1/* $NetBSD: vfs_syscalls.c,v 1.505 2016/07/31 20:34:04 dholland 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 1.505 2016/07/31 20:34:04 dholland 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/specfs/specdev.h>
112
113#include <nfs/rpcv2.h>
114#include <nfs/nfsproto.h>
115#include <nfs/nfs.h>
116#include <nfs/nfs_var.h>
117
118/* XXX this shouldn't be here */
119#ifndef OFF_T_MAX
120#define OFF_T_MAX __type_max(off_t)
121#endif
122
123static int change_flags(struct vnode *, u_long, struct lwp *);
124static int change_mode(struct vnode *, int, struct lwp *);
125static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
126static int do_sys_openat(lwp_t *, int, const char *, int, int, int *);
127static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t,
128 enum uio_seg);
129static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t);
130static int do_sys_symlinkat(struct lwp *, const char *, int, const char *,
131 enum uio_seg);
132static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *,
133 enum uio_seg, int);
134static int do_sys_readlinkat(struct lwp *, int, const char *, char *,
135 size_t, register_t *);
136static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg);
137
138static int fd_nameiat(struct lwp *, int, struct nameidata *);
139static int fd_nameiat_simple_user(struct lwp *, int, const char *,
140 namei_simple_flags_t, struct vnode **);
141
142
143/*
144 * This table is used to maintain compatibility with 4.3BSD
145 * and NetBSD 0.9 mount syscalls - and possibly other systems.
146 * Note, the order is important!
147 *
148 * Do not modify this table. It should only contain filesystems
149 * supported by NetBSD 0.9 and 4.3BSD.
150 */
151const char * const mountcompatnames[] = {
152 NULL, /* 0 = MOUNT_NONE */
153 MOUNT_FFS, /* 1 = MOUNT_UFS */
154 MOUNT_NFS, /* 2 */
155 MOUNT_MFS, /* 3 */
156 MOUNT_MSDOS, /* 4 */
157 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */
158 MOUNT_FDESC, /* 6 */
159 MOUNT_KERNFS, /* 7 */
160 NULL, /* 8 = MOUNT_DEVFS */
161 MOUNT_AFS, /* 9 */
162};
163
164const int nmountcompatnames = __arraycount(mountcompatnames);
165
166static int
167fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp)
168{
169 file_t *dfp;
170 int error;
171
172 if (fdat != AT_FDCWD) {
173 if ((error = fd_getvnode(fdat, &dfp)) != 0)
174 goto out;
175
176 NDAT(ndp, dfp->f_vnode);
177 }
178
179 error = namei(ndp);
180
181 if (fdat != AT_FDCWD)
182 fd_putfile(fdat);
183out:
184 return error;
185}
186
187static int
188fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path,
189 namei_simple_flags_t sflags, struct vnode **vp_ret)
190{
191 file_t *dfp;
192 struct vnode *dvp;
193 int error;
194
195 if (fdat != AT_FDCWD) {
196 if ((error = fd_getvnode(fdat, &dfp)) != 0)
197 goto out;
198
199 dvp = dfp->f_vnode;
200 } else {
201 dvp = NULL;
202 }
203
204 error = nameiat_simple_user(dvp, path, sflags, vp_ret);
205
206 if (fdat != AT_FDCWD)
207 fd_putfile(fdat);
208out:
209 return error;
210}
211
212static int
213open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags)
214{
215 int error;
216
217 fp->f_flag = flags & FMASK;
218 fp->f_type = DTYPE_VNODE;
219 fp->f_ops = &vnops;
220 fp->f_vnode = vp;
221
222 if (flags & (O_EXLOCK | O_SHLOCK)) {
223 struct flock lf;
224 int type;
225
226 lf.l_whence = SEEK_SET;
227 lf.l_start = 0;
228 lf.l_len = 0;
229 if (flags & O_EXLOCK)
230 lf.l_type = F_WRLCK;
231 else
232 lf.l_type = F_RDLCK;
233 type = F_FLOCK;
234 if ((flags & FNONBLOCK) == 0)
235 type |= F_WAIT;
236 VOP_UNLOCK(vp);
237 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
238 if (error) {
239 (void) vn_close(vp, fp->f_flag, fp->f_cred);
240 fd_abort(l->l_proc, fp, indx);
241 return error;
242 }
243 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
244 atomic_or_uint(&fp->f_flag, FHASLOCK);
245 }
246 if (flags & O_CLOEXEC)
247 fd_set_exclose(l, indx, true);
248 return 0;
249}
250
251static int
252mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
253 void *data, size_t *data_len)
254{
255 struct mount *mp;
256 int error = 0, saved_flags;
257
258 mp = vp->v_mount;
259 saved_flags = mp->mnt_flag;
260
261 /* We can operate only on VV_ROOT nodes. */
262 if ((vp->v_vflag & VV_ROOT) == 0) {
263 error = EINVAL;
264 goto out;
265 }
266
267 /*
268 * We only allow the filesystem to be reloaded if it
269 * is currently mounted read-only. Additionally, we
270 * prevent read-write to read-only downgrades.
271 */
272 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 &&
273 (mp->mnt_flag & MNT_RDONLY) == 0 &&
274 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) {
275 error = EOPNOTSUPP; /* Needs translation */
276 goto out;
277 }
278
279 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
280 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
281 if (error)
282 goto out;
283
284 if (vfs_busy(mp, NULL)) {
285 error = EPERM;
286 goto out;
287 }
288
289 mutex_enter(&mp->mnt_updating);
290
291 mp->mnt_flag &= ~MNT_OP_FLAGS;
292 mp->mnt_flag |= flags & MNT_OP_FLAGS;
293
294 /*
295 * Set the mount level flags.
296 */
297 if (flags & MNT_RDONLY)
298 mp->mnt_flag |= MNT_RDONLY;
299 else if (mp->mnt_flag & MNT_RDONLY)
300 mp->mnt_iflag |= IMNT_WANTRDWR;
301 mp->mnt_flag &= ~MNT_BASIC_FLAGS;
302 mp->mnt_flag |= flags & MNT_BASIC_FLAGS;
303 error = VFS_MOUNT(mp, path, data, data_len);
304
305 if (error && data != NULL) {
306 int error2;
307
308 /*
309 * Update failed; let's try and see if it was an
310 * export request. For compat with 3.0 and earlier.
311 */
312 error2 = vfs_hooks_reexport(mp, path, data);
313
314 /*
315 * Only update error code if the export request was
316 * understood but some problem occurred while
317 * processing it.
318 */
319 if (error2 != EJUSTRETURN)
320 error = error2;
321 }
322
323 if (mp->mnt_iflag & IMNT_WANTRDWR)
324 mp->mnt_flag &= ~MNT_RDONLY;
325 if (error)
326 mp->mnt_flag = saved_flags;
327 mp->mnt_flag &= ~MNT_OP_FLAGS;
328 mp->mnt_iflag &= ~IMNT_WANTRDWR;
329 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
330 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0)
331 vfs_syncer_add_to_worklist(mp);
332 } else {
333 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0)
334 vfs_syncer_remove_from_worklist(mp);
335 }
336 mutex_exit(&mp->mnt_updating);
337 vfs_unbusy(mp, false, NULL);
338
339 if ((error == 0) && !(saved_flags & MNT_EXTATTR) &&
340 (flags & MNT_EXTATTR)) {
341 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START,
342 NULL, 0, NULL) != 0) {
343 printf("%s: failed to start extattr, error = %d",
344 mp->mnt_stat.f_mntonname, error);
345 mp->mnt_flag &= ~MNT_EXTATTR;
346 }
347 }
348
349 if ((error == 0) && (saved_flags & MNT_EXTATTR) &&
350 !(flags & MNT_EXTATTR)) {
351 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP,
352 NULL, 0, NULL) != 0) {
353 printf("%s: failed to stop extattr, error = %d",
354 mp->mnt_stat.f_mntonname, error);
355 mp->mnt_flag |= MNT_RDONLY;
356 }
357 }
358 out:
359 return (error);
360}
361
362static int
363mount_get_vfsops(const char *fstype, enum uio_seg type_seg,
364 struct vfsops **vfsops)
365{
366 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
367 int error;
368
369 if (type_seg == UIO_USERSPACE) {
370 /* Copy file-system type from userspace. */
371 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
372 } else {
373 error = copystr(fstype, fstypename, sizeof(fstypename), NULL);
374 KASSERT(error == 0);
375 }
376
377 if (error) {
378 /*
379 * Historically, filesystem types were identified by numbers.
380 * If we get an integer for the filesystem type instead of a
381 * string, we check to see if it matches one of the historic
382 * filesystem types.
383 */
384 u_long fsindex = (u_long)fstype;
385 if (fsindex >= nmountcompatnames ||
386 mountcompatnames[fsindex] == NULL)
387 return ENODEV;
388 strlcpy(fstypename, mountcompatnames[fsindex],
389 sizeof(fstypename));
390 }
391
392 /* Accept `ufs' as an alias for `ffs', for compatibility. */
393 if (strcmp(fstypename, "ufs") == 0)
394 fstypename[0] = 'f';
395
396 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
397 return 0;
398
399 /* If we can autoload a vfs module, try again */
400 (void)module_autoload(fstypename, MODULE_CLASS_VFS);
401
402 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
403 return 0;
404
405 return ENODEV;
406}
407
408static int
409mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
410 void *data, size_t *data_len)
411{
412 struct mount *mp;
413 int error;
414
415 /* If MNT_GETARGS is specified, it should be the only flag. */
416 if (flags & ~MNT_GETARGS)
417 return EINVAL;
418
419 mp = vp->v_mount;
420
421 /* XXX: probably some notion of "can see" here if we want isolation. */
422 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
423 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
424 if (error)
425 return error;
426
427 if ((vp->v_vflag & VV_ROOT) == 0)
428 return EINVAL;
429
430 if (vfs_busy(mp, NULL))
431 return EPERM;
432
433 mutex_enter(&mp->mnt_updating);
434 mp->mnt_flag &= ~MNT_OP_FLAGS;
435 mp->mnt_flag |= MNT_GETARGS;
436 error = VFS_MOUNT(mp, path, data, data_len);
437 mp->mnt_flag &= ~MNT_OP_FLAGS;
438 mutex_exit(&mp->mnt_updating);
439
440 vfs_unbusy(mp, false, NULL);
441 return (error);
442}
443
444int
445sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
446{
447 /* {
448 syscallarg(const char *) type;
449 syscallarg(const char *) path;
450 syscallarg(int) flags;
451 syscallarg(void *) data;
452 syscallarg(size_t) data_len;
453 } */
454
455 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path),
456 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
457 SCARG(uap, data_len), retval);
458}
459
460int
461do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg,
462 const char *path, int flags, void *data, enum uio_seg data_seg,
463 size_t data_len, register_t *retval)
464{
465 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */
466 struct vnode *vp;
467 void *data_buf = data;
468 bool vfsopsrele = false;
469 size_t alloc_sz = 0;
470 int error;
471
472 /*
473 * Get vnode to be covered
474 */
475 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
476 if (error != 0) {
477 vp = NULL;
478 goto done;
479 }
480
481 if (flags & (MNT_GETARGS | MNT_UPDATE)) {
482 vfsops = vp->v_mount->mnt_op;
483 } else {
484 /* 'type' is userspace */
485 error = mount_get_vfsops(type, type_seg, &vfsops);
486 if (error != 0)
487 goto done;
488 vfsopsrele = true;
489 }
490
491 /*
492 * We allow data to be NULL, even for userspace. Some fs's don't need
493 * it. The others will handle NULL.
494 */
495 if (data != NULL && data_seg == UIO_USERSPACE) {
496 if (data_len == 0) {
497 /* No length supplied, use default for filesystem */
498 data_len = vfsops->vfs_min_mount_data;
499
500 /*
501 * Hopefully a longer buffer won't make copyin() fail.
502 * For compatibility with 3.0 and earlier.
503 */
504 if (flags & MNT_UPDATE
505 && data_len < sizeof (struct mnt_export_args30))
506 data_len = sizeof (struct mnt_export_args30);
507 }
508 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) {
509 error = EINVAL;
510 goto done;
511 }
512 alloc_sz = data_len;
513 data_buf = kmem_alloc(alloc_sz, KM_SLEEP);
514
515 /* NFS needs the buffer even for mnt_getargs .... */
516 error = copyin(data, data_buf, data_len);
517 if (error != 0)
518 goto done;
519 }
520
521 if (flags & MNT_GETARGS) {
522 if (data_len == 0) {
523 error = EINVAL;
524 goto done;
525 }
526 error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
527 if (error != 0)
528 goto done;
529 if (data_seg == UIO_USERSPACE)
530 error = copyout(data_buf, data, data_len);
531 *retval = data_len;
532 } else if (flags & MNT_UPDATE) {
533 error = mount_update(l, vp, path, flags, data_buf, &data_len);
534 } else {
535 /* Locking is handled internally in mount_domount(). */
536 KASSERT(vfsopsrele == true);
537 error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
538 &data_len);
539 vfsopsrele = false;
540 }
541
542 done:
543 if (vfsopsrele)
544 vfs_delref(vfsops);
545 if (vp != NULL) {
546 vrele(vp);
547 }
548 if (data_buf != data)
549 kmem_free(data_buf, alloc_sz);
550 return (error);
551}
552
553/*
554 * Unmount a file system.
555 *
556 * Note: unmount takes a path to the vnode mounted on as argument,
557 * not special file (as before).
558 */
559/* ARGSUSED */
560int
561sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
562{
563 /* {
564 syscallarg(const char *) path;
565 syscallarg(int) flags;
566 } */
567 struct vnode *vp;
568 struct mount *mp;
569 int error;
570 struct pathbuf *pb;
571 struct nameidata nd;
572
573 error = pathbuf_copyin(SCARG(uap, path), &pb);
574 if (error) {
575 return error;
576 }
577
578 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb);
579 if ((error = namei(&nd)) != 0) {
580 pathbuf_destroy(pb);
581 return error;
582 }
583 vp = nd.ni_vp;
584 pathbuf_destroy(pb);
585
586 mp = vp->v_mount;
587 atomic_inc_uint(&mp->mnt_refcnt);
588 VOP_UNLOCK(vp);
589
590 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
591 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
592 if (error) {
593 vrele(vp);
594 vfs_destroy(mp);
595 return (error);
596 }
597
598 /*
599 * Don't allow unmounting the root file system.
600 */
601 if (mp->mnt_flag & MNT_ROOTFS) {
602 vrele(vp);
603 vfs_destroy(mp);
604 return (EINVAL);
605 }
606
607 /*
608 * Must be the root of the filesystem
609 */
610 if ((vp->v_vflag & VV_ROOT) == 0) {
611 vrele(vp);
612 vfs_destroy(mp);
613 return (EINVAL);
614 }
615
616 vrele(vp);
617 error = dounmount(mp, SCARG(uap, flags), l);
618 vfs_destroy(mp);
619 return error;
620}
621
622/*
623 * Sync each mounted filesystem.
624 */
625#ifdef DEBUG
626int syncprt = 0;
627struct ctldebug debug0 = { "syncprt", &syncprt };
628#endif
629
630void
631do_sys_sync(struct lwp *l)
632{
633 struct mount *mp, *nmp;
634 int asyncflag;
635
636 mutex_enter(&mountlist_lock);
637 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
638 if (vfs_busy(mp, &nmp)) {
639 continue;
640 }
641 mutex_enter(&mp->mnt_updating);
642 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
643 asyncflag = mp->mnt_flag & MNT_ASYNC;
644 mp->mnt_flag &= ~MNT_ASYNC;
645 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
646 if (asyncflag)
647 mp->mnt_flag |= MNT_ASYNC;
648 }
649 mutex_exit(&mp->mnt_updating);
650 vfs_unbusy(mp, false, &nmp);
651 }
652 mutex_exit(&mountlist_lock);
653#ifdef DEBUG
654 if (syncprt)
655 vfs_bufstats();
656#endif /* DEBUG */
657}
658
659/* ARGSUSED */
660int
661sys_sync(struct lwp *l, const void *v, register_t *retval)
662{
663 do_sys_sync(l);
664 return (0);
665}
666
667
668/*
669 * Access or change filesystem quotas.
670 *
671 * (this is really 14 different calls bundled into one)
672 */
673
674static int
675do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u)
676{
677 struct quotastat info_k;
678 int error;
679
680 /* ensure any padding bytes are cleared */
681 memset(&info_k, 0, sizeof(info_k));
682
683 error = vfs_quotactl_stat(mp, &info_k);
684 if (error) {
685 return error;
686 }
687
688 return copyout(&info_k, info_u, sizeof(info_k));
689}
690
691static int
692do_sys_quotactl_idtypestat(struct mount *mp, int idtype,
693 struct quotaidtypestat *info_u)
694{
695 struct quotaidtypestat info_k;
696 int error;
697
698 /* ensure any padding bytes are cleared */
699 memset(&info_k, 0, sizeof(info_k));
700
701 error = vfs_quotactl_idtypestat(mp, idtype, &info_k);
702 if (error) {
703 return error;
704 }
705
706 return copyout(&info_k, info_u, sizeof(info_k));
707}
708
709static int
710do_sys_quotactl_objtypestat(struct mount *mp, int objtype,
711 struct quotaobjtypestat *info_u)
712{
713 struct quotaobjtypestat info_k;
714 int error;
715
716 /* ensure any padding bytes are cleared */
717 memset(&info_k, 0, sizeof(info_k));
718
719 error = vfs_quotactl_objtypestat(mp, objtype, &info_k);
720 if (error) {
721 return error;
722 }
723
724 return copyout(&info_k, info_u, sizeof(info_k));
725}
726
727static int
728do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u,
729 struct quotaval *val_u)
730{
731 struct quotakey key_k;
732 struct quotaval val_k;
733 int error;
734
735 /* ensure any padding bytes are cleared */
736 memset(&val_k, 0, sizeof(val_k));
737
738 error = copyin(key_u, &key_k, sizeof(key_k));
739 if (error) {
740 return error;
741 }
742
743 error = vfs_quotactl_get(mp, &key_k, &val_k);
744 if (error) {
745 return error;
746 }
747
748 return copyout(&val_k, val_u, sizeof(val_k));
749}
750
751static int
752do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u,
753 const struct quotaval *val_u)
754{
755 struct quotakey key_k;
756 struct quotaval val_k;
757 int error;
758
759 error = copyin(key_u, &key_k, sizeof(key_k));
760 if (error) {
761 return error;
762 }
763
764 error = copyin(val_u, &val_k, sizeof(val_k));
765 if (error) {
766 return error;
767 }
768
769 return vfs_quotactl_put(mp, &key_k, &val_k);
770}
771
772static int
773do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u)
774{
775 struct quotakey key_k;
776 int error;
777
778 error = copyin(key_u, &key_k, sizeof(key_k));
779 if (error) {
780 return error;
781 }
782
783 return vfs_quotactl_del(mp, &key_k);
784}
785
786static int
787do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u)
788{
789 struct quotakcursor cursor_k;
790 int error;
791
792 /* ensure any padding bytes are cleared */
793 memset(&cursor_k, 0, sizeof(cursor_k));
794
795 error = vfs_quotactl_cursoropen(mp, &cursor_k);
796 if (error) {
797 return error;
798 }
799
800 return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
801}
802
803static int
804do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u)
805{
806 struct quotakcursor cursor_k;
807 int error;
808
809 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
810 if (error) {
811 return error;
812 }
813
814 return vfs_quotactl_cursorclose(mp, &cursor_k);
815}
816
817static int
818do_sys_quotactl_cursorskipidtype(struct mount *mp,
819 struct quotakcursor *cursor_u, int idtype)
820{
821 struct quotakcursor cursor_k;
822 int error;
823
824 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
825 if (error) {
826 return error;
827 }
828
829 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype);
830 if (error) {
831 return error;
832 }
833
834 return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
835}
836
837static int
838do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u,
839 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum,
840 unsigned *ret_u)
841{
842#define CGET_STACK_MAX 8
843 struct quotakcursor cursor_k;
844 struct quotakey stackkeys[CGET_STACK_MAX];
845 struct quotaval stackvals[CGET_STACK_MAX];
846 struct quotakey *keys_k;
847 struct quotaval *vals_k;
848 unsigned ret_k;
849 int error;
850
851 if (maxnum > 128) {
852 maxnum = 128;
853 }
854
855 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
856 if (error) {
857 return error;
858 }
859
860 if (maxnum <= CGET_STACK_MAX) {
861 keys_k = stackkeys;
862 vals_k = stackvals;
863 /* ensure any padding bytes are cleared */
864 memset(keys_k, 0, maxnum * sizeof(keys_k[0]));
865 memset(vals_k, 0, maxnum * sizeof(vals_k[0]));
866 } else {
867 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP);
868 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP);
869 }
870
871 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum,
872 &ret_k);
873 if (error) {
874 goto fail;
875 }
876
877 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0]));
878 if (error) {
879 goto fail;
880 }
881
882 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0]));
883 if (error) {
884 goto fail;
885 }
886
887 error = copyout(&ret_k, ret_u, sizeof(ret_k));
888 if (error) {
889 goto fail;
890 }
891
892 /* do last to maximize the chance of being able to recover a failure */
893 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k));
894
895fail:
896 if (keys_k != stackkeys) {
897 kmem_free(keys_k, maxnum * sizeof(keys_k[0]));
898 }
899 if (vals_k != stackvals) {
900 kmem_free(vals_k, maxnum * sizeof(vals_k[0]));
901 }
902 return error;
903}
904
905static int
906do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u,
907 int *ret_u)
908{
909 struct quotakcursor cursor_k;
910 int ret_k;
911 int error;
912
913 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
914 if (error) {
915 return error;
916 }
917
918 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k);
919 if (error) {
920 return error;
921 }
922
923 error = copyout(&ret_k, ret_u, sizeof(ret_k));
924 if (error) {
925 return error;
926 }
927
928 return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
929}
930
931static int
932do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u)
933{
934 struct quotakcursor cursor_k;
935 int error;
936
937 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k));
938 if (error) {
939 return error;
940 }
941
942 error = vfs_quotactl_cursorrewind(mp, &cursor_k);
943 if (error) {
944 return error;
945 }
946
947 return copyout(&cursor_k, cursor_u, sizeof(cursor_k));
948}
949
950static int
951do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u)
952{
953 char *path_k;
954 int error;
955
956 /* XXX this should probably be a struct pathbuf */
957 path_k = PNBUF_GET();
958 error = copyin(path_u, path_k, PATH_MAX);
959 if (error) {
960 PNBUF_PUT(path_k);
961 return error;
962 }
963
964 error = vfs_quotactl_quotaon(mp, idtype, path_k);
965
966 PNBUF_PUT(path_k);
967 return error;
968}
969
970static int
971do_sys_quotactl_quotaoff(struct mount *mp, int idtype)
972{
973 return vfs_quotactl_quotaoff(mp, idtype);
974}
975
976int
977do_sys_quotactl(const char *path_u, const struct quotactl_args *args)
978{
979 struct mount *mp;
980 struct vnode *vp;
981 int error;
982
983 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp);
984 if (error != 0)
985 return (error);
986 mp = vp->v_mount;
987
988 switch (args->qc_op) {
989 case QUOTACTL_STAT:
990 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info);
991 break;
992 case QUOTACTL_IDTYPESTAT:
993 error = do_sys_quotactl_idtypestat(mp,
994 args->u.idtypestat.qc_idtype,
995 args->u.idtypestat.qc_info);
996 break;
997 case QUOTACTL_OBJTYPESTAT:
998 error = do_sys_quotactl_objtypestat(mp,
999 args->u.objtypestat.qc_objtype,
1000 args->u.objtypestat.qc_info);
1001 break;
1002 case QUOTACTL_GET:
1003 error = do_sys_quotactl_get(mp,
1004 args->u.get.qc_key,
1005 args->u.get.qc_val);
1006 break;
1007 case QUOTACTL_PUT:
1008 error = do_sys_quotactl_put(mp,
1009 args->u.put.qc_key,
1010 args->u.put.qc_val);
1011 break;
1012 case QUOTACTL_DEL:
1013 error = do_sys_quotactl_del(mp, args->u.del.qc_key);
1014 break;
1015 case QUOTACTL_CURSOROPEN:
1016 error = do_sys_quotactl_cursoropen(mp,
1017 args->u.cursoropen.qc_cursor);
1018 break;
1019 case QUOTACTL_CURSORCLOSE:
1020 error = do_sys_quotactl_cursorclose(mp,
1021 args->u.cursorclose.qc_cursor);
1022 break;
1023 case QUOTACTL_CURSORSKIPIDTYPE:
1024 error = do_sys_quotactl_cursorskipidtype(mp,
1025 args->u.cursorskipidtype.qc_cursor,
1026 args->u.cursorskipidtype.qc_idtype);
1027 break;
1028 case QUOTACTL_CURSORGET:
1029 error = do_sys_quotactl_cursorget(mp,
1030 args->u.cursorget.qc_cursor,
1031 args->u.cursorget.qc_keys,
1032 args->u.cursorget.qc_vals,
1033 args->u.cursorget.qc_maxnum,
1034 args->u.cursorget.qc_ret);
1035 break;
1036 case QUOTACTL_CURSORATEND:
1037 error = do_sys_quotactl_cursoratend(mp,
1038 args->u.cursoratend.qc_cursor,
1039 args->u.cursoratend.qc_ret);
1040 break;
1041 case QUOTACTL_CURSORREWIND:
1042 error = do_sys_quotactl_cursorrewind(mp,
1043 args->u.cursorrewind.qc_cursor);
1044 break;
1045 case QUOTACTL_QUOTAON:
1046 error = do_sys_quotactl_quotaon(mp,
1047 args->u.quotaon.qc_idtype,
1048 args->u.quotaon.qc_quotafile);
1049 break;
1050 case QUOTACTL_QUOTAOFF:
1051 error = do_sys_quotactl_quotaoff(mp,
1052 args->u.quotaoff.qc_idtype);
1053 break;
1054 default:
1055 error = EINVAL;
1056 break;
1057 }
1058
1059 vrele(vp);
1060 return error;
1061}
1062
1063/* ARGSUSED */
1064int
1065sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap,
1066 register_t *retval)
1067{
1068 /* {
1069 syscallarg(const char *) path;
1070 syscallarg(struct quotactl_args *) args;
1071 } */
1072 struct quotactl_args args;
1073 int error;
1074
1075 error = copyin(SCARG(uap, args), &args, sizeof(args));
1076 if (error) {
1077 return error;
1078 }
1079
1080 return do_sys_quotactl(SCARG(uap, path), &args);
1081}
1082
1083int
1084dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
1085 int root)
1086{
1087 struct cwdinfo *cwdi = l->l_proc->p_cwdi;
1088 int error = 0;
1089
1090 /*
1091 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1092 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
1093 * overrides MNT_NOWAIT.
1094 */
1095 if (flags == MNT_NOWAIT || flags == MNT_LAZY ||
1096 (flags != MNT_WAIT && flags != 0)) {
1097 memcpy(sp, &mp->mnt_stat, sizeof(*sp));
1098 goto done;
1099 }
1100
1101 /* Get the filesystem stats now */
1102 memset(sp, 0, sizeof(*sp));
1103 if ((error = VFS_STATVFS(mp, sp)) != 0) {
1104 return error;
1105 }
1106
1107 if (cwdi->cwdi_rdir == NULL)
1108 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
1109done:
1110 if (cwdi->cwdi_rdir != NULL) {
1111 size_t len;
1112 char *bp;
1113 char c;
1114 char *path = PNBUF_GET();
1115
1116 bp = path + MAXPATHLEN;
1117 *--bp = '\0';
1118 rw_enter(&cwdi->cwdi_lock, RW_READER);
1119 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
1120 MAXPATHLEN / 2, 0, l);
1121 rw_exit(&cwdi->cwdi_lock);
1122 if (error) {
1123 PNBUF_PUT(path);
1124 return error;
1125 }
1126 len = strlen(bp);
1127 if (len != 1) {
1128 /*
1129 * for mount points that are below our root, we can see
1130 * them, so we fix up the pathname and return them. The
1131 * rest we cannot see, so we don't allow viewing the
1132 * data.
1133 */
1134 if (strncmp(bp, sp->f_mntonname, len) == 0 &&
1135 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
1136 (void)strlcpy(sp->f_mntonname,
1137 c == '\0' ? "/" : &sp->f_mntonname[len],
1138 sizeof(sp->f_mntonname));
1139 } else {
1140 if (root)
1141 (void)strlcpy(sp->f_mntonname, "/",
1142 sizeof(sp->f_mntonname));
1143 else
1144 error = EPERM;
1145 }
1146 }
1147 PNBUF_PUT(path);
1148 }
1149 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
1150 return error;
1151}
1152
1153/*
1154 * Get filesystem statistics by path.
1155 */
1156int
1157do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
1158{
1159 struct mount *mp;
1160 int error;
1161 struct vnode *vp;
1162
1163 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp);
1164 if (error != 0)
1165 return error;
1166 mp = vp->v_mount;
1167 error = dostatvfs(mp, sb, l, flags, 1);
1168 vrele(vp);
1169 return error;
1170}
1171
1172/* ARGSUSED */
1173int
1174sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
1175{
1176 /* {
1177 syscallarg(const char *) path;
1178 syscallarg(struct statvfs *) buf;
1179 syscallarg(int) flags;
1180 } */
1181 struct statvfs *sb;
1182 int error;
1183
1184 sb = STATVFSBUF_GET();
1185 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
1186 if (error == 0)
1187 error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1188 STATVFSBUF_PUT(sb);
1189 return error;
1190}
1191
1192/*
1193 * Get filesystem statistics by fd.
1194 */
1195int
1196do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
1197{
1198 file_t *fp;
1199 struct mount *mp;
1200 int error;
1201
1202 /* fd_getvnode() will use the descriptor for us */
1203 if ((error = fd_getvnode(fd, &fp)) != 0)
1204 return (error);
1205 mp = fp->f_vnode->v_mount;
1206 error = dostatvfs(mp, sb, curlwp, flags, 1);
1207 fd_putfile(fd);
1208 return error;
1209}
1210
1211/* ARGSUSED */
1212int
1213sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
1214{
1215 /* {
1216 syscallarg(int) fd;
1217 syscallarg(struct statvfs *) buf;
1218 syscallarg(int) flags;
1219 } */
1220 struct statvfs *sb;
1221 int error;
1222
1223 sb = STATVFSBUF_GET();
1224 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
1225 if (error == 0)
1226 error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1227 STATVFSBUF_PUT(sb);
1228 return error;
1229}
1230
1231
1232/*
1233 * Get statistics on all filesystems.
1234 */
1235int
1236do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
1237 int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
1238 register_t *retval)
1239{
1240 int root = 0;
1241 struct proc *p = l->l_proc;
1242 struct mount *mp, *nmp;
1243 struct statvfs *sb;
1244 size_t count, maxcount;
1245 int error = 0;
1246
1247 sb = STATVFSBUF_GET();
1248 maxcount = bufsize / entry_sz;
1249 mutex_enter(&mountlist_lock);
1250 count = 0;
1251 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
1252 if (vfs_busy(mp, &nmp)) {
1253 continue;
1254 }
1255 if (sfsp && count < maxcount) {
1256 error = dostatvfs(mp, sb, l, flags, 0);
1257 if (error) {
1258 vfs_unbusy(mp, false, &nmp);
1259 error = 0;
1260 continue;
1261 }
1262 error = copyfn(sb, sfsp, entry_sz);
1263 if (error) {
1264 vfs_unbusy(mp, false, NULL);
1265 goto out;
1266 }
1267 sfsp = (char *)sfsp + entry_sz;
1268 root |= strcmp(sb->f_mntonname, "/") == 0;
1269 }
1270 count++;
1271 vfs_unbusy(mp, false, &nmp);
1272 }
1273 mutex_exit(&mountlist_lock);
1274
1275 if (root == 0 && p->p_cwdi->cwdi_rdir) {
1276 /*
1277 * fake a root entry
1278 */
1279 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1280 sb, l, flags, 1);
1281 if (error != 0)
1282 goto out;
1283 if (sfsp) {
1284 error = copyfn(sb, sfsp, entry_sz);
1285 if (error != 0)
1286 goto out;
1287 }
1288 count++;
1289 }
1290 if (sfsp && count > maxcount)
1291 *retval = maxcount;
1292 else
1293 *retval = count;
1294out:
1295 STATVFSBUF_PUT(sb);
1296 return error;
1297}
1298
1299int
1300sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
1301{
1302 /* {
1303 syscallarg(struct statvfs *) buf;
1304 syscallarg(size_t) bufsize;
1305 syscallarg(int) flags;
1306 } */
1307
1308 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1309 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1310}
1311
1312/*
1313 * Change current working directory to a given file descriptor.
1314 */
1315/* ARGSUSED */
1316int
1317sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
1318{
1319 /* {
1320 syscallarg(int) fd;
1321 } */
1322 struct proc *p = l->l_proc;
1323 struct cwdinfo *cwdi;
1324 struct vnode *vp, *tdp;
1325 struct mount *mp;
1326 file_t *fp;
1327 int error, fd;
1328
1329 /* fd_getvnode() will use the descriptor for us */
1330 fd = SCARG(uap, fd);
1331 if ((error = fd_getvnode(fd, &fp)) != 0)
1332 return (error);
1333 vp = fp->f_vnode;
1334
1335 vref(vp);
1336 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1337 if (vp->v_type != VDIR)
1338 error = ENOTDIR;
1339 else
1340 error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1341 if (error) {
1342 vput(vp);
1343 goto out;
1344 }
1345 while ((mp = vp->v_mountedhere) != NULL) {
1346 error = vfs_busy(mp, NULL);
1347 vput(vp);
1348 if (error != 0)
1349 goto out;
1350 error = VFS_ROOT(mp, &tdp);
1351 vfs_unbusy(mp, false, NULL);
1352 if (error)
1353 goto out;
1354 vp = tdp;
1355 }
1356 VOP_UNLOCK(vp);
1357
1358 /*
1359 * Disallow changing to a directory not under the process's
1360 * current root directory (if there is one).
1361 */
1362 cwdi = p->p_cwdi;
1363 rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1364 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1365 vrele(vp);
1366 error = EPERM; /* operation not permitted */
1367 } else {
1368 vrele(cwdi->cwdi_cdir);
1369 cwdi->cwdi_cdir = vp;
1370 }
1371 rw_exit(&cwdi->cwdi_lock);
1372
1373 out:
1374 fd_putfile(fd);
1375 return (error);
1376}
1377
1378/*
1379 * Change this process's notion of the root directory to a given file
1380 * descriptor.
1381 */
1382int
1383sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
1384{
1385 struct proc *p = l->l_proc;
1386 struct vnode *vp;
1387 file_t *fp;
1388 int error, fd = SCARG(uap, fd);
1389
1390 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1391 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1392 return error;
1393 /* fd_getvnode() will use the descriptor for us */
1394 if ((error = fd_getvnode(fd, &fp)) != 0)
1395 return error;
1396 vp = fp->f_vnode;
1397 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1398 if (vp->v_type != VDIR)
1399 error = ENOTDIR;
1400 else
1401 error = VOP_ACCESS(vp, VEXEC, l->l_cred);
1402 VOP_UNLOCK(vp);
1403 if (error)
1404 goto out;
1405 vref(vp);
1406
1407 change_root(p->p_cwdi, vp, l);
1408
1409 out:
1410 fd_putfile(fd);
1411 return (error);
1412}
1413
1414/*
1415 * Change current working directory (``.'').
1416 */
1417/* ARGSUSED */
1418int
1419sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
1420{
1421 /* {
1422 syscallarg(const char *) path;
1423 } */
1424 struct proc *p = l->l_proc;
1425 struct cwdinfo *cwdi;
1426 int error;
1427 struct vnode *vp;
1428
1429 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1430 &vp, l)) != 0)
1431 return (error);
1432 cwdi = p->p_cwdi;
1433 rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1434 vrele(cwdi->cwdi_cdir);
1435 cwdi->cwdi_cdir = vp;
1436 rw_exit(&cwdi->cwdi_lock);
1437 return (0);
1438}
1439
1440/*
1441 * Change notion of root (``/'') directory.
1442 */
1443/* ARGSUSED */
1444int
1445sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
1446{
1447 /* {
1448 syscallarg(const char *) path;
1449 } */
1450 struct proc *p = l->l_proc;
1451 int error;
1452 struct vnode *vp;
1453
1454 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1455 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1456 return (error);
1457 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE,
1458 &vp, l)) != 0)
1459 return (error);
1460
1461 change_root(p->p_cwdi, vp, l);
1462
1463 return (0);
1464}
1465
1466/*
1467 * Common routine for chroot and fchroot.
1468 * NB: callers need to properly authorize the change root operation.
1469 */
1470void
1471change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l)
1472{
1473 struct proc *p = l->l_proc;
1474 kauth_cred_t ncred;
1475
1476 ncred = kauth_cred_alloc();
1477
1478 rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1479 if (cwdi->cwdi_rdir != NULL)
1480 vrele(cwdi->cwdi_rdir);
1481 cwdi->cwdi_rdir = vp;
1482
1483 /*
1484 * Prevent escaping from chroot by putting the root under
1485 * the working directory. Silently chdir to / if we aren't
1486 * already there.
1487 */
1488 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1489 /*
1490 * XXX would be more failsafe to change directory to a
1491 * deadfs node here instead
1492 */
1493 vrele(cwdi->cwdi_cdir);
1494 vref(vp);
1495 cwdi->cwdi_cdir = vp;
1496 }
1497 rw_exit(&cwdi->cwdi_lock);
1498
1499 /* Get a write lock on the process credential. */
1500 proc_crmod_enter();
1501
1502 kauth_cred_clone(p->p_cred, ncred);
1503 kauth_proc_chroot(ncred, p->p_cwdi);
1504
1505 /* Broadcast our credentials to the process and other LWPs. */
1506 proc_crmod_leave(ncred, p->p_cred, true);
1507}
1508
1509/*
1510 * Common routine for chroot and chdir.
1511 * XXX "where" should be enum uio_seg
1512 */
1513int
1514chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l)
1515{
1516 struct pathbuf *pb;
1517 struct nameidata nd;
1518 int error;
1519
1520 error = pathbuf_maybe_copyin(path, where, &pb);
1521 if (error) {
1522 return error;
1523 }
1524 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
1525 if ((error = namei(&nd)) != 0) {
1526 pathbuf_destroy(pb);
1527 return error;
1528 }
1529 *vpp = nd.ni_vp;
1530 pathbuf_destroy(pb);
1531
1532 if ((*vpp)->v_type != VDIR)
1533 error = ENOTDIR;
1534 else
1535 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred);
1536
1537 if (error)
1538 vput(*vpp);
1539 else
1540 VOP_UNLOCK(*vpp);
1541 return (error);
1542}
1543
1544/*
1545 * Internals of sys_open - path has already been converted into a pathbuf
1546 * (so we can easily reuse this function from other parts of the kernel,
1547 * like posix_spawn post-processing).
1548 */
1549int
1550do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags,
1551 int open_mode, int *fd)
1552{
1553 struct proc *p = l->l_proc;
1554 struct cwdinfo *cwdi = p->p_cwdi;
1555 file_t *fp;
1556 struct vnode *vp;
1557 int flags, cmode;
1558 int indx, error;
1559 struct nameidata nd;
1560
1561 if (open_flags & O_SEARCH) {
1562 open_flags &= ~(int)O_SEARCH;
1563 }
1564
1565 flags = FFLAGS(open_flags);
1566 if ((flags & (FREAD | FWRITE)) == 0)
1567 return EINVAL;
1568
1569 if ((error = fd_allocfile(&fp, &indx)) != 0) {
1570 return error;
1571 }
1572
1573 /* We're going to read cwdi->cwdi_cmask unlocked here. */
1574 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1575 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb);
1576 if (dvp != NULL)
1577 NDAT(&nd, dvp);
1578
1579 l->l_dupfd = -indx - 1; /* XXX check for fdopen */
1580 if ((error = vn_open(&nd, flags, cmode)) != 0) {
1581 fd_abort(p, fp, indx);
1582 if ((error == EDUPFD || error == EMOVEFD) &&
1583 l->l_dupfd >= 0 && /* XXX from fdopen */
1584 (error =
1585 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
1586 *fd = indx;
1587 return 0;
1588 }
1589 if (error == ERESTART)
1590 error = EINTR;
1591 return error;
1592 }
1593
1594 l->l_dupfd = 0;
1595 vp = nd.ni_vp;
1596
1597 if ((error = open_setfp(l, fp, vp, indx, flags)))
1598 return error;
1599
1600 VOP_UNLOCK(vp);
1601 *fd = indx;
1602 fd_affix(p, fp, indx);
1603 return 0;
1604}
1605
1606int
1607fd_open(const char *path, int open_flags, int open_mode, int *fd)
1608{
1609 struct pathbuf *pb;
1610 int error, oflags;
1611
1612 oflags = FFLAGS(open_flags);
1613 if ((oflags & (FREAD | FWRITE)) == 0)
1614 return EINVAL;
1615
1616 pb = pathbuf_create(path);
1617 if (pb == NULL)
1618 return ENOMEM;
1619
1620 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd);
1621 pathbuf_destroy(pb);
1622
1623 return error;
1624}
1625
1626/*
1627 * Check permissions, allocate an open file structure,
1628 * and call the device open routine if any.
1629 */
1630static int
1631do_sys_openat(lwp_t *l, int fdat, const char *path, int flags,
1632 int mode, int *fd)
1633{
1634 file_t *dfp = NULL;
1635 struct vnode *dvp = NULL;
1636 struct pathbuf *pb;
1637 int error;
1638
1639#ifdef COMPAT_10 /* XXX: and perhaps later */
1640 if (path == NULL) {
1641 pb = pathbuf_create(".");
1642 if (pb == NULL)
1643 return ENOMEM;
1644 } else
1645#endif
1646 {
1647 error = pathbuf_copyin(path, &pb);
1648 if (error)
1649 return error;
1650 }
1651
1652 if (fdat != AT_FDCWD) {
1653 /* fd_getvnode() will use the descriptor for us */
1654 if ((error = fd_getvnode(fdat, &dfp)) != 0)
1655 goto out;
1656
1657 dvp = dfp->f_vnode;
1658 }
1659
1660 error = do_open(l, dvp, pb, flags, mode, fd);
1661
1662 if (dfp != NULL)
1663 fd_putfile(fdat);
1664out:
1665 pathbuf_destroy(pb);
1666 return error;
1667}
1668
1669int
1670sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
1671{
1672 /* {
1673 syscallarg(const char *) path;
1674 syscallarg(int) flags;
1675 syscallarg(int) mode;
1676 } */
1677 int error;
1678 int fd;
1679
1680 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path),
1681 SCARG(uap, flags), SCARG(uap, mode), &fd);
1682
1683 if (error == 0)
1684 *retval = fd;
1685
1686 return error;
1687}
1688
1689int
1690sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval)
1691{
1692 /* {
1693 syscallarg(int) fd;
1694 syscallarg(const char *) path;
1695 syscallarg(int) oflags;
1696 syscallarg(int) mode;
1697 } */
1698 int error;
1699 int fd;
1700
1701 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path),
1702 SCARG(uap, oflags), SCARG(uap, mode), &fd);
1703
1704 if (error == 0)
1705 *retval = fd;
1706
1707 return error;
1708}
1709
1710static void
1711vfs__fhfree(fhandle_t *fhp)
1712{
1713 size_t fhsize;
1714
1715 fhsize = FHANDLE_SIZE(fhp);
1716 kmem_free(fhp, fhsize);
1717}
1718
1719/*
1720 * vfs_composefh: compose a filehandle.
1721 */
1722
1723int
1724vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1725{
1726 struct mount *mp;
1727 struct fid *fidp;
1728 int error;
1729 size_t needfhsize;
1730 size_t fidsize;
1731
1732 mp = vp->v_mount;
1733 fidp = NULL;
1734 if (*fh_size < FHANDLE_SIZE_MIN) {
1735 fidsize = 0;
1736 } else {
1737 fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1738 if (fhp != NULL) {
1739 memset(fhp, 0, *fh_size);
1740 fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1741 fidp = &fhp->fh_fid;
1742 }
1743 }
1744 error = VFS_VPTOFH(vp, fidp, &fidsize);
1745 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1746 if (error == 0 && *fh_size < needfhsize) {
1747 error = E2BIG;
1748 }
1749 *fh_size = needfhsize;
1750 return error;
1751}
1752
1753int
1754vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1755{
1756 struct mount *mp;
1757 fhandle_t *fhp;
1758 size_t fhsize;
1759 size_t fidsize;
1760 int error;
1761
1762 mp = vp->v_mount;
1763 fidsize = 0;
1764 error = VFS_VPTOFH(vp, NULL, &fidsize);
1765 KASSERT(error != 0);
1766 if (error != E2BIG) {
1767 goto out;
1768 }
1769 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1770 fhp = kmem_zalloc(fhsize, KM_SLEEP);
1771 if (fhp == NULL) {
1772 error = ENOMEM;
1773 goto out;
1774 }
1775 fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1776 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1777 if (error == 0) {
1778 KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1779 FHANDLE_FILEID(fhp)->fid_len == fidsize));
1780 *fhpp = fhp;
1781 } else {
1782 kmem_free(fhp, fhsize);
1783 }
1784out:
1785 return error;
1786}
1787
1788void
1789vfs_composefh_free(fhandle_t *fhp)
1790{
1791
1792 vfs__fhfree(fhp);
1793}
1794
1795/*
1796 * vfs_fhtovp: lookup a vnode by a filehandle.
1797 */
1798
1799int
1800vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1801{
1802 struct mount *mp;
1803 int error;
1804
1805 *vpp = NULL;
1806 mp = vfs_getvfs(FHANDLE_FSID(fhp));
1807 if (mp == NULL) {
1808 error = ESTALE;
1809 goto out;
1810 }
1811 if (mp->mnt_op->vfs_fhtovp == NULL) {
1812 error = EOPNOTSUPP;
1813 goto out;
1814 }
1815 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1816out:
1817 return error;
1818}
1819
1820/*
1821 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1822 * the needed size.
1823 */
1824
1825int
1826vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1827{
1828 fhandle_t *fhp;
1829 int error;
1830
1831 if (fhsize > FHANDLE_SIZE_MAX) {
1832 return EINVAL;
1833 }
1834 if (fhsize < FHANDLE_SIZE_MIN) {
1835 return EINVAL;
1836 }
1837again:
1838 fhp = kmem_alloc(fhsize, KM_SLEEP);
1839 if (fhp == NULL) {
1840 return ENOMEM;
1841 }
1842 error = copyin(ufhp, fhp, fhsize);
1843 if (error == 0) {
1844 /* XXX this check shouldn't be here */
1845 if (FHANDLE_SIZE(fhp) == fhsize) {
1846 *fhpp = fhp;
1847 return 0;
1848 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1849 /*
1850 * a kludge for nfsv2 padded handles.
1851 */
1852 size_t sz;
1853
1854 sz = FHANDLE_SIZE(fhp);
1855 kmem_free(fhp, fhsize);
1856 fhsize = sz;
1857 goto again;
1858 } else {
1859 /*
1860 * userland told us wrong size.
1861 */
1862 error = EINVAL;
1863 }
1864 }
1865 kmem_free(fhp, fhsize);
1866 return error;
1867}
1868
1869void
1870vfs_copyinfh_free(fhandle_t *fhp)
1871{
1872
1873 vfs__fhfree(fhp);
1874}
1875
1876/*
1877 * Get file handle system call
1878 */
1879int
1880sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
1881{
1882 /* {
1883 syscallarg(char *) fname;
1884 syscallarg(fhandle_t *) fhp;
1885 syscallarg(size_t *) fh_size;
1886 } */
1887 struct vnode *vp;
1888 fhandle_t *fh;
1889 int error;
1890 struct pathbuf *pb;
1891 struct nameidata nd;
1892 size_t sz;
1893 size_t usz;
1894
1895 /*
1896 * Must be super user
1897 */
1898 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1899 0, NULL, NULL, NULL);
1900 if (error)
1901 return (error);
1902
1903 error = pathbuf_copyin(SCARG(uap, fname), &pb);
1904 if (error) {
1905 return error;
1906 }
1907 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
1908 error = namei(&nd);
1909 if (error) {
1910 pathbuf_destroy(pb);
1911 return error;
1912 }
1913 vp = nd.ni_vp;
1914 pathbuf_destroy(pb);
1915
1916 error = vfs_composefh_alloc(vp, &fh);
1917 vput(vp);
1918 if (error != 0) {
1919 return error;
1920 }
1921 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1922 if (error != 0) {
1923 goto out;
1924 }
1925 sz = FHANDLE_SIZE(fh);
1926 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1927 if (error != 0) {
1928 goto out;
1929 }
1930 if (usz >= sz) {
1931 error = copyout(fh, SCARG(uap, fhp), sz);
1932 } else {
1933 error = E2BIG;
1934 }
1935out:
1936 vfs_composefh_free(fh);
1937 return (error);
1938}
1939
1940/*
1941 * Open a file given a file handle.
1942 *
1943 * Check permissions, allocate an open file structure,
1944 * and call the device open routine if any.
1945 */
1946
1947int
1948dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1949 register_t *retval)
1950{
1951 file_t *fp;
1952 struct vnode *vp = NULL;
1953 kauth_cred_t cred = l->l_cred;
1954 file_t *nfp;
1955 int indx, error;
1956 struct vattr va;
1957 fhandle_t *fh;
1958 int flags;
1959 proc_t *p;
1960
1961 p = curproc;
1962
1963 /*
1964 * Must be super user
1965 */
1966 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1967 0, NULL, NULL, NULL)))
1968 return (error);
1969
1970 if (oflags & O_SEARCH) {
1971 oflags &= ~(int)O_SEARCH;
1972 }
1973
1974 flags = FFLAGS(oflags);
1975 if ((flags & (FREAD | FWRITE)) == 0)
1976 return (EINVAL);
1977 if ((flags & O_CREAT))
1978 return (EINVAL);
1979 if ((error = fd_allocfile(&nfp, &indx)) != 0)
1980 return (error);
1981 fp = nfp;
1982 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1983 if (error != 0) {
1984 goto bad;
1985 }
1986 error = vfs_fhtovp(fh, &vp);
1987 vfs_copyinfh_free(fh);
1988 if (error != 0) {
1989 goto bad;
1990 }
1991
1992 /* Now do an effective vn_open */
1993
1994 if (vp->v_type == VSOCK) {
1995 error = EOPNOTSUPP;
1996 goto bad;
1997 }
1998 error = vn_openchk(vp, cred, flags);
1999 if (error != 0)
2000 goto bad;
2001 if (flags & O_TRUNC) {
2002 VOP_UNLOCK(vp); /* XXX */
2003 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
2004 vattr_null(&va);
2005 va.va_size = 0;
2006 error = VOP_SETATTR(vp, &va, cred);
2007 if (error)
2008 goto bad;
2009 }
2010 if ((error = VOP_OPEN(vp, flags, cred)) != 0)
2011 goto bad;
2012 if (flags & FWRITE) {
2013 mutex_enter(vp->v_interlock);
2014 vp->v_writecount++;
2015 mutex_exit(vp->v_interlock);
2016 }
2017
2018 /* done with modified vn_open, now finish what sys_open does. */
2019 if ((error = open_setfp(l, fp, vp, indx, flags)))
2020 return error;
2021
2022 VOP_UNLOCK(vp);
2023 *retval = indx;
2024 fd_affix(p, fp, indx);
2025 return (0);
2026
2027bad:
2028 fd_abort(p, fp, indx);
2029 if (vp != NULL)
2030 vput(vp);
2031 return (error);
2032}
2033
2034int
2035sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
2036{
2037 /* {
2038 syscallarg(const void *) fhp;
2039 syscallarg(size_t) fh_size;
2040 syscallarg(int) flags;
2041 } */
2042
2043 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
2044 SCARG(uap, flags), retval);
2045}
2046
2047int
2048do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
2049{
2050 int error;
2051 fhandle_t *fh;
2052 struct vnode *vp;
2053
2054 /*
2055 * Must be super user
2056 */
2057 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
2058 0, NULL, NULL, NULL)))
2059 return (error);
2060
2061 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
2062 if (error != 0)
2063 return error;
2064
2065 error = vfs_fhtovp(fh, &vp);
2066 vfs_copyinfh_free(fh);
2067 if (error != 0)
2068 return error;
2069
2070 error = vn_stat(vp, sb);
2071 vput(vp);
2072 return error;
2073}
2074
2075
2076/* ARGSUSED */
2077int
2078sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval)
2079{
2080 /* {
2081 syscallarg(const void *) fhp;
2082 syscallarg(size_t) fh_size;
2083 syscallarg(struct stat *) sb;
2084 } */
2085 struct stat sb;
2086 int error;
2087
2088 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
2089 if (error)
2090 return error;
2091 return copyout(&sb, SCARG(uap, sb), sizeof(sb));
2092}
2093
2094int
2095do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
2096 int flags)
2097{
2098 fhandle_t *fh;
2099 struct mount *mp;
2100 struct vnode *vp;
2101 int error;
2102
2103 /*
2104 * Must be super user
2105 */
2106 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
2107 0, NULL, NULL, NULL)))
2108 return error;
2109
2110 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
2111 if (error != 0)
2112 return error;
2113
2114 error = vfs_fhtovp(fh, &vp);
2115 vfs_copyinfh_free(fh);
2116 if (error != 0)
2117 return error;
2118
2119 mp = vp->v_mount;
2120 error = dostatvfs(mp, sb, l, flags, 1);
2121 vput(vp);
2122 return error;
2123}
2124
2125/* ARGSUSED */
2126int
2127sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
2128{
2129 /* {
2130 syscallarg(const void *) fhp;
2131 syscallarg(size_t) fh_size;
2132 syscallarg(struct statvfs *) buf;
2133 syscallarg(int) flags;
2134 } */
2135 struct statvfs *sb = STATVFSBUF_GET();
2136 int error;
2137
2138 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
2139 SCARG(uap, flags));
2140 if (error == 0)
2141 error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
2142 STATVFSBUF_PUT(sb);
2143 return error;
2144}
2145
2146/*
2147 * Create a special file.
2148 */
2149/* ARGSUSED */
2150int
2151sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap,
2152 register_t *retval)
2153{
2154 /* {
2155 syscallarg(const char *) path;
2156 syscallarg(mode_t) mode;
2157 syscallarg(dev_t) dev;
2158 } */
2159 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode),
2160 SCARG(uap, dev), retval, UIO_USERSPACE);
2161}
2162
2163int
2164sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap,
2165 register_t *retval)
2166{
2167 /* {
2168 syscallarg(int) fd;
2169 syscallarg(const char *) path;
2170 syscallarg(mode_t) mode;
2171 syscallarg(int) pad;
2172 syscallarg(dev_t) dev;
2173 } */
2174
2175 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path),
2176 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE);
2177}
2178
2179int
2180do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev,
2181 register_t *retval, enum uio_seg seg)
2182{
2183 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg);
2184}
2185
2186int
2187do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode,
2188 dev_t dev, register_t *retval, enum uio_seg seg)
2189{
2190 struct proc *p = l->l_proc;
2191 struct vnode *vp;
2192 struct vattr vattr;
2193 int error, optype;
2194 struct pathbuf *pb;
2195 struct nameidata nd;
2196 const char *pathstring;
2197
2198 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
2199 0, NULL, NULL, NULL)) != 0)
2200 return (error);
2201
2202 optype = VOP_MKNOD_DESCOFFSET;
2203
2204 error = pathbuf_maybe_copyin(pathname, seg, &pb);
2205 if (error) {
2206 return error;
2207 }
2208 pathstring = pathbuf_stringcopy_get(pb);
2209 if (pathstring == NULL) {
2210 pathbuf_destroy(pb);
2211 return ENOMEM;
2212 }
2213
2214 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb);
2215
2216 if ((error = fd_nameiat(l, fdat, &nd)) != 0)
2217 goto out;
2218 vp = nd.ni_vp;
2219
2220 if (vp != NULL)
2221 error = EEXIST;
2222 else {
2223 vattr_null(&vattr);
2224 /* We will read cwdi->cwdi_cmask unlocked. */
2225 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
2226 vattr.va_rdev = dev;
2227
2228 switch (mode & S_IFMT) {
2229 case S_IFMT: /* used by badsect to flag bad sectors */
2230 vattr.va_type = VBAD;
2231 break;
2232 case S_IFCHR:
2233 vattr.va_type = VCHR;
2234 break;
2235 case S_IFBLK:
2236 vattr.va_type = VBLK;
2237 break;
2238 case S_IFWHT:
2239 optype = VOP_WHITEOUT_DESCOFFSET;
2240 break;
2241 case S_IFREG:
2242#if NVERIEXEC > 0
2243 error = veriexec_openchk(l, nd.ni_vp, pathstring,
2244 O_CREAT);
2245#endif /* NVERIEXEC > 0 */
2246 vattr.va_type = VREG;
2247 vattr.va_rdev = VNOVAL;
2248 optype = VOP_CREATE_DESCOFFSET;
2249 break;
2250 default:
2251 error = EINVAL;
2252 break;
2253 }
2254 }
2255 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET
2256 && vattr.va_rdev == VNOVAL)
2257 error = EINVAL;
2258 if (!error) {
2259 switch (optype) {
2260 case VOP_WHITEOUT_DESCOFFSET:
2261 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
2262 if (error)
2263 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2264 vput(nd.ni_dvp);
2265 break;
2266
2267 case VOP_MKNOD_DESCOFFSET:
2268 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
2269 &nd.ni_cnd, &vattr);
2270 if (error == 0)
2271 vrele(nd.ni_vp);
2272 vput(nd.ni_dvp);
2273 break;
2274
2275 case VOP_CREATE_DESCOFFSET:
2276 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
2277 &nd.ni_cnd, &vattr);
2278 if (error == 0)
2279 vrele(nd.ni_vp);
2280 vput(nd.ni_dvp);
2281 break;
2282 }
2283 } else {
2284 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2285 if (nd.ni_dvp == vp)
2286 vrele(nd.ni_dvp);
2287 else
2288 vput(nd.ni_dvp);
2289 if (vp)
2290 vrele(vp);
2291 }
2292out:
2293 pathbuf_stringcopy_put(pb, pathstring);
2294 pathbuf_destroy(pb);
2295 return (error);
2296}
2297
2298/*
2299 * Create a named pipe.
2300 */
2301/* ARGSUSED */
2302int
2303sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
2304{
2305 /* {
2306 syscallarg(const char *) path;
2307 syscallarg(int) mode;
2308 } */
2309 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode));
2310}
2311
2312int
2313sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap,
2314 register_t *retval)
2315{
2316 /* {
2317 syscallarg(int) fd;
2318 syscallarg(const char *) path;
2319 syscallarg(int) mode;
2320 } */
2321
2322 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path),
2323 SCARG(uap, mode));
2324}
2325
2326static int
2327do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode)
2328{
2329 struct proc *p = l->l_proc;
2330 struct vattr vattr;
2331 int error;
2332 struct pathbuf *pb;
2333 struct nameidata nd;
2334
2335 error = pathbuf_copyin(path, &pb);
2336 if (error) {
2337 return error;
2338 }
2339 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb);
2340
2341 if ((error = fd_nameiat(l, fdat, &nd)) != 0) {
2342 pathbuf_destroy(pb);
2343 return error;
2344 }
2345 if (nd.ni_vp != NULL) {
2346 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2347 if (nd.ni_dvp == nd.ni_vp)
2348 vrele(nd.ni_dvp);
2349 else
2350 vput(nd.ni_dvp);
2351 vrele(nd.ni_vp);
2352 pathbuf_destroy(pb);
2353 return (EEXIST);
2354 }
2355 vattr_null(&vattr);
2356 vattr.va_type = VFIFO;
2357 /* We will read cwdi->cwdi_cmask unlocked. */
2358 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
2359 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2360 if (error == 0)
2361 vrele(nd.ni_vp);
2362 vput(nd.ni_dvp);
2363 pathbuf_destroy(pb);
2364 return (error);
2365}
2366
2367/*
2368 * Make a hard file link.
2369 */
2370/* ARGSUSED */
2371int
2372do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink,
2373 const char *link, int follow, register_t *retval)
2374{
2375 struct vnode *vp;
2376 struct pathbuf *linkpb;
2377 struct nameidata nd;
2378 namei_simple_flags_t ns_flags;
2379 int error;
2380
2381 if (follow & AT_SYMLINK_FOLLOW)
2382 ns_flags = NSM_FOLLOW_TRYEMULROOT;
2383 else
2384 ns_flags = NSM_NOFOLLOW_TRYEMULROOT;
2385
2386 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp);
2387 if (error != 0)
2388 return (error);
2389 error = pathbuf_copyin(link, &linkpb);
2390 if (error) {
2391 goto out1;
2392 }
2393 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb);
2394 if ((error = fd_nameiat(l, fdlink, &nd)) != 0)
2395 goto out2;
2396 if (nd.ni_vp) {
2397 error = EEXIST;
2398 goto abortop;
2399 }
2400 /* Prevent hard links on directories. */
2401 if (vp->v_type == VDIR) {
2402 error = EPERM;
2403 goto abortop;
2404 }
2405 /* Prevent cross-mount operation. */
2406 if (nd.ni_dvp->v_mount != vp->v_mount) {
2407 error = EXDEV;
2408 goto abortop;
2409 }
2410 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2411 VOP_UNLOCK(nd.ni_dvp);
2412 vrele(nd.ni_dvp);
2413out2:
2414 pathbuf_destroy(linkpb);
2415out1:
2416 vrele(vp);
2417 return (error);
2418abortop:
2419 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2420 if (nd.ni_dvp == nd.ni_vp)
2421 vrele(nd.ni_dvp);
2422 else
2423 vput(nd.ni_dvp);
2424 if (nd.ni_vp != NULL)
2425 vrele(nd.ni_vp);
2426 goto out2;
2427}
2428
2429int
2430sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
2431{
2432 /* {
2433 syscallarg(const char *) path;
2434 syscallarg(const char *) link;
2435 } */
2436 const char *path = SCARG(uap, path);
2437 const char *link = SCARG(uap, link);
2438
2439 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link,
2440 AT_SYMLINK_FOLLOW, retval);
2441}
2442
2443int
2444sys_linkat(struct lwp *l, const struct sys_linkat_args *uap,
2445 register_t *retval)
2446{
2447 /* {
2448 syscallarg(int) fd1;
2449 syscallarg(const char *) name1;
2450 syscallarg(int) fd2;
2451 syscallarg(const char *) name2;
2452 syscallarg(int) flags;
2453 } */
2454 int fd1 = SCARG(uap, fd1);
2455 const char *name1 = SCARG(uap, name1);
2456 int fd2 = SCARG(uap, fd2);
2457 const char *name2 = SCARG(uap, name2);
2458 int follow;
2459
2460 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW;
2461
2462 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval);
2463}
2464
2465
2466int
2467do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg)
2468{
2469 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg);
2470}
2471
2472static int
2473do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat,
2474 const char *link, enum uio_seg seg)
2475{
2476 struct proc *p = curproc;
2477 struct vattr vattr;
2478 char *path;
2479 int error;
2480 struct pathbuf *linkpb;
2481 struct nameidata nd;
2482
2483 KASSERT(l != NULL || fdat == AT_FDCWD);
2484
2485 path = PNBUF_GET();
2486 if (seg == UIO_USERSPACE) {
2487 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0)
2488 goto out1;
2489 if ((error = pathbuf_copyin(link, &linkpb)) != 0)
2490 goto out1;
2491 } else {
2492 KASSERT(strlen(patharg) < MAXPATHLEN);
2493 strcpy(path, patharg);
2494 linkpb = pathbuf_create(link);
2495 if (linkpb == NULL) {
2496 error = ENOMEM;
2497 goto out1;
2498 }
2499 }
2500 ktrkuser("symlink-target", path, strlen(path));
2501
2502 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb);
2503 if ((error = fd_nameiat(l, fdat, &nd)) != 0)
2504 goto out2;
2505 if (nd.ni_vp) {
2506 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2507 if (nd.ni_dvp == nd.ni_vp)
2508 vrele(nd.ni_dvp);
2509 else
2510 vput(nd.ni_dvp);
2511 vrele(nd.ni_vp);
2512 error = EEXIST;
2513 goto out2;
2514 }
2515 vattr_null(&vattr);
2516 vattr.va_type = VLNK;
2517 /* We will read cwdi->cwdi_cmask unlocked. */
2518 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
2519 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
2520 if (error == 0)
2521 vrele(nd.ni_vp);
2522 vput(nd.ni_dvp);
2523out2:
2524 pathbuf_destroy(linkpb);
2525out1:
2526 PNBUF_PUT(path);
2527 return (error);
2528}
2529
2530/*
2531 * Make a symbolic link.
2532 */
2533/* ARGSUSED */
2534int
2535sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
2536{
2537 /* {
2538 syscallarg(const char *) path;
2539 syscallarg(const char *) link;
2540 } */
2541
2542 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link),
2543 UIO_USERSPACE);
2544}
2545
2546int
2547sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap,
2548 register_t *retval)
2549{
2550 /* {
2551 syscallarg(const char *) path1;
2552 syscallarg(int) fd;
2553 syscallarg(const char *) path2;
2554 } */
2555
2556 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd),
2557 SCARG(uap, path2), UIO_USERSPACE);
2558}
2559
2560/*
2561 * Delete a whiteout from the filesystem.
2562 */
2563/* ARGSUSED */
2564int
2565sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
2566{
2567 /* {
2568 syscallarg(const char *) path;
2569 } */
2570 int error;
2571 struct pathbuf *pb;
2572 struct nameidata nd;
2573
2574 error = pathbuf_copyin(SCARG(uap, path), &pb);
2575 if (error) {
2576 return error;
2577 }
2578
2579 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb);
2580 error = namei(&nd);
2581 if (error) {
2582 pathbuf_destroy(pb);
2583 return (error);
2584 }
2585
2586 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2587 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2588 if (nd.ni_dvp == nd.ni_vp)
2589 vrele(nd.ni_dvp);
2590 else
2591 vput(nd.ni_dvp);
2592 if (nd.ni_vp)
2593 vrele(nd.ni_vp);
2594 pathbuf_destroy(pb);
2595 return (EEXIST);
2596 }
2597 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2598 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2599 vput(nd.ni_dvp);
2600 pathbuf_destroy(pb);
2601 return (error);
2602}
2603
2604/*
2605 * Delete a name from the filesystem.
2606 */
2607/* ARGSUSED */
2608int
2609sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
2610{
2611 /* {
2612 syscallarg(const char *) path;
2613 } */
2614
2615 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE);
2616}
2617
2618int
2619sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap,
2620 register_t *retval)
2621{
2622 /* {
2623 syscallarg(int) fd;
2624 syscallarg(const char *) path;
2625 syscallarg(int) flag;
2626 } */
2627
2628 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path),
2629 SCARG(uap, flag), UIO_USERSPACE);
2630}
2631
2632int
2633do_sys_unlink(const char *arg, enum uio_seg seg)
2634{
2635 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg);
2636}
2637
2638static int
2639do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags,
2640 enum uio_seg seg)
2641{
2642 struct vnode *vp;
2643 int error;
2644 struct pathbuf *pb;
2645 struct nameidata nd;
2646 const char *pathstring;
2647
2648 KASSERT(l != NULL || fdat == AT_FDCWD);
2649
2650 error = pathbuf_maybe_copyin(arg, seg, &pb);
2651 if (error) {
2652 return error;
2653 }
2654 pathstring = pathbuf_stringcopy_get(pb);
2655 if (pathstring == NULL) {
2656 pathbuf_destroy(pb);
2657 return ENOMEM;
2658 }
2659
2660 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb);
2661 if ((error = fd_nameiat(l, fdat, &nd)) != 0)
2662 goto out;
2663 vp = nd.ni_vp;
2664
2665 /*
2666 * The root of a mounted filesystem cannot be deleted.
2667 */
2668 if ((vp->v_vflag & VV_ROOT) != 0) {
2669 error = EBUSY;
2670 goto abort;
2671 }
2672
2673 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) {
2674 error = EBUSY;
2675 goto abort;
2676 }
2677
2678 /*
2679 * No rmdir "." please.
2680 */
2681 if (nd.ni_dvp == vp) {
2682 error = EINVAL;
2683 goto abort;
2684 }
2685
2686 /*
2687 * AT_REMOVEDIR is required to remove a directory
2688 */
2689 if (vp->v_type == VDIR) {
2690 if (!(flags & AT_REMOVEDIR)) {
2691 error = EPERM;
2692 goto abort;
2693 } else {
2694 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2695 goto out;
2696 }
2697 }
2698
2699 /*
2700 * Starting here we only deal with non directories.
2701 */
2702 if (flags & AT_REMOVEDIR) {
2703 error = ENOTDIR;
2704 goto abort;
2705 }
2706
2707#if NVERIEXEC > 0
2708 /* Handle remove requests for veriexec entries. */
2709 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) {
2710 goto abort;
2711 }
2712#endif /* NVERIEXEC > 0 */
2713
2714#ifdef FILEASSOC
2715 (void)fileassoc_file_delete(vp);
2716#endif /* FILEASSOC */
2717 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2718 goto out;
2719
2720abort:
2721 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2722 if (nd.ni_dvp == vp)
2723 vrele(nd.ni_dvp);
2724 else
2725 vput(nd.ni_dvp);
2726 vput(vp);
2727
2728out:
2729 pathbuf_stringcopy_put(pb, pathstring);
2730 pathbuf_destroy(pb);
2731 return (error);
2732}
2733
2734/*
2735 * Reposition read/write file offset.
2736 */
2737int
2738sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
2739{
2740 /* {
2741 syscallarg(int) fd;
2742 syscallarg(int) pad;
2743 syscallarg(off_t) offset;
2744 syscallarg(int) whence;
2745 } */
2746 kauth_cred_t cred = l->l_cred;
2747 file_t *fp;
2748 struct vnode *vp;
2749 struct vattr vattr;
2750 off_t newoff;
2751 int error, fd;
2752
2753 fd = SCARG(uap, fd);
2754
2755 if ((fp = fd_getfile(fd)) == NULL)
2756 return (EBADF);
2757
2758 vp = fp->f_vnode;
2759 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2760 error = ESPIPE;
2761 goto out;
2762 }
2763
2764 vn_lock(vp, LK_SHARED | LK_RETRY);
2765
2766 switch (SCARG(uap, whence)) {
2767 case SEEK_CUR:
2768 newoff = fp->f_offset + SCARG(uap, offset);
2769 break;
2770 case SEEK_END:
2771 error = VOP_GETATTR(vp, &vattr, cred);
2772 if (error) {
2773 VOP_UNLOCK(vp);
2774 goto out;
2775 }
2776 newoff = SCARG(uap, offset) + vattr.va_size;
2777 break;
2778 case SEEK_SET:
2779 newoff = SCARG(uap, offset);
2780 break;
2781 default:
2782 error = EINVAL;
2783 VOP_UNLOCK(vp);
2784 goto out;
2785 }
2786 VOP_UNLOCK(vp);
2787 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2788 *(off_t *)retval = fp->f_offset = newoff;
2789 }
2790 out:
2791 fd_putfile(fd);
2792 return (error);
2793}
2794
2795/*
2796 * Positional read system call.
2797 */
2798int
2799sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
2800{
2801 /* {
2802 syscallarg(int) fd;
2803 syscallarg(void *) buf;
2804 syscallarg(size_t) nbyte;
2805 syscallarg(off_t) offset;
2806 } */
2807 file_t *fp;
2808 struct vnode *vp;
2809 off_t offset;
2810 int error, fd = SCARG(uap, fd);
2811
2812 if ((fp = fd_getfile(fd)) == NULL)
2813 return (EBADF);
2814
2815 if ((fp->f_flag & FREAD) == 0) {
2816 fd_putfile(fd);
2817 return (EBADF);
2818 }
2819
2820 vp = fp->f_vnode;
2821 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2822 error = ESPIPE;
2823 goto out;
2824 }
2825
2826 offset = SCARG(uap, offset);
2827
2828 /*
2829 * XXX This works because no file systems actually
2830 * XXX take any action on the seek operation.
2831 */
2832 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2833 goto out;
2834
2835 /* dofileread() will unuse the descriptor for us */
2836 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2837 &offset, 0, retval));
2838
2839 out:
2840 fd_putfile(fd);
2841 return (error);
2842}
2843
2844/*
2845 * Positional scatter read system call.
2846 */
2847int
2848sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
2849{
2850 /* {
2851 syscallarg(int) fd;
2852 syscallarg(const struct iovec *) iovp;
2853 syscallarg(int) iovcnt;
2854 syscallarg(off_t) offset;
2855 } */
2856 off_t offset = SCARG(uap, offset);
2857
2858 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2859 SCARG(uap, iovcnt), &offset, 0, retval);
2860}
2861
2862/*
2863 * Positional write system call.
2864 */
2865int
2866sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
2867{
2868 /* {
2869 syscallarg(int) fd;
2870 syscallarg(const void *) buf;
2871 syscallarg(size_t) nbyte;
2872 syscallarg(off_t) offset;
2873 } */
2874 file_t *fp;
2875 struct vnode *vp;
2876 off_t offset;
2877 int error, fd = SCARG(uap, fd);
2878
2879 if ((fp = fd_getfile(fd)) == NULL)
2880 return (EBADF);
2881
2882 if ((fp->f_flag & FWRITE) == 0) {
2883 fd_putfile(fd);
2884 return (EBADF);
2885 }
2886
2887 vp = fp->f_vnode;
2888 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2889 error = ESPIPE;
2890 goto out;
2891 }
2892
2893 offset = SCARG(uap, offset);
2894
2895 /*
2896 * XXX This works because no file systems actually
2897 * XXX take any action on the seek operation.
2898 */
2899 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2900 goto out;
2901
2902 /* dofilewrite() will unuse the descriptor for us */
2903 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2904 &offset, 0, retval));
2905
2906 out:
2907 fd_putfile(fd);
2908 return (error);
2909}
2910
2911/*
2912 * Positional gather write system call.
2913 */
2914int
2915sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
2916{
2917 /* {
2918 syscallarg(int) fd;
2919 syscallarg(const struct iovec *) iovp;
2920 syscallarg(int) iovcnt;
2921 syscallarg(off_t) offset;
2922 } */
2923 off_t offset = SCARG(uap, offset);
2924
2925 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2926 SCARG(uap, iovcnt), &offset, 0, retval);
2927}
2928
2929/*
2930 * Check access permissions.
2931 */
2932int
2933sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
2934{
2935 /* {
2936 syscallarg(const char *) path;
2937 syscallarg(int) flags;
2938 } */
2939
2940 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path),
2941 SCARG(uap, flags), 0);
2942}
2943
2944int
2945do_sys_accessat(struct lwp *l, int fdat, const char *path,
2946 int mode, int flags)
2947{
2948 kauth_cred_t cred;
2949 struct vnode *vp;
2950 int error, nd_flag, vmode;
2951 struct pathbuf *pb;
2952 struct nameidata nd;
2953
2954 CTASSERT(F_OK == 0);
2955 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) {
2956 /* nonsense mode */
2957 return EINVAL;
2958 }
2959
2960 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT;
2961 if (flags & AT_SYMLINK_NOFOLLOW)
2962 nd_flag &= ~FOLLOW;
2963
2964 error = pathbuf_copyin(path, &pb);
2965 if (error)
2966 return error;
2967
2968 NDINIT(&nd, LOOKUP, nd_flag, pb);
2969
2970 /* Override default credentials */
2971 cred = kauth_cred_dup(l->l_cred);
2972 if (!(flags & AT_EACCESS)) {
2973 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2974 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2975 }
2976 nd.ni_cnd.cn_cred = cred;
2977
2978 if ((error = fd_nameiat(l, fdat, &nd)) != 0) {
2979 pathbuf_destroy(pb);
2980 goto out;
2981 }
2982 vp = nd.ni_vp;
2983 pathbuf_destroy(pb);
2984
2985 /* Flags == 0 means only check for existence. */
2986 if (mode) {
2987 vmode = 0;
2988 if (mode & R_OK)
2989 vmode |= VREAD;
2990 if (mode & W_OK)
2991 vmode |= VWRITE;
2992 if (mode & X_OK)
2993 vmode |= VEXEC;
2994
2995 error = VOP_ACCESS(vp, vmode, cred);
2996 if (!error && (vmode & VWRITE))
2997 error = vn_writechk(vp);
2998 }
2999 vput(vp);
3000out:
3001 kauth_cred_free(cred);
3002 return (error);
3003}
3004
3005int
3006sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap,
3007 register_t *retval)
3008{
3009 /* {
3010 syscallarg(int) fd;
3011 syscallarg(const char *) path;
3012 syscallarg(int) amode;
3013 syscallarg(int) flag;
3014 } */
3015
3016 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path),
3017 SCARG(uap, amode), SCARG(uap, flag));
3018}
3019
3020/*
3021 * Common code for all sys_stat functions, including compat versions.
3022 */
3023int
3024do_sys_stat(const char *userpath, unsigned int nd_flag,
3025 struct stat *sb)
3026{
3027 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb);
3028}
3029
3030int
3031do_sys_statat(struct lwp *l, int fdat, const char *userpath,
3032 unsigned int nd_flag, struct stat *sb)
3033{
3034 int error;
3035 struct pathbuf *pb;
3036 struct nameidata nd;
3037
3038 KASSERT(l != NULL || fdat == AT_FDCWD);
3039
3040 error = pathbuf_copyin(userpath, &pb);
3041 if (error) {
3042 return error;
3043 }
3044
3045 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb);
3046
3047 error = fd_nameiat(l, fdat, &nd);
3048 if (error != 0) {
3049 pathbuf_destroy(pb);
3050 return error;
3051 }
3052 error = vn_stat(nd.ni_vp, sb);
3053 vput(nd.ni_vp);
3054 pathbuf_destroy(pb);
3055 return error;
3056}
3057
3058/*
3059 * Get file status; this version follows links.
3060 */
3061/* ARGSUSED */
3062int
3063sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval)
3064{
3065 /* {
3066 syscallarg(const char *) path;
3067 syscallarg(struct stat *) ub;
3068 } */
3069 struct stat sb;
3070 int error;
3071
3072 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb);
3073 if (error)
3074 return error;
3075 return copyout(&sb, SCARG(uap, ub), sizeof(sb));
3076}
3077
3078/*
3079 * Get file status; this version does not follow links.
3080 */
3081/* ARGSUSED */
3082int
3083sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval)
3084{
3085 /* {
3086 syscallarg(const char *) path;
3087 syscallarg(struct stat *) ub;
3088 } */
3089 struct stat sb;
3090 int error;
3091
3092 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb);
3093 if (error)
3094 return error;
3095 return copyout(&sb, SCARG(uap, ub), sizeof(sb));
3096}
3097
3098int
3099sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap,
3100 register_t *retval)
3101{
3102 /* {
3103 syscallarg(int) fd;
3104 syscallarg(const char *) path;
3105 syscallarg(struct stat *) buf;
3106 syscallarg(int) flag;
3107 } */
3108 unsigned int nd_flag;
3109 struct stat sb;
3110 int error;
3111
3112 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW)
3113 nd_flag = NOFOLLOW;
3114 else
3115 nd_flag = FOLLOW;
3116
3117 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag,
3118 &sb);
3119 if (error)
3120 return error;
3121 return copyout(&sb, SCARG(uap, buf), sizeof(sb));
3122}
3123
3124/*
3125 * Get configurable pathname variables.
3126 */
3127/* ARGSUSED */
3128int
3129sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
3130{
3131 /* {
3132 syscallarg(const char *) path;
3133 syscallarg(int) name;
3134 } */
3135 int error;
3136 struct pathbuf *pb;
3137 struct nameidata nd;
3138
3139 error = pathbuf_copyin(SCARG(uap, path), &pb);
3140 if (error) {
3141 return error;
3142 }
3143 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb);
3144 if ((error = namei(&nd)) != 0) {
3145 pathbuf_destroy(pb);
3146 return (error);
3147 }
3148 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
3149 vput(nd.ni_vp);
3150 pathbuf_destroy(pb);
3151 return (error);
3152}
3153
3154/*
3155 * Return target name of a symbolic link.
3156 */
3157/* ARGSUSED */
3158int
3159sys_readlink(struct lwp *l, const struct sys_readlink_args *uap,
3160 register_t *retval)
3161{
3162 /* {
3163 syscallarg(const char *) path;
3164 syscallarg(char *) buf;
3165 syscallarg(size_t) count;
3166 } */
3167 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path),
3168 SCARG(uap, buf), SCARG(uap, count), retval);
3169}
3170
3171static int
3172do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf,
3173 size_t count, register_t *retval)
3174{
3175 struct vnode *vp;
3176 struct iovec aiov;
3177 struct uio auio;
3178 int error;
3179 struct pathbuf *pb;
3180 struct nameidata nd;
3181
3182 error = pathbuf_copyin(path, &pb);
3183 if (error) {
3184 return error;
3185 }
3186 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb);
3187 if ((error = fd_nameiat(l, fdat, &nd)) != 0) {
3188 pathbuf_destroy(pb);
3189 return error;
3190 }
3191 vp = nd.ni_vp;
3192 pathbuf_destroy(pb);
3193 if (vp->v_type != VLNK)
3194 error = EINVAL;
3195 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
3196 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
3197 aiov.iov_base = buf;
3198 aiov.iov_len = count;
3199 auio.uio_iov = &aiov;
3200 auio.uio_iovcnt = 1;
3201 auio.uio_offset = 0;
3202 auio.uio_rw = UIO_READ;
3203 KASSERT(l == curlwp);
3204 auio.uio_vmspace = l->l_proc->p_vmspace;
3205 auio.uio_resid = count;
3206 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0)
3207 *retval = count - auio.uio_resid;
3208 }
3209 vput(vp);
3210 return (error);
3211}
3212
3213int
3214sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap,
3215 register_t *retval)
3216{
3217 /* {
3218 syscallarg(int) fd;
3219 syscallarg(const char *) path;
3220 syscallarg(char *) buf;
3221 syscallarg(size_t) bufsize;
3222 } */
3223
3224 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path),
3225 SCARG(uap, buf), SCARG(uap, bufsize), retval);
3226}
3227
3228/*
3229 * Change flags of a file given a path name.
3230 */
3231/* ARGSUSED */
3232int
3233sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
3234{
3235 /* {
3236 syscallarg(const char *) path;
3237 syscallarg(u_long) flags;
3238 } */
3239 struct vnode *vp;
3240 int error;
3241
3242 error = namei_simple_user(SCARG(uap, path),
3243 NSM_FOLLOW_TRYEMULROOT, &vp);
3244 if (error != 0)
3245 return (error);
3246 error = change_flags(vp, SCARG(uap, flags), l);
3247 vput(vp);
3248 return (error);
3249}
3250
3251/*
3252 * Change flags of a file given a file descriptor.
3253 */
3254/* ARGSUSED */
3255int
3256sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
3257{
3258 /* {
3259 syscallarg(int) fd;
3260 syscallarg(u_long) flags;
3261 } */
3262 struct vnode *vp;
3263 file_t *fp;
3264 int error;
3265
3266 /* fd_getvnode() will use the descriptor for us */
3267 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3268 return (error);
3269 vp = fp->f_vnode;
3270 error = change_flags(vp, SCARG(uap, flags), l);
3271 VOP_UNLOCK(vp);
3272 fd_putfile(SCARG(uap, fd));
3273 return (error);
3274}
3275
3276/*
3277 * Change flags of a file given a path name; this version does
3278 * not follow links.
3279 */
3280int
3281sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
3282{
3283 /* {
3284 syscallarg(const char *) path;
3285 syscallarg(u_long) flags;
3286 } */
3287 struct vnode *vp;
3288 int error;
3289
3290 error = namei_simple_user(SCARG(uap, path),
3291 NSM_NOFOLLOW_TRYEMULROOT, &vp);
3292 if (error != 0)
3293 return (error);
3294 error = change_flags(vp, SCARG(uap, flags), l);
3295 vput(vp);
3296 return (error);
3297}
3298
3299/*
3300 * Common routine to change flags of a file.
3301 */
3302int
3303change_flags(struct vnode *vp, u_long flags, struct lwp *l)
3304{
3305 struct vattr vattr;
3306 int error;
3307
3308 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3309
3310 vattr_null(&vattr);
3311 vattr.va_flags = flags;
3312 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3313
3314 return (error);
3315}
3316
3317/*
3318 * Change mode of a file given path name; this version follows links.
3319 */
3320/* ARGSUSED */
3321int
3322sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
3323{
3324 /* {
3325 syscallarg(const char *) path;
3326 syscallarg(int) mode;
3327 } */
3328 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path),
3329 SCARG(uap, mode), 0);
3330}
3331
3332int
3333do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags)
3334{
3335 int error;
3336 struct vnode *vp;
3337 namei_simple_flags_t ns_flag;
3338
3339 if (flags & AT_SYMLINK_NOFOLLOW)
3340 ns_flag = NSM_NOFOLLOW_TRYEMULROOT;
3341 else
3342 ns_flag = NSM_FOLLOW_TRYEMULROOT;
3343
3344 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp);
3345 if (error != 0)
3346 return error;
3347
3348 error = change_mode(vp, mode, l);
3349
3350 vrele(vp);
3351
3352 return (error);
3353}
3354
3355/*
3356 * Change mode of a file given a file descriptor.
3357 */
3358/* ARGSUSED */
3359int
3360sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
3361{
3362 /* {
3363 syscallarg(int) fd;
3364 syscallarg(int) mode;
3365 } */
3366 file_t *fp;
3367 int error;
3368
3369 /* fd_getvnode() will use the descriptor for us */
3370 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3371 return (error);
3372 error = change_mode(fp->f_vnode, SCARG(uap, mode), l);
3373 fd_putfile(SCARG(uap, fd));
3374 return (error);
3375}
3376
3377int
3378sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap,
3379 register_t *retval)
3380{
3381 /* {
3382 syscallarg(int) fd;
3383 syscallarg(const char *) path;
3384 syscallarg(int) mode;
3385 syscallarg(int) flag;
3386 } */
3387
3388 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path),
3389 SCARG(uap, mode), SCARG(uap, flag));
3390}
3391
3392/*
3393 * Change mode of a file given path name; this version does not follow links.
3394 */
3395/* ARGSUSED */
3396int
3397sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
3398{
3399 /* {
3400 syscallarg(const char *) path;
3401 syscallarg(int) mode;
3402 } */
3403 int error;
3404 struct vnode *vp;
3405
3406 error = namei_simple_user(SCARG(uap, path),
3407 NSM_NOFOLLOW_TRYEMULROOT, &vp);
3408 if (error != 0)
3409 return (error);
3410
3411 error = change_mode(vp, SCARG(uap, mode), l);
3412
3413 vrele(vp);
3414 return (error);
3415}
3416
3417/*
3418 * Common routine to set mode given a vnode.
3419 */
3420static int
3421change_mode(struct vnode *vp, int mode, struct lwp *l)
3422{
3423 struct vattr vattr;
3424 int error;
3425
3426 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3427 vattr_null(&vattr);
3428 vattr.va_mode = mode & ALLPERMS;
3429 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3430 VOP_UNLOCK(vp);
3431 return (error);
3432}
3433
3434/*
3435 * Set ownership given a path name; this version follows links.
3436 */
3437/* ARGSUSED */
3438int
3439sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
3440{
3441 /* {
3442 syscallarg(const char *) path;
3443 syscallarg(uid_t) uid;
3444 syscallarg(gid_t) gid;
3445 } */
3446 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid),
3447 SCARG(uap, gid), 0);
3448}
3449
3450int
3451do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid,
3452 gid_t gid, int flags)
3453{
3454 int error;
3455 struct vnode *vp;
3456 namei_simple_flags_t ns_flag;
3457
3458 if (flags & AT_SYMLINK_NOFOLLOW)
3459 ns_flag = NSM_NOFOLLOW_TRYEMULROOT;
3460 else
3461 ns_flag = NSM_FOLLOW_TRYEMULROOT;
3462
3463 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp);
3464 if (error != 0)
3465 return error;
3466
3467 error = change_owner(vp, uid, gid, l, 0);
3468
3469 vrele(vp);
3470
3471 return (error);
3472}
3473
3474/*
3475 * Set ownership given a path name; this version follows links.
3476 * Provides POSIX semantics.
3477 */
3478/* ARGSUSED */
3479int
3480sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
3481{
3482 /* {
3483 syscallarg(const char *) path;
3484 syscallarg(uid_t) uid;
3485 syscallarg(gid_t) gid;
3486 } */
3487 int error;
3488 struct vnode *vp;
3489
3490 error = namei_simple_user(SCARG(uap, path),
3491 NSM_FOLLOW_TRYEMULROOT, &vp);
3492 if (error != 0)
3493 return (error);
3494
3495 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
3496
3497 vrele(vp);
3498 return (error);
3499}
3500
3501/*
3502 * Set ownership given a file descriptor.
3503 */
3504/* ARGSUSED */
3505int
3506sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
3507{
3508 /* {
3509 syscallarg(int) fd;
3510 syscallarg(uid_t) uid;
3511 syscallarg(gid_t) gid;
3512 } */
3513 int error;
3514 file_t *fp;
3515
3516 /* fd_getvnode() will use the descriptor for us */
3517 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3518 return (error);
3519 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid),
3520 l, 0);
3521 fd_putfile(SCARG(uap, fd));
3522 return (error);
3523}
3524
3525int
3526sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap,
3527 register_t *retval)
3528{
3529 /* {
3530 syscallarg(int) fd;
3531 syscallarg(const char *) path;
3532 syscallarg(uid_t) owner;
3533 syscallarg(gid_t) group;
3534 syscallarg(int) flag;
3535 } */
3536
3537 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path),
3538 SCARG(uap, owner), SCARG(uap, group),
3539 SCARG(uap, flag));
3540}
3541
3542/*
3543 * Set ownership given a file descriptor, providing POSIX/XPG semantics.
3544 */
3545/* ARGSUSED */
3546int
3547sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
3548{
3549 /* {
3550 syscallarg(int) fd;
3551 syscallarg(uid_t) uid;
3552 syscallarg(gid_t) gid;
3553 } */
3554 int error;
3555 file_t *fp;
3556
3557 /* fd_getvnode() will use the descriptor for us */
3558 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3559 return (error);
3560 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid),
3561 l, 1);
3562 fd_putfile(SCARG(uap, fd));
3563 return (error);
3564}
3565
3566/*
3567 * Set ownership given a path name; this version does not follow links.
3568 */
3569/* ARGSUSED */
3570int
3571sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
3572{
3573 /* {
3574 syscallarg(const char *) path;
3575 syscallarg(uid_t) uid;
3576 syscallarg(gid_t) gid;
3577 } */
3578 int error;
3579 struct vnode *vp;
3580
3581 error = namei_simple_user(SCARG(uap, path),
3582 NSM_NOFOLLOW_TRYEMULROOT, &vp);
3583 if (error != 0)
3584 return (error);
3585
3586 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
3587
3588 vrele(vp);
3589 return (error);
3590}
3591
3592/*
3593 * Set ownership given a path name; this version does not follow links.
3594 * Provides POSIX/XPG semantics.
3595 */
3596/* ARGSUSED */
3597int
3598sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
3599{
3600 /* {
3601 syscallarg(const char *) path;
3602 syscallarg(uid_t) uid;
3603 syscallarg(gid_t) gid;
3604 } */
3605 int error;
3606 struct vnode *vp;
3607
3608 error = namei_simple_user(SCARG(uap, path),
3609 NSM_NOFOLLOW_TRYEMULROOT, &vp);
3610 if (error != 0)
3611 return (error);
3612
3613 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
3614
3615 vrele(vp);
3616 return (error);
3617}
3618
3619/*
3620 * Common routine to set ownership given a vnode.
3621 */
3622static int
3623change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
3624 int posix_semantics)
3625{
3626 struct vattr vattr;
3627 mode_t newmode;
3628 int error;
3629
3630 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3631 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
3632 goto out;
3633
3634#define CHANGED(x) ((int)(x) != -1)
3635 newmode = vattr.va_mode;
3636 if (posix_semantics) {
3637 /*
3638 * POSIX/XPG semantics: if the caller is not the super-user,
3639 * clear set-user-id and set-group-id bits. Both POSIX and
3640 * the XPG consider the behaviour for calls by the super-user
3641 * implementation-defined; we leave the set-user-id and set-
3642 * group-id settings intact in that case.
3643 */
3644 if (vattr.va_mode & S_ISUID) {
3645 if (kauth_authorize_vnode(l->l_cred,
3646 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0)
3647 newmode &= ~S_ISUID;
3648 }
3649 if (vattr.va_mode & S_ISGID) {
3650 if (kauth_authorize_vnode(l->l_cred,
3651 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0)
3652 newmode &= ~S_ISGID;
3653 }
3654 } else {
3655 /*
3656 * NetBSD semantics: when changing owner and/or group,
3657 * clear the respective bit(s).
3658 */
3659 if (CHANGED(uid))
3660 newmode &= ~S_ISUID;
3661 if (CHANGED(gid))
3662 newmode &= ~S_ISGID;
3663 }
3664 /* Update va_mode iff altered. */
3665 if (vattr.va_mode == newmode)
3666 newmode = VNOVAL;
3667
3668 vattr_null(&vattr);
3669 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
3670 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
3671 vattr.va_mode = newmode;
3672 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3673#undef CHANGED
3674
3675out:
3676 VOP_UNLOCK(vp);
3677 return (error);
3678}
3679
3680/*
3681 * Set the access and modification times given a path name; this
3682 * version follows links.
3683 */
3684/* ARGSUSED */
3685int
3686sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap,
3687 register_t *retval)
3688{
3689 /* {
3690 syscallarg(const char *) path;
3691 syscallarg(const struct timeval *) tptr;
3692 } */
3693
3694 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
3695 SCARG(uap, tptr), UIO_USERSPACE);
3696}
3697
3698/*
3699 * Set the access and modification times given a file descriptor.
3700 */
3701/* ARGSUSED */
3702int
3703sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap,
3704 register_t *retval)
3705{
3706 /* {
3707 syscallarg(int) fd;
3708 syscallarg(const struct timeval *) tptr;
3709 } */
3710 int error;
3711 file_t *fp;
3712
3713 /* fd_getvnode() will use the descriptor for us */
3714 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3715 return (error);
3716 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr),
3717 UIO_USERSPACE);
3718 fd_putfile(SCARG(uap, fd));
3719 return (error);
3720}
3721
3722int
3723sys_futimens(struct lwp *l, const struct sys_futimens_args *uap,
3724 register_t *retval)
3725{
3726 /* {
3727 syscallarg(int) fd;
3728 syscallarg(const struct timespec *) tptr;
3729 } */
3730 int error;
3731 file_t *fp;
3732
3733 /* fd_getvnode() will use the descriptor for us */
3734 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3735 return (error);
3736 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0,
3737 SCARG(uap, tptr), UIO_USERSPACE);
3738 fd_putfile(SCARG(uap, fd));
3739 return (error);
3740}
3741
3742/*
3743 * Set the access and modification times given a path name; this
3744 * version does not follow links.
3745 */
3746int
3747sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap,
3748 register_t *retval)
3749{
3750 /* {
3751 syscallarg(const char *) path;
3752 syscallarg(const struct timeval *) tptr;
3753 } */
3754
3755 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
3756 SCARG(uap, tptr), UIO_USERSPACE);
3757}
3758
3759int
3760sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap,
3761 register_t *retval)
3762{
3763 /* {
3764 syscallarg(int) fd;
3765 syscallarg(const char *) path;
3766 syscallarg(const struct timespec *) tptr;
3767 syscallarg(int) flag;
3768 } */
3769 int follow;
3770 const struct timespec *tptr;
3771 int error;
3772
3773 tptr = SCARG(uap, tptr);
3774 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3775
3776 error = do_sys_utimensat(l, SCARG(uap, fd), NULL,
3777 SCARG(uap, path), follow, tptr, UIO_USERSPACE);
3778
3779 return error;
3780}
3781
3782/*
3783 * Common routine to set access and modification times given a vnode.
3784 */
3785int
3786do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag,
3787 const struct timespec *tptr, enum uio_seg seg)
3788{
3789 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg);
3790}
3791
3792int
3793do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp,
3794 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg)
3795{
3796 struct vattr vattr;
3797 int error, dorele = 0;
3798 namei_simple_flags_t sflags;
3799 bool vanull, setbirthtime;
3800 struct timespec ts[2];
3801
3802 KASSERT(l != NULL || fdat == AT_FDCWD);
3803
3804 /*
3805 * I have checked all callers and they pass either FOLLOW,
3806 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW
3807 * is 0. More to the point, they don't pass anything else.
3808 * Let's keep it that way at least until the namei interfaces
3809 * are fully sanitized.
3810 */
3811 KASSERT(flag == NOFOLLOW || flag == FOLLOW);
3812 sflags = (flag == FOLLOW) ?
3813 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT;
3814
3815 if (tptr == NULL) {
3816 vanull = true;
3817 nanotime(&ts[0]);
3818 ts[1] = ts[0];
3819 } else {
3820 vanull = false;
3821 if (seg != UIO_SYSSPACE) {
3822 error = copyin(tptr, ts, sizeof (ts));
3823 if (error != 0)
3824 return error;
3825 } else {
3826 ts[0] = tptr[0];
3827 ts[1] = tptr[1];
3828 }
3829 }
3830
3831 if (ts[0].tv_nsec == UTIME_NOW) {
3832 nanotime(&ts[0]);
3833 if (ts[1].tv_nsec == UTIME_NOW) {
3834 vanull = true;
3835 ts[1] = ts[0];
3836 }
3837 } else if (ts[1].tv_nsec == UTIME_NOW)
3838 nanotime(&ts[1]);
3839
3840 if (vp == NULL) {
3841 /* note: SEG describes TPTR, not PATH; PATH is always user */
3842 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp);
3843 if (error != 0)
3844 return error;
3845 dorele = 1;
3846 }
3847
3848 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3849 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
3850 timespeccmp(&ts[1], &vattr.va_birthtime, <));
3851 vattr_null(&vattr);
3852
3853 if (ts[0].tv_nsec != UTIME_OMIT)
3854 vattr.va_atime = ts[0];
3855
3856 if (ts[1].tv_nsec != UTIME_OMIT) {
3857 vattr.va_mtime = ts[1];
3858 if (setbirthtime)
3859 vattr.va_birthtime = ts[1];
3860 }
3861
3862 if (vanull)
3863 vattr.va_vaflags |= VA_UTIMES_NULL;
3864 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3865 VOP_UNLOCK(vp);
3866
3867 if (dorele != 0)
3868 vrele(vp);
3869
3870 return error;
3871}
3872
3873int
3874do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
3875 const struct timeval *tptr, enum uio_seg seg)
3876{
3877 struct timespec ts[2];
3878 struct timespec *tsptr = NULL;
3879 int error;
3880
3881 if (tptr != NULL) {
3882 struct timeval tv[2];
3883
3884 if (seg != UIO_SYSSPACE) {
3885 error = copyin(tptr, tv, sizeof (tv));
3886 if (error != 0)
3887 return error;
3888 tptr = tv;
3889 }
3890
3891 if ((tv[0].tv_usec == UTIME_NOW) ||
3892 (tv[0].tv_usec == UTIME_OMIT))
3893 ts[0].tv_nsec = tv[0].tv_usec;
3894 else
3895 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
3896
3897 if ((tv[1].tv_usec == UTIME_NOW) ||
3898 (tv[1].tv_usec == UTIME_OMIT))
3899 ts[1].tv_nsec = tv[1].tv_usec;
3900 else
3901 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
3902
3903 tsptr = &ts[0];
3904 }
3905
3906 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE);
3907}
3908
3909/*
3910 * Truncate a file given its path name.
3911 */
3912/* ARGSUSED */
3913int
3914sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
3915{
3916 /* {
3917 syscallarg(const char *) path;
3918 syscallarg(int) pad;
3919 syscallarg(off_t) length;
3920 } */
3921 struct vnode *vp;
3922 struct vattr vattr;
3923 int error;
3924
3925 if (SCARG(uap, length) < 0)
3926 return EINVAL;
3927
3928 error = namei_simple_user(SCARG(uap, path),
3929 NSM_FOLLOW_TRYEMULROOT, &vp);
3930 if (error != 0)
3931 return (error);
3932 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3933 if (vp->v_type == VDIR)
3934 error = EISDIR;
3935 else if ((error = vn_writechk(vp)) == 0 &&
3936 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
3937 vattr_null(&vattr);
3938 vattr.va_size = SCARG(uap, length);
3939 error = VOP_SETATTR(vp, &vattr, l->l_cred);
3940 }
3941 vput(vp);
3942 return (error);
3943}
3944
3945/*
3946 * Truncate a file given a file descriptor.
3947 */
3948/* ARGSUSED */
3949int
3950sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
3951{
3952 /* {
3953 syscallarg(int) fd;
3954 syscallarg(int) pad;
3955 syscallarg(off_t) length;
3956 } */
3957 struct vattr vattr;
3958 struct vnode *vp;
3959 file_t *fp;
3960 int error;
3961
3962 if (SCARG(uap, length) < 0)
3963 return EINVAL;
3964
3965 /* fd_getvnode() will use the descriptor for us */
3966 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
3967 return (error);
3968 if ((fp->f_flag & FWRITE) == 0) {
3969 error = EINVAL;
3970 goto out;
3971 }
3972 vp = fp->f_vnode;
3973 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3974 if (vp->v_type == VDIR)
3975 error = EISDIR;
3976 else if ((error = vn_writechk(vp)) == 0) {
3977 vattr_null(&vattr);
3978 vattr.va_size = SCARG(uap, length);
3979 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
3980 }
3981 VOP_UNLOCK(vp);
3982 out:
3983 fd_putfile(SCARG(uap, fd));
3984 return (error);
3985}
3986
3987/*
3988 * Sync an open file.
3989 */
3990/* ARGSUSED */
3991int
3992sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
3993{
3994 /* {
3995 syscallarg(int) fd;
3996 } */
3997 struct vnode *vp;
3998 file_t *fp;
3999 int error;
4000
4001 /* fd_getvnode() will use the descriptor for us */
4002 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
4003 return (error);
4004 vp = fp->f_vnode;
4005 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4006 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
4007 VOP_UNLOCK(vp);
4008 fd_putfile(SCARG(uap, fd));
4009 return (error);
4010}
4011
4012/*
4013 * Sync a range of file data. API modeled after that found in AIX.
4014 *
4015 * FDATASYNC indicates that we need only save enough metadata to be able
4016 * to re-read the written data. Note we duplicate AIX's requirement that
4017 * the file be open for writing.
4018 */
4019/* ARGSUSED */
4020int
4021sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
4022{
4023 /* {
4024 syscallarg(int) fd;
4025 syscallarg(int) flags;
4026 syscallarg(off_t) start;
4027 syscallarg(off_t) length;
4028 } */
4029 struct vnode *vp;
4030 file_t *fp;
4031 int flags, nflags;
4032 off_t s, e, len;
4033 int error;
4034
4035 /* fd_getvnode() will use the descriptor for us */
4036 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
4037 return (error);
4038
4039 if ((fp->f_flag & FWRITE) == 0) {
4040 error = EBADF;
4041 goto out;
4042 }
4043
4044 flags = SCARG(uap, flags);
4045 if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
4046 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
4047 error = EINVAL;
4048 goto out;
4049 }
4050 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */
4051 if (flags & FDATASYNC)
4052 nflags = FSYNC_DATAONLY | FSYNC_WAIT;
4053 else
4054 nflags = FSYNC_WAIT;
4055 if (flags & FDISKSYNC)
4056 nflags |= FSYNC_CACHE;
4057
4058 len = SCARG(uap, length);
4059 /* If length == 0, we do the whole file, and s = e = 0 will do that */
4060 if (len) {
4061 s = SCARG(uap, start);
4062 e = s + len;
4063 if (e < s) {
4064 error = EINVAL;
4065 goto out;
4066 }
4067 } else {
4068 e = 0;
4069 s = 0;
4070 }
4071
4072 vp = fp->f_vnode;
4073 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4074 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
4075 VOP_UNLOCK(vp);
4076out:
4077 fd_putfile(SCARG(uap, fd));
4078 return (error);
4079}
4080
4081/*
4082 * Sync the data of an open file.
4083 */
4084/* ARGSUSED */
4085int
4086sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
4087{
4088 /* {
4089 syscallarg(int) fd;
4090 } */
4091 struct vnode *vp;
4092 file_t *fp;
4093 int error;
4094
4095 /* fd_getvnode() will use the descriptor for us */
4096 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
4097 return (error);
4098 if ((fp->f_flag & FWRITE) == 0) {
4099 fd_putfile(SCARG(uap, fd));
4100 return (EBADF);
4101 }
4102 vp = fp->f_vnode;
4103 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4104 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
4105 VOP_UNLOCK(vp);
4106 fd_putfile(SCARG(uap, fd));
4107 return (error);
4108}
4109
4110/*
4111 * Rename files, (standard) BSD semantics frontend.
4112 */
4113/* ARGSUSED */
4114int
4115sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
4116{
4117 /* {
4118 syscallarg(const char *) from;
4119 syscallarg(const char *) to;
4120 } */
4121
4122 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD,
4123 SCARG(uap, to), UIO_USERSPACE, 0));
4124}
4125
4126int
4127sys_renameat(struct lwp *l, const struct sys_renameat_args *uap,
4128 register_t *retval)
4129{
4130 /* {
4131 syscallarg(int) fromfd;
4132 syscallarg(const char *) from;
4133 syscallarg(int) tofd;
4134 syscallarg(const char *) to;
4135 } */
4136
4137 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from),
4138 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0));
4139}
4140
4141/*
4142 * Rename files, POSIX semantics frontend.
4143 */
4144/* ARGSUSED */
4145int
4146sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
4147{
4148 /* {
4149 syscallarg(const char *) from;
4150 syscallarg(const char *) to;
4151 } */
4152
4153 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD,
4154 SCARG(uap, to), UIO_USERSPACE, 1));
4155}
4156
4157/*
4158 * Rename files. Source and destination must either both be directories,
4159 * or both not be directories. If target is a directory, it must be empty.
4160 * If `from' and `to' refer to the same object, the value of the `retain'
4161 * argument is used to determine whether `from' will be
4162 *
4163 * (retain == 0) deleted unless `from' and `to' refer to the same
4164 * object in the file system's name space (BSD).
4165 * (retain == 1) always retained (POSIX).
4166 *
4167 * XXX Synchronize with nfsrv_rename in nfs_serv.c.
4168 */
4169int
4170do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
4171{
4172 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain);
4173}
4174
4175static int
4176do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd,
4177 const char *to, enum uio_seg seg, int retain)
4178{
4179 struct pathbuf *fpb, *tpb;
4180 struct nameidata fnd, tnd;
4181 struct vnode *fdvp, *fvp;
4182 struct vnode *tdvp, *tvp;
4183 struct mount *mp, *tmp;
4184 int error;
4185
4186 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD));
4187
4188 error = pathbuf_maybe_copyin(from, seg, &fpb);
4189 if (error)
4190 goto out0;
4191 KASSERT(fpb != NULL);
4192
4193 error = pathbuf_maybe_copyin(to, seg, &tpb);
4194 if (error)
4195 goto out1;
4196 KASSERT(tpb != NULL);
4197
4198 /*
4199 * Lookup from.
4200 *
4201 * XXX LOCKPARENT is wrong because we don't actually want it
4202 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is
4203 * insane, so for the time being we need to leave it like this.
4204 */
4205 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb);
4206 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0)
4207 goto out2;
4208
4209 /*
4210 * Pull out the important results of the lookup, fdvp and fvp.
4211 * Of course, fvp is bogus because we're about to unlock fdvp.
4212 */
4213 fdvp = fnd.ni_dvp;
4214 fvp = fnd.ni_vp;
4215 KASSERT(fdvp != NULL);
4216 KASSERT(fvp != NULL);
4217 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE));
4218
4219 /*
4220 * Make sure neither fdvp nor fvp is locked.
4221 */
4222 if (fdvp != fvp)
4223 VOP_UNLOCK(fdvp);
4224 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */
4225 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */
4226
4227 /*
4228 * Reject renaming `.' and `..'. Can't do this until after
4229 * namei because we need namei's parsing to find the final
4230 * component name. (namei should just leave us with the final
4231 * component name and not look it up itself, but anyway...)
4232 *
4233 * This was here before because we used to relookup from
4234 * instead of to and relookup requires the caller to check
4235 * this, but now file systems may depend on this check, so we
4236 * must retain it until the file systems are all rototilled.
4237 */
4238 if (((fnd.ni_cnd.cn_namelen == 1) &&
4239 (fnd.ni_cnd.cn_nameptr[0] == '.')) ||
4240 ((fnd.ni_cnd.cn_namelen == 2) &&
4241 (fnd.ni_cnd.cn_nameptr[0] == '.') &&
4242 (fnd.ni_cnd.cn_nameptr[1] == '.'))) {
4243 error = EINVAL; /* XXX EISDIR? */
4244 goto abort0;
4245 }
4246
4247 /*
4248 * Lookup to.
4249 *
4250 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using
4251 * fvp here to decide whether to add CREATEDIR is a load of
4252 * bollocks because fvp might be the wrong node by now, since
4253 * fdvp is unlocked.
4254 *
4255 * XXX Why not pass CREATEDIR always?
4256 */
4257 NDINIT(&tnd, RENAME,
4258 (LOCKPARENT | NOCACHE | TRYEMULROOT |
4259 ((fvp->v_type == VDIR)? CREATEDIR : 0)),
4260 tpb);
4261 if ((error = fd_nameiat(l, tofd, &tnd)) != 0)
4262 goto abort0;
4263
4264 /*
4265 * Pull out the important results of the lookup, tdvp and tvp.
4266 * Of course, tvp is bogus because we're about to unlock tdvp.
4267 */
4268 tdvp = tnd.ni_dvp;
4269 tvp = tnd.ni_vp;
4270 KASSERT(tdvp != NULL);
4271 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE));
4272
4273 /*
4274 * Make sure neither tdvp nor tvp is locked.
4275 */
4276 if (tdvp != tvp)
4277 VOP_UNLOCK(tdvp);
4278 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */
4279 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */
4280
4281 /*
4282 * Reject renaming onto `.' or `..'. relookup is unhappy with
4283 * these, which is why we must do this here. Once upon a time
4284 * we relooked up from instead of to, and consequently didn't
4285 * need this check, but now that we relookup to instead of
4286 * from, we need this; and we shall need it forever forward
4287 * until the VOP_RENAME protocol changes, because file systems
4288 * will no doubt begin to depend on this check.
4289 */
4290 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) {
4291 error = EISDIR;
4292 goto abort1;
4293 }
4294 if ((tnd.ni_cnd.cn_namelen == 2) &&
4295 (tnd.ni_cnd.cn_nameptr[0] == '.') &&
4296 (tnd.ni_cnd.cn_nameptr[1] == '.')) {
4297 error = EINVAL;
4298 goto abort1;
4299 }
4300
4301 /*
4302 * Get the mount point. If the file system has been unmounted,
4303 * which it may be because we're not holding any vnode locks,
4304 * then v_mount will be NULL. We're not really supposed to
4305 * read v_mount without holding the vnode lock, but since we
4306 * have fdvp referenced, if fdvp->v_mount changes then at worst
4307 * it will be set to NULL, not changed to another mount point.
4308 * And, of course, since it is up to the file system to
4309 * determine the real lock order, we can't lock both fdvp and
4310 * tdvp at the same time.
4311 */
4312 mp = fdvp->v_mount;
4313 if (mp == NULL) {
4314 error = ENOENT;
4315 goto abort1;
4316 }
4317
4318 /*
4319 * Make sure the mount points match. Again, although we don't
4320 * hold any vnode locks, the v_mount fields may change -- but
4321 * at worst they will change to NULL, so this will never become
4322 * a cross-device rename, because we hold vnode references.
4323 *
4324 * XXX Because nothing is locked and the compiler may reorder
4325 * things here, unmounting the file system at an inopportune
4326 * moment may cause rename to fail with EXDEV when it really
4327 * should fail with ENOENT.
4328 */
4329 tmp = tdvp->v_mount;
4330 if (tmp == NULL) {
4331 error = ENOENT;
4332 goto abort1;
4333 }
4334
4335 if (mp != tmp) {
4336 error = EXDEV;
4337 goto abort1;
4338 }
4339
4340 /*
4341 * Take the vfs rename lock to avoid cross-directory screw cases.
4342 * Nothing is locked currently, so taking this lock is safe.
4343 */
4344 error = VFS_RENAMELOCK_ENTER(mp);
4345 if (error)
4346 goto abort1;
4347
4348 /*
4349 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced,
4350 * and nothing is locked except for the vfs rename lock.
4351 *
4352 * The next step is a little rain dance to conform to the
4353 * insane lock protocol, even though it does nothing to ward
4354 * off race conditions.
4355 *
4356 * We need tdvp and tvp to be locked. However, because we have
4357 * unlocked tdvp in order to hold no locks while we take the
4358 * vfs rename lock, tvp may be wrong here, and we can't safely
4359 * lock it even if the sensible file systems will just unlock
4360 * it straight away. Consequently, we must lock tdvp and then
4361 * relookup tvp to get it locked.
4362 *
4363 * Finally, because the VOP_RENAME protocol is brain-damaged
4364 * and various file systems insanely depend on the semantics of
4365 * this brain damage, the lookup of to must be the last lookup
4366 * before VOP_RENAME.
4367 */
4368 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY);
4369 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0);
4370 if (error)
4371 goto abort2;
4372
4373 /*
4374 * Drop the old tvp and pick up the new one -- which might be
4375 * the same, but that doesn't matter to us. After this, tdvp
4376 * and tvp should both be locked.
4377 */
4378 if (tvp != NULL)
4379 vrele(tvp);
4380 tvp = tnd.ni_vp;
4381 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE);
4382 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE));
4383
4384 /*
4385 * The old do_sys_rename had various consistency checks here
4386 * involving fvp and tvp. fvp is bogus already here, and tvp
4387 * will become bogus soon in any sensible file system, so the
4388 * only purpose in putting these checks here is to give lip
4389 * service to these screw cases and to acknowledge that they
4390 * exist, not actually to handle them, but here you go
4391 * anyway...
4392 */
4393
4394 /*
4395 * Acknowledge that directories and non-directories aren't
4396 * suposed to mix.
4397 */
4398 if (tvp != NULL) {
4399 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) {
4400 error = ENOTDIR;
4401 goto abort3;
4402 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) {
4403 error = EISDIR;
4404 goto abort3;
4405 }
4406 }
4407
4408 /*
4409 * Acknowledge some random screw case, among the dozens that
4410 * might arise.
4411 */
4412 if (fvp == tdvp) {
4413 error = EINVAL;
4414 goto abort3;
4415 }
4416
4417 /*
4418 * Acknowledge that POSIX has a wacky screw case.
4419 *
4420 * XXX Eventually the retain flag needs to be passed on to
4421 * VOP_RENAME.
4422 */
4423 if (fvp == tvp) {
4424 if (retain) {
4425 error = 0;
4426 goto abort3;
4427 } else if ((fdvp == tdvp) &&
4428 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) &&
4429 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr,
4430 fnd.ni_cnd.cn_namelen))) {
4431 error = 0;
4432 goto abort3;
4433 }
4434 }
4435
4436 /*
4437 * Make sure veriexec can screw us up. (But a race can screw
4438 * up veriexec, of course -- remember, fvp and (soon) tvp are
4439 * bogus.)
4440 */
4441#if NVERIEXEC > 0
4442 {
4443 char *f1, *f2;
4444 size_t f1_len;
4445 size_t f2_len;
4446
4447 f1_len = fnd.ni_cnd.cn_namelen + 1;
4448 f1 = kmem_alloc(f1_len, KM_SLEEP);
4449 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len);
4450
4451 f2_len = tnd.ni_cnd.cn_namelen + 1;
4452 f2 = kmem_alloc(f2_len, KM_SLEEP);
4453 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len);
4454
4455 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2);
4456
4457 kmem_free(f1, f1_len);
4458 kmem_free(f2, f2_len);
4459
4460 if (error)
4461 goto abort3;
4462 }
4463#endif /* NVERIEXEC > 0 */
4464
4465 /*
4466 * All ready. Incant the rename vop.
4467 */
4468 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */
4469 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */
4470 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE);
4471 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE));
4472 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd);
4473
4474 /*
4475 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks
4476 * tdvp and tvp. But we can't assert any of that.
4477 */
4478 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */
4479 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */
4480 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */
4481 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */
4482
4483 /*
4484 * So all we have left to do is to drop the rename lock and
4485 * destroy the pathbufs.
4486 */
4487 VFS_RENAMELOCK_EXIT(mp);
4488 goto out2;
4489
4490abort3: if ((tvp != NULL) && (tvp != tdvp))
4491 VOP_UNLOCK(tvp);
4492abort2: VOP_UNLOCK(tdvp);
4493 VFS_RENAMELOCK_EXIT(mp);
4494abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd);
4495 vrele(tdvp);
4496 if (tvp != NULL)
4497 vrele(tvp);
4498abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd);
4499 vrele(fdvp);
4500 vrele(fvp);
4501out2: pathbuf_destroy(tpb);
4502out1: pathbuf_destroy(fpb);
4503out0: return error;
4504}
4505
4506/*
4507 * Make a directory file.
4508 */
4509/* ARGSUSED */
4510int
4511sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
4512{
4513 /* {
4514 syscallarg(const char *) path;
4515 syscallarg(int) mode;
4516 } */
4517
4518 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path),
4519 SCARG(uap, mode), UIO_USERSPACE);
4520}
4521
4522int
4523sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap,
4524 register_t *retval)
4525{
4526 /* {
4527 syscallarg(int) fd;
4528 syscallarg(const char *) path;
4529 syscallarg(int) mode;
4530 } */
4531
4532 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path),
4533 SCARG(uap, mode), UIO_USERSPACE);
4534}
4535
4536
4537int
4538do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg)
4539{
4540 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE);
4541}
4542
4543static int
4544do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode,
4545 enum uio_seg seg)
4546{
4547 struct proc *p = curlwp->l_proc;
4548 struct vnode *vp;
4549 struct vattr vattr;
4550 int error;
4551 struct pathbuf *pb;
4552 struct nameidata nd;
4553
4554 KASSERT(l != NULL || fdat == AT_FDCWD);
4555
4556 /* XXX bollocks, should pass in a pathbuf */
4557 error = pathbuf_maybe_copyin(path, seg, &pb);
4558 if (error) {
4559 return error;
4560 }
4561
4562 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb);
4563
4564 if ((error = fd_nameiat(l, fdat, &nd)) != 0) {
4565 pathbuf_destroy(pb);
4566 return (error);
4567 }
4568 vp = nd.ni_vp;
4569 if (vp != NULL) {
4570 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
4571 if (nd.ni_dvp == vp)
4572 vrele(nd.ni_dvp);
4573 else
4574 vput(nd.ni_dvp);
4575 vrele(vp);
4576 pathbuf_destroy(pb);
4577 return (EEXIST);
4578 }
4579 vattr_null(&vattr);
4580 vattr.va_type = VDIR;
4581 /* We will read cwdi->cwdi_cmask unlocked. */
4582 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
4583 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
4584 if (!error)
4585 vrele(nd.ni_vp);
4586 vput(nd.ni_dvp);
4587 pathbuf_destroy(pb);
4588 return (error);
4589}
4590
4591/*
4592 * Remove a directory file.
4593 */
4594/* ARGSUSED */
4595int
4596sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
4597{
4598 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path),
4599 AT_REMOVEDIR, UIO_USERSPACE);
4600}
4601
4602/*
4603 * Read a block of directory entries in a file system independent format.
4604 */
4605int
4606sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
4607{
4608 /* {
4609 syscallarg(int) fd;
4610 syscallarg(char *) buf;
4611 syscallarg(size_t) count;
4612 } */
4613 file_t *fp;
4614 int error, done;
4615
4616 /* fd_getvnode() will use the descriptor for us */
4617 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
4618 return (error);
4619 if ((fp->f_flag & FREAD) == 0) {
4620 error = EBADF;
4621 goto out;
4622 }
4623 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
4624 SCARG(uap, count), &done, l, 0, 0);
4625 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
4626 *retval = done;
4627 out:
4628 fd_putfile(SCARG(uap, fd));
4629 return (error);
4630}
4631
4632/*
4633 * Set the mode mask for creation of filesystem nodes.
4634 */
4635int
4636sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
4637{
4638 /* {
4639 syscallarg(mode_t) newmask;
4640 } */
4641 struct proc *p = l->l_proc;
4642 struct cwdinfo *cwdi;
4643
4644 /*
4645 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's
4646 * important is that we serialize changes to the mask. The
4647 * rw_exit() will issue a write memory barrier on our behalf,
4648 * and force the changes out to other CPUs (as it must use an
4649 * atomic operation, draining the local CPU's store buffers).
4650 */
4651 cwdi = p->p_cwdi;
4652 rw_enter(&cwdi->cwdi_lock, RW_WRITER);
4653 *retval = cwdi->cwdi_cmask;
4654 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
4655 rw_exit(&cwdi->cwdi_lock);
4656
4657 return (0);
4658}
4659
4660int
4661dorevoke(struct vnode *vp, kauth_cred_t cred)
4662{
4663 struct vattr vattr;
4664 int error, fs_decision;
4665
4666 vn_lock(vp, LK_SHARED | LK_RETRY);
4667 error = VOP_GETATTR(vp, &vattr, cred);
4668 VOP_UNLOCK(vp);
4669 if (error != 0)
4670 return error;
4671 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM;
4672 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL,
4673 fs_decision);
4674 if (!error)
4675 VOP_REVOKE(vp, REVOKEALL);
4676 return (error);
4677}
4678
4679/*
4680 * Void all references to file by ripping underlying filesystem
4681 * away from vnode.
4682 */
4683/* ARGSUSED */
4684int
4685sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
4686{
4687 /* {
4688 syscallarg(const char *) path;
4689 } */
4690 struct vnode *vp;
4691 int error;
4692
4693 error = namei_simple_user(SCARG(uap, path),
4694 NSM_FOLLOW_TRYEMULROOT, &vp);
4695 if (error != 0)
4696 return (error);
4697 error = dorevoke(vp, l->l_cred);
4698 vrele(vp);
4699 return (error);
4700}
4701
4702/*
4703 * Allocate backing store for a file, filling a hole without having to
4704 * explicitly write anything out.
4705 */
4706/* ARGSUSED */
4707int
4708sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap,
4709 register_t *retval)
4710{
4711 /* {
4712 syscallarg(int) fd;
4713 syscallarg(off_t) pos;
4714 syscallarg(off_t) len;
4715 } */
4716 int fd;
4717 off_t pos, len;
4718 struct file *fp;
4719 struct vnode *vp;
4720 int error;
4721
4722 fd = SCARG(uap, fd);
4723 pos = SCARG(uap, pos);
4724 len = SCARG(uap, len);
4725
4726 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) {
4727 *retval = EINVAL;
4728 return 0;
4729 }
4730
4731 error = fd_getvnode(fd, &fp);
4732 if (error) {
4733 *retval = error;
4734 return 0;
4735 }
4736 if ((fp->f_flag & FWRITE) == 0) {
4737 error = EBADF;
4738 goto fail;
4739 }
4740 vp = fp->f_vnode;
4741
4742 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4743 if (vp->v_type == VDIR) {
4744 error = EISDIR;
4745 } else {
4746 error = VOP_FALLOCATE(vp, pos, len);
4747 }
4748 VOP_UNLOCK(vp);
4749
4750fail:
4751 fd_putfile(fd);
4752 *retval = error;
4753 return 0;
4754}
4755
4756/*
4757 * Deallocate backing store for a file, creating a hole. Also used for
4758 * invoking TRIM on disks.
4759 */
4760/* ARGSUSED */
4761int
4762sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap,
4763 register_t *retval)
4764{
4765 /* {
4766 syscallarg(int) fd;
4767 syscallarg(off_t) pos;
4768 syscallarg(off_t) len;
4769 } */
4770 int fd;
4771 off_t pos, len;
4772 struct file *fp;
4773 struct vnode *vp;
4774 int error;
4775
4776 fd = SCARG(uap, fd);
4777 pos = SCARG(uap, pos);
4778 len = SCARG(uap, len);
4779
4780 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) {
4781 return EINVAL;
4782 }
4783
4784 error = fd_getvnode(fd, &fp);
4785 if (error) {
4786 return error;
4787 }
4788 if ((fp->f_flag & FWRITE) == 0) {
4789 error = EBADF;
4790 goto fail;
4791 }
4792 vp = fp->f_vnode;
4793
4794 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4795 if (vp->v_type == VDIR) {
4796 error = EISDIR;
4797 } else {
4798 error = VOP_FDISCARD(vp, pos, len);
4799 }
4800 VOP_UNLOCK(vp);
4801
4802fail:
4803 fd_putfile(fd);
4804 return error;
4805}
4806