vfs_syscalls.c revision 1.449.2.2.6.1 1 /* $NetBSD: vfs_syscalls.c,v 1.449.2.2.6.1 2014/04/21 10:17:47 bouyer Exp $ */
2
3 /*-
4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Andrew Doran.
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 *
19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
32 /*
33 * Copyright (c) 1989, 1993
34 * The Regents of the University of California. All rights reserved.
35 * (c) UNIX System Laboratories, Inc.
36 * All or some portions of this file are derived from material licensed
37 * to the University of California by American Telephone and Telegraph
38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
39 * the permission of UNIX System Laboratories, Inc.
40 *
41 * Redistribution and use in source and binary forms, with or without
42 * modification, are permitted provided that the following conditions
43 * are met:
44 * 1. Redistributions of source code must retain the above copyright
45 * notice, this list of conditions and the following disclaimer.
46 * 2. Redistributions in binary form must reproduce the above copyright
47 * notice, this list of conditions and the following disclaimer in the
48 * documentation and/or other materials provided with the distribution.
49 * 3. Neither the name of the University nor the names of its contributors
50 * may be used to endorse or promote products derived from this software
51 * without specific prior written permission.
52 *
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * SUCH DAMAGE.
64 *
65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95
66 */
67
68 /*
69 * Virtual File System System Calls
70 */
71
72 #include <sys/cdefs.h>
73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.449.2.2.6.1 2014/04/21 10:17:47 bouyer Exp $");
74
75 #ifdef _KERNEL_OPT
76 #include "opt_fileassoc.h"
77 #include "veriexec.h"
78 #endif
79
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/namei.h>
83 #include <sys/filedesc.h>
84 #include <sys/kernel.h>
85 #include <sys/file.h>
86 #include <sys/fcntl.h>
87 #include <sys/stat.h>
88 #include <sys/vnode.h>
89 #include <sys/mount.h>
90 #include <sys/proc.h>
91 #include <sys/uio.h>
92 #include <sys/kmem.h>
93 #include <sys/dirent.h>
94 #include <sys/sysctl.h>
95 #include <sys/syscallargs.h>
96 #include <sys/vfs_syscalls.h>
97 #include <sys/quota.h>
98 #include <sys/quotactl.h>
99 #include <sys/ktrace.h>
100 #ifdef FILEASSOC
101 #include <sys/fileassoc.h>
102 #endif /* FILEASSOC */
103 #include <sys/extattr.h>
104 #include <sys/verified_exec.h>
105 #include <sys/kauth.h>
106 #include <sys/atomic.h>
107 #include <sys/module.h>
108 #include <sys/buf.h>
109
110 #include <miscfs/genfs/genfs.h>
111 #include <miscfs/syncfs/syncfs.h>
112 #include <miscfs/specfs/specdev.h>
113
114 #include <nfs/rpcv2.h>
115 #include <nfs/nfsproto.h>
116 #include <nfs/nfs.h>
117 #include <nfs/nfs_var.h>
118
119 static int change_flags(struct vnode *, u_long, struct lwp *);
120 static int change_mode(struct vnode *, int, struct lwp *l);
121 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
122 static int do_open(lwp_t *, struct pathbuf *, int, int, int *);
123
124 /*
125 * This table is used to maintain compatibility with 4.3BSD
126 * and NetBSD 0.9 mount syscalls - and possibly other systems.
127 * Note, the order is important!
128 *
129 * Do not modify this table. It should only contain filesystems
130 * supported by NetBSD 0.9 and 4.3BSD.
131 */
132 const char * const mountcompatnames[] = {
133 NULL, /* 0 = MOUNT_NONE */
134 MOUNT_FFS, /* 1 = MOUNT_UFS */
135 MOUNT_NFS, /* 2 */
136 MOUNT_MFS, /* 3 */
137 MOUNT_MSDOS, /* 4 */
138 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */
139 MOUNT_FDESC, /* 6 */
140 MOUNT_KERNFS, /* 7 */
141 NULL, /* 8 = MOUNT_DEVFS */
142 MOUNT_AFS, /* 9 */
143 };
144
145 const int nmountcompatnames = __arraycount(mountcompatnames);
146
147 static int
148 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags)
149 {
150 int error;
151
152 fp->f_flag = flags & FMASK;
153 fp->f_type = DTYPE_VNODE;
154 fp->f_ops = &vnops;
155 fp->f_data = vp;
156
157 if (flags & (O_EXLOCK | O_SHLOCK)) {
158 struct flock lf;
159 int type;
160
161 lf.l_whence = SEEK_SET;
162 lf.l_start = 0;
163 lf.l_len = 0;
164 if (flags & O_EXLOCK)
165 lf.l_type = F_WRLCK;
166 else
167 lf.l_type = F_RDLCK;
168 type = F_FLOCK;
169 if ((flags & FNONBLOCK) == 0)
170 type |= F_WAIT;
171 VOP_UNLOCK(vp);
172 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
173 if (error) {
174 (void) vn_close(vp, fp->f_flag, fp->f_cred);
175 fd_abort(l->l_proc, fp, indx);
176 return error;
177 }
178 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
179 atomic_or_uint(&fp->f_flag, FHASLOCK);
180 }
181 if (flags & O_CLOEXEC)
182 fd_set_exclose(l, indx, true);
183 return 0;
184 }
185
186 static int
187 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
188 void *data, size_t *data_len)
189 {
190 struct mount *mp;
191 int error = 0, saved_flags;
192
193 mp = vp->v_mount;
194 saved_flags = mp->mnt_flag;
195
196 /* We can operate only on VV_ROOT nodes. */
197 if ((vp->v_vflag & VV_ROOT) == 0) {
198 error = EINVAL;
199 goto out;
200 }
201
202 /*
203 * We only allow the filesystem to be reloaded if it
204 * is currently mounted read-only. Additionally, we
205 * prevent read-write to read-only downgrades.
206 */
207 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 &&
208 (mp->mnt_flag & MNT_RDONLY) == 0 &&
209 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) {
210 error = EOPNOTSUPP; /* Needs translation */
211 goto out;
212 }
213
214 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
215 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
216 if (error)
217 goto out;
218
219 if (vfs_busy(mp, NULL)) {
220 error = EPERM;
221 goto out;
222 }
223
224 mutex_enter(&mp->mnt_updating);
225
226 mp->mnt_flag &= ~MNT_OP_FLAGS;
227 mp->mnt_flag |= flags & MNT_OP_FLAGS;
228
229 /*
230 * Set the mount level flags.
231 */
232 if (flags & MNT_RDONLY)
233 mp->mnt_flag |= MNT_RDONLY;
234 else if (mp->mnt_flag & MNT_RDONLY)
235 mp->mnt_iflag |= IMNT_WANTRDWR;
236 mp->mnt_flag &= ~MNT_BASIC_FLAGS;
237 mp->mnt_flag |= flags & MNT_BASIC_FLAGS;
238 error = VFS_MOUNT(mp, path, data, data_len);
239
240 if (error && data != NULL) {
241 int error2;
242
243 /*
244 * Update failed; let's try and see if it was an
245 * export request. For compat with 3.0 and earlier.
246 */
247 error2 = vfs_hooks_reexport(mp, path, data);
248
249 /*
250 * Only update error code if the export request was
251 * understood but some problem occurred while
252 * processing it.
253 */
254 if (error2 != EJUSTRETURN)
255 error = error2;
256 }
257
258 if (mp->mnt_iflag & IMNT_WANTRDWR)
259 mp->mnt_flag &= ~MNT_RDONLY;
260 if (error)
261 mp->mnt_flag = saved_flags;
262 mp->mnt_flag &= ~MNT_OP_FLAGS;
263 mp->mnt_iflag &= ~IMNT_WANTRDWR;
264 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
265 if (mp->mnt_syncer == NULL)
266 error = vfs_allocate_syncvnode(mp);
267 } else {
268 if (mp->mnt_syncer != NULL)
269 vfs_deallocate_syncvnode(mp);
270 }
271 mutex_exit(&mp->mnt_updating);
272 vfs_unbusy(mp, false, NULL);
273
274 if ((error == 0) && !(saved_flags & MNT_EXTATTR) &&
275 (flags & MNT_EXTATTR)) {
276 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START,
277 NULL, 0, NULL) != 0) {
278 printf("%s: failed to start extattr, error = %d",
279 mp->mnt_stat.f_mntonname, error);
280 mp->mnt_flag &= ~MNT_EXTATTR;
281 }
282 }
283
284 if ((error == 0) && (saved_flags & MNT_EXTATTR) &&
285 !(flags & MNT_EXTATTR)) {
286 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP,
287 NULL, 0, NULL) != 0) {
288 printf("%s: failed to stop extattr, error = %d",
289 mp->mnt_stat.f_mntonname, error);
290 mp->mnt_flag |= MNT_RDONLY;
291 }
292 }
293 out:
294 return (error);
295 }
296
297 static int
298 mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
299 {
300 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
301 int error;
302
303 /* Copy file-system type from userspace. */
304 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
305 if (error) {
306 /*
307 * Historically, filesystem types were identified by numbers.
308 * If we get an integer for the filesystem type instead of a
309 * string, we check to see if it matches one of the historic
310 * filesystem types.
311 */
312 u_long fsindex = (u_long)fstype;
313 if (fsindex >= nmountcompatnames ||
314 mountcompatnames[fsindex] == NULL)
315 return ENODEV;
316 strlcpy(fstypename, mountcompatnames[fsindex],
317 sizeof(fstypename));
318 }
319
320 /* Accept `ufs' as an alias for `ffs', for compatibility. */
321 if (strcmp(fstypename, "ufs") == 0)
322 fstypename[0] = 'f';
323
324 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
325 return 0;
326
327 /* If we can autoload a vfs module, try again */
328 (void)module_autoload(fstypename, MODULE_CLASS_VFS);
329
330 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
331 return 0;
332
333 return ENODEV;
334 }
335
336 static int
337 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
338 void *data, size_t *data_len)
339 {
340 struct mount *mp;
341 int error;
342
343 /* If MNT_GETARGS is specified, it should be the only flag. */
344 if (flags & ~MNT_GETARGS)
345 return EINVAL;
346
347 mp = vp->v_mount;
348
349 /* XXX: probably some notion of "can see" here if we want isolation. */
350 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
351 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
352 if (error)
353 return error;
354
355 if ((vp->v_vflag & VV_ROOT) == 0)
356 return EINVAL;
357
358 if (vfs_busy(mp, NULL))
359 return EPERM;
360
361 mutex_enter(&mp->mnt_updating);
362 mp->mnt_flag &= ~MNT_OP_FLAGS;
363 mp->mnt_flag |= MNT_GETARGS;
364 error = VFS_MOUNT(mp, path, data, data_len);
365 mp->mnt_flag &= ~MNT_OP_FLAGS;
366 mutex_exit(&mp->mnt_updating);
367
368 vfs_unbusy(mp, false, NULL);
369 return (error);
370 }
371
372 int
373 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
374 {
375 /* {
376 syscallarg(const char *) type;
377 syscallarg(const char *) path;
378 syscallarg(int) flags;
379 syscallarg(void *) data;
380 syscallarg(size_t) data_len;
381 } */
382
383 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
384 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
385 SCARG(uap, data_len), retval);
386 }
387
388 int
389 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
390 const char *path, int flags, void *data, enum uio_seg data_seg,
391 size_t data_len, register_t *retval)
392 {
393 struct vnode *vp;
394 void *data_buf = data;
395 bool vfsopsrele = false;
396 size_t alloc_sz = 0;
397 int error;
398
399 /* XXX: The calling convention of this routine is totally bizarre */
400 if (vfsops)
401 vfsopsrele = true;
402
403 /*
404 * Get vnode to be covered
405 */
406 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
407 if (error != 0) {
408 vp = NULL;
409 goto done;
410 }
411
412 if (vfsops == NULL) {
413 if (flags & (MNT_GETARGS | MNT_UPDATE)) {
414 vfsops = vp->v_mount->mnt_op;
415 } else {
416 /* 'type' is userspace */
417 error = mount_get_vfsops(type, &vfsops);
418 if (error != 0)
419 goto done;
420 vfsopsrele = true;
421 }
422 }
423
424 /*
425 * We allow data to be NULL, even for userspace. Some fs's don't need
426 * it. The others will handle NULL.
427 */
428 if (data != NULL && data_seg == UIO_USERSPACE) {
429 if (data_len == 0) {
430 /* No length supplied, use default for filesystem */
431 data_len = vfsops->vfs_min_mount_data;
432
433 /*
434 * Hopefully a longer buffer won't make copyin() fail.
435 * For compatibility with 3.0 and earlier.
436 */
437 if (flags & MNT_UPDATE
438 && data_len < sizeof (struct mnt_export_args30))
439 data_len = sizeof (struct mnt_export_args30);
440 }
441 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) {
442 error = EINVAL;
443 goto done;
444 }
445 alloc_sz = data_len;
446 data_buf = kmem_alloc(alloc_sz, KM_SLEEP);
447
448 /* NFS needs the buffer even for mnt_getargs .... */
449 error = copyin(data, data_buf, data_len);
450 if (error != 0)
451 goto done;
452 }
453
454 if (flags & MNT_GETARGS) {
455 if (data_len == 0) {
456 error = EINVAL;
457 goto done;
458 }
459 error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
460 if (error != 0)
461 goto done;
462 if (data_seg == UIO_USERSPACE)
463 error = copyout(data_buf, data, data_len);
464 *retval = data_len;
465 } else if (flags & MNT_UPDATE) {
466 error = mount_update(l, vp, path, flags, data_buf, &data_len);
467 } else {
468 /* Locking is handled internally in mount_domount(). */
469 KASSERT(vfsopsrele == true);
470 error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
471 &data_len);
472 vfsopsrele = false;
473 }
474
475 done:
476 if (vfsopsrele)
477 vfs_delref(vfsops);
478 if (vp != NULL) {
479 vrele(vp);
480 }
481 if (data_buf != data)
482 kmem_free(data_buf, alloc_sz);
483 return (error);
484 }
485
486 /*
487 * Unmount a file system.
488 *
489 * Note: unmount takes a path to the vnode mounted on as argument,
490 * not special file (as before).
491 */
492 /* ARGSUSED */
493 int
494 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
495 {
496 /* {
497 syscallarg(const char *) path;
498 syscallarg(int) flags;
499 } */
500 struct vnode *vp;
501 struct mount *mp;
502 int error;
503 struct pathbuf *pb;
504 struct nameidata nd;
505
506 error = pathbuf_copyin(SCARG(uap, path), &pb);
507 if (error) {
508 return error;
509 }
510
511 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
512 if ((error = namei(&nd)) != 0) {
513 pathbuf_destroy(pb);
514 return error;
515 }
516 vp = nd.ni_vp;
517 pathbuf_destroy(pb);
518
519 mp = vp->v_mount;
520 atomic_inc_uint(&mp->mnt_refcnt);
521 VOP_UNLOCK(vp);
522
523 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
524 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
525 if (error) {
526 vrele(vp);
527 vfs_destroy(mp);
528 return (error);
529 }
530
531 /*
532 * Don't allow unmounting the root file system.
533 */
534 if (mp->mnt_flag & MNT_ROOTFS) {
535 vrele(vp);
536 vfs_destroy(mp);
537 return (EINVAL);
538 }
539
540 /*
541 * Must be the root of the filesystem
542 */
543 if ((vp->v_vflag & VV_ROOT) == 0) {
544 vrele(vp);
545 vfs_destroy(mp);
546 return (EINVAL);
547 }
548
549 vrele(vp);
550 error = dounmount(mp, SCARG(uap, flags), l);
551 vfs_destroy(mp);
552 return error;
553 }
554
555 /*
556 * Sync each mounted filesystem.
557 */
558 #ifdef DEBUG
559 int syncprt = 0;
560 struct ctldebug debug0 = { "syncprt", &syncprt };
561 #endif
562
563 void
564 do_sys_sync(struct lwp *l)
565 {
566 struct mount *mp, *nmp;
567 int asyncflag;
568
569 mutex_enter(&mountlist_lock);
570 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
571 mp = nmp) {
572 if (vfs_busy(mp, &nmp)) {
573 continue;
574 }
575 mutex_enter(&mp->mnt_updating);
576 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
577 asyncflag = mp->mnt_flag & MNT_ASYNC;
578 mp->mnt_flag &= ~MNT_ASYNC;
579 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
580 if (asyncflag)
581 mp->mnt_flag |= MNT_ASYNC;
582 }
583 mutex_exit(&mp->mnt_updating);
584 vfs_unbusy(mp, false, &nmp);
585 }
586 mutex_exit(&mountlist_lock);
587 #ifdef DEBUG
588 if (syncprt)
589 vfs_bufstats();
590 #endif /* DEBUG */
591 }
592
593 /* ARGSUSED */
594 int
595 sys_sync(struct lwp *l, const void *v, register_t *retval)
596 {
597 do_sys_sync(l);
598 return (0);
599 }
600
601
602 /*
603 * Access or change filesystem quotas.
604 *
605 * (this is really 14 different calls bundled into one)
606 */
607
608 static int
609 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u)
610 {
611 struct quotastat info_k;
612 int error;
613
614 /* ensure any padding bytes are cleared */
615 memset(&info_k, 0, sizeof(info_k));
616
617 error = vfs_quotactl_stat(mp, &info_k);
618 if (error) {
619 return error;
620 }
621
622 return copyout(&info_k, info_u, sizeof(info_k));
623 }
624
625 static int
626 do_sys_quotactl_idtypestat(struct mount *mp, int idtype,
627 struct quotaidtypestat *info_u)
628 {
629 struct quotaidtypestat info_k;
630 int error;
631
632 /* ensure any padding bytes are cleared */
633 memset(&info_k, 0, sizeof(info_k));
634
635 error = vfs_quotactl_idtypestat(mp, idtype, &info_k);
636 if (error) {
637 return error;
638 }
639
640 return copyout(&info_k, info_u, sizeof(info_k));
641 }
642
643 static int
644 do_sys_quotactl_objtypestat(struct mount *mp, int objtype,
645 struct quotaobjtypestat *info_u)
646 {
647 struct quotaobjtypestat info_k;
648 int error;
649
650 /* ensure any padding bytes are cleared */
651 memset(&info_k, 0, sizeof(info_k));
652
653 error = vfs_quotactl_objtypestat(mp, objtype, &info_k);
654 if (error) {
655 return error;
656 }
657
658 return copyout(&info_k, info_u, sizeof(info_k));
659 }
660
661 static int
662 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u,
663 struct quotaval *val_u)
664 {
665 struct quotakey key_k;
666 struct quotaval val_k;
667 int error;
668
669 /* ensure any padding bytes are cleared */
670 memset(&val_k, 0, sizeof(val_k));
671
672 error = copyin(key_u, &key_k, sizeof(key_k));
673 if (error) {
674 return error;
675 }
676
677 error = vfs_quotactl_get(mp, &key_k, &val_k);
678 if (error) {
679 return error;
680 }
681
682 return copyout(&val_k, val_u, sizeof(val_k));
683 }
684
685 static int
686 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u,
687 const struct quotaval *val_u)
688 {
689 struct quotakey key_k;
690 struct quotaval val_k;
691 int error;
692
693 error = copyin(key_u, &key_k, sizeof(key_k));
694 if (error) {
695 return error;
696 }
697
698 error = copyin(val_u, &val_k, sizeof(val_k));
699 if (error) {
700 return error;
701 }
702
703 return vfs_quotactl_put(mp, &key_k, &val_k);
704 }
705
706 static int
707 do_sys_quotactl_delete(struct mount *mp, const struct quotakey *key_u)
708 {
709 struct quotakey key_k;
710 int error;
711
712 error = copyin(key_u, &key_k, sizeof(key_k));
713 if (error) {
714 return error;
715 }
716
717 return vfs_quotactl_delete(mp, &key_k);
718 }
719
720 static int
721 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u)
722 {
723 struct quotakcursor cursor_k;
724 int error;
725
726 /* ensure any padding bytes are cleared */
727 memset(&cursor_k, 0, sizeof(cursor_k));
728
729 error = vfs_quotactl_cursoropen(mp, &cursor_k);
730 if (error) {
731 return error;
732 }
733
734 return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
735 }
736
737 static int
738 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u)
739 {
740 struct quotakcursor cursor_k;
741 int error;
742
743 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
744 if (error) {
745 return error;
746 }
747
748 return vfs_quotactl_cursorclose(mp, &cursor_k);
749 }
750
751 static int
752 do_sys_quotactl_cursorskipidtype(struct mount *mp,
753 struct quotakcursor *cursor_u, int idtype)
754 {
755 struct quotakcursor cursor_k;
756 int error;
757
758 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
759 if (error) {
760 return error;
761 }
762
763 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype);
764 if (error) {
765 return error;
766 }
767
768 return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
769 }
770
771 static int
772 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u,
773 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum,
774 unsigned *ret_u)
775 {
776 #define CGET_STACK_MAX 8
777 struct quotakcursor cursor_k;
778 struct quotakey stackkeys[CGET_STACK_MAX];
779 struct quotaval stackvals[CGET_STACK_MAX];
780 struct quotakey *keys_k;
781 struct quotaval *vals_k;
782 unsigned ret_k;
783 int error;
784
785 if (maxnum > 128) {
786 maxnum = 128;
787 }
788
789 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
790 if (error) {
791 return error;
792 }
793
794 if (maxnum <= CGET_STACK_MAX) {
795 keys_k = stackkeys;
796 vals_k = stackvals;
797 /* ensure any padding bytes are cleared */
798 memset(keys_k, 0, maxnum * sizeof(keys_k[0]));
799 memset(vals_k, 0, maxnum * sizeof(vals_k[0]));
800 } else {
801 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP);
802 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP);
803 }
804
805 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum,
806 &ret_k);
807 if (error) {
808 goto fail;
809 }
810
811 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0]));
812 if (error) {
813 goto fail;
814 }
815
816 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0]));
817 if (error) {
818 goto fail;
819 }
820
821 error = copyout(&ret_k, ret_u, sizeof(ret_k));
822 if (error) {
823 goto fail;
824 }
825
826 /* do last to maximize the chance of being able to recover a failure */
827 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k));
828
829 fail:
830 if (keys_k != stackkeys) {
831 kmem_free(keys_k, maxnum * sizeof(keys_k[0]));
832 }
833 if (vals_k != stackvals) {
834 kmem_free(vals_k, maxnum * sizeof(vals_k[0]));
835 }
836 return error;
837 }
838
839 static int
840 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u,
841 int *ret_u)
842 {
843 struct quotakcursor cursor_k;
844 int ret_k;
845 int error;
846
847 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
848 if (error) {
849 return error;
850 }
851
852 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k);
853 if (error) {
854 return error;
855 }
856
857 error = copyout(&ret_k, ret_u, sizeof(ret_k));
858 if (error) {
859 return error;
860 }
861
862 return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
863 }
864
865 static int
866 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u)
867 {
868 struct quotakcursor cursor_k;
869 int error;
870
871 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
872 if (error) {
873 return error;
874 }
875
876 error = vfs_quotactl_cursorrewind(mp, &cursor_k);
877 if (error) {
878 return error;
879 }
880
881 return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
882 }
883
884 static int
885 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u)
886 {
887 char *path_k;
888 int error;
889
890 /* XXX this should probably be a struct pathbuf */
891 path_k = PNBUF_GET();
892 error = copyin(path_u, path_k, PATH_MAX);
893 if (error) {
894 PNBUF_PUT(path_k);
895 return error;
896 }
897
898 error = vfs_quotactl_quotaon(mp, idtype, path_k);
899
900 PNBUF_PUT(path_k);
901 return error;
902 }
903
904 static int
905 do_sys_quotactl_quotaoff(struct mount *mp, int idtype)
906 {
907 return vfs_quotactl_quotaoff(mp, idtype);
908 }
909
910 int
911 do_sys_quotactl(const char *path_u, const struct quotactl_args *args)
912 {
913 struct mount *mp;
914 struct vnode *vp;
915 int error;
916
917 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp);
918 if (error != 0)
919 return (error);
920 mp = vp->v_mount;
921
922 switch (args->qc_op) {
923 case QUOTACTL_STAT:
924 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info);
925 break;
926 case QUOTACTL_IDTYPESTAT:
927 error = do_sys_quotactl_idtypestat(mp,
928 args->u.idtypestat.qc_idtype,
929 args->u.idtypestat.qc_info);
930 break;
931 case QUOTACTL_OBJTYPESTAT:
932 error = do_sys_quotactl_objtypestat(mp,
933 args->u.objtypestat.qc_objtype,
934 args->u.objtypestat.qc_info);
935 break;
936 case QUOTACTL_GET:
937 error = do_sys_quotactl_get(mp,
938 args->u.get.qc_key,
939 args->u.get.qc_val);
940 break;
941 case QUOTACTL_PUT:
942 error = do_sys_quotactl_put(mp,
943 args->u.put.qc_key,
944 args->u.put.qc_val);
945 break;
946 case QUOTACTL_DELETE:
947 error = do_sys_quotactl_delete(mp, args->u.delete.qc_key);
948 break;
949 case QUOTACTL_CURSOROPEN:
950 error = do_sys_quotactl_cursoropen(mp,
951 args->u.cursoropen.qc_cursor);
952 break;
953 case QUOTACTL_CURSORCLOSE:
954 error = do_sys_quotactl_cursorclose(mp,
955 args->u.cursorclose.qc_cursor);
956 break;
957 case QUOTACTL_CURSORSKIPIDTYPE:
958 error = do_sys_quotactl_cursorskipidtype(mp,
959 args->u.cursorskipidtype.qc_cursor,
960 args->u.cursorskipidtype.qc_idtype);
961 break;
962 case QUOTACTL_CURSORGET:
963 error = do_sys_quotactl_cursorget(mp,
964 args->u.cursorget.qc_cursor,
965 args->u.cursorget.qc_keys,
966 args->u.cursorget.qc_vals,
967 args->u.cursorget.qc_maxnum,
968 args->u.cursorget.qc_ret);
969 break;
970 case QUOTACTL_CURSORATEND:
971 error = do_sys_quotactl_cursoratend(mp,
972 args->u.cursoratend.qc_cursor,
973 args->u.cursoratend.qc_ret);
974 break;
975 case QUOTACTL_CURSORREWIND:
976 error = do_sys_quotactl_cursorrewind(mp,
977 args->u.cursorrewind.qc_cursor);
978 break;
979 case QUOTACTL_QUOTAON:
980 error = do_sys_quotactl_quotaon(mp,
981 args->u.quotaon.qc_idtype,
982 args->u.quotaon.qc_quotafile);
983 break;
984 case QUOTACTL_QUOTAOFF:
985 error = do_sys_quotactl_quotaoff(mp,
986 args->u.quotaoff.qc_idtype);
987 break;
988 default:
989 error = EINVAL;
990 break;
991 }
992
993 vrele(vp);
994 return error;
995 }
996
997 /* ARGSUSED */
998 int
999 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap,
1000 register_t *retval)
1001 {
1002 /* {
1003 syscallarg(const char *) path;
1004 syscallarg(struct quotactl_args *) args;
1005 } */
1006 struct quotactl_args args;
1007 int error;
1008
1009 error = copyin(SCARG(uap, args), &args, sizeof(args));
1010 if (error) {
1011 return error;
1012 }
1013
1014 return do_sys_quotactl(SCARG(uap, path), &args);
1015 }
1016
1017 int
1018 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
1019 int root)
1020 {
1021 struct cwdinfo *cwdi = l->l_proc->p_cwdi;
1022 int error = 0;
1023
1024 /*
1025 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1026 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
1027 * overrides MNT_NOWAIT.
1028 */
1029 if (flags == MNT_NOWAIT || flags == MNT_LAZY ||
1030 (flags != MNT_WAIT && flags != 0)) {
1031 memcpy(sp, &mp->mnt_stat, sizeof(*sp));
1032 goto done;
1033 }
1034
1035 /* Get the filesystem stats now */
1036 memset(sp, 0, sizeof(*sp));
1037 if ((error = VFS_STATVFS(mp, sp)) != 0) {
1038 return error;
1039 }
1040
1041 if (cwdi->cwdi_rdir == NULL)
1042 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
1043 done:
1044 if (cwdi->cwdi_rdir != NULL) {
1045 size_t len;
1046 char *bp;
1047 char c;
1048 char *path = PNBUF_GET();
1049
1050 bp = path + MAXPATHLEN;
1051 *--bp = '\0';
1052 rw_enter(&cwdi->cwdi_lock, RW_READER);
1053 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
1054 MAXPATHLEN / 2, 0, l);
1055 rw_exit(&cwdi->cwdi_lock);
1056 if (error) {
1057 PNBUF_PUT(path);
1058 return error;
1059 }
1060 len = strlen(bp);
1061 if (len != 1) {
1062 /*
1063 * for mount points that are below our root, we can see
1064 * them, so we fix up the pathname and return them. The
1065 * rest we cannot see, so we don't allow viewing the
1066 * data.
1067 */
1068 if (strncmp(bp, sp->f_mntonname, len) == 0 &&
1069 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
1070 (void)strlcpy(sp->f_mntonname,
1071 c == '\0' ? "/" : &sp->f_mntonname[len],
1072 sizeof(sp->f_mntonname));
1073 } else {
1074 if (root)
1075 (void)strlcpy(sp->f_mntonname, "/",
1076 sizeof(sp->f_mntonname));
1077 else
1078 error = EPERM;
1079 }
1080 }
1081 PNBUF_PUT(path);
1082 }
1083 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
1084 return error;
1085 }
1086
1087 /*
1088 * Get filesystem statistics by path.
1089 */
1090 int
1091 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
1092 {
1093 struct mount *mp;
1094 int error;
1095 struct vnode *vp;
1096
1097 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
1098 if (error != 0)
1099 return error;
1100 mp = vp->v_mount;
1101 error = dostatvfs(mp, sb, l, flags, 1);
1102 vrele(vp);
1103 return error;
1104 }
1105
1106 /* ARGSUSED */
1107 int
1108 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
1109 {
1110 /* {
1111 syscallarg(const char *) path;
1112 syscallarg(struct statvfs *) buf;
1113 syscallarg(int) flags;
1114 } */
1115 struct statvfs *sb;
1116 int error;
1117
1118 sb = STATVFSBUF_GET();
1119 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
1120 if (error == 0)
1121 error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1122 STATVFSBUF_PUT(sb);
1123 return error;
1124 }
1125
1126 /*
1127 * Get filesystem statistics by fd.
1128 */
1129 int
1130 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
1131 {
1132 file_t *fp;
1133 struct mount *mp;
1134 int error;
1135
1136 /* fd_getvnode() will use the descriptor for us */
1137 if ((error = fd_getvnode(fd, &fp)) != 0)
1138 return (error);
1139 mp = ((struct vnode *)fp->f_data)->v_mount;
1140 error = dostatvfs(mp, sb, curlwp, flags, 1);
1141 fd_putfile(fd);
1142 return error;
1143 }
1144
1145 /* ARGSUSED */
1146 int
1147 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
1148 {
1149 /* {
1150 syscallarg(int) fd;
1151 syscallarg(struct statvfs *) buf;
1152 syscallarg(int) flags;
1153 } */
1154 struct statvfs *sb;
1155 int error;
1156
1157 sb = STATVFSBUF_GET();
1158 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
1159 if (error == 0)
1160 error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1161 STATVFSBUF_PUT(sb);
1162 return error;
1163 }
1164
1165
1166 /*
1167 * Get statistics on all filesystems.
1168 */
1169 int
1170 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
1171 int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
1172 register_t *retval)
1173 {
1174 int root = 0;
1175 struct proc *p = l->l_proc;
1176 struct mount *mp, *nmp;
1177 struct statvfs *sb;
1178 size_t count, maxcount;
1179 int error = 0;
1180
1181 sb = STATVFSBUF_GET();
1182 maxcount = bufsize / entry_sz;
1183 mutex_enter(&mountlist_lock);
1184 count = 0;
1185 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
1186 mp = nmp) {
1187 if (vfs_busy(mp, &nmp)) {
1188 continue;
1189 }
1190 if (sfsp && count < maxcount) {
1191 error = dostatvfs(mp, sb, l, flags, 0);
1192 if (error) {
1193 vfs_unbusy(mp, false, &nmp);
1194 error = 0;
1195 continue;
1196 }
1197 error = copyfn(sb, sfsp, entry_sz);
1198 if (error) {
1199 vfs_unbusy(mp, false, NULL);
1200 goto out;
1201 }
1202 sfsp = (char *)sfsp + entry_sz;
1203 root |= strcmp(sb->f_mntonname, "/") == 0;
1204 }
1205 count++;
1206 vfs_unbusy(mp, false, &nmp);
1207 }
1208 mutex_exit(&mountlist_lock);
1209
1210 if (root == 0 && p->p_cwdi->cwdi_rdir) {
1211 /*
1212 * fake a root entry
1213 */
1214 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1215 sb, l, flags, 1);
1216 if (error != 0)
1217 goto out;
1218 if (sfsp) {
1219 error = copyfn(sb, sfsp, entry_sz);
1220 if (error != 0)
1221 goto out;
1222 }
1223 count++;
1224 }
1225 if (sfsp && count > maxcount)
1226 *retval = maxcount;
1227 else
1228 *retval = count;
1229 out:
1230 STATVFSBUF_PUT(sb);
1231 return error;
1232 }
1233
1234 int
1235 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
1236 {
1237 /* {
1238 syscallarg(struct statvfs *) buf;
1239 syscallarg(size_t) bufsize;
1240 syscallarg(int) flags;
1241 } */
1242
1243 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1244 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1245 }
1246
1247 /*
1248 * Change current working directory to a given file descriptor.
1249 */
1250 /* ARGSUSED */
1251 int
1252 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
1253 {
1254 /* {
1255 syscallarg(int) fd;
1256 } */
1257 struct proc *p = l->l_proc;
1258 struct cwdinfo *cwdi;
1259 struct vnode *vp, *tdp;
1260 struct mount *mp;
1261 file_t *fp;
1262 int error, fd;
1263
1264 /* fd_getvnode() will use the descriptor for us */
1265 fd = SCARG(uap, fd);
1266 if ((error = fd_getvnode(fd, &fp)) != 0)
1267 return (error);
1268 vp = fp->f_data;
1269
1270 vref(vp);
1271 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1272 if (vp->v_type != VDIR)
1273 error = ENOTDIR;
1274 else
1275 error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1276 if (error) {
1277 vput(vp);
1278 goto out;
1279 }
1280 while ((mp = vp->v_mountedhere) != NULL) {
1281 error = vfs_busy(mp, NULL);
1282 vput(vp);
1283 if (error != 0)
1284 goto out;
1285 error = VFS_ROOT(mp, &tdp);
1286 vfs_unbusy(mp, false, NULL);
1287 if (error)
1288 goto out;
1289 vp = tdp;
1290 }
1291 VOP_UNLOCK(vp);
1292
1293 /*
1294 * Disallow changing to a directory not under the process's
1295 * current root directory (if there is one).
1296 */
1297 cwdi = p->p_cwdi;
1298 rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1299 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1300 vrele(vp);
1301 error = EPERM; /* operation not permitted */
1302 } else {
1303 vrele(cwdi->cwdi_cdir);
1304 cwdi->cwdi_cdir = vp;
1305 }
1306 rw_exit(&cwdi->cwdi_lock);
1307
1308 out:
1309 fd_putfile(fd);
1310 return (error);
1311 }
1312
1313 /*
1314 * Change this process's notion of the root directory to a given file
1315 * descriptor.
1316 */
1317 int
1318 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
1319 {
1320 struct proc *p = l->l_proc;
1321 struct vnode *vp;
1322 file_t *fp;
1323 int error, fd = SCARG(uap, fd);
1324
1325 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1326 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1327 return error;
1328 /* fd_getvnode() will use the descriptor for us */
1329 if ((error = fd_getvnode(fd, &fp)) != 0)
1330 return error;
1331 vp = fp->f_data;
1332 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1333 if (vp->v_type != VDIR)
1334 error = ENOTDIR;
1335 else
1336 error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1337 VOP_UNLOCK(vp);
1338 if (error)
1339 goto out;
1340 vref(vp);
1341
1342 change_root(p->p_cwdi, vp, l);
1343
1344 out:
1345 fd_putfile(fd);
1346 return (error);
1347 }
1348
1349 /*
1350 * Change current working directory (``.'').
1351 */
1352 /* ARGSUSED */
1353 int
1354 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
1355 {
1356 /* {
1357 syscallarg(const char *) path;
1358 } */
1359 struct proc *p = l->l_proc;
1360 struct cwdinfo *cwdi;
1361 int error;
1362 struct vnode *vp;
1363
1364 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1365 &vp, l)) != 0)
1366 return (error);
1367 cwdi = p->p_cwdi;
1368 rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1369 vrele(cwdi->cwdi_cdir);
1370 cwdi->cwdi_cdir = vp;
1371 rw_exit(&cwdi->cwdi_lock);
1372 return (0);
1373 }
1374
1375 /*
1376 * Change notion of root (``/'') directory.
1377 */
1378 /* ARGSUSED */
1379 int
1380 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1381 {
1382 /* {
1383 syscallarg(const char *) path;
1384 } */
1385 struct proc *p = l->l_proc;
1386 int error;
1387 struct vnode *vp;
1388
1389 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1390 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1391 return (error);
1392 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1393 &vp, l)) != 0)
1394 return (error);
1395
1396 change_root(p->p_cwdi, vp, l);
1397
1398 return (0);
1399 }
1400
1401 /*
1402 * Common routine for chroot and fchroot.
1403 * NB: callers need to properly authorize the change root operation.
1404 */
1405 void
1406 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l)
1407 {
1408
1409 rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1410 if (cwdi->cwdi_rdir != NULL)
1411 vrele(cwdi->cwdi_rdir);
1412 cwdi->cwdi_rdir = vp;
1413
1414 /*
1415 * Prevent escaping from chroot by putting the root under
1416 * the working directory. Silently chdir to / if we aren't
1417 * already there.
1418 */
1419 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1420 /*
1421 * XXX would be more failsafe to change directory to a
1422 * deadfs node here instead
1423 */
1424 vrele(cwdi->cwdi_cdir);
1425 vref(vp);
1426 cwdi->cwdi_cdir = vp;
1427 }
1428 rw_exit(&cwdi->cwdi_lock);
1429 }
1430
1431 /*
1432 * Common routine for chroot and chdir.
1433 * XXX "where" should be enum uio_seg
1434 */
1435 int
1436 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l)
1437 {
1438 struct pathbuf *pb;
1439 struct nameidata nd;
1440 int error;
1441
1442 error = pathbuf_maybe_copyin(path, where, &pb);
1443 if (error) {
1444 return error;
1445 }
1446 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
1447 if ((error = namei(&nd)) != 0) {
1448 pathbuf_destroy(pb);
1449 return error;
1450 }
1451 *vpp = nd.ni_vp;
1452 pathbuf_destroy(pb);
1453
1454 if ((*vpp)->v_type != VDIR)
1455 error = ENOTDIR;
1456 else
1457 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred);
1458
1459 if (error)
1460 vput(*vpp);
1461 else
1462 VOP_UNLOCK(*vpp);
1463 return (error);
1464 }
1465
1466 /*
1467 * Internals of sys_open - path has already been converted into a pathbuf
1468 * (so we can easily reuse this function from other parts of the kernel,
1469 * like posix_spawn post-processing).
1470 */
1471 static int
1472 do_open(lwp_t *l, struct pathbuf *pb, int open_flags, int open_mode, int *fd)
1473 {
1474 struct proc *p = l->l_proc;
1475 struct cwdinfo *cwdi = p->p_cwdi;
1476 file_t *fp;
1477 struct vnode *vp;
1478 int flags, cmode;
1479 int indx, error;
1480 struct nameidata nd;
1481
1482 flags = FFLAGS(open_flags);
1483 if ((flags & (FREAD | FWRITE)) == 0)
1484 return EINVAL;
1485
1486 if ((error = fd_allocfile(&fp, &indx)) != 0) {
1487 return error;
1488 }
1489
1490 /* We're going to read cwdi->cwdi_cmask unlocked here. */
1491 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1492 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb);
1493 l->l_dupfd = -indx - 1; /* XXX check for fdopen */
1494 if ((error = vn_open(&nd, flags, cmode)) != 0) {
1495 fd_abort(p, fp, indx);
1496 if ((error == EDUPFD || error == EMOVEFD) &&
1497 l->l_dupfd >= 0 && /* XXX from fdopen */
1498 (error =
1499 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1500 *fd = indx;
1501 return 0;
1502 }
1503 if (error == ERESTART)
1504 error = EINTR;
1505 return error;
1506 }
1507
1508 l->l_dupfd = 0;
1509 vp = nd.ni_vp;
1510
1511 if ((error = open_setfp(l, fp, vp, indx, flags)))
1512 return error;
1513
1514 VOP_UNLOCK(vp);
1515 *fd = indx;
1516 fd_affix(p, fp, indx);
1517 return 0;
1518 }
1519
1520 int
1521 fd_open(const char *path, int open_flags, int open_mode, int *fd)
1522 {
1523 struct pathbuf *pb;
1524 int error, oflags;
1525
1526 oflags = FFLAGS(open_flags);
1527 if ((oflags & (FREAD | FWRITE)) == 0)
1528 return EINVAL;
1529
1530 pb = pathbuf_create(path);
1531 if (pb == NULL)
1532 return ENOMEM;
1533
1534 error = do_open(curlwp, pb, open_flags, open_mode, fd);
1535 pathbuf_destroy(pb);
1536
1537 return error;
1538 }
1539
1540 /*
1541 * Check permissions, allocate an open file structure,
1542 * and call the device open routine if any.
1543 */
1544 int
1545 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1546 {
1547 /* {
1548 syscallarg(const char *) path;
1549 syscallarg(int) flags;
1550 syscallarg(int) mode;
1551 } */
1552 struct pathbuf *pb;
1553 int result, flags, error;
1554
1555 flags = FFLAGS(SCARG(uap, flags));
1556 if ((flags & (FREAD | FWRITE)) == 0)
1557 return EINVAL;
1558
1559 error = pathbuf_copyin(SCARG(uap, path), &pb);
1560 if (error)
1561 return error;
1562
1563 error = do_open(l, pb, SCARG(uap, flags), SCARG(uap, mode), &result);
1564 pathbuf_destroy(pb);
1565
1566 *retval = result;
1567 return error;
1568 }
1569
1570 int
1571 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval)
1572 {
1573 /* {
1574 syscallarg(int) fd;
1575 syscallarg(const char *) path;
1576 syscallarg(int) flags;
1577 syscallarg(int) mode;
1578 } */
1579
1580 return ENOSYS;
1581 }
1582
1583 static void
1584 vfs__fhfree(fhandle_t *fhp)
1585 {
1586 size_t fhsize;
1587
1588 if (fhp == NULL) {
1589 return;
1590 }
1591 fhsize = FHANDLE_SIZE(fhp);
1592 kmem_free(fhp, fhsize);
1593 }
1594
1595 /*
1596 * vfs_composefh: compose a filehandle.
1597 */
1598
1599 int
1600 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1601 {
1602 struct mount *mp;
1603 struct fid *fidp;
1604 int error;
1605 size_t needfhsize;
1606 size_t fidsize;
1607
1608 mp = vp->v_mount;
1609 fidp = NULL;
1610 if (*fh_size < FHANDLE_SIZE_MIN) {
1611 fidsize = 0;
1612 } else {
1613 fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1614 if (fhp != NULL) {
1615 memset(fhp, 0, *fh_size);
1616 fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1617 fidp = &fhp->fh_fid;
1618 }
1619 }
1620 error = VFS_VPTOFH(vp, fidp, &fidsize);
1621 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1622 if (error == 0 && *fh_size < needfhsize) {
1623 error = E2BIG;
1624 }
1625 *fh_size = needfhsize;
1626 return error;
1627 }
1628
1629 int
1630 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1631 {
1632 struct mount *mp;
1633 fhandle_t *fhp;
1634 size_t fhsize;
1635 size_t fidsize;
1636 int error;
1637
1638 *fhpp = NULL;
1639 mp = vp->v_mount;
1640 fidsize = 0;
1641 error = VFS_VPTOFH(vp, NULL, &fidsize);
1642 KASSERT(error != 0);
1643 if (error != E2BIG) {
1644 goto out;
1645 }
1646 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1647 fhp = kmem_zalloc(fhsize, KM_SLEEP);
1648 if (fhp == NULL) {
1649 error = ENOMEM;
1650 goto out;
1651 }
1652 fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1653 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1654 if (error == 0) {
1655 KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1656 FHANDLE_FILEID(fhp)->fid_len == fidsize));
1657 *fhpp = fhp;
1658 } else {
1659 kmem_free(fhp, fhsize);
1660 }
1661 out:
1662 return error;
1663 }
1664
1665 void
1666 vfs_composefh_free(fhandle_t *fhp)
1667 {
1668
1669 vfs__fhfree(fhp);
1670 }
1671
1672 /*
1673 * vfs_fhtovp: lookup a vnode by a filehandle.
1674 */
1675
1676 int
1677 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1678 {
1679 struct mount *mp;
1680 int error;
1681
1682 *vpp = NULL;
1683 mp = vfs_getvfs(FHANDLE_FSID(fhp));
1684 if (mp == NULL) {
1685 error = ESTALE;
1686 goto out;
1687 }
1688 if (mp->mnt_op->vfs_fhtovp == NULL) {
1689 error = EOPNOTSUPP;
1690 goto out;
1691 }
1692 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1693 out:
1694 return error;
1695 }
1696
1697 /*
1698 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1699 * the needed size.
1700 */
1701
1702 int
1703 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1704 {
1705 fhandle_t *fhp;
1706 int error;
1707
1708 *fhpp = NULL;
1709 if (fhsize > FHANDLE_SIZE_MAX) {
1710 return EINVAL;
1711 }
1712 if (fhsize < FHANDLE_SIZE_MIN) {
1713 return EINVAL;
1714 }
1715 again:
1716 fhp = kmem_alloc(fhsize, KM_SLEEP);
1717 if (fhp == NULL) {
1718 return ENOMEM;
1719 }
1720 error = copyin(ufhp, fhp, fhsize);
1721 if (error == 0) {
1722 /* XXX this check shouldn't be here */
1723 if (FHANDLE_SIZE(fhp) == fhsize) {
1724 *fhpp = fhp;
1725 return 0;
1726 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1727 /*
1728 * a kludge for nfsv2 padded handles.
1729 */
1730 size_t sz;
1731
1732 sz = FHANDLE_SIZE(fhp);
1733 kmem_free(fhp, fhsize);
1734 fhsize = sz;
1735 goto again;
1736 } else {
1737 /*
1738 * userland told us wrong size.
1739 */
1740 error = EINVAL;
1741 }
1742 }
1743 kmem_free(fhp, fhsize);
1744 return error;
1745 }
1746
1747 void
1748 vfs_copyinfh_free(fhandle_t *fhp)
1749 {
1750
1751 vfs__fhfree(fhp);
1752 }
1753
1754 /*
1755 * Get file handle system call
1756 */
1757 int
1758 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1759 {
1760 /* {
1761 syscallarg(char *) fname;
1762 syscallarg(fhandle_t *) fhp;
1763 syscallarg(size_t *) fh_size;
1764 } */
1765 struct vnode *vp;
1766 fhandle_t *fh;
1767 int error;
1768 struct pathbuf *pb;
1769 struct nameidata nd;
1770 size_t sz;
1771 size_t usz;
1772
1773 /*
1774 * Must be super user
1775 */
1776 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1777 0, NULL, NULL, NULL);
1778 if (error)
1779 return (error);
1780
1781 error = pathbuf_copyin(SCARG(uap, fname), &pb);
1782 if (error) {
1783 return error;
1784 }
1785 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
1786 error = namei(&nd);
1787 if (error) {
1788 pathbuf_destroy(pb);
1789 return error;
1790 }
1791 vp = nd.ni_vp;
1792 pathbuf_destroy(pb);
1793
1794 error = vfs_composefh_alloc(vp, &fh);
1795 vput(vp);
1796 if (error != 0) {
1797 goto out;
1798 }
1799 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1800 if (error != 0) {
1801 goto out;
1802 }
1803 sz = FHANDLE_SIZE(fh);
1804 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1805 if (error != 0) {
1806 goto out;
1807 }
1808 if (usz >= sz) {
1809 error = copyout(fh, SCARG(uap, fhp), sz);
1810 } else {
1811 error = E2BIG;
1812 }
1813 out:
1814 vfs_composefh_free(fh);
1815 return (error);
1816 }
1817
1818 /*
1819 * Open a file given a file handle.
1820 *
1821 * Check permissions, allocate an open file structure,
1822 * and call the device open routine if any.
1823 */
1824
1825 int
1826 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1827 register_t *retval)
1828 {
1829 file_t *fp;
1830 struct vnode *vp = NULL;
1831 kauth_cred_t cred = l->l_cred;
1832 file_t *nfp;
1833 int indx, error = 0;
1834 struct vattr va;
1835 fhandle_t *fh;
1836 int flags;
1837 proc_t *p;
1838
1839 p = curproc;
1840
1841 /*
1842 * Must be super user
1843 */
1844 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1845 0, NULL, NULL, NULL)))
1846 return (error);
1847
1848 flags = FFLAGS(oflags);
1849 if ((flags & (FREAD | FWRITE)) == 0)
1850 return (EINVAL);
1851 if ((flags & O_CREAT))
1852 return (EINVAL);
1853 if ((error = fd_allocfile(&nfp, &indx)) != 0)
1854 return (error);
1855 fp = nfp;
1856 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1857 if (error != 0) {
1858 goto bad;
1859 }
1860 error = vfs_fhtovp(fh, &vp);
1861 if (error != 0) {
1862 goto bad;
1863 }
1864
1865 /* Now do an effective vn_open */
1866
1867 if (vp->v_type == VSOCK) {
1868 error = EOPNOTSUPP;
1869 goto bad;
1870 }
1871 error = vn_openchk(vp, cred, flags);
1872 if (error != 0)
1873 goto bad;
1874 if (flags & O_TRUNC) {
1875 VOP_UNLOCK(vp); /* XXX */
1876 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
1877 vattr_null(&va);
1878 va.va_size = 0;
1879 error = VOP_SETATTR(vp, &va, cred);
1880 if (error)
1881 goto bad;
1882 }
1883 if ((error = VOP_OPEN(vp, flags, cred)) != 0)
1884 goto bad;
1885 if (flags & FWRITE) {
1886 mutex_enter(vp->v_interlock);
1887 vp->v_writecount++;
1888 mutex_exit(vp->v_interlock);
1889 }
1890
1891 /* done with modified vn_open, now finish what sys_open does. */
1892 if ((error = open_setfp(l, fp, vp, indx, flags)))
1893 return error;
1894
1895 VOP_UNLOCK(vp);
1896 *retval = indx;
1897 fd_affix(p, fp, indx);
1898 vfs_copyinfh_free(fh);
1899 return (0);
1900
1901 bad:
1902 fd_abort(p, fp, indx);
1903 if (vp != NULL)
1904 vput(vp);
1905 vfs_copyinfh_free(fh);
1906 return (error);
1907 }
1908
1909 int
1910 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
1911 {
1912 /* {
1913 syscallarg(const void *) fhp;
1914 syscallarg(size_t) fh_size;
1915 syscallarg(int) flags;
1916 } */
1917
1918 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1919 SCARG(uap, flags), retval);
1920 }
1921
1922 int
1923 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1924 {
1925 int error;
1926 fhandle_t *fh;
1927 struct vnode *vp;
1928
1929 /*
1930 * Must be super user
1931 */
1932 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1933 0, NULL, NULL, NULL)))
1934 return (error);
1935
1936 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1937 if (error != 0)
1938 return error;
1939
1940 error = vfs_fhtovp(fh, &vp);
1941 vfs_copyinfh_free(fh);
1942 if (error != 0)
1943 return error;
1944
1945 error = vn_stat(vp, sb);
1946 vput(vp);
1947 return error;
1948 }
1949
1950
1951 /* ARGSUSED */
1952 int
1953 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval)
1954 {
1955 /* {
1956 syscallarg(const void *) fhp;
1957 syscallarg(size_t) fh_size;
1958 syscallarg(struct stat *) sb;
1959 } */
1960 struct stat sb;
1961 int error;
1962
1963 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1964 if (error)
1965 return error;
1966 return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1967 }
1968
1969 int
1970 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1971 int flags)
1972 {
1973 fhandle_t *fh;
1974 struct mount *mp;
1975 struct vnode *vp;
1976 int error;
1977
1978 /*
1979 * Must be super user
1980 */
1981 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1982 0, NULL, NULL, NULL)))
1983 return error;
1984
1985 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1986 if (error != 0)
1987 return error;
1988
1989 error = vfs_fhtovp(fh, &vp);
1990 vfs_copyinfh_free(fh);
1991 if (error != 0)
1992 return error;
1993
1994 mp = vp->v_mount;
1995 error = dostatvfs(mp, sb, l, flags, 1);
1996 vput(vp);
1997 return error;
1998 }
1999
2000 /* ARGSUSED */
2001 int
2002 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
2003 {
2004 /* {
2005 syscallarg(const void *) fhp;
2006 syscallarg(size_t) fh_size;
2007 syscallarg(struct statvfs *) buf;
2008 syscallarg(int) flags;
2009 } */
2010 struct statvfs *sb = STATVFSBUF_GET();
2011 int error;
2012
2013 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
2014 SCARG(uap, flags));
2015 if (error == 0)
2016 error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
2017 STATVFSBUF_PUT(sb);
2018 return error;
2019 }
2020
2021 /*
2022 * Create a special file.
2023 */
2024 /* ARGSUSED */
2025 int
2026 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap,
2027 register_t *retval)
2028 {
2029 /* {
2030 syscallarg(const char *) path;
2031 syscallarg(mode_t) mode;
2032 syscallarg(dev_t) dev;
2033 } */
2034 return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode),
2035 SCARG(uap, dev), retval, UIO_USERSPACE);
2036 }
2037
2038 int
2039 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap,
2040 register_t *retval)
2041 {
2042 /* {
2043 syscallarg(int) fd;
2044 syscallarg(const char *) path;
2045 syscallarg(mode_t) mode;
2046 syscallarg(uint32_t) dev;
2047 } */
2048
2049 return ENOSYS;
2050 }
2051
2052 int
2053 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev,
2054 register_t *retval, enum uio_seg seg)
2055 {
2056 struct proc *p = l->l_proc;
2057 struct vnode *vp;
2058 struct vattr vattr;
2059 int error, optype;
2060 struct pathbuf *pb;
2061 struct nameidata nd;
2062 const char *pathstring;
2063
2064 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
2065 0, NULL, NULL, NULL)) != 0)
2066 return (error);
2067
2068 optype = VOP_MKNOD_DESCOFFSET;
2069
2070 error = pathbuf_maybe_copyin(pathname, seg, &pb);
2071 if (error) {
2072 return error;
2073 }
2074 pathstring = pathbuf_stringcopy_get(pb);
2075 if (pathstring == NULL) {
2076 pathbuf_destroy(pb);
2077 return ENOMEM;
2078 }
2079
2080 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb);
2081 if ((error = namei(&nd)) != 0)
2082 goto out;
2083 vp = nd.ni_vp;
2084
2085 if (vp != NULL)
2086 error = EEXIST;
2087 else {
2088 vattr_null(&vattr);
2089 /* We will read cwdi->cwdi_cmask unlocked. */
2090 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
2091 vattr.va_rdev = dev;
2092
2093 switch (mode & S_IFMT) {
2094 case S_IFMT: /* used by badsect to flag bad sectors */
2095 vattr.va_type = VBAD;
2096 break;
2097 case S_IFCHR:
2098 vattr.va_type = VCHR;
2099 break;
2100 case S_IFBLK:
2101 vattr.va_type = VBLK;
2102 break;
2103 case S_IFWHT:
2104 optype = VOP_WHITEOUT_DESCOFFSET;
2105 break;
2106 case S_IFREG:
2107 #if NVERIEXEC > 0
2108 error = veriexec_openchk(l, nd.ni_vp, pathstring,
2109 O_CREAT);
2110 #endif /* NVERIEXEC > 0 */
2111 vattr.va_type = VREG;
2112 vattr.va_rdev = VNOVAL;
2113 optype = VOP_CREATE_DESCOFFSET;
2114 break;
2115 default:
2116 error = EINVAL;
2117 break;
2118 }
2119 }
2120 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET
2121 && vattr.va_rdev == VNOVAL)
2122 error = EINVAL;
2123 if (!error) {
2124 switch (optype) {
2125 case VOP_WHITEOUT_DESCOFFSET:
2126 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
2127 if (error)
2128 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2129 vput(nd.ni_dvp);
2130 break;
2131
2132 case VOP_MKNOD_DESCOFFSET:
2133 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
2134 &nd.ni_cnd, &vattr);
2135 if (error == 0)
2136 vput(nd.ni_vp);
2137 break;
2138
2139 case VOP_CREATE_DESCOFFSET:
2140 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
2141 &nd.ni_cnd, &vattr);
2142 if (error == 0)
2143 vput(nd.ni_vp);
2144 break;
2145 }
2146 } else {
2147 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2148 if (nd.ni_dvp == vp)
2149 vrele(nd.ni_dvp);
2150 else
2151 vput(nd.ni_dvp);
2152 if (vp)
2153 vrele(vp);
2154 }
2155 out:
2156 pathbuf_stringcopy_put(pb, pathstring);
2157 pathbuf_destroy(pb);
2158 return (error);
2159 }
2160
2161 /*
2162 * Create a named pipe.
2163 */
2164 /* ARGSUSED */
2165 int
2166 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
2167 {
2168 /* {
2169 syscallarg(const char *) path;
2170 syscallarg(int) mode;
2171 } */
2172 struct proc *p = l->l_proc;
2173 struct vattr vattr;
2174 int error;
2175 struct pathbuf *pb;
2176 struct nameidata nd;
2177
2178 error = pathbuf_copyin(SCARG(uap, path), &pb);
2179 if (error) {
2180 return error;
2181 }
2182 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb);
2183 if ((error = namei(&nd)) != 0) {
2184 pathbuf_destroy(pb);
2185 return error;
2186 }
2187 if (nd.ni_vp != NULL) {
2188 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2189 if (nd.ni_dvp == nd.ni_vp)
2190 vrele(nd.ni_dvp);
2191 else
2192 vput(nd.ni_dvp);
2193 vrele(nd.ni_vp);
2194 pathbuf_destroy(pb);
2195 return (EEXIST);
2196 }
2197 vattr_null(&vattr);
2198 vattr.va_type = VFIFO;
2199 /* We will read cwdi->cwdi_cmask unlocked. */
2200 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
2201 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2202 if (error == 0)
2203 vput(nd.ni_vp);
2204 pathbuf_destroy(pb);
2205 return (error);
2206 }
2207
2208 int
2209 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap,
2210 register_t *retval)
2211 {
2212 /* {
2213 syscallarg(int) fd;
2214 syscallarg(const char *) path;
2215 syscallarg(int) mode;
2216 } */
2217
2218 return ENOSYS;
2219 }
2220 /*
2221 * Make a hard file link.
2222 */
2223 /* ARGSUSED */
2224 static int
2225 do_sys_link(struct lwp *l, const char *path, const char *link,
2226 int follow, register_t *retval)
2227 {
2228 struct vnode *vp;
2229 struct pathbuf *linkpb;
2230 struct nameidata nd;
2231 namei_simple_flags_t namei_simple_flags;
2232 int error;
2233
2234 if (follow)
2235 namei_simple_flags = NSM_FOLLOW_TRYEMULROOT;
2236 else
2237 namei_simple_flags = NSM_NOFOLLOW_TRYEMULROOT;
2238
2239 error = namei_simple_user(path, namei_simple_flags, &vp);
2240 if (error != 0)
2241 return (error);
2242 error = pathbuf_copyin(link, &linkpb);
2243 if (error) {
2244 goto out1;
2245 }
2246 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb);
2247 if ((error = namei(&nd)) != 0)
2248 goto out2;
2249 if (nd.ni_vp) {
2250 error = EEXIST;
2251 goto abortop;
2252 }
2253 /* Prevent hard links on directories. */
2254 if (vp->v_type == VDIR) {
2255 error = EPERM;
2256 goto abortop;
2257 }
2258 /* Prevent cross-mount operation. */
2259 if (nd.ni_dvp->v_mount != vp->v_mount) {
2260 error = EXDEV;
2261 goto abortop;
2262 }
2263 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2264 out2:
2265 pathbuf_destroy(linkpb);
2266 out1:
2267 vrele(vp);
2268 return (error);
2269 abortop:
2270 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2271 if (nd.ni_dvp == nd.ni_vp)
2272 vrele(nd.ni_dvp);
2273 else
2274 vput(nd.ni_dvp);
2275 if (nd.ni_vp != NULL)
2276 vrele(nd.ni_vp);
2277 goto out2;
2278 }
2279
2280 int
2281 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
2282 {
2283 /* {
2284 syscallarg(const char *) path;
2285 syscallarg(const char *) link;
2286 } */
2287 const char *path = SCARG(uap, path);
2288 const char *link = SCARG(uap, link);
2289
2290 return do_sys_link(l, path, link, 1, retval);
2291 }
2292
2293 int
2294 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap,
2295 register_t *retval)
2296 {
2297 /* {
2298 syscallarg(int) fd1;
2299 syscallarg(const char *) name1;
2300 syscallarg(int) fd2;
2301 syscallarg(const char *) name2;
2302 syscallarg(int) flags;
2303 } */
2304 const char *name1 = SCARG(uap, name1);
2305 const char *name2 = SCARG(uap, name2);
2306 int follow;
2307
2308 /*
2309 * Specified fd1 and fd2 are not yet implemented
2310 */
2311 if ((SCARG(uap, fd1) != AT_FDCWD) || (SCARG(uap, fd2) != AT_FDCWD))
2312 return ENOSYS;
2313
2314 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW;
2315
2316 return do_sys_link(l, name1, name2, follow, retval);
2317 }
2318
2319
2320 int
2321 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg)
2322 {
2323 struct proc *p = curproc;
2324 struct vattr vattr;
2325 char *path;
2326 int error;
2327 struct pathbuf *linkpb;
2328 struct nameidata nd;
2329
2330 path = PNBUF_GET();
2331 if (seg == UIO_USERSPACE) {
2332 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0)
2333 goto out1;
2334 if ((error = pathbuf_copyin(link, &linkpb)) != 0)
2335 goto out1;
2336 } else {
2337 KASSERT(strlen(patharg) < MAXPATHLEN);
2338 strcpy(path, patharg);
2339 linkpb = pathbuf_create(link);
2340 if (linkpb == NULL) {
2341 error = ENOMEM;
2342 goto out1;
2343 }
2344 }
2345 ktrkuser("symlink-target", path, strlen(path));
2346
2347 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb);
2348 if ((error = namei(&nd)) != 0)
2349 goto out2;
2350 if (nd.ni_vp) {
2351 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2352 if (nd.ni_dvp == nd.ni_vp)
2353 vrele(nd.ni_dvp);
2354 else
2355 vput(nd.ni_dvp);
2356 vrele(nd.ni_vp);
2357 error = EEXIST;
2358 goto out2;
2359 }
2360 vattr_null(&vattr);
2361 vattr.va_type = VLNK;
2362 /* We will read cwdi->cwdi_cmask unlocked. */
2363 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
2364 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
2365 if (error == 0)
2366 vput(nd.ni_vp);
2367 out2:
2368 pathbuf_destroy(linkpb);
2369 out1:
2370 PNBUF_PUT(path);
2371 return (error);
2372 }
2373
2374 /*
2375 * Make a symbolic link.
2376 */
2377 /* ARGSUSED */
2378 int
2379 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
2380 {
2381 /* {
2382 syscallarg(const char *) path;
2383 syscallarg(const char *) link;
2384 } */
2385
2386 return do_sys_symlink(SCARG(uap, path), SCARG(uap, link),
2387 UIO_USERSPACE);
2388 }
2389
2390 int
2391 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap,
2392 register_t *retval)
2393 {
2394 /* {
2395 syscallarg(int) fd;
2396 syscallarg(const char *) path;
2397 syscallarg(const char *) link;
2398 } */
2399
2400 return ENOSYS;
2401 }
2402
2403 /*
2404 * Delete a whiteout from the filesystem.
2405 */
2406 /* ARGSUSED */
2407 int
2408 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
2409 {
2410 /* {
2411 syscallarg(const char *) path;
2412 } */
2413 int error;
2414 struct pathbuf *pb;
2415 struct nameidata nd;
2416
2417 error = pathbuf_copyin(SCARG(uap, path), &pb);
2418 if (error) {
2419 return error;
2420 }
2421
2422 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb);
2423 error = namei(&nd);
2424 if (error) {
2425 pathbuf_destroy(pb);
2426 return (error);
2427 }
2428
2429 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2430 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2431 if (nd.ni_dvp == nd.ni_vp)
2432 vrele(nd.ni_dvp);
2433 else
2434 vput(nd.ni_dvp);
2435 if (nd.ni_vp)
2436 vrele(nd.ni_vp);
2437 pathbuf_destroy(pb);
2438 return (EEXIST);
2439 }
2440 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2441 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2442 vput(nd.ni_dvp);
2443 pathbuf_destroy(pb);
2444 return (error);
2445 }
2446
2447 /*
2448 * Delete a name from the filesystem.
2449 */
2450 /* ARGSUSED */
2451 int
2452 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
2453 {
2454 /* {
2455 syscallarg(const char *) path;
2456 } */
2457
2458 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
2459 }
2460
2461 int
2462 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap,
2463 register_t *retval)
2464 {
2465 /* {
2466 syscallarg(int) fd;
2467 syscallarg(const char *) path;
2468 } */
2469
2470 return ENOSYS;
2471 }
2472
2473 int
2474 do_sys_unlink(const char *arg, enum uio_seg seg)
2475 {
2476 struct vnode *vp;
2477 int error;
2478 struct pathbuf *pb;
2479 struct nameidata nd;
2480 const char *pathstring;
2481
2482 error = pathbuf_maybe_copyin(arg, seg, &pb);
2483 if (error) {
2484 return error;
2485 }
2486 pathstring = pathbuf_stringcopy_get(pb);
2487 if (pathstring == NULL) {
2488 pathbuf_destroy(pb);
2489 return ENOMEM;
2490 }
2491
2492 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb);
2493 if ((error = namei(&nd)) != 0)
2494 goto out;
2495 vp = nd.ni_vp;
2496
2497 /*
2498 * The root of a mounted filesystem cannot be deleted.
2499 */
2500 if (vp->v_vflag & VV_ROOT) {
2501 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2502 if (nd.ni_dvp == vp)
2503 vrele(nd.ni_dvp);
2504 else
2505 vput(nd.ni_dvp);
2506 vput(vp);
2507 error = EBUSY;
2508 goto out;
2509 }
2510
2511 #if NVERIEXEC > 0
2512 /* Handle remove requests for veriexec entries. */
2513 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) {
2514 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2515 if (nd.ni_dvp == vp)
2516 vrele(nd.ni_dvp);
2517 else
2518 vput(nd.ni_dvp);
2519 vput(vp);
2520 goto out;
2521 }
2522 #endif /* NVERIEXEC > 0 */
2523
2524 #ifdef FILEASSOC
2525 (void)fileassoc_file_delete(vp);
2526 #endif /* FILEASSOC */
2527 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2528 out:
2529 pathbuf_stringcopy_put(pb, pathstring);
2530 pathbuf_destroy(pb);
2531 return (error);
2532 }
2533
2534 /*
2535 * Reposition read/write file offset.
2536 */
2537 int
2538 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
2539 {
2540 /* {
2541 syscallarg(int) fd;
2542 syscallarg(int) pad;
2543 syscallarg(off_t) offset;
2544 syscallarg(int) whence;
2545 } */
2546 kauth_cred_t cred = l->l_cred;
2547 file_t *fp;
2548 struct vnode *vp;
2549 struct vattr vattr;
2550 off_t newoff;
2551 int error, fd;
2552
2553 fd = SCARG(uap, fd);
2554
2555 if ((fp = fd_getfile(fd)) == NULL)
2556 return (EBADF);
2557
2558 vp = fp->f_data;
2559 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2560 error = ESPIPE;
2561 goto out;
2562 }
2563
2564 switch (SCARG(uap, whence)) {
2565 case SEEK_CUR:
2566 newoff = fp->f_offset + SCARG(uap, offset);
2567 break;
2568 case SEEK_END:
2569 vn_lock(vp, LK_SHARED | LK_RETRY);
2570 error = VOP_GETATTR(vp, &vattr, cred);
2571 VOP_UNLOCK(vp);
2572 if (error) {
2573 goto out;
2574 }
2575 newoff = SCARG(uap, offset) + vattr.va_size;
2576 break;
2577 case SEEK_SET:
2578 newoff = SCARG(uap, offset);
2579 break;
2580 default:
2581 error = EINVAL;
2582 goto out;
2583 }
2584 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2585 *(off_t *)retval = fp->f_offset = newoff;
2586 }
2587 out:
2588 fd_putfile(fd);
2589 return (error);
2590 }
2591
2592 /*
2593 * Positional read system call.
2594 */
2595 int
2596 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
2597 {
2598 /* {
2599 syscallarg(int) fd;
2600 syscallarg(void *) buf;
2601 syscallarg(size_t) nbyte;
2602 syscallarg(off_t) offset;
2603 } */
2604 file_t *fp;
2605 struct vnode *vp;
2606 off_t offset;
2607 int error, fd = SCARG(uap, fd);
2608
2609 if ((fp = fd_getfile(fd)) == NULL)
2610 return (EBADF);
2611
2612 if ((fp->f_flag & FREAD) == 0) {
2613 fd_putfile(fd);
2614 return (EBADF);
2615 }
2616
2617 vp = fp->f_data;
2618 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2619 error = ESPIPE;
2620 goto out;
2621 }
2622
2623 offset = SCARG(uap, offset);
2624
2625 /*
2626 * XXX This works because no file systems actually
2627 * XXX take any action on the seek operation.
2628 */
2629 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2630 goto out;
2631
2632 /* dofileread() will unuse the descriptor for us */
2633 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2634 &offset, 0, retval));
2635
2636 out:
2637 fd_putfile(fd);
2638 return (error);
2639 }
2640
2641 /*
2642 * Positional scatter read system call.
2643 */
2644 int
2645 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
2646 {
2647 /* {
2648 syscallarg(int) fd;
2649 syscallarg(const struct iovec *) iovp;
2650 syscallarg(int) iovcnt;
2651 syscallarg(off_t) offset;
2652 } */
2653 off_t offset = SCARG(uap, offset);
2654
2655 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2656 SCARG(uap, iovcnt), &offset, 0, retval);
2657 }
2658
2659 /*
2660 * Positional write system call.
2661 */
2662 int
2663 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
2664 {
2665 /* {
2666 syscallarg(int) fd;
2667 syscallarg(const void *) buf;
2668 syscallarg(size_t) nbyte;
2669 syscallarg(off_t) offset;
2670 } */
2671 file_t *fp;
2672 struct vnode *vp;
2673 off_t offset;
2674 int error, fd = SCARG(uap, fd);
2675
2676 if ((fp = fd_getfile(fd)) == NULL)
2677 return (EBADF);
2678
2679 if ((fp->f_flag & FWRITE) == 0) {
2680 fd_putfile(fd);
2681 return (EBADF);
2682 }
2683
2684 vp = fp->f_data;
2685 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2686 error = ESPIPE;
2687 goto out;
2688 }
2689
2690 offset = SCARG(uap, offset);
2691
2692 /*
2693 * XXX This works because no file systems actually
2694 * XXX take any action on the seek operation.
2695 */
2696 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2697 goto out;
2698
2699 /* dofilewrite() will unuse the descriptor for us */
2700 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2701 &offset, 0, retval));
2702
2703 out:
2704 fd_putfile(fd);
2705 return (error);
2706 }
2707
2708 /*
2709 * Positional gather write system call.
2710 */
2711 int
2712 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
2713 {
2714 /* {
2715 syscallarg(int) fd;
2716 syscallarg(const struct iovec *) iovp;
2717 syscallarg(int) iovcnt;
2718 syscallarg(off_t) offset;
2719 } */
2720 off_t offset = SCARG(uap, offset);
2721
2722 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2723 SCARG(uap, iovcnt), &offset, 0, retval);
2724 }
2725
2726 /*
2727 * Check access permissions.
2728 */
2729 int
2730 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
2731 {
2732 /* {
2733 syscallarg(const char *) path;
2734 syscallarg(int) flags;
2735 } */
2736 kauth_cred_t cred;
2737 struct vnode *vp;
2738 int error, flags;
2739 struct pathbuf *pb;
2740 struct nameidata nd;
2741
2742 CTASSERT(F_OK == 0);
2743 if ((SCARG(uap, flags) & ~(R_OK | W_OK | X_OK)) != 0) {
2744 /* nonsense flags */
2745 return EINVAL;
2746 }
2747
2748 error = pathbuf_copyin(SCARG(uap, path), &pb);
2749 if (error) {
2750 return error;
2751 }
2752 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
2753
2754 /* Override default credentials */
2755 cred = kauth_cred_dup(l->l_cred);
2756 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2757 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2758 nd.ni_cnd.cn_cred = cred;
2759
2760 if ((error = namei(&nd)) != 0) {
2761 pathbuf_destroy(pb);
2762 goto out;
2763 }
2764 vp = nd.ni_vp;
2765 pathbuf_destroy(pb);
2766
2767 /* Flags == 0 means only check for existence. */
2768 if (SCARG(uap, flags)) {
2769 flags = 0;
2770 if (SCARG(uap, flags) & R_OK)
2771 flags |= VREAD;
2772 if (SCARG(uap, flags) & W_OK)
2773 flags |= VWRITE;
2774 if (SCARG(uap, flags) & X_OK)
2775 flags |= VEXEC;
2776
2777 error = VOP_ACCESS(vp, flags, cred);
2778 if (!error && (flags & VWRITE))
2779 error = vn_writechk(vp);
2780 }
2781 vput(vp);
2782 out:
2783 kauth_cred_free(cred);
2784 return (error);
2785 }
2786
2787 int
2788 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap,
2789 register_t *retval)
2790 {
2791 /* {
2792 syscallarg(int) fd;
2793 syscallarg(const char *) path;
2794 syscallarg(int) amode;
2795 syscallarg(int) flag;
2796 } */
2797
2798 return ENOSYS;
2799 }
2800
2801 /*
2802 * Common code for all sys_stat functions, including compat versions.
2803 */
2804 int
2805 do_sys_stat(const char *userpath, unsigned int nd_flags, struct stat *sb)
2806 {
2807 int error;
2808 struct pathbuf *pb;
2809 struct nameidata nd;
2810
2811 error = pathbuf_copyin(userpath, &pb);
2812 if (error) {
2813 return error;
2814 }
2815 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, pb);
2816 error = namei(&nd);
2817 if (error != 0) {
2818 pathbuf_destroy(pb);
2819 return error;
2820 }
2821 error = vn_stat(nd.ni_vp, sb);
2822 vput(nd.ni_vp);
2823 pathbuf_destroy(pb);
2824 return error;
2825 }
2826
2827 /*
2828 * Get file status; this version follows links.
2829 */
2830 /* ARGSUSED */
2831 int
2832 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval)
2833 {
2834 /* {
2835 syscallarg(const char *) path;
2836 syscallarg(struct stat *) ub;
2837 } */
2838 struct stat sb;
2839 int error;
2840
2841 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
2842 if (error)
2843 return error;
2844 return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2845 }
2846
2847 /*
2848 * Get file status; this version does not follow links.
2849 */
2850 /* ARGSUSED */
2851 int
2852 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval)
2853 {
2854 /* {
2855 syscallarg(const char *) path;
2856 syscallarg(struct stat *) ub;
2857 } */
2858 struct stat sb;
2859 int error;
2860
2861 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
2862 if (error)
2863 return error;
2864 return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2865 }
2866
2867 int
2868 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap,
2869 register_t *retval)
2870 {
2871 /* {
2872 syscallarg(int) fd;
2873 syscallarg(const char *) path;
2874 syscallarg(struct stat *) ub;
2875 syscallarg(int) flag;
2876 } */
2877
2878 return ENOSYS;
2879 }
2880 /*
2881 * Get configurable pathname variables.
2882 */
2883 /* ARGSUSED */
2884 int
2885 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
2886 {
2887 /* {
2888 syscallarg(const char *) path;
2889 syscallarg(int) name;
2890 } */
2891 int error;
2892 struct pathbuf *pb;
2893 struct nameidata nd;
2894
2895 error = pathbuf_copyin(SCARG(uap, path), &pb);
2896 if (error) {
2897 return error;
2898 }
2899 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
2900 if ((error = namei(&nd)) != 0) {
2901 pathbuf_destroy(pb);
2902 return (error);
2903 }
2904 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2905 vput(nd.ni_vp);
2906 pathbuf_destroy(pb);
2907 return (error);
2908 }
2909
2910 /*
2911 * Return target name of a symbolic link.
2912 */
2913 /* ARGSUSED */
2914 int
2915 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
2916 {
2917 /* {
2918 syscallarg(const char *) path;
2919 syscallarg(char *) buf;
2920 syscallarg(size_t) count;
2921 } */
2922 struct vnode *vp;
2923 struct iovec aiov;
2924 struct uio auio;
2925 int error;
2926 struct pathbuf *pb;
2927 struct nameidata nd;
2928
2929 error = pathbuf_copyin(SCARG(uap, path), &pb);
2930 if (error) {
2931 return error;
2932 }
2933 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb);
2934 if ((error = namei(&nd)) != 0) {
2935 pathbuf_destroy(pb);
2936 return error;
2937 }
2938 vp = nd.ni_vp;
2939 pathbuf_destroy(pb);
2940 if (vp->v_type != VLNK)
2941 error = EINVAL;
2942 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2943 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
2944 aiov.iov_base = SCARG(uap, buf);
2945 aiov.iov_len = SCARG(uap, count);
2946 auio.uio_iov = &aiov;
2947 auio.uio_iovcnt = 1;
2948 auio.uio_offset = 0;
2949 auio.uio_rw = UIO_READ;
2950 KASSERT(l == curlwp);
2951 auio.uio_vmspace = l->l_proc->p_vmspace;
2952 auio.uio_resid = SCARG(uap, count);
2953 error = VOP_READLINK(vp, &auio, l->l_cred);
2954 }
2955 vput(vp);
2956 *retval = SCARG(uap, count) - auio.uio_resid;
2957 return (error);
2958 }
2959
2960 int
2961 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap,
2962 register_t *retval)
2963 {
2964 /* {
2965 syscallarg(int) fd;
2966 syscallarg(const char *) path;
2967 syscallarg(char *) buf;
2968 syscallarg(size_t) count;
2969 } */
2970
2971 return ENOSYS;
2972 }
2973
2974 /*
2975 * Change flags of a file given a path name.
2976 */
2977 /* ARGSUSED */
2978 int
2979 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
2980 {
2981 /* {
2982 syscallarg(const char *) path;
2983 syscallarg(u_long) flags;
2984 } */
2985 struct vnode *vp;
2986 int error;
2987
2988 error = namei_simple_user(SCARG(uap, path),
2989 NSM_FOLLOW_TRYEMULROOT, &vp);
2990 if (error != 0)
2991 return (error);
2992 error = change_flags(vp, SCARG(uap, flags), l);
2993 vput(vp);
2994 return (error);
2995 }
2996
2997 /*
2998 * Change flags of a file given a file descriptor.
2999 */
3000 /* ARGSUSED */
3001 int
3002 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
3003 {
3004 /* {
3005 syscallarg(int) fd;
3006 syscallarg(u_long) flags;
3007 } */
3008 struct vnode *vp;
3009 file_t *fp;
3010 int error;
3011
3012 /* fd_getvnode() will use the descriptor for us */
3013 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3014 return (error);
3015 vp = fp->f_data;
3016 error = change_flags(vp, SCARG(uap, flags), l);
3017 VOP_UNLOCK(vp);
3018 fd_putfile(SCARG(uap, fd));
3019 return (error);
3020 }
3021
3022 /*
3023 * Change flags of a file given a path name; this version does
3024 * not follow links.
3025 */
3026 int
3027 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
3028 {
3029 /* {
3030 syscallarg(const char *) path;
3031 syscallarg(u_long) flags;
3032 } */
3033 struct vnode *vp;
3034 int error;
3035
3036 error = namei_simple_user(SCARG(uap, path),
3037 NSM_NOFOLLOW_TRYEMULROOT, &vp);
3038 if (error != 0)
3039 return (error);
3040 error = change_flags(vp, SCARG(uap, flags), l);
3041 vput(vp);
3042 return (error);
3043 }
3044
3045 /*
3046 * Common routine to change flags of a file.
3047 */
3048 int
3049 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
3050 {
3051 struct vattr vattr;
3052 int error;
3053
3054 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3055 /*
3056 * Non-superusers cannot change the flags on devices, even if they
3057 * own them.
3058 */
3059 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
3060 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
3061 goto out;
3062 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
3063 error = EINVAL;
3064 goto out;
3065 }
3066 }
3067 vattr_null(&vattr);
3068 vattr.va_flags = flags;
3069 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3070 out:
3071 return (error);
3072 }
3073
3074 /*
3075 * Change mode of a file given path name; this version follows links.
3076 */
3077 /* ARGSUSED */
3078 int
3079 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
3080 {
3081 /* {
3082 syscallarg(const char *) path;
3083 syscallarg(int) mode;
3084 } */
3085 int error;
3086 struct vnode *vp;
3087
3088 error = namei_simple_user(SCARG(uap, path),
3089 NSM_FOLLOW_TRYEMULROOT, &vp);
3090 if (error != 0)
3091 return (error);
3092
3093 error = change_mode(vp, SCARG(uap, mode), l);
3094
3095 vrele(vp);
3096 return (error);
3097 }
3098
3099 /*
3100 * Change mode of a file given a file descriptor.
3101 */
3102 /* ARGSUSED */
3103 int
3104 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
3105 {
3106 /* {
3107 syscallarg(int) fd;
3108 syscallarg(int) mode;
3109 } */
3110 file_t *fp;
3111 int error;
3112
3113 /* fd_getvnode() will use the descriptor for us */
3114 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3115 return (error);
3116 error = change_mode(fp->f_data, SCARG(uap, mode), l);
3117 fd_putfile(SCARG(uap, fd));
3118 return (error);
3119 }
3120
3121 int
3122 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap,
3123 register_t *retval)
3124 {
3125 /* {
3126 syscallarg(int) fd;
3127 syscallarg(const char *) path;
3128 syscallarg(int) mode;
3129 syscallarg(int) flag;
3130 } */
3131
3132 return ENOSYS;
3133 }
3134
3135 /*
3136 * Change mode of a file given path name; this version does not follow links.
3137 */
3138 /* ARGSUSED */
3139 int
3140 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
3141 {
3142 /* {
3143 syscallarg(const char *) path;
3144 syscallarg(int) mode;
3145 } */
3146 int error;
3147 struct vnode *vp;
3148
3149 error = namei_simple_user(SCARG(uap, path),
3150 NSM_NOFOLLOW_TRYEMULROOT, &vp);
3151 if (error != 0)
3152 return (error);
3153
3154 error = change_mode(vp, SCARG(uap, mode), l);
3155
3156 vrele(vp);
3157 return (error);
3158 }
3159
3160 /*
3161 * Common routine to set mode given a vnode.
3162 */
3163 static int
3164 change_mode(struct vnode *vp, int mode, struct lwp *l)
3165 {
3166 struct vattr vattr;
3167 int error;
3168
3169 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3170 vattr_null(&vattr);
3171 vattr.va_mode = mode & ALLPERMS;
3172 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3173 VOP_UNLOCK(vp);
3174 return (error);
3175 }
3176
3177 /*
3178 * Set ownership given a path name; this version follows links.
3179 */
3180 /* ARGSUSED */
3181 int
3182 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
3183 {
3184 /* {
3185 syscallarg(const char *) path;
3186 syscallarg(uid_t) uid;
3187 syscallarg(gid_t) gid;
3188 } */
3189 int error;
3190 struct vnode *vp;
3191
3192 error = namei_simple_user(SCARG(uap, path),
3193 NSM_FOLLOW_TRYEMULROOT, &vp);
3194 if (error != 0)
3195 return (error);
3196
3197 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
3198
3199 vrele(vp);
3200 return (error);
3201 }
3202
3203 /*
3204 * Set ownership given a path name; this version follows links.
3205 * Provides POSIX semantics.
3206 */
3207 /* ARGSUSED */
3208 int
3209 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
3210 {
3211 /* {
3212 syscallarg(const char *) path;
3213 syscallarg(uid_t) uid;
3214 syscallarg(gid_t) gid;
3215 } */
3216 int error;
3217 struct vnode *vp;
3218
3219 error = namei_simple_user(SCARG(uap, path),
3220 NSM_FOLLOW_TRYEMULROOT, &vp);
3221 if (error != 0)
3222 return (error);
3223
3224 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
3225
3226 vrele(vp);
3227 return (error);
3228 }
3229
3230 /*
3231 * Set ownership given a file descriptor.
3232 */
3233 /* ARGSUSED */
3234 int
3235 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
3236 {
3237 /* {
3238 syscallarg(int) fd;
3239 syscallarg(uid_t) uid;
3240 syscallarg(gid_t) gid;
3241 } */
3242 int error;
3243 file_t *fp;
3244
3245 /* fd_getvnode() will use the descriptor for us */
3246 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3247 return (error);
3248 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
3249 l, 0);
3250 fd_putfile(SCARG(uap, fd));
3251 return (error);
3252 }
3253
3254 int
3255 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap,
3256 register_t *retval)
3257 {
3258 /* {
3259 syscallarg(int) fd;
3260 syscallarg(const char *) path;
3261 syscallarg(uid_t) uid;
3262 syscallarg(gid_t) gid;
3263 syscallarg(int) flag;
3264 } */
3265
3266 return ENOSYS;
3267 }
3268
3269 /*
3270 * Set ownership given a file descriptor, providing POSIX/XPG semantics.
3271 */
3272 /* ARGSUSED */
3273 int
3274 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
3275 {
3276 /* {
3277 syscallarg(int) fd;
3278 syscallarg(uid_t) uid;
3279 syscallarg(gid_t) gid;
3280 } */
3281 int error;
3282 file_t *fp;
3283
3284 /* fd_getvnode() will use the descriptor for us */
3285 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3286 return (error);
3287 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
3288 l, 1);
3289 fd_putfile(SCARG(uap, fd));
3290 return (error);
3291 }
3292
3293 /*
3294 * Set ownership given a path name; this version does not follow links.
3295 */
3296 /* ARGSUSED */
3297 int
3298 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
3299 {
3300 /* {
3301 syscallarg(const char *) path;
3302 syscallarg(uid_t) uid;
3303 syscallarg(gid_t) gid;
3304 } */
3305 int error;
3306 struct vnode *vp;
3307
3308 error = namei_simple_user(SCARG(uap, path),
3309 NSM_NOFOLLOW_TRYEMULROOT, &vp);
3310 if (error != 0)
3311 return (error);
3312
3313 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
3314
3315 vrele(vp);
3316 return (error);
3317 }
3318
3319 /*
3320 * Set ownership given a path name; this version does not follow links.
3321 * Provides POSIX/XPG semantics.
3322 */
3323 /* ARGSUSED */
3324 int
3325 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
3326 {
3327 /* {
3328 syscallarg(const char *) path;
3329 syscallarg(uid_t) uid;
3330 syscallarg(gid_t) gid;
3331 } */
3332 int error;
3333 struct vnode *vp;
3334
3335 error = namei_simple_user(SCARG(uap, path),
3336 NSM_NOFOLLOW_TRYEMULROOT, &vp);
3337 if (error != 0)
3338 return (error);
3339
3340 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
3341
3342 vrele(vp);
3343 return (error);
3344 }
3345
3346 /*
3347 * Common routine to set ownership given a vnode.
3348 */
3349 static int
3350 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
3351 int posix_semantics)
3352 {
3353 struct vattr vattr;
3354 mode_t newmode;
3355 int error;
3356
3357 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3358 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
3359 goto out;
3360
3361 #define CHANGED(x) ((int)(x) != -1)
3362 newmode = vattr.va_mode;
3363 if (posix_semantics) {
3364 /*
3365 * POSIX/XPG semantics: if the caller is not the super-user,
3366 * clear set-user-id and set-group-id bits. Both POSIX and
3367 * the XPG consider the behaviour for calls by the super-user
3368 * implementation-defined; we leave the set-user-id and set-
3369 * group-id settings intact in that case.
3370 */
3371 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
3372 NULL) != 0)
3373 newmode &= ~(S_ISUID | S_ISGID);
3374 } else {
3375 /*
3376 * NetBSD semantics: when changing owner and/or group,
3377 * clear the respective bit(s).
3378 */
3379 if (CHANGED(uid))
3380 newmode &= ~S_ISUID;
3381 if (CHANGED(gid))
3382 newmode &= ~S_ISGID;
3383 }
3384 /* Update va_mode iff altered. */
3385 if (vattr.va_mode == newmode)
3386 newmode = VNOVAL;
3387
3388 vattr_null(&vattr);
3389 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
3390 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
3391 vattr.va_mode = newmode;
3392 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3393 #undef CHANGED
3394
3395 out:
3396 VOP_UNLOCK(vp);
3397 return (error);
3398 }
3399
3400 /*
3401 * Set the access and modification times given a path name; this
3402 * version follows links.
3403 */
3404 /* ARGSUSED */
3405 int
3406 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap,
3407 register_t *retval)
3408 {
3409 /* {
3410 syscallarg(const char *) path;
3411 syscallarg(const struct timeval *) tptr;
3412 } */
3413
3414 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
3415 SCARG(uap, tptr), UIO_USERSPACE);
3416 }
3417
3418 /*
3419 * Set the access and modification times given a file descriptor.
3420 */
3421 /* ARGSUSED */
3422 int
3423 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap,
3424 register_t *retval)
3425 {
3426 /* {
3427 syscallarg(int) fd;
3428 syscallarg(const struct timeval *) tptr;
3429 } */
3430 int error;
3431 file_t *fp;
3432
3433 /* fd_getvnode() will use the descriptor for us */
3434 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3435 return (error);
3436 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
3437 UIO_USERSPACE);
3438 fd_putfile(SCARG(uap, fd));
3439 return (error);
3440 }
3441
3442 int
3443 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap,
3444 register_t *retval)
3445 {
3446 /* {
3447 syscallarg(int) fd;
3448 syscallarg(const struct timespec *) tptr;
3449 } */
3450 int error;
3451 file_t *fp;
3452
3453 /* fd_getvnode() will use the descriptor for us */
3454 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3455 return (error);
3456 error = do_sys_utimens(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
3457 UIO_USERSPACE);
3458 fd_putfile(SCARG(uap, fd));
3459 return (error);
3460 }
3461
3462 /*
3463 * Set the access and modification times given a path name; this
3464 * version does not follow links.
3465 */
3466 int
3467 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap,
3468 register_t *retval)
3469 {
3470 /* {
3471 syscallarg(const char *) path;
3472 syscallarg(const struct timeval *) tptr;
3473 } */
3474
3475 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
3476 SCARG(uap, tptr), UIO_USERSPACE);
3477 }
3478
3479 int
3480 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap,
3481 register_t *retval)
3482 {
3483 /* {
3484 syscallarg(int) fd;
3485 syscallarg(const char *) path;
3486 syscallarg(const struct timespec *) tptr;
3487 syscallarg(int) flag;
3488 } */
3489 int follow;
3490 const struct timespec *tptr;
3491
3492 /*
3493 * Specified fd is not yet implemented
3494 */
3495 if (SCARG(uap, fd) != AT_FDCWD)
3496 return ENOSYS;
3497
3498 tptr = SCARG(uap, tptr);
3499 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3500
3501 return do_sys_utimens(l, NULL, SCARG(uap, path), follow,
3502 tptr, UIO_USERSPACE);
3503 }
3504
3505 /*
3506 * Common routine to set access and modification times given a vnode.
3507 */
3508 int
3509 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag,
3510 const struct timespec *tptr, enum uio_seg seg)
3511 {
3512 struct vattr vattr;
3513 int error, dorele = 0;
3514 namei_simple_flags_t sflags;
3515
3516 bool vanull, setbirthtime;
3517 struct timespec ts[2];
3518
3519 /*
3520 * I have checked all callers and they pass either FOLLOW,
3521 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW
3522 * is 0. More to the point, they don't pass anything else.
3523 * Let's keep it that way at least until the namei interfaces
3524 * are fully sanitized.
3525 */
3526 KASSERT(flag == NOFOLLOW || flag == FOLLOW);
3527 sflags = (flag == FOLLOW) ?
3528 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT;
3529
3530 if (tptr == NULL) {
3531 vanull = true;
3532 nanotime(&ts[0]);
3533 ts[1] = ts[0];
3534 } else {
3535 vanull = false;
3536 if (seg != UIO_SYSSPACE) {
3537 error = copyin(tptr, ts, sizeof (ts));
3538 if (error != 0)
3539 return error;
3540 } else {
3541 ts[0] = tptr[0];
3542 ts[1] = tptr[1];
3543 }
3544 }
3545
3546 if (ts[0].tv_nsec == UTIME_NOW) {
3547 nanotime(&ts[0]);
3548 if (ts[1].tv_nsec == UTIME_NOW) {
3549 vanull = true;
3550 ts[1] = ts[0];
3551 }
3552 } else if (ts[1].tv_nsec == UTIME_NOW)
3553 nanotime(&ts[1]);
3554
3555 if (vp == NULL) {
3556 /* note: SEG describes TPTR, not PATH; PATH is always user */
3557 error = namei_simple_user(path, sflags, &vp);
3558 if (error != 0)
3559 return error;
3560 dorele = 1;
3561 }
3562
3563 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3564 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
3565 timespeccmp(&ts[1], &vattr.va_birthtime, <));
3566 vattr_null(&vattr);
3567
3568 if (ts[0].tv_nsec != UTIME_OMIT)
3569 vattr.va_atime = ts[0];
3570
3571 if (ts[1].tv_nsec != UTIME_OMIT) {
3572 vattr.va_mtime = ts[1];
3573 if (setbirthtime)
3574 vattr.va_birthtime = ts[1];
3575 }
3576
3577 if (vanull)
3578 vattr.va_vaflags |= VA_UTIMES_NULL;
3579 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3580 VOP_UNLOCK(vp);
3581
3582 if (dorele != 0)
3583 vrele(vp);
3584
3585 return error;
3586 }
3587
3588 int
3589 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
3590 const struct timeval *tptr, enum uio_seg seg)
3591 {
3592 struct timespec ts[2];
3593 struct timespec *tsptr = NULL;
3594 int error;
3595
3596 if (tptr != NULL) {
3597 struct timeval tv[2];
3598
3599 if (seg != UIO_SYSSPACE) {
3600 error = copyin(tptr, tv, sizeof (tv));
3601 if (error != 0)
3602 return error;
3603 tptr = tv;
3604 }
3605
3606 if ((tv[0].tv_usec == UTIME_NOW) ||
3607 (tv[0].tv_usec == UTIME_OMIT))
3608 ts[0].tv_nsec = tv[0].tv_usec;
3609 else
3610 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
3611
3612 if ((tv[1].tv_usec == UTIME_NOW) ||
3613 (tv[1].tv_usec == UTIME_OMIT))
3614 ts[1].tv_nsec = tv[1].tv_usec;
3615 else
3616 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
3617
3618 tsptr = &ts[0];
3619 }
3620
3621 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE);
3622 }
3623
3624 /*
3625 * Truncate a file given its path name.
3626 */
3627 /* ARGSUSED */
3628 int
3629 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
3630 {
3631 /* {
3632 syscallarg(const char *) path;
3633 syscallarg(int) pad;
3634 syscallarg(off_t) length;
3635 } */
3636 struct vnode *vp;
3637 struct vattr vattr;
3638 int error;
3639
3640 error = namei_simple_user(SCARG(uap, path),
3641 NSM_FOLLOW_TRYEMULROOT, &vp);
3642 if (error != 0)
3643 return (error);
3644 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3645 if (vp->v_type == VDIR)
3646 error = EISDIR;
3647 else if ((error = vn_writechk(vp)) == 0 &&
3648 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
3649 vattr_null(&vattr);
3650 vattr.va_size = SCARG(uap, length);
3651 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3652 }
3653 vput(vp);
3654 return (error);
3655 }
3656
3657 /*
3658 * Truncate a file given a file descriptor.
3659 */
3660 /* ARGSUSED */
3661 int
3662 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
3663 {
3664 /* {
3665 syscallarg(int) fd;
3666 syscallarg(int) pad;
3667 syscallarg(off_t) length;
3668 } */
3669 struct vattr vattr;
3670 struct vnode *vp;
3671 file_t *fp;
3672 int error;
3673
3674 /* fd_getvnode() will use the descriptor for us */
3675 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3676 return (error);
3677 if ((fp->f_flag & FWRITE) == 0) {
3678 error = EINVAL;
3679 goto out;
3680 }
3681 vp = fp->f_data;
3682 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3683 if (vp->v_type == VDIR)
3684 error = EISDIR;
3685 else if ((error = vn_writechk(vp)) == 0) {
3686 vattr_null(&vattr);
3687 vattr.va_size = SCARG(uap, length);
3688 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3689 }
3690 VOP_UNLOCK(vp);
3691 out:
3692 fd_putfile(SCARG(uap, fd));
3693 return (error);
3694 }
3695
3696 /*
3697 * Sync an open file.
3698 */
3699 /* ARGSUSED */
3700 int
3701 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
3702 {
3703 /* {
3704 syscallarg(int) fd;
3705 } */
3706 struct vnode *vp;
3707 file_t *fp;
3708 int error;
3709
3710 /* fd_getvnode() will use the descriptor for us */
3711 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3712 return (error);
3713 vp = fp->f_data;
3714 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3715 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
3716 VOP_UNLOCK(vp);
3717 fd_putfile(SCARG(uap, fd));
3718 return (error);
3719 }
3720
3721 /*
3722 * Sync a range of file data. API modeled after that found in AIX.
3723 *
3724 * FDATASYNC indicates that we need only save enough metadata to be able
3725 * to re-read the written data. Note we duplicate AIX's requirement that
3726 * the file be open for writing.
3727 */
3728 /* ARGSUSED */
3729 int
3730 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
3731 {
3732 /* {
3733 syscallarg(int) fd;
3734 syscallarg(int) flags;
3735 syscallarg(off_t) start;
3736 syscallarg(off_t) length;
3737 } */
3738 struct vnode *vp;
3739 file_t *fp;
3740 int flags, nflags;
3741 off_t s, e, len;
3742 int error;
3743
3744 /* fd_getvnode() will use the descriptor for us */
3745 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3746 return (error);
3747
3748 if ((fp->f_flag & FWRITE) == 0) {
3749 error = EBADF;
3750 goto out;
3751 }
3752
3753 flags = SCARG(uap, flags);
3754 if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3755 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3756 error = EINVAL;
3757 goto out;
3758 }
3759 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3760 if (flags & FDATASYNC)
3761 nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3762 else
3763 nflags = FSYNC_WAIT;
3764 if (flags & FDISKSYNC)
3765 nflags |= FSYNC_CACHE;
3766
3767 len = SCARG(uap, length);
3768 /* If length == 0, we do the whole file, and s = e = 0 will do that */
3769 if (len) {
3770 s = SCARG(uap, start);
3771 e = s + len;
3772 if (e < s) {
3773 error = EINVAL;
3774 goto out;
3775 }
3776 } else {
3777 e = 0;
3778 s = 0;
3779 }
3780
3781 vp = fp->f_data;
3782 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3783 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
3784 VOP_UNLOCK(vp);
3785 out:
3786 fd_putfile(SCARG(uap, fd));
3787 return (error);
3788 }
3789
3790 /*
3791 * Sync the data of an open file.
3792 */
3793 /* ARGSUSED */
3794 int
3795 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
3796 {
3797 /* {
3798 syscallarg(int) fd;
3799 } */
3800 struct vnode *vp;
3801 file_t *fp;
3802 int error;
3803
3804 /* fd_getvnode() will use the descriptor for us */
3805 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3806 return (error);
3807 if ((fp->f_flag & FWRITE) == 0) {
3808 fd_putfile(SCARG(uap, fd));
3809 return (EBADF);
3810 }
3811 vp = fp->f_data;
3812 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3813 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
3814 VOP_UNLOCK(vp);
3815 fd_putfile(SCARG(uap, fd));
3816 return (error);
3817 }
3818
3819 /*
3820 * Rename files, (standard) BSD semantics frontend.
3821 */
3822 /* ARGSUSED */
3823 int
3824 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
3825 {
3826 /* {
3827 syscallarg(const char *) from;
3828 syscallarg(const char *) to;
3829 } */
3830
3831 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
3832 }
3833
3834 int
3835 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap,
3836 register_t *retval)
3837 {
3838 /* {
3839 syscallarg(int) fromfd;
3840 syscallarg(const char *) from;
3841 syscallarg(int) tofd;
3842 syscallarg(const char *) to;
3843 } */
3844
3845 return ENOSYS;
3846 }
3847
3848 /*
3849 * Rename files, POSIX semantics frontend.
3850 */
3851 /* ARGSUSED */
3852 int
3853 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
3854 {
3855 /* {
3856 syscallarg(const char *) from;
3857 syscallarg(const char *) to;
3858 } */
3859
3860 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
3861 }
3862
3863 /*
3864 * Rename files. Source and destination must either both be directories,
3865 * or both not be directories. If target is a directory, it must be empty.
3866 * If `from' and `to' refer to the same object, the value of the `retain'
3867 * argument is used to determine whether `from' will be
3868 *
3869 * (retain == 0) deleted unless `from' and `to' refer to the same
3870 * object in the file system's name space (BSD).
3871 * (retain == 1) always retained (POSIX).
3872 */
3873 int
3874 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
3875 {
3876 struct vnode *tvp, *fvp, *tdvp;
3877 struct pathbuf *frompb, *topb;
3878 struct nameidata fromnd, tond;
3879 struct mount *fs;
3880 int error;
3881
3882 error = pathbuf_maybe_copyin(from, seg, &frompb);
3883 if (error) {
3884 return error;
3885 }
3886 error = pathbuf_maybe_copyin(to, seg, &topb);
3887 if (error) {
3888 pathbuf_destroy(frompb);
3889 return error;
3890 }
3891
3892 NDINIT(&fromnd, DELETE, LOCKPARENT | TRYEMULROOT | INRENAME,
3893 frompb);
3894 if ((error = namei(&fromnd)) != 0) {
3895 pathbuf_destroy(frompb);
3896 pathbuf_destroy(topb);
3897 return (error);
3898 }
3899 if (fromnd.ni_dvp != fromnd.ni_vp)
3900 VOP_UNLOCK(fromnd.ni_dvp);
3901 fvp = fromnd.ni_vp;
3902
3903 fs = fvp->v_mount;
3904 error = VFS_RENAMELOCK_ENTER(fs);
3905 if (error) {
3906 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3907 vrele(fromnd.ni_dvp);
3908 vrele(fvp);
3909 goto out1;
3910 }
3911
3912 /*
3913 * close, partially, yet another race - ideally we should only
3914 * go as far as getting fromnd.ni_dvp before getting the per-fs
3915 * lock, and then continue to get fromnd.ni_vp, but we can't do
3916 * that with namei as it stands.
3917 *
3918 * This still won't prevent rmdir from nuking fromnd.ni_vp
3919 * under us. The real fix is to get the locks in the right
3920 * order and do the lookups in the right places, but that's a
3921 * major rototill.
3922 *
3923 * Note: this logic (as well as this whole function) is cloned
3924 * in nfs_serv.c. Proceed accordingly.
3925 */
3926 vrele(fvp);
3927 if ((fromnd.ni_cnd.cn_namelen == 1 &&
3928 fromnd.ni_cnd.cn_nameptr[0] == '.') ||
3929 (fromnd.ni_cnd.cn_namelen == 2 &&
3930 fromnd.ni_cnd.cn_nameptr[0] == '.' &&
3931 fromnd.ni_cnd.cn_nameptr[1] == '.')) {
3932 error = EINVAL;
3933 VFS_RENAMELOCK_EXIT(fs);
3934 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3935 vrele(fromnd.ni_dvp);
3936 goto out1;
3937 }
3938 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
3939 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd, 0);
3940 if (error) {
3941 VOP_UNLOCK(fromnd.ni_dvp);
3942 VFS_RENAMELOCK_EXIT(fs);
3943 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3944 vrele(fromnd.ni_dvp);
3945 goto out1;
3946 }
3947 VOP_UNLOCK(fromnd.ni_vp);
3948 if (fromnd.ni_dvp != fromnd.ni_vp)
3949 VOP_UNLOCK(fromnd.ni_dvp);
3950 fvp = fromnd.ni_vp;
3951
3952 NDINIT(&tond, RENAME,
3953 LOCKPARENT | LOCKLEAF | NOCACHE | TRYEMULROOT
3954 | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0),
3955 topb);
3956 if ((error = namei(&tond)) != 0) {
3957 VFS_RENAMELOCK_EXIT(fs);
3958 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3959 vrele(fromnd.ni_dvp);
3960 vrele(fvp);
3961 goto out1;
3962 }
3963 tdvp = tond.ni_dvp;
3964 tvp = tond.ni_vp;
3965
3966 if (tvp != NULL) {
3967 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3968 error = ENOTDIR;
3969 goto out;
3970 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3971 error = EISDIR;
3972 goto out;
3973 }
3974 }
3975
3976 if (fvp == tdvp)
3977 error = EINVAL;
3978
3979 /*
3980 * Source and destination refer to the same object.
3981 */
3982 if (fvp == tvp) {
3983 if (retain)
3984 error = -1;
3985 else if (fromnd.ni_dvp == tdvp &&
3986 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3987 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
3988 fromnd.ni_cnd.cn_namelen))
3989 error = -1;
3990 }
3991 /*
3992 * Prevent cross-mount operation.
3993 */
3994 if (error == 0) {
3995 if (tond.ni_dvp->v_mount != fromnd.ni_dvp->v_mount) {
3996 error = EXDEV;
3997 }
3998 }
3999 #if NVERIEXEC > 0
4000 if (!error) {
4001 char *f1, *f2;
4002 size_t f1_len;
4003 size_t f2_len;
4004
4005 f1_len = fromnd.ni_cnd.cn_namelen + 1;
4006 f1 = kmem_alloc(f1_len, KM_SLEEP);
4007 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len);
4008
4009 f2_len = tond.ni_cnd.cn_namelen + 1;
4010 f2 = kmem_alloc(f2_len, KM_SLEEP);
4011 strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len);
4012
4013 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2);
4014
4015 kmem_free(f1, f1_len);
4016 kmem_free(f2, f2_len);
4017 }
4018 #endif /* NVERIEXEC > 0 */
4019
4020 out:
4021 if (!error) {
4022 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
4023 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
4024 VFS_RENAMELOCK_EXIT(fs);
4025 } else {
4026 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
4027 if (tdvp == tvp)
4028 vrele(tdvp);
4029 else
4030 vput(tdvp);
4031 if (tvp)
4032 vput(tvp);
4033 VFS_RENAMELOCK_EXIT(fs);
4034 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
4035 vrele(fromnd.ni_dvp);
4036 vrele(fvp);
4037 }
4038 out1:
4039 pathbuf_destroy(frompb);
4040 pathbuf_destroy(topb);
4041 return (error == -1 ? 0 : error);
4042 }
4043
4044 /*
4045 * Make a directory file.
4046 */
4047 /* ARGSUSED */
4048 int
4049 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
4050 {
4051 /* {
4052 syscallarg(const char *) path;
4053 syscallarg(int) mode;
4054 } */
4055
4056 return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE);
4057 }
4058
4059 int
4060 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap,
4061 register_t *retval)
4062 {
4063 /* {
4064 syscallarg(int) fd;
4065 syscallarg(const char *) path;
4066 syscallarg(int) mode;
4067 } */
4068
4069 return ENOSYS;
4070 }
4071
4072
4073 int
4074 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg)
4075 {
4076 struct proc *p = curlwp->l_proc;
4077 struct vnode *vp;
4078 struct vattr vattr;
4079 int error;
4080 struct pathbuf *pb;
4081 struct nameidata nd;
4082
4083 /* XXX bollocks, should pass in a pathbuf */
4084 error = pathbuf_maybe_copyin(path, seg, &pb);
4085 if (error) {
4086 return error;
4087 }
4088
4089 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb);
4090 if ((error = namei(&nd)) != 0) {
4091 pathbuf_destroy(pb);
4092 return (error);
4093 }
4094 vp = nd.ni_vp;
4095 if (vp != NULL) {
4096 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
4097 if (nd.ni_dvp == vp)
4098 vrele(nd.ni_dvp);
4099 else
4100 vput(nd.ni_dvp);
4101 vrele(vp);
4102 pathbuf_destroy(pb);
4103 return (EEXIST);
4104 }
4105 vattr_null(&vattr);
4106 vattr.va_type = VDIR;
4107 /* We will read cwdi->cwdi_cmask unlocked. */
4108 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
4109 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
4110 if (!error)
4111 vput(nd.ni_vp);
4112 pathbuf_destroy(pb);
4113 return (error);
4114 }
4115
4116 /*
4117 * Remove a directory file.
4118 */
4119 /* ARGSUSED */
4120 int
4121 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
4122 {
4123 /* {
4124 syscallarg(const char *) path;
4125 } */
4126 struct vnode *vp;
4127 int error;
4128 struct pathbuf *pb;
4129 struct nameidata nd;
4130
4131 error = pathbuf_copyin(SCARG(uap, path), &pb);
4132 if (error) {
4133 return error;
4134 }
4135 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb);
4136 if ((error = namei(&nd)) != 0) {
4137 pathbuf_destroy(pb);
4138 return error;
4139 }
4140 vp = nd.ni_vp;
4141 if (vp->v_type != VDIR) {
4142 error = ENOTDIR;
4143 goto out;
4144 }
4145 /*
4146 * No rmdir "." please.
4147 */
4148 if (nd.ni_dvp == vp) {
4149 error = EINVAL;
4150 goto out;
4151 }
4152 /*
4153 * The root of a mounted filesystem cannot be deleted.
4154 */
4155 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
4156 error = EBUSY;
4157 goto out;
4158 }
4159 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
4160 pathbuf_destroy(pb);
4161 return (error);
4162
4163 out:
4164 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
4165 if (nd.ni_dvp == vp)
4166 vrele(nd.ni_dvp);
4167 else
4168 vput(nd.ni_dvp);
4169 vput(vp);
4170 pathbuf_destroy(pb);
4171 return (error);
4172 }
4173
4174 /*
4175 * Read a block of directory entries in a file system independent format.
4176 */
4177 int
4178 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
4179 {
4180 /* {
4181 syscallarg(int) fd;
4182 syscallarg(char *) buf;
4183 syscallarg(size_t) count;
4184 } */
4185 file_t *fp;
4186 int error, done;
4187
4188 /* fd_getvnode() will use the descriptor for us */
4189 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
4190 return (error);
4191 if ((fp->f_flag & FREAD) == 0) {
4192 error = EBADF;
4193 goto out;
4194 }
4195 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
4196 SCARG(uap, count), &done, l, 0, 0);
4197 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
4198 *retval = done;
4199 out:
4200 fd_putfile(SCARG(uap, fd));
4201 return (error);
4202 }
4203
4204 /*
4205 * Set the mode mask for creation of filesystem nodes.
4206 */
4207 int
4208 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
4209 {
4210 /* {
4211 syscallarg(mode_t) newmask;
4212 } */
4213 struct proc *p = l->l_proc;
4214 struct cwdinfo *cwdi;
4215
4216 /*
4217 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's
4218 * important is that we serialize changes to the mask. The
4219 * rw_exit() will issue a write memory barrier on our behalf,
4220 * and force the changes out to other CPUs (as it must use an
4221 * atomic operation, draining the local CPU's store buffers).
4222 */
4223 cwdi = p->p_cwdi;
4224 rw_enter(&cwdi->cwdi_lock, RW_WRITER);
4225 *retval = cwdi->cwdi_cmask;
4226 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
4227 rw_exit(&cwdi->cwdi_lock);
4228
4229 return (0);
4230 }
4231
4232 int
4233 dorevoke(struct vnode *vp, kauth_cred_t cred)
4234 {
4235 struct vattr vattr;
4236 int error;
4237
4238 vn_lock(vp, LK_SHARED | LK_RETRY);
4239 error = VOP_GETATTR(vp, &vattr, cred);
4240 VOP_UNLOCK(vp);
4241 if (error != 0)
4242 return error;
4243 if (kauth_cred_geteuid(cred) == vattr.va_uid ||
4244 (error = kauth_authorize_generic(cred,
4245 KAUTH_GENERIC_ISSUSER, NULL)) == 0)
4246 VOP_REVOKE(vp, REVOKEALL);
4247 return (error);
4248 }
4249
4250 /*
4251 * Void all references to file by ripping underlying filesystem
4252 * away from vnode.
4253 */
4254 /* ARGSUSED */
4255 int
4256 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
4257 {
4258 /* {
4259 syscallarg(const char *) path;
4260 } */
4261 struct vnode *vp;
4262 int error;
4263
4264 error = namei_simple_user(SCARG(uap, path),
4265 NSM_FOLLOW_TRYEMULROOT, &vp);
4266 if (error != 0)
4267 return (error);
4268 error = dorevoke(vp, l->l_cred);
4269 vrele(vp);
4270 return (error);
4271 }
4272