUDIT
2920 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2921 AUDIT_ARG_VNODE1(fp->f_vnode);
2922 VOP_UNLOCK(fp->f_vnode, 0);
2923 #endif
2924 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
2925 fdrop(fp, td);
2926 return (error);
2927 }
2928
2929 int
2930 sys_futimens(struct thread *td, struct futimens_args *uap)
2931 {
2932
2933 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
2934 }
2935
2936 int
2937 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
2938 enum uio_seg tptrseg)
2939 {
2940 struct timespec ts[2];
2941 struct file *fp;
2942 cap_rights_t rights;
2943 int error, flags;
2944
2945 AUDIT_ARG_FD(fd);
2946 error = getutimens(tptr, tptrseg, ts, &flags);
2947 if (error != 0)
2948 return (error);
2949 if (flags & UTIMENS_EXIT)
2950 return (0);
2951 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
2952 if (error != 0)
2953 return (error);
2954 #ifdef AUDIT
2955 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2956 AUDIT_ARG_VNODE1(fp->f_vnode);
2957 VOP_UNLOCK(fp->f_vnode, 0);
2958 #endif
2959 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
2960 fdrop(fp, td);
2961 return (error);
2962 }
2963
2964 int
2965 sys_utimensat(struct thread *td, struct utimensat_args *uap)
2966 {
2967
2968 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
2969 uap->times, UIO_USERSPACE, uap->flag));
2970 }
2971
2972 int
2973 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2974 struct timespec *tptr, enum uio_seg tptrseg, int flag)
2975 {
2976 struct nameidata nd;
2977 struct timespec ts[2];
2978 cap_rights_t rights;
2979 int error, flags;
2980
2981 if (flag & ~AT_SYMLINK_NOFOLLOW)
2982 return (EINVAL);
2983
2984 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
2985 return (error);
2986 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2987 FOLLOW) | AUDITVNODE1, pathseg, path, fd,
2988 cap_rights_init(&rights, CAP_FUTIMES), td);
2989 if ((error = namei(&nd)) != 0)
2990 return (error);
2991 /*
2992 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
2993 * POSIX states:
2994 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
2995 * "Search permission is denied by a component of the path prefix."
2996 */
2997 NDFREE(&nd, NDF_ONLY_PNBUF);
2998 if ((flags & UTIMENS_EXIT) == 0)
2999 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3000 vrele(nd.ni_vp);
3001 return (error);
3002 }
3003
3004 /*
3005 * Truncate a file given its path name.
3006 */
3007 #ifndef _SYS_SYSPROTO_H_
3008 struct truncate_args {
3009 char *path;
3010 int pad;
3011 off_t length;
3012 };
3013 #endif
3014 int
3015 sys_truncate(struct thread *td, struct truncate_args *uap)
3016 {
3017
3018 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3019 }
3020
3021 int
3022 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3023 {
3024 struct mount *mp;
3025 struct vnode *vp;
3026 void *rl_cookie;
3027 struct vattr vattr;
3028 struct nameidata nd;
3029 int error;
3030
3031 if (length < 0)
3032 return(EINVAL);
3033 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3034 if ((error = namei(&nd)) != 0)
3035 return (error);
3036 vp = nd.ni_vp;
3037 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3038 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3039 vn_rangelock_unlock(vp, rl_cookie);
3040 vrele(vp);
3041 return (error);
3042 }
3043 NDFREE(&nd, NDF_ONLY_PNBUF);
3044 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3045 if (vp->v_type == VDIR)
3046 error = EISDIR;
3047 #ifdef MAC
3048 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3049 }
3050 #endif
3051 else if ((error = vn_writechk(vp)) == 0 &&
3052 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3053 VATTR_NULL(&vattr);
3054 vattr.va_size = length;
3055 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3056 }
3057 VOP_UNLOCK(vp, 0);
3058 vn_finished_write(mp);
3059 vn_rangelock_unlock(vp, rl_cookie);
3060 vrele(vp);
3061 return (error);
3062 }
3063
3064 #if defined(COMPAT_43)
3065 /*
3066 * Truncate a file given its path name.
3067 */
3068 #ifndef _SYS_SYSPROTO_H_
3069 struct otruncate_args {
3070 char *path;
3071 long length;
3072 };
3073 #endif
3074 int
3075 otruncate(struct thread *td, struct otruncate_args *uap)
3076 {
3077
3078 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3079 }
3080 #endif /* COMPAT_43 */
3081
3082 #if defined(COMPAT_FREEBSD6)
3083 /* Versions with the pad argument */
3084 int
3085 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3086 {
3087
3088 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3089 }
3090
3091 int
3092 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3093 {
3094
3095 return (kern_ftruncate(td, uap->fd, uap->length));
3096 }
3097 #endif
3098
3099 int
3100 kern_fsync(struct thread *td, int fd, bool fullsync)
3101 {
3102 struct vnode *vp;
3103 struct mount *mp;
3104 struct file *fp;
3105 cap_rights_t rights;
3106 int error, lock_flags;
3107
3108 AUDIT_ARG_FD(fd);
3109 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3110 if (error != 0)
3111 return (error);
3112 vp = fp->f_vnode;
3113 #if 0
3114 if (!fullsync)
3115 /* XXXKIB: compete outstanding aio writes */;
3116 #endif
3117 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3118 if (error != 0)
3119 goto drop;
3120 if (MNT_SHARED_WRITES(mp) ||
3121 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3122 lock_flags = LK_SHARED;
3123 } else {
3124 lock_flags = LK_EXCLUSIVE;
3125 }
3126 vn_lock(vp, lock_flags | LK_RETRY);
3127 AUDIT_ARG_VNODE1(vp);
3128 if (vp->v_object != NULL) {
3129 VM_OBJECT_WLOCK(vp->v_object);
3130 vm_object_page_clean(vp->v_object, 0, 0, 0);
3131 VM_OBJECT_WUNLOCK(vp->v_object);
3132 }
3133 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3134 VOP_UNLOCK(vp, 0);
3135 vn_finished_write(mp);
3136 drop:
3137 fdrop(fp, td);
3138 return (error);
3139 }
3140
3141 /*
3142 * Sync an open file.
3143 */
3144 #ifndef _SYS_SYSPROTO_H_
3145 struct fsync_args {
3146 int fd;
3147 };
3148 #endif
3149 int
3150 sys_fsync(struct thread *td, struct fsync_args *uap)
3151 {
3152
3153 return (kern_fsync(td, uap->fd, true));
3154 }
3155
3156 int
3157 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3158 {
3159
3160 return (kern_fsync(td, uap->fd, false));
3161 }
3162
3163 /*
3164 * Rename files. Source and destination must either both be directories, or
3165 * both not be directories. If target is a directory, it must be empty.
3166 */
3167 #ifndef _SYS_SYSPROTO_H_
3168 struct rename_args {
3169 char *from;
3170 char *to;
3171 };
3172 #endif
3173 int
3174 sys_rename(struct thread *td, struct rename_args *uap)
3175 {
3176
3177 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3178 uap->to, UIO_USERSPACE));
3179 }
3180
3181 #ifndef _SYS_SYSPROTO_H_
3182 struct renameat_args {
3183 int oldfd;
3184 char *old;
3185 int newfd;
3186 char *new;
3187 };
3188 #endif
3189 int
3190 sys_renameat(struct thread *td, struct renameat_args *uap)
3191 {
3192
3193 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3194 UIO_USERSPACE));
3195 }
3196
3197 int
3198 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3199 enum uio_seg pathseg)
3200 {
3201 struct mount *mp = NULL;
3202 struct vnode *tvp, *fvp, *tdvp;
3203 struct nameidata fromnd, tond;
3204 cap_rights_t rights;
3205 int error;
3206
3207 again:
3208 bwillwrite();
3209 #ifdef MAC
3210 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3211 AUDITVNODE1, pathseg, old, oldfd,
3212 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3213 #else
3214 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3215 pathseg, old, oldfd,
3216 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3217 #endif
3218
3219 if ((error = namei(&fromnd)) != 0)
3220 return (error);
3221 #ifdef MAC
3222 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3223 fromnd.ni_vp, &fromnd.ni_cnd);
3224 VOP_UNLOCK(fromnd.ni_dvp, 0);
3225 if (fromnd.ni_dvp != fromnd.ni_vp)
3226 VOP_UNLOCK(fromnd.ni_vp, 0);
3227 #endif
3228 fvp = fromnd.ni_vp;
3229 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3230 SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3231 cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3232 if (fromnd.ni_vp->v_type == VDIR)
3233 tond.ni_cnd.cn_flags |= WILLBEDIR;
3234 if ((error = namei(&tond)) != 0) {
3235 /* Translate error code for rename("dir1", "dir2/."). */
3236 if (error == EISDIR && fvp->v_type == VDIR)
3237 error = EINVAL;
3238 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3239 vrele(fromnd.ni_dvp);
3240 vrele(fvp);
3241 goto out1;
3242 }
3243 tdvp = tond.ni_dvp;
3244 tvp = tond.ni_vp;
3245 error = vn_start_write(fvp, &mp, V_NOWAIT);
3246 if (error != 0) {
3247 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3248 NDFREE(&tond, NDF_ONLY_PNBUF);
3249 if (tvp != NULL)
3250 vput(tvp);
3251 if (tdvp == tvp)
3252 vrele(tdvp);
3253 else
3254 vput(tdvp);
3255 vrele(fromnd.ni_dvp);
3256 vrele(fvp);
3257 vrele(tond.ni_startdir);
3258 if (fromnd.ni_startdir != NULL)
3259 vrele(fromnd.ni_startdir);
3260 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3261 if (error != 0)
3262 return (error);
3263 goto again;
3264 }
3265 if (tvp != NULL) {
3266 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3267 error = ENOTDIR;
3268 goto out;
3269 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3270 error = EISDIR;
3271 goto out;
3272 }
3273 #ifdef CAPABILITIES
3274 if (newfd != AT_FDCWD) {
3275 /*
3276 * If the target already exists we require CAP_UNLINKAT
3277 * from 'newfd'.
3278 */
3279 error = cap_check(&tond.ni_filecaps.fc_rights,
3280 cap_rights_init(&rights, CAP_UNLINKAT));
3281 if (error != 0)
3282 goto out;
3283 }
3284 #endif
3285 }
3286 if (fvp == tdvp) {
3287 error = EINVAL;
3288 goto out;
3289 }
3290 /*
3291 * If the source is the same as the destination (that is, if they
3292 * are links to the same vnode), then there is nothing to do.
3293 */
3294 if (fvp == tvp)
3295 error = -1;
3296 #ifdef MAC
3297 else
3298 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3299 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3300 #endif
3301 out:
3302 if (error == 0) {
3303 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3304 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3305 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3306 NDFREE(&tond, NDF_ONLY_PNBUF);
3307 } else {
3308 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3309 NDFREE(&tond, NDF_ONLY_PNBUF);
3310 if (tvp != NULL)
3311 vput(tvp);
3312 if (tdvp == tvp)
3313 vrele(tdvp);
3314 else
3315 vput(tdvp);
3316 vrele(fromnd.ni_dvp);
3317 vrele(fvp);
3318 }
3319 vrele(tond.ni_startdir);
3320 vn_finished_write(mp);
3321 out1:
3322 if (fromnd.ni_startdir)
3323 vrele(fromnd.ni_startdir);
3324 if (error == -1)
3325 return (0);
3326 return (error);
3327 }
3328
3329 /*
3330 * Make a directory file.
3331 */
3332 #ifndef _SYS_SYSPROTO_H_
3333 struct mkdir_args {
3334 char *path;
3335 int mode;
3336 };
3337 #endif
3338 int
3339 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3340 {
3341
3342 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3343 uap->mode));
3344 }
3345
3346 #ifndef _SYS_SYSPROTO_H_
3347 struct mkdirat_args {
3348 int fd;
3349 char *path;
3350 mode_t mode;
3351 };
3352 #endif
3353 int
3354 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3355 {
3356
3357 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3358 }
3359
3360 int
3361 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3362 int mode)
3363 {
3364 struct mount *mp;
3365 struct vnode *vp;
3366 struct vattr vattr;
3367 struct nameidata nd;
3368 cap_rights_t rights;
3369 int error;
3370
3371 AUDIT_ARG_MODE(mode);
3372 restart:
3373 bwillwrite();
3374 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3375 NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3376 td);
3377 nd.ni_cnd.cn_flags |= WILLBEDIR;
3378 if ((error = namei(&nd)) != 0)
3379 return (error);
3380 vp = nd.ni_vp;
3381 if (vp != NULL) {
3382 NDFREE(&nd, NDF_ONLY_PNBUF);
3383 /*
3384 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3385 * the strange behaviour of leaving the vnode unlocked
3386 * if the target is the same vnode as the parent.
3387 */
3388 if (vp == nd.ni_dvp)
3389 vrele(nd.ni_dvp);
3390 else
3391 vput(nd.ni_dvp);
3392 vrele(vp);
3393 return (EEXIST);
3394 }
3395 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3396 NDFREE(&nd, NDF_ONLY_PNBUF);
3397 vput(nd.ni_dvp);
3398 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3399 return (error);
3400 goto restart;
3401 }
3402 VATTR_NULL(&vattr);
3403 vattr.va_type = VDIR;
3404 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3405 #ifdef MAC
3406 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3407 &vattr);
3408 if (error != 0)
3409 goto out;
3410 #endif
3411 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3412 #ifdef MAC
3413 out:
3414 #endif
3415 NDFREE(&nd, NDF_ONLY_PNBUF);
3416 vput(nd.ni_dvp);
3417 if (error == 0)
3418 vput(nd.ni_vp);
3419 vn_finished_write(mp);
3420 return (error);
3421 }
3422
3423 /*
3424 * Remove a directory file.
3425 */
3426 #ifndef _SYS_SYSPROTO_H_
3427 struct rmdir_args {
3428 char *path;
3429 };
3430 #endif
3431 int
3432 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3433 {
3434
3435 return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3436 }
3437
3438 int
3439 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3440 {
3441 struct mount *mp;
3442 struct vnode *vp;
3443 struct nameidata nd;
3444 cap_rights_t rights;
3445 int error;
3446
3447 restart:
3448 bwillwrite();
3449 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3450 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3451 if ((error = namei(&nd)) != 0)
3452 return (error);
3453 vp = nd.ni_vp;
3454 if (vp->v_type != VDIR) {
3455 error = ENOTDIR;
3456 goto out;
3457 }
3458 /*
3459 * No rmdir "." please.
3460 */
3461 if (nd.ni_dvp == vp) {
3462 error = EINVAL;
3463 goto out;
3464 }
3465 /*
3466 * The root of a mounted filesystem cannot be deleted.
3467 */
3468 if (vp->v_vflag & VV_ROOT) {
3469 error = EBUSY;
3470 goto out;
3471 }
3472 #ifdef MAC
3473 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3474 &nd.ni_cnd);
3475 if (error != 0)
3476 goto out;
3477 #endif
3478 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3479 NDFREE(&nd, NDF_ONLY_PNBUF);
3480 vput(vp);
3481 if (nd.ni_dvp == vp)
3482 vrele(nd.ni_dvp);
3483 else
3484 vput(nd.ni_dvp);
3485 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3486 return (error);
3487 goto restart;
3488 }
3489 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3490 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3491 vn_finished_write(mp);
3492 out:
3493 NDFREE(&nd, NDF_ONLY_PNBUF);
3494 vput(vp);
3495 if (nd.ni_dvp == vp)
3496 vrele(nd.ni_dvp);
3497 else
3498 vput(nd.ni_dvp);
3499 return (error);
3500 }
3501
3502 #ifdef COMPAT_43
3503 /*
3504 * Read a block of directory entries in a filesystem independent format.
3505 */
3506 #ifndef _SYS_SYSPROTO_H_
3507 struct ogetdirentries_args {
3508 int fd;
3509 char *buf;
3510 u_int count;
3511 long *basep;
3512 };
3513 #endif
3514 int
3515 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3516 {
3517 long loff;
3518 int error;
3519
3520 error = kern_ogetdirentries(td, uap, &loff);
3521 if (error == 0)
3522 error = copyout(&loff, uap->basep, sizeof(long));
3523 return (error);
3524 }
3525
3526 int
3527 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3528 long *ploff)
3529 {
3530 struct vnode *vp;
3531 struct file *fp;
3532 struct uio auio, kuio;
3533 struct iovec aiov, kiov;
3534 struct dirent *dp, *edp;
3535 cap_rights_t rights;
3536 caddr_t dirbuf;
3537 int error, eofflag, readcnt;
3538 long loff;
3539 off_t foffset;
3540
3541 /* XXX arbitrary sanity limit on `count'. */
3542 if (uap->count > 64 * 1024)
3543 return (EINVAL);
3544 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp);
3545 if (error != 0)
3546 return (error);
3547 if ((fp->f_flag & FREAD) == 0) {
3548 fdrop(fp, td);
3549 return (EBADF);
3550 }
3551 vp = fp->f_vnode;
3552 foffset = foffset_lock(fp, 0);
3553 unionread:
3554 if (vp->v_type != VDIR) {
3555 foffset_unlock(fp, foffset, 0);
3556 fdrop(fp, td);
3557 return (EINVAL);
3558 }
3559 aiov.iov_base = uap->buf;
3560 aiov.iov_len = uap->count;
3561 auio.uio_iov = &aiov;
3562 auio.uio_iovcnt = 1;
3563 auio.uio_rw = UIO_READ;
3564 auio.uio_segflg = UIO_USERSPACE;
3565 auio.uio_td = td;
3566 auio.uio_resid = uap->count;
3567 vn_lock(vp, LK_SHARED | LK_RETRY);
3568 loff = auio.uio_offset = foffset;
3569 #ifdef MAC
3570 error = mac_vnode_check_readdir(td->td_ucred, vp);
3571 if (error != 0) {
3572 VOP_UNLOCK(vp, 0);
3573 foffset_unlock(fp, foffset, FOF_NOUPDATE);
3574 fdrop(fp, td);
3575 return (error);
3576 }
3577 #endif
3578 # if (BYTE_ORDER != LITTLE_ENDIAN)
3579 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3580 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3581 NULL, NULL);
3582 foffset = auio.uio_offset;
3583 } else
3584 # endif
3585 {
3586 kuio = auio;
3587 kuio.uio_iov = &kiov;
3588 kuio.uio_segflg = UIO_SYSSPACE;
3589 kiov.iov_len = uap->count;
3590 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3591 kiov.iov_base = dirbuf;
3592 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3593 NULL, NULL);
3594 foffset = kuio.uio_offset;
3595 if (error == 0) {
3596 readcnt = uap->count - kuio.uio_resid;
3597 edp = (struct dirent *)&dirbuf[readcnt];
3598 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3599 # if (BYTE_ORDER == LITTLE_ENDIAN)
3600 /*
3601 * The expected low byte of
3602 * dp->d_namlen is our dp->d_type.
3603 * The high MBZ byte of dp->d_namlen
3604 * is our dp->d_namlen.
3605 */
3606 dp->d_type = dp->d_namlen;
3607 dp->d_namlen = 0;
3608 # else
3609 /*
3610 * The dp->d_type is the high byte
3611 * of the expected dp->d_namlen,
3612 * so must be zero'ed.
3613 */
3614 dp->d_type = 0;
3615 # endif
3616 if (dp->d_reclen > 0) {
3617 dp = (struct dirent *)
3618 ((char *)dp + dp->d_reclen);
3619 } else {
3620 error = EIO;
3621 break;
3622 }
3623 }
3624 if (dp >= edp)
3625 error = uiomove(dirbuf, readcnt, &auio);
3626 }
3627 free(dirbuf, M_TEMP);
3628 }
3629 if (error != 0) {
3630 VOP_UNLOCK(vp, 0);
3631 foffset_unlock(fp, foffset, 0);
3632 fdrop(fp, td);
3633 return (error);
3634 }
3635 if (uap->count == auio.uio_resid &&
3636 (vp->v_vflag & VV_ROOT) &&
3637 (vp->v_mount->mnt_flag & MNT_UNION)) {
3638 struct vnode *tvp = vp;
3639 vp = vp->v_mount->mnt_vnodecovered;
3640 VREF(vp);
3641 fp->f_vnode = vp;
3642 fp->f_data = vp;
3643 foffset = 0;
3644 vput(tvp);
3645 goto unionread;
3646 }
3647 VOP_UNLOCK(vp, 0);
3648 foffset_unlock(fp, foffset, 0);
3649 fdrop(fp, td);
3650 td->td_retval[0] = uap->count - auio.uio_resid;
3651 if (error == 0)
3652 *ploff = loff;
3653 return (error);
3654 }
3655 #endif /* COMPAT_43 */
3656
3657 /*
3658 * Read a block of directory entries in a filesystem independent format.
3659 */
3660 #ifndef _SYS_SYSPROTO_H_
3661 struct getdirentries_args {
3662 int fd;
3663 char *buf;
3664 u_int count;
3665 long *basep;
3666 };
3667 #endif
3668 int
3669 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
3670 {
3671 long base;
3672 int error;
3673
3674 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3675 NULL, UIO_USERSPACE);
3676 if (error != 0)
3677 return (error);
3678 if (uap->basep != NULL)
3679 error = copyout(&base, uap->basep, sizeof(long));
3680 return (error);
3681 }
3682
3683 int
3684 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3685 long *basep, ssize_t *residp, enum uio_seg bufseg)
3686 {
3687 struct vnode *vp;
3688 struct file *fp;
3689 struct uio auio;
3690 struct iovec aiov;
3691 cap_rights_t rights;
3692 long loff;
3693 int error, eofflag;
3694 off_t foffset;
3695
3696 AUDIT_ARG_FD(fd);
3697 if (count > IOSIZE_MAX)
3698 return (EINVAL);
3699 auio.uio_resid = count;
3700 error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3701 if (error != 0)
3702 return (error);
3703 if ((fp->f_flag & FREAD) == 0) {
3704 fdrop(fp, td);
3705 return (EBADF);
3706 }
3707 vp = fp->f_vnode;
3708 foffset = foffset_lock(fp, 0);
3709 unionread:
3710 if (vp->v_type != VDIR) {
3711 error = EINVAL;
3712 goto fail;
3713 }
3714 aiov.iov_base = buf;
3715 aiov.iov_len = count;
3716 auio.uio_iov = &aiov;
3717 auio.uio_iovcnt = 1;
3718 auio.uio_rw = UIO_READ;
3719 auio.uio_segflg = bufseg;
3720 auio.uio_td = td;
3721 vn_lock(vp, LK_SHARED | LK_RETRY);
3722 AUDIT_ARG_VNODE1(vp);
3723 loff = auio.uio_offset = foffset;
3724 #ifdef MAC
3725 error = mac_vnode_check_readdir(td->td_ucred, vp);
3726 if (error == 0)
3727 #endif
3728 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3729 NULL);
3730 foffset = auio.uio_offset;
3731 if (error != 0) {
3732 VOP_UNLOCK(vp, 0);
3733 goto fail;
3734 }
3735 if (count == auio.uio_resid &&
3736 (vp->v_vflag & VV_ROOT) &&
3737 (vp->v_mount->mnt_flag & MNT_UNION)) {
3738 struct vnode *tvp = vp;
3739
3740 vp = vp->v_mount->mnt_vnodecovered;
3741 VREF(vp);
3742 fp->f_vnode = vp;
3743 fp->f_data = vp;
3744 foffset = 0;
3745 vput(tvp);
3746 goto unionread;
3747 }
3748 VOP_UNLOCK(vp, 0);
3749 *basep = loff;
3750 if (residp != NULL)
3751 *residp = auio.uio_resid;
3752 td->td_retval[0] = count - auio.uio_resid;
3753 fail:
3754 foffset_unlock(fp, foffset, 0);
3755 fdrop(fp, td);
3756 return (error);
3757 }
3758
3759 #ifndef _SYS_SYSPROTO_H_
3760 struct getdents_args {
3761 int fd;
3762 char *buf;
3763 size_t count;
3764 };
3765 #endif
3766 int
3767 sys_getdents(struct thread *td, struct getdents_args *uap)
3768 {
3769 struct getdirentries_args ap;
3770
3771 ap.fd = uap->fd;
3772 ap.buf = uap->buf;
3773 ap.count = uap->count;
3774 ap.basep = NULL;
3775 return (sys_getdirentries(td, &ap));
3776 }
3777
3778 /*
3779 * Set the mode mask for creation of filesystem nodes.
3780 */
3781 #ifndef _SYS_SYSPROTO_H_
3782 struct umask_args {
3783 int newmask;
3784 };
3785 #endif
3786 int
3787 sys_umask(struct thread *td, struct umask_args *uap)
3788 {
3789 struct filedesc *fdp;
3790
3791 fdp = td->td_proc->p_fd;
3792 FILEDESC_XLOCK(fdp);
3793 td->td_retval[0] = fdp->fd_cmask;
3794 fdp->fd_cmask = uap->newmask & ALLPERMS;
3795 FILEDESC_XUNLOCK(fdp);
3796 return (0);
3797 }
3798
3799 /*
3800 * Void all references to file by ripping underlying filesystem away from
3801 * vnode.
3802 */
3803 #ifndef _SYS_SYSPROTO_H_
3804 struct revoke_args {
3805 char *path;
3806 };
3807 #endif
3808 int
3809 sys_revoke(struct thread *td, struct revoke_args *uap)
3810 {
3811 struct vnode *vp;
3812 struct vattr vattr;
3813 struct nameidata nd;
3814 int error;
3815
3816 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
3817 uap->path, td);
3818 if ((error = namei(&nd)) != 0)
3819 return (error);
3820 vp = nd.ni_vp;
3821 NDFREE(&nd, NDF_ONLY_PNBUF);
3822 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
3823 error = EINVAL;
3824 goto out;
3825 }
3826 #ifdef MAC
3827 error = mac_vnode_check_revoke(td->td_ucred, vp);
3828 if (error != 0)
3829 goto out;
3830 #endif
3831 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
3832 if (error != 0)
3833 goto out;
3834 if (td->td_ucred->cr_uid != vattr.va_uid) {
3835 error = priv_check(td, PRIV_VFS_ADMIN);
3836 if (error != 0)
3837 goto out;
3838 }
3839 if (vcount(vp) > 1)
3840 VOP_REVOKE(vp, REVOKEALL);
3841 out:
3842 vput(vp);
3843 return (error);
3844 }
3845
3846 /*
3847 * Convert a user file descriptor to a kernel file entry and check that, if it
3848 * is a capability, the correct rights are present. A reference on the file
3849 * entry is held upon returning.
3850 */
3851 int
3852 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
3853 {
3854 struct file *fp;
3855 int error;
3856
3857 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
3858 if (error != 0)
3859 return (error);
3860
3861 /*
3862 * The file could be not of the vnode type, or it may be not
3863 * yet fully initialized, in which case the f_vnode pointer
3864 * may be set, but f_ops is still badfileops. E.g.,
3865 * devfs_open() transiently create such situation to
3866 * facilitate csw d_fdopen().
3867 *
3868 * Dupfdopen() handling in kern_openat() installs the
3869 * half-baked file into the process descriptor table, allowing
3870 * other thread to dereference it. Guard against the race by
3871 * checking f_ops.
3872 */
3873 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
3874 fdrop(fp, td);
3875 return (EINVAL);
3876 }
3877 *fpp = fp;
3878 return (0);
3879 }
3880
3881
3882 /*
3883 * Get an (NFS) file handle.
3884 */
3885 #ifndef _SYS_SYSPROTO_H_
3886 struct lgetfh_args {
3887 char *fname;
3888 fhandle_t *fhp;
3889 };
3890 #endif
3891 int
3892 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
3893 {
3894 struct nameidata nd;
3895 fhandle_t fh;
3896 struct vnode *vp;
3897 int error;
3898
3899 error = priv_check(td, PRIV_VFS_GETFH);
3900 if (error != 0)
3901 return (error);
3902 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
3903 uap->fname, td);
3904 error = namei(&nd);
3905 if (error != 0)
3906 return (error);
3907 NDFREE(&nd, NDF_ONLY_PNBUF);
3908 vp = nd.ni_vp;
3909 bzero(&fh, sizeof(fh));
3910 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3911 error = VOP_VPTOFH(vp, &fh.fh_fid);
3912 vput(vp);
3913 if (error == 0)
3914 error = copyout(&fh, uap->fhp, sizeof (fh));
3915 return (error);
3916 }
3917
3918 #ifndef _SYS_SYSPROTO_H_
3919 struct getfh_args {
3920 char *fname;
3921 fhandle_t *fhp;
3922 };
3923 #endif
3924 int
3925 sys_getfh(struct thread *td, struct getfh_args *uap)
3926 {
3927 struct nameidata nd;
3928 fhandle_t fh;
3929 struct vnode *vp;
3930 int error;
3931
3932 error = priv_check(td, PRIV_VFS_GETFH);
3933 if (error != 0)
3934 return (error);
3935 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
3936 uap->fname, td);
3937 error = namei(&nd);
3938 if (error != 0)
3939 return (error);
3940 NDFREE(&nd, NDF_ONLY_PNBUF);
3941 vp = nd.ni_vp;
3942 bzero(&fh, sizeof(fh));
3943 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3944 error = VOP_VPTOFH(vp, &fh.fh_fid);
3945 vput(vp);
3946 if (error == 0)
3947 error = copyout(&fh, uap->fhp, sizeof (fh));
3948 return (error);
3949 }
3950
3951 /*
3952 * syscall for the rpc.lockd to use to translate a NFS file handle into an
3953 * open descriptor.
3954 *
3955 * warning: do not remove the priv_check() call or this becomes one giant
3956 * security hole.
3957 */
3958 #ifndef _SYS_SYSPROTO_H_
3959 struct fhopen_args {
3960 const struct fhandle *u_fhp;
3961 int flags;
3962 };
3963 #endif
3964 int
3965 sys_fhopen(struct thread *td, struct fhopen_args *uap)
3966 {
3967 struct mount *mp;
3968 struct vnode *vp;
3969 struct fhandle fhp;
3970 struct file *fp;
3971 int fmode, error;
3972 int indx;
3973
3974 error = priv_check(td, PRIV_VFS_FHOPEN);
3975 if (error != 0)
3976 return (error);
3977 indx = -1;
3978 fmode = FFLAGS(uap->flags);
3979 /* why not allow a non-read/write open for our lockd? */
3980 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3981 return (EINVAL);
3982 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
3983 if (error != 0)
3984 return(error);
3985 /* find the mount point */
3986 mp = vfs_busyfs(&fhp.fh_fsid);
3987 if (mp == NULL)
3988 return (ESTALE);
3989 /* now give me my vnode, it gets returned to me locked */
3990 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
3991 vfs_unbusy(mp);
3992 if (error != 0)
3993 return (error);
3994
3995 error = falloc_noinstall(td, &fp);
3996 if (error != 0) {
3997 vput(vp);
3998 return (error);
3999 }
4000 /*
4001 * An extra reference on `fp' has been held for us by
4002 * falloc_noinstall().
4003 */
4004
4005 #ifdef INVARIANTS
4006 td->td_dupfd = -1;
4007 #endif
4008 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4009 if (error != 0) {
4010 KASSERT(fp->f_ops == &badfileops,
4011 ("VOP_OPEN in fhopen() set f_ops"));
4012 KASSERT(td->td_dupfd < 0,
4013 ("fhopen() encountered fdopen()"));
4014
4015 vput(vp);
4016 goto bad;
4017 }
4018 #ifdef INVARIANTS
4019 td->td_dupfd = 0;
4020 #endif
4021 fp->f_vnode = vp;
4022 fp->f_seqcount = 1;
4023 finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4024 &vnops);
4025 VOP_UNLOCK(vp, 0);
4026 if ((fmode & O_TRUNC) != 0) {
4027 error = fo_truncate(fp, 0, td->td_ucred, td);
4028 if (error != 0)
4029 goto bad;
4030 }
4031
4032 error = finstall(td, fp, &indx, fmode, NULL);
4033 bad:
4034 fdrop(fp, td);
4035 td->td_retval[0] = indx;
4036 return (error);
4037 }
4038
4039 /*
4040 * Stat an (NFS) file handle.
4041 */
4042 #ifndef _SYS_SYSPROTO_H_
4043 struct fhstat_args {
4044 struct fhandle *u_fhp;
4045 struct stat *sb;
4046 };
4047 #endif
4048 int
4049 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4050 {
4051 struct stat sb;
4052 struct fhandle fh;
4053 int error;
4054
4055 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4056 if (error != 0)
4057 return (error);
4058 error = kern_fhstat(td, fh, &sb);
4059 if (error == 0)
4060 error = copyout(&sb, uap->sb, sizeof(sb));
4061 return (error);
4062 }
4063
4064 int
4065 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4066 {
4067 struct mount *mp;
4068 struct vnode *vp;
4069 int error;
4070
4071 error = priv_check(td, PRIV_VFS_FHSTAT);
4072 if (error != 0)
4073 return (error);
4074 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4075 return (ESTALE);
4076 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4077 vfs_unbusy(mp);
4078 if (error != 0)
4079 return (error);
4080 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4081 vput(vp);
4082 return (error);
4083 }
4084
4085 /*
4086 * Implement fstatfs() for (NFS) file handles.
4087 */
4088 #ifndef _SYS_SYSPROTO_H_
4089 struct fhstatfs_args {
4090 struct fhandle *u_fhp;
4091 struct statfs *buf;
4092 };
4093 #endif
4094 int
4095 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4096 {
4097 struct statfs *sfp;
4098 fhandle_t fh;
4099 int error;
4100
4101 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4102 if (error != 0)
4103 return (error);
4104 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4105 error = kern_fhstatfs(td, fh, sfp);
4106 if (error == 0)
4107 error = copyout(sfp, uap->buf, sizeof(*sfp));
4108 free(sfp, M_STATFS);
4109 return (error);
4110 }
4111
4112 int
4113 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4114 {
4115 struct statfs *sp;
4116 struct mount *mp;
4117 struct vnode *vp;
4118 int error;
4119
4120 error = priv_check(td, PRIV_VFS_FHSTATFS);
4121 if (error != 0)
4122 return (error);
4123 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4124 return (ESTALE);
4125 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4126 if (error != 0) {
4127 vfs_unbusy(mp);
4128 return (error);
4129 }
4130 vput(vp);
4131 error = prison_canseemount(td->td_ucred, mp);
4132 if (error != 0)
4133 goto out;
4134 #ifdef MAC
4135 error = mac_mount_check_stat(td->td_ucred, mp);
4136 if (error != 0)
4137 goto out;
4138 #endif
4139 /*
4140 * Set these in case the underlying filesystem fails to do so.
4141 */
4142 sp = &mp->mnt_stat;
4143 sp->f_version = STATFS_VERSION;
4144 sp->f_namemax = NAME_MAX;
4145 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4146 error = VFS_STATFS(mp, sp);
4147 if (error == 0)
4148 *buf = *sp;
4149 out:
4150 vfs_unbusy(mp);
4151 return (error);
4152 }
4153
4154 int
4155 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4156 {
4157 struct file *fp;
4158 struct mount *mp;
4159 struct vnode *vp;
4160 cap_rights_t rights;
4161 off_t olen, ooffset;
4162 int error;
4163
4164 if (offset < 0 || len <= 0)
4165 return (EINVAL);
4166 /* Check for wrap. */
4167 if (offset > OFF_MAX - len)
4168 return (EFBIG);
4169 error = fget(td, fd, cap_rights_init(&rights, CAP_PWRITE), &fp);
4170 if (error != 0)
4171 return (error);
4172 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4173 error = ESPIPE;
4174 goto out;
4175 }
4176 if ((fp->f_flag & FWRITE) == 0) {
4177 error = EBADF;
4178 goto out;
4179 }
4180 if (fp->f_type != DTYPE_VNODE) {
4181 error = ENODEV;
4182 goto out;
4183 }
4184 vp = fp->f_vnode;
4185 if (vp->v_type != VREG) {
4186 error = ENODEV;
4187 goto out;
4188 }
4189
4190 /* Allocating blocks may take a long time, so iterate. */
4191 for (;;) {
4192 olen = len;
4193 ooffset = offset;
4194
4195 bwillwrite();
4196 mp = NULL;
4197 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4198 if (error != 0)
4199 break;
4200 error = vn_lock(vp, LK_EXCLUSIVE);
4201 if (error != 0) {
4202 vn_finished_write(mp);
4203 break;
4204 }
4205 #ifdef MAC
4206 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4207 if (error == 0)
4208 #endif
4209 error = VOP_ALLOCATE(vp, &offset, &len);
4210 VOP_UNLOCK(vp, 0);
4211 vn_finished_write(mp);
4212
4213 if (olen + ooffset != offset + len) {
4214 panic("offset + len changed from %jx/%jx to %jx/%jx",
4215 ooffset, olen, offset, len);
4216 }
4217 if (error != 0 || len == 0)
4218 break;
4219 KASSERT(olen > len, ("Iteration did not make progress?"));
4220 maybe_yield();
4221 }
4222 out:
4223 fdrop(fp, td);
4224 return (error);
4225 }
4226
4227 int
4228 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4229 {
4230 int error;
4231
4232 error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4233 return (kern_posix_error(td, error));
4234 }
4235
4236 /*
4237 * Unlike madvise(2), we do not make a best effort to remember every
4238 * possible caching hint. Instead, we remember the last setting with
4239 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4240 * region of any current setting.
4241 */
4242 int
4243 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4244 int advice)
4245 {
4246 struct fadvise_info *fa, *new;
4247 struct file *fp;
4248 struct vnode *vp;
4249 cap_rights_t rights;
4250 off_t end;
4251 int error;
4252
4253 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4254 return (EINVAL);
4255 switch (advice) {
4256 case POSIX_FADV_SEQUENTIAL:
4257 case POSIX_FADV_RANDOM:
4258 case POSIX_FADV_NOREUSE:
4259 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4260 break;
4261 case POSIX_FADV_NORMAL:
4262 case POSIX_FADV_WILLNEED:
4263 case POSIX_FADV_DONTNEED:
4264 new = NULL;
4265 break;
4266 default:
4267 return (EINVAL);
4268 }
4269 /* XXX: CAP_POSIX_FADVISE? */
4270 error = fget(td, fd, cap_rights_init(&rights), &fp);
4271 if (error != 0)
4272 goto out;
4273 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4274 error = ESPIPE;
4275 goto out;
4276 }
4277 if (fp->f_type != DTYPE_VNODE) {
4278 error = ENODEV;
4279 goto out;
4280 }
4281 vp = fp->f_vnode;
4282 if (vp->v_type != VREG) {
4283 error = ENODEV;
4284 goto out;
4285 }
4286 if (len == 0)
4287 end = OFF_MAX;
4288 else
4289 end = offset + len - 1;
4290 switch (advice) {
4291 case POSIX_FADV_SEQUENTIAL:
4292 case POSIX_FADV_RANDOM:
4293 case POSIX_FADV_NOREUSE:
4294 /*
4295 * Try to merge any existing non-standard region with
4296 * this new region if possible, otherwise create a new
4297 * non-standard region for this request.
4298 */
4299 mtx_pool_lock(mtxpool_sleep, fp);
4300 fa = fp->f_advice;
4301 if (fa != NULL && fa->fa_advice == advice &&
4302 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4303 (end != OFF_MAX && fa->fa_start == end + 1) ||
4304 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4305 if (offset < fa->fa_start)
4306 fa->fa_start = offset;
4307 if (end > fa->fa_end)
4308 fa->fa_end = end;
4309 } else {
4310 new->fa_advice = advice;
4311 new->fa_start = offset;
4312 new->fa_end = end;
4313 fp->f_advice = new;
4314 new = fa;
4315 }
4316 mtx_pool_unlock(mtxpool_sleep, fp);
4317 break;
4318 case POSIX_FADV_NORMAL:
4319 /*
4320 * If a the "normal" region overlaps with an existing
4321 * non-standard region, trim or remove the
4322 * non-standard region.
4323 */
4324 mtx_pool_lock(mtxpool_sleep, fp);
4325 fa = fp->f_advice;
4326 if (fa != NULL) {
4327 if (offset <= fa->fa_start && end >= fa->fa_end) {
4328 new = fa;
4329 fp->f_advice = NULL;
4330 } else if (offset <= fa->fa_start &&
4331 end >= fa->fa_start)
4332 fa->fa_start = end + 1;
4333 else if (offset <= fa->fa_end && end >= fa->fa_end)
4334 fa->fa_end = offset - 1;
4335 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4336 /*
4337 * If the "normal" region is a middle
4338 * portion of the existing
4339 * non-standard region, just remove
4340 * the whole thing rather than picking
4341 * one side or the other to
4342 * preserve.
4343 */
4344 new = fa;
4345 fp->f_advice = NULL;
4346 }
4347 }
4348 mtx_pool_unlock(mtxpool_sleep, fp);
4349 break;
4350 case POSIX_FADV_WILLNEED:
4351 case POSIX_FADV_DONTNEED:
4352 error = VOP_ADVISE(vp, offset, end, advice);
4353 break;
4354 }
4355 out:
4356 if (fp != NULL)
4357 fdrop(fp, td);
4358 free(new, M_FADVISE);
4359 return (error);
4360 }
4361
4362 int
4363 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4364 {
4365 int error;
4366
4367 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4368 uap->advice);
4369 return (kern_posix_error(td, error));
4370 }
Cache object: d2ca41e51f251c87b329dd28f97f5a3b
|