vfs_syscalls.c revision 1.164.2.8 1 /* $NetBSD: vfs_syscalls.c,v 1.164.2.8 2002/05/29 21:33:19 nathanw Exp $ */
2
3 /*
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. All advertising materials mentioning features or use of this software
21 * must display the following acknowledgement:
22 * This product includes software developed by the University of
23 * California, Berkeley and its contributors.
24 * 4. Neither the name of the University nor the names of its contributors
25 * may be used to endorse or promote products derived from this software
26 * without specific prior written permission.
27 *
28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * SUCH DAMAGE.
39 *
40 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95
41 */
42
43 #include <sys/cdefs.h>
44 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.164.2.8 2002/05/29 21:33:19 nathanw Exp $");
45
46 #include "opt_compat_netbsd.h"
47 #include "opt_compat_43.h"
48
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/namei.h>
52 #include <sys/filedesc.h>
53 #include <sys/kernel.h>
54 #include <sys/file.h>
55 #include <sys/stat.h>
56 #include <sys/vnode.h>
57 #include <sys/mount.h>
58 #include <sys/lwp.h>
59 #include <sys/proc.h>
60 #include <sys/uio.h>
61 #include <sys/malloc.h>
62 #include <sys/dirent.h>
63 #include <sys/sysctl.h>
64 #include <sys/sa.h>
65 #include <sys/syscallargs.h>
66
67 #include <miscfs/genfs/genfs.h>
68 #include <miscfs/syncfs/syncfs.h>
69
70 static int change_dir __P((struct nameidata *, struct proc *));
71 static int change_flags __P((struct vnode *, u_long, struct proc *));
72 static int change_mode __P((struct vnode *, int, struct proc *p));
73 static int change_owner __P((struct vnode *, uid_t, gid_t, struct proc *,
74 int));
75 static int change_utimes __P((struct vnode *vp, const struct timeval *,
76 struct proc *p));
77 static int rename_files __P((const char *, const char *, struct proc *, int));
78
79 void checkdirs __P((struct vnode *));
80
81 int dovfsusermount = 0;
82
83 /*
84 * Virtual File System System Calls
85 */
86
87 /*
88 * Mount a file system.
89 */
90
91 #if defined(COMPAT_09) || defined(COMPAT_43)
92 /*
93 * This table is used to maintain compatibility with 4.3BSD
94 * and NetBSD 0.9 mount syscalls. Note, the order is important!
95 *
96 * Do not modify this table. It should only contain filesystems
97 * supported by NetBSD 0.9 and 4.3BSD.
98 */
99 const char * const mountcompatnames[] = {
100 NULL, /* 0 = MOUNT_NONE */
101 MOUNT_FFS, /* 1 = MOUNT_UFS */
102 MOUNT_NFS, /* 2 */
103 MOUNT_MFS, /* 3 */
104 MOUNT_MSDOS, /* 4 */
105 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */
106 MOUNT_FDESC, /* 6 */
107 MOUNT_KERNFS, /* 7 */
108 NULL, /* 8 = MOUNT_DEVFS */
109 MOUNT_AFS, /* 9 */
110 };
111 const int nmountcompatnames = sizeof(mountcompatnames) /
112 sizeof(mountcompatnames[0]);
113 #endif /* COMPAT_09 || COMPAT_43 */
114
115 /* ARGSUSED */
116 int
117 sys_mount(l, v, retval)
118 struct lwp *l;
119 void *v;
120 register_t *retval;
121 {
122 struct sys_mount_args /* {
123 syscallarg(const char *) type;
124 syscallarg(const char *) path;
125 syscallarg(int) flags;
126 syscallarg(void *) data;
127 } */ *uap = v;
128 struct proc *p = l->l_proc;
129 struct vnode *vp;
130 struct mount *mp;
131 int error, flag = 0;
132 char fstypename[MFSNAMELEN];
133 struct vattr va;
134 struct nameidata nd;
135 struct vfsops *vfs;
136
137 if (dovfsusermount == 0 && (error = suser(p->p_ucred, &p->p_acflag)))
138 return (error);
139 /*
140 * Get vnode to be covered
141 */
142 NDINIT(&nd, LOOKUP, FOLLOW , UIO_USERSPACE,
143 SCARG(uap, path), p);
144 if ((error = namei(&nd)) != 0)
145 return (error);
146 vp = nd.ni_vp;
147 /*
148 * A lookup in VFS_MOUNT might result in an attempt to
149 * lock this vnode again, so make the lock resursive.
150 */
151 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
152 if (SCARG(uap, flags) & MNT_UPDATE) {
153 if ((vp->v_flag & VROOT) == 0) {
154 vput(vp);
155 return (EINVAL);
156 }
157 mp = vp->v_mount;
158 flag = mp->mnt_flag;
159 vfs = mp->mnt_op;
160 /*
161 * We only allow the filesystem to be reloaded if it
162 * is currently mounted read-only.
163 */
164 if ((SCARG(uap, flags) & MNT_RELOAD) &&
165 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
166 vput(vp);
167 return (EOPNOTSUPP); /* Needs translation */
168 }
169 /*
170 * In "highly secure" mode, don't let the caller do anything
171 * but downgrade a filesystem from read-write to read-only.
172 * (see also below; MNT_UPDATE is required.)
173 */
174 if (securelevel >= 2 &&
175 (SCARG(uap, flags) !=
176 (mp->mnt_flag | MNT_RDONLY |
177 MNT_RELOAD | MNT_FORCE | MNT_UPDATE))) {
178 vput(vp);
179 return (EPERM);
180 }
181 mp->mnt_flag |=
182 SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
183 /*
184 * Only root, or the user that did the original mount is
185 * permitted to update it.
186 */
187 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
188 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
189 vput(vp);
190 return (error);
191 }
192 /*
193 * Do not allow NFS export by non-root users. For non-root
194 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
195 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
196 */
197 if (p->p_ucred->cr_uid != 0) {
198 if (SCARG(uap, flags) & MNT_EXPORTED) {
199 vput(vp);
200 return (EPERM);
201 }
202 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
203 if (flag & MNT_NOEXEC)
204 SCARG(uap, flags) |= MNT_NOEXEC;
205 }
206 if (vfs_busy(mp, LK_NOWAIT, 0)) {
207 vput(vp);
208 return (EPERM);
209 }
210 VOP_UNLOCK(vp, 0);
211 goto update;
212 } else {
213 if (securelevel >= 2) {
214 vput(vp);
215 return (EPERM);
216 }
217 }
218 /*
219 * If the user is not root, ensure that they own the directory
220 * onto which we are attempting to mount.
221 */
222 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
223 (va.va_uid != p->p_ucred->cr_uid &&
224 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
225 vput(vp);
226 return (error);
227 }
228 /*
229 * Do not allow NFS export by non-root users. For non-root users,
230 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
231 * mount point is already MNT_NOEXEC.
232 */
233 if (p->p_ucred->cr_uid != 0) {
234 if (SCARG(uap, flags) & MNT_EXPORTED) {
235 vput(vp);
236 return (EPERM);
237 }
238 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
239 if (vp->v_mount->mnt_flag & MNT_NOEXEC)
240 SCARG(uap, flags) |= MNT_NOEXEC;
241 }
242 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
243 return (error);
244 if (vp->v_type != VDIR) {
245 vput(vp);
246 return (ENOTDIR);
247 }
248 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
249 if (error) {
250 #if defined(COMPAT_09) || defined(COMPAT_43)
251 /*
252 * Historically filesystem types were identified by number.
253 * If we get an integer for the filesystem type instead of a
254 * string, we check to see if it matches one of the historic
255 * filesystem types.
256 */
257 u_long fsindex = (u_long)SCARG(uap, type);
258 if (fsindex >= nmountcompatnames ||
259 mountcompatnames[fsindex] == NULL) {
260 vput(vp);
261 return (ENODEV);
262 }
263 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
264 #else
265 vput(vp);
266 return (error);
267 #endif
268 }
269 #ifdef COMPAT_10
270 /* Accept `ufs' as an alias for `ffs'. */
271 if (!strncmp(fstypename, "ufs", MFSNAMELEN))
272 strncpy(fstypename, "ffs", MFSNAMELEN);
273 #endif
274 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
275 vput(vp);
276 return (ENODEV);
277 }
278 if (vp->v_mountedhere != NULL) {
279 vput(vp);
280 return (EBUSY);
281 }
282
283 /*
284 * Allocate and initialize the file system.
285 */
286 mp = (struct mount *)malloc((u_long)sizeof(struct mount),
287 M_MOUNT, M_WAITOK);
288 memset((char *)mp, 0, (u_long)sizeof(struct mount));
289 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
290 (void)vfs_busy(mp, LK_NOWAIT, 0);
291 mp->mnt_op = vfs;
292 vfs->vfs_refcount++;
293 mp->mnt_vnodecovered = vp;
294 mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
295 mp->mnt_unmounter = NULL;
296 update:
297 /*
298 * Set the mount level flags.
299 */
300 if (SCARG(uap, flags) & MNT_RDONLY)
301 mp->mnt_flag |= MNT_RDONLY;
302 else if (mp->mnt_flag & MNT_RDONLY)
303 mp->mnt_flag |= MNT_WANTRDWR;
304 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
305 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
306 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
307 mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
308 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
309 MNT_NOCOREDUMP | MNT_IGNORE | MNT_NOATIME | MNT_NODEVMTIME |
310 MNT_SYMPERM | MNT_SOFTDEP);
311 /*
312 * Mount the filesystem.
313 */
314 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
315 if (mp->mnt_flag & MNT_UPDATE) {
316 vrele(vp);
317 if (mp->mnt_flag & MNT_WANTRDWR)
318 mp->mnt_flag &= ~MNT_RDONLY;
319 mp->mnt_flag &=~
320 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR);
321 if (error)
322 mp->mnt_flag = flag;
323 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
324 if (mp->mnt_syncer == NULL)
325 error = vfs_allocate_syncvnode(mp);
326 } else {
327 if (mp->mnt_syncer != NULL)
328 vfs_deallocate_syncvnode(mp);
329 }
330 vfs_unbusy(mp);
331 return (error);
332 }
333 /*
334 * Put the new filesystem on the mount list after root.
335 */
336 cache_purge(vp);
337 if (!error) {
338 vp->v_mountedhere = mp;
339 simple_lock(&mountlist_slock);
340 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
341 simple_unlock(&mountlist_slock);
342 checkdirs(vp);
343 VOP_UNLOCK(vp, 0);
344 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
345 error = vfs_allocate_syncvnode(mp);
346 vfs_unbusy(mp);
347 (void) VFS_STATFS(mp, &mp->mnt_stat, p);
348 if ((error = VFS_START(mp, 0, p)))
349 vrele(vp);
350 } else {
351 vp->v_mountedhere = (struct mount *)0;
352 vfs->vfs_refcount--;
353 vfs_unbusy(mp);
354 free((caddr_t)mp, M_MOUNT);
355 vput(vp);
356 }
357 return (error);
358 }
359
360 /*
361 * Scan all active processes to see if any of them have a current
362 * or root directory onto which the new filesystem has just been
363 * mounted. If so, replace them with the new mount point.
364 */
365 void
366 checkdirs(olddp)
367 struct vnode *olddp;
368 {
369 struct cwdinfo *cwdi;
370 struct vnode *newdp;
371 struct proc *p;
372
373 if (olddp->v_usecount == 1)
374 return;
375 if (VFS_ROOT(olddp->v_mountedhere, &newdp))
376 panic("mount: lost mount");
377 proclist_lock_read();
378 for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
379 cwdi = p->p_cwdi;
380 if (cwdi->cwdi_cdir == olddp) {
381 vrele(cwdi->cwdi_cdir);
382 VREF(newdp);
383 cwdi->cwdi_cdir = newdp;
384 }
385 if (cwdi->cwdi_rdir == olddp) {
386 vrele(cwdi->cwdi_rdir);
387 VREF(newdp);
388 cwdi->cwdi_rdir = newdp;
389 }
390 }
391 proclist_unlock_read();
392 if (rootvnode == olddp) {
393 vrele(rootvnode);
394 VREF(newdp);
395 rootvnode = newdp;
396 }
397 vput(newdp);
398 }
399
400 /*
401 * Unmount a file system.
402 *
403 * Note: unmount takes a path to the vnode mounted on as argument,
404 * not special file (as before).
405 */
406 /* ARGSUSED */
407 int
408 sys_unmount(l, v, retval)
409 struct lwp *l;
410 void *v;
411 register_t *retval;
412 {
413 struct sys_unmount_args /* {
414 syscallarg(const char *) path;
415 syscallarg(int) flags;
416 } */ *uap = v;
417 struct proc *p = l->l_proc;
418 struct vnode *vp;
419 struct mount *mp;
420 int error;
421 struct nameidata nd;
422
423 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
424 SCARG(uap, path), p);
425 if ((error = namei(&nd)) != 0)
426 return (error);
427 vp = nd.ni_vp;
428 mp = vp->v_mount;
429
430 /*
431 * Only root, or the user that did the original mount is
432 * permitted to unmount this filesystem.
433 */
434 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
435 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
436 vput(vp);
437 return (error);
438 }
439
440 /*
441 * Don't allow unmounting the root file system.
442 */
443 if (mp->mnt_flag & MNT_ROOTFS) {
444 vput(vp);
445 return (EINVAL);
446 }
447
448 /*
449 * Must be the root of the filesystem
450 */
451 if ((vp->v_flag & VROOT) == 0) {
452 vput(vp);
453 return (EINVAL);
454 }
455 vput(vp);
456
457 /*
458 * XXX Freeze syncer. Must do this before locking the
459 * mount point. See dounmount() for details.
460 */
461 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
462
463 if (vfs_busy(mp, 0, 0)) {
464 lockmgr(&syncer_lock, LK_RELEASE, NULL);
465 return (EBUSY);
466 }
467
468 return (dounmount(mp, SCARG(uap, flags), p));
469 }
470
471 /*
472 * Do the actual file system unmount. File system is assumed to have been
473 * marked busy by the caller.
474 */
475 int
476 dounmount(mp, flags, p)
477 struct mount *mp;
478 int flags;
479 struct proc *p;
480 {
481 struct vnode *coveredvp;
482 int error;
483 int async;
484 int used_syncer;
485
486 simple_lock(&mountlist_slock);
487 vfs_unbusy(mp);
488 used_syncer = (mp->mnt_syncer != NULL);
489
490 /*
491 * XXX Syncer must be frozen when we get here. This should really
492 * be done on a per-mountpoint basis, but especially the softdep
493 * code possibly called from the syncer doens't exactly work on a
494 * per-mountpoint basis, so the softdep code would become a maze
495 * of vfs_busy() calls.
496 *
497 * The caller of dounmount() must acquire syncer_lock because
498 * the syncer itself acquires locks in syncer_lock -> vfs_busy
499 * order, and we must preserve that order to avoid deadlock.
500 *
501 * So, if the file system did not use the syncer, now is
502 * the time to release the syncer_lock.
503 */
504 if (used_syncer == 0)
505 lockmgr(&syncer_lock, LK_RELEASE, NULL);
506
507 mp->mnt_flag |= MNT_UNMOUNT;
508 mp->mnt_unmounter = p;
509 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
510 if (mp->mnt_flag & MNT_EXPUBLIC)
511 vfs_setpublicfs(NULL, NULL, NULL);
512 async = mp->mnt_flag & MNT_ASYNC;
513 mp->mnt_flag &= ~MNT_ASYNC;
514 cache_purgevfs(mp); /* remove cache entries for this file sys */
515 if (mp->mnt_syncer != NULL)
516 vfs_deallocate_syncvnode(mp);
517 if (((mp->mnt_flag & MNT_RDONLY) ||
518 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
519 (flags & MNT_FORCE))
520 error = VFS_UNMOUNT(mp, flags, p);
521 simple_lock(&mountlist_slock);
522 if (error) {
523 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
524 (void) vfs_allocate_syncvnode(mp);
525 mp->mnt_flag &= ~MNT_UNMOUNT;
526 mp->mnt_unmounter = NULL;
527 mp->mnt_flag |= async;
528 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
529 &mountlist_slock);
530 if (used_syncer)
531 lockmgr(&syncer_lock, LK_RELEASE, NULL);
532 while (mp->mnt_wcnt > 0) {
533 wakeup((caddr_t)mp);
534 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0);
535 }
536 return (error);
537 }
538 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
539 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
540 coveredvp->v_mountedhere = NULL;
541 vrele(coveredvp);
542 }
543 mp->mnt_op->vfs_refcount--;
544 if (mp->mnt_vnodelist.lh_first != NULL)
545 panic("unmount: dangling vnode");
546 mp->mnt_flag |= MNT_GONE;
547 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
548 if (used_syncer)
549 lockmgr(&syncer_lock, LK_RELEASE, NULL);
550 while(mp->mnt_wcnt > 0) {
551 wakeup((caddr_t)mp);
552 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0);
553 }
554 free((caddr_t)mp, M_MOUNT);
555 return (0);
556 }
557
558 /*
559 * Sync each mounted filesystem.
560 */
561 #ifdef DEBUG
562 int syncprt = 0;
563 struct ctldebug debug0 = { "syncprt", &syncprt };
564 #endif
565
566 /* ARGSUSED */
567 int
568 sys_sync(l, v, retval)
569 struct lwp *l;
570 void *v;
571 register_t *retval;
572 {
573 struct mount *mp, *nmp;
574 int asyncflag;
575 struct proc *p = l->l_proc;
576
577 simple_lock(&mountlist_slock);
578 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
579 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
580 nmp = mp->mnt_list.cqe_prev;
581 continue;
582 }
583 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
584 asyncflag = mp->mnt_flag & MNT_ASYNC;
585 mp->mnt_flag &= ~MNT_ASYNC;
586 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
587 if (asyncflag)
588 mp->mnt_flag |= MNT_ASYNC;
589 }
590 simple_lock(&mountlist_slock);
591 nmp = mp->mnt_list.cqe_prev;
592 vfs_unbusy(mp);
593
594 }
595 simple_unlock(&mountlist_slock);
596 #ifdef DEBUG
597 if (syncprt)
598 vfs_bufstats();
599 #endif /* DEBUG */
600 return (0);
601 }
602
603 /*
604 * Change filesystem quotas.
605 */
606 /* ARGSUSED */
607 int
608 sys_quotactl(l, v, retval)
609 struct lwp *l;
610 void *v;
611 register_t *retval;
612 {
613 struct sys_quotactl_args /* {
614 syscallarg(const char *) path;
615 syscallarg(int) cmd;
616 syscallarg(int) uid;
617 syscallarg(caddr_t) arg;
618 } */ *uap = v;
619 struct proc *p = l->l_proc;
620 struct mount *mp;
621 int error;
622 struct nameidata nd;
623
624 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
625 if ((error = namei(&nd)) != 0)
626 return (error);
627 mp = nd.ni_vp->v_mount;
628 vrele(nd.ni_vp);
629 return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
630 SCARG(uap, arg), p));
631 }
632
633 /*
634 * Get filesystem statistics.
635 */
636 /* ARGSUSED */
637 int
638 sys_statfs(l, v, retval)
639 struct lwp *l;
640 void *v;
641 register_t *retval;
642 {
643 struct sys_statfs_args /* {
644 syscallarg(const char *) path;
645 syscallarg(struct statfs *) buf;
646 } */ *uap = v;
647 struct proc *p = l->l_proc;
648 struct mount *mp;
649 struct statfs *sp;
650 int error;
651 struct nameidata nd;
652
653 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
654 if ((error = namei(&nd)) != 0)
655 return (error);
656 mp = nd.ni_vp->v_mount;
657 sp = &mp->mnt_stat;
658 vrele(nd.ni_vp);
659 if ((error = VFS_STATFS(mp, sp, p)) != 0)
660 return (error);
661 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
662 sp->f_oflags = sp->f_flags & 0xffff;
663 return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
664 }
665
666 /*
667 * Get filesystem statistics.
668 */
669 /* ARGSUSED */
670 int
671 sys_fstatfs(l, v, retval)
672 struct lwp *l;
673 void *v;
674 register_t *retval;
675 {
676 struct sys_fstatfs_args /* {
677 syscallarg(int) fd;
678 syscallarg(struct statfs *) buf;
679 } */ *uap = v;
680 struct proc *p = l->l_proc;
681 struct file *fp;
682 struct mount *mp;
683 struct statfs *sp;
684 int error;
685
686 /* getvnode() will use the descriptor for us */
687 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
688 return (error);
689 mp = ((struct vnode *)fp->f_data)->v_mount;
690 sp = &mp->mnt_stat;
691 if ((error = VFS_STATFS(mp, sp, p)) != 0)
692 goto out;
693 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
694 sp->f_oflags = sp->f_flags & 0xffff;
695 error = copyout(sp, SCARG(uap, buf), sizeof(*sp));
696 out:
697 FILE_UNUSE(fp, p);
698 return (error);
699 }
700
701 /*
702 * Get statistics on all filesystems.
703 */
704 int
705 sys_getfsstat(l, v, retval)
706 struct lwp *l;
707 void *v;
708 register_t *retval;
709 {
710 struct sys_getfsstat_args /* {
711 syscallarg(struct statfs *) buf;
712 syscallarg(long) bufsize;
713 syscallarg(int) flags;
714 } */ *uap = v;
715 struct proc *p = l->l_proc;
716 struct mount *mp, *nmp;
717 struct statfs *sp;
718 caddr_t sfsp;
719 long count, maxcount, error;
720
721 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
722 sfsp = (caddr_t)SCARG(uap, buf);
723 simple_lock(&mountlist_slock);
724 count = 0;
725 for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
726 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
727 nmp = mp->mnt_list.cqe_next;
728 continue;
729 }
730 if (sfsp && count < maxcount) {
731 sp = &mp->mnt_stat;
732 /*
733 * If MNT_NOWAIT or MNT_LAZY is specified, do not
734 * refresh the fsstat cache. MNT_WAIT or MNT_LAXY
735 * overrides MNT_NOWAIT.
736 */
737 if (SCARG(uap, flags) != MNT_NOWAIT &&
738 SCARG(uap, flags) != MNT_LAZY &&
739 (SCARG(uap, flags) == MNT_WAIT ||
740 SCARG(uap, flags) == 0) &&
741 (error = VFS_STATFS(mp, sp, p)) != 0) {
742 simple_lock(&mountlist_slock);
743 nmp = mp->mnt_list.cqe_next;
744 vfs_unbusy(mp);
745 continue;
746 }
747 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
748 sp->f_oflags = sp->f_flags & 0xffff;
749 error = copyout(sp, sfsp, sizeof(*sp));
750 if (error) {
751 vfs_unbusy(mp);
752 return (error);
753 }
754 sfsp += sizeof(*sp);
755 }
756 count++;
757 simple_lock(&mountlist_slock);
758 nmp = mp->mnt_list.cqe_next;
759 vfs_unbusy(mp);
760 }
761 simple_unlock(&mountlist_slock);
762 if (sfsp && count > maxcount)
763 *retval = maxcount;
764 else
765 *retval = count;
766 return (0);
767 }
768
769 /*
770 * Change current working directory to a given file descriptor.
771 */
772 /* ARGSUSED */
773 int
774 sys_fchdir(l, v, retval)
775 struct lwp *l;
776 void *v;
777 register_t *retval;
778 {
779 struct sys_fchdir_args /* {
780 syscallarg(int) fd;
781 } */ *uap = v;
782 struct proc *p = l->l_proc;
783 struct filedesc *fdp = p->p_fd;
784 struct cwdinfo *cwdi = p->p_cwdi;
785 struct vnode *vp, *tdp;
786 struct mount *mp;
787 struct file *fp;
788 int error;
789
790 /* getvnode() will use the descriptor for us */
791 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
792 return (error);
793 vp = (struct vnode *)fp->f_data;
794
795 VREF(vp);
796 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
797 if (vp->v_type != VDIR)
798 error = ENOTDIR;
799 else
800 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
801 while (!error && (mp = vp->v_mountedhere) != NULL) {
802 if (vfs_busy(mp, 0, 0))
803 continue;
804 error = VFS_ROOT(mp, &tdp);
805 vfs_unbusy(mp);
806 if (error)
807 break;
808 vput(vp);
809 vp = tdp;
810 }
811 if (error) {
812 vput(vp);
813 goto out;
814 }
815 VOP_UNLOCK(vp, 0);
816
817 /*
818 * Disallow changing to a directory not under the process's
819 * current root directory (if there is one).
820 */
821 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
822 vrele(vp);
823 error = EPERM; /* operation not permitted */
824 goto out;
825 }
826
827 vrele(cwdi->cwdi_cdir);
828 cwdi->cwdi_cdir = vp;
829 out:
830 FILE_UNUSE(fp, p);
831 return (error);
832 }
833
834 /*
835 * Change this process's notion of the root directory to a given file descriptor.
836 */
837
838 int
839 sys_fchroot(l, v, retval)
840 struct lwp *l;
841 void *v;
842 register_t *retval;
843 {
844 struct sys_fchroot_args *uap = v;
845 struct proc *p = l->l_proc;
846 struct filedesc *fdp = p->p_fd;
847 struct cwdinfo *cwdi = p->p_cwdi;
848 struct vnode *vp;
849 struct file *fp;
850 int error;
851
852 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
853 return error;
854 /* getvnode() will use the descriptor for us */
855 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
856 return error;
857 vp = (struct vnode *) fp->f_data;
858 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
859 if (vp->v_type != VDIR)
860 error = ENOTDIR;
861 else
862 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
863 VOP_UNLOCK(vp, 0);
864 if (error)
865 goto out;
866 VREF(vp);
867
868 /*
869 * Prevent escaping from chroot by putting the root under
870 * the working directory. Silently chdir to / if we aren't
871 * already there.
872 */
873 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
874 /*
875 * XXX would be more failsafe to change directory to a
876 * deadfs node here instead
877 */
878 vrele(cwdi->cwdi_cdir);
879 VREF(vp);
880 cwdi->cwdi_cdir = vp;
881 }
882
883 if (cwdi->cwdi_rdir != NULL)
884 vrele(cwdi->cwdi_rdir);
885 cwdi->cwdi_rdir = vp;
886 out:
887 FILE_UNUSE(fp, p);
888 return (error);
889 }
890
891
892
893 /*
894 * Change current working directory (``.'').
895 */
896 /* ARGSUSED */
897 int
898 sys_chdir(l, v, retval)
899 struct lwp *l;
900 void *v;
901 register_t *retval;
902 {
903 struct sys_chdir_args /* {
904 syscallarg(const char *) path;
905 } */ *uap = v;
906 struct proc *p = l->l_proc;
907 struct cwdinfo *cwdi = p->p_cwdi;
908 int error;
909 struct nameidata nd;
910
911 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
912 SCARG(uap, path), p);
913 if ((error = change_dir(&nd, p)) != 0)
914 return (error);
915 vrele(cwdi->cwdi_cdir);
916 cwdi->cwdi_cdir = nd.ni_vp;
917 return (0);
918 }
919
920 /*
921 * Change notion of root (``/'') directory.
922 */
923 /* ARGSUSED */
924 int
925 sys_chroot(l, v, retval)
926 struct lwp *l;
927 void *v;
928 register_t *retval;
929 {
930 struct sys_chroot_args /* {
931 syscallarg(const char *) path;
932 } */ *uap = v;
933 struct proc *p = l->l_proc;
934 struct cwdinfo *cwdi = p->p_cwdi;
935 struct vnode *vp;
936 int error;
937 struct nameidata nd;
938
939 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
940 return (error);
941 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
942 SCARG(uap, path), p);
943 if ((error = change_dir(&nd, p)) != 0)
944 return (error);
945 if (cwdi->cwdi_rdir != NULL)
946 vrele(cwdi->cwdi_rdir);
947 vp = nd.ni_vp;
948 cwdi->cwdi_rdir = vp;
949
950 /*
951 * Prevent escaping from chroot by putting the root under
952 * the working directory. Silently chdir to / if we aren't
953 * already there.
954 */
955 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
956 /*
957 * XXX would be more failsafe to change directory to a
958 * deadfs node here instead
959 */
960 vrele(cwdi->cwdi_cdir);
961 VREF(vp);
962 cwdi->cwdi_cdir = vp;
963 }
964
965 return (0);
966 }
967
968 /*
969 * Common routine for chroot and chdir.
970 */
971 static int
972 change_dir(ndp, p)
973 struct nameidata *ndp;
974 struct proc *p;
975 {
976 struct vnode *vp;
977 int error;
978
979 if ((error = namei(ndp)) != 0)
980 return (error);
981 vp = ndp->ni_vp;
982 if (vp->v_type != VDIR)
983 error = ENOTDIR;
984 else
985 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
986
987 if (error)
988 vput(vp);
989 else
990 VOP_UNLOCK(vp, 0);
991 return (error);
992 }
993
994 /*
995 * Check permissions, allocate an open file structure,
996 * and call the device open routine if any.
997 */
998 int
999 sys_open(l, v, retval)
1000 struct lwp *l;
1001 void *v;
1002 register_t *retval;
1003 {
1004 struct sys_open_args /* {
1005 syscallarg(const char *) path;
1006 syscallarg(int) flags;
1007 syscallarg(int) mode;
1008 } */ *uap = v;
1009 struct proc *p = l->l_proc;
1010 struct cwdinfo *cwdi = p->p_cwdi;
1011 struct filedesc *fdp = p->p_fd;
1012 struct file *fp;
1013 struct vnode *vp;
1014 int flags, cmode;
1015 int type, indx, error;
1016 struct flock lf;
1017 struct nameidata nd;
1018
1019 flags = FFLAGS(SCARG(uap, flags));
1020 if ((flags & (FREAD | FWRITE)) == 0)
1021 return (EINVAL);
1022 /* falloc() will use the file descriptor for us */
1023 if ((error = falloc(p, &fp, &indx)) != 0)
1024 return (error);
1025 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1026 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1027 p->p_dupfd = -indx - 1; /* XXX check for fdopen */
1028 if ((error = vn_open(&nd, flags, cmode)) != 0) {
1029 FILE_UNUSE(fp, p);
1030 ffree(fp);
1031 if ((error == ENODEV || error == ENXIO) &&
1032 p->p_dupfd >= 0 && /* XXX from fdopen */
1033 (error =
1034 dupfdopen(p, indx, p->p_dupfd, flags, error)) == 0) {
1035 *retval = indx;
1036 return (0);
1037 }
1038 if (error == ERESTART)
1039 error = EINTR;
1040 fdremove(fdp, indx);
1041 return (error);
1042 }
1043 p->p_dupfd = 0;
1044 vp = nd.ni_vp;
1045 fp->f_flag = flags & FMASK;
1046 fp->f_type = DTYPE_VNODE;
1047 fp->f_ops = &vnops;
1048 fp->f_data = (caddr_t)vp;
1049 if (flags & (O_EXLOCK | O_SHLOCK)) {
1050 lf.l_whence = SEEK_SET;
1051 lf.l_start = 0;
1052 lf.l_len = 0;
1053 if (flags & O_EXLOCK)
1054 lf.l_type = F_WRLCK;
1055 else
1056 lf.l_type = F_RDLCK;
1057 type = F_FLOCK;
1058 if ((flags & FNONBLOCK) == 0)
1059 type |= F_WAIT;
1060 VOP_UNLOCK(vp, 0);
1061 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1062 if (error) {
1063 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1064 FILE_UNUSE(fp, p);
1065 ffree(fp);
1066 fdremove(fdp, indx);
1067 return (error);
1068 }
1069 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1070 fp->f_flag |= FHASLOCK;
1071 }
1072 VOP_UNLOCK(vp, 0);
1073 *retval = indx;
1074 FILE_SET_MATURE(fp);
1075 FILE_UNUSE(fp, p);
1076 return (0);
1077 }
1078
1079 /*
1080 * Get file handle system call
1081 */
1082 int
1083 sys_getfh(l, v, retval)
1084 struct lwp *l;
1085 void *v;
1086 register_t *retval;
1087 {
1088 struct sys_getfh_args /* {
1089 syscallarg(char *) fname;
1090 syscallarg(fhandle_t *) fhp;
1091 } */ *uap = v;
1092 struct proc *p = l->l_proc;
1093 struct vnode *vp;
1094 fhandle_t fh;
1095 int error;
1096 struct nameidata nd;
1097
1098 /*
1099 * Must be super user
1100 */
1101 error = suser(p->p_ucred, &p->p_acflag);
1102 if (error)
1103 return (error);
1104 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1105 SCARG(uap, fname), p);
1106 error = namei(&nd);
1107 if (error)
1108 return (error);
1109 vp = nd.ni_vp;
1110 memset((caddr_t)&fh, 0, sizeof(fh));
1111 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1112 error = VFS_VPTOFH(vp, &fh.fh_fid);
1113 vput(vp);
1114 if (error)
1115 return (error);
1116 error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1117 return (error);
1118 }
1119
1120 /*
1121 * Open a file given a file handle.
1122 *
1123 * Check permissions, allocate an open file structure,
1124 * and call the device open routine if any.
1125 */
1126 int
1127 sys_fhopen(l, v, retval)
1128 struct lwp *l;
1129 void *v;
1130 register_t *retval;
1131 {
1132 struct sys_fhopen_args /* {
1133 syscallarg(const fhandle_t *) fhp;
1134 syscallarg(int) flags;
1135 } */ *uap = v;
1136 struct proc *p = l->l_proc;
1137 struct filedesc *fdp = p->p_fd;
1138 struct file *fp;
1139 struct vnode *vp = NULL;
1140 struct mount *mp;
1141 struct ucred *cred = p->p_ucred;
1142 int flags;
1143 struct file *nfp;
1144 int type, indx, error=0;
1145 struct flock lf;
1146 struct vattr va;
1147 fhandle_t fh;
1148
1149 /*
1150 * Must be super user
1151 */
1152 if ((error = suser(p->p_ucred, &p->p_acflag)))
1153 return (error);
1154
1155 flags = FFLAGS(SCARG(uap, flags));
1156 if ((flags & (FREAD | FWRITE)) == 0)
1157 return (EINVAL);
1158 if ((flags & O_CREAT))
1159 return (EINVAL);
1160 /* falloc() will use the file descriptor for us */
1161 if ((error = falloc(p, &nfp, &indx)) != 0)
1162 return (error);
1163 fp = nfp;
1164 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1165 goto bad;
1166
1167 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1168 error = ESTALE;
1169 goto bad;
1170 }
1171
1172 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1173 vp = NULL; /* most likely unnecessary sanity for bad: */
1174 goto bad;
1175 }
1176
1177 /* Now do an effective vn_open */
1178
1179 if (vp->v_type == VSOCK) {
1180 error = EOPNOTSUPP;
1181 goto bad;
1182 }
1183 if (flags & FREAD) {
1184 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1185 goto bad;
1186 }
1187 if (flags & (FWRITE | O_TRUNC)) {
1188 if (vp->v_type == VDIR) {
1189 error = EISDIR;
1190 goto bad;
1191 }
1192 if ((error = vn_writechk(vp)) != 0 ||
1193 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1194 goto bad;
1195 }
1196 if (flags & O_TRUNC) {
1197 VOP_UNLOCK(vp, 0); /* XXX */
1198 VOP_LEASE(vp, p, cred, LEASE_WRITE);
1199 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
1200 VATTR_NULL(&va);
1201 va.va_size = 0;
1202 if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1203 goto bad;
1204 }
1205 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1206 goto bad;
1207 if (vp->v_type == VREG &&
1208 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1209 error = EIO;
1210 goto bad;
1211 }
1212 if (flags & FWRITE)
1213 vp->v_writecount++;
1214
1215 /* done with modified vn_open, now finish what sys_open does. */
1216
1217 fp->f_flag = flags & FMASK;
1218 fp->f_type = DTYPE_VNODE;
1219 fp->f_ops = &vnops;
1220 fp->f_data = (caddr_t)vp;
1221 if (flags & (O_EXLOCK | O_SHLOCK)) {
1222 lf.l_whence = SEEK_SET;
1223 lf.l_start = 0;
1224 lf.l_len = 0;
1225 if (flags & O_EXLOCK)
1226 lf.l_type = F_WRLCK;
1227 else
1228 lf.l_type = F_RDLCK;
1229 type = F_FLOCK;
1230 if ((flags & FNONBLOCK) == 0)
1231 type |= F_WAIT;
1232 VOP_UNLOCK(vp, 0);
1233 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1234 if (error) {
1235 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1236 FILE_UNUSE(fp, p);
1237 ffree(fp);
1238 fdremove(fdp, indx);
1239 return (error);
1240 }
1241 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1242 fp->f_flag |= FHASLOCK;
1243 }
1244 VOP_UNLOCK(vp, 0);
1245 *retval = indx;
1246 FILE_SET_MATURE(fp);
1247 FILE_UNUSE(fp, p);
1248 return (0);
1249
1250 bad:
1251 FILE_UNUSE(fp, p);
1252 ffree(fp);
1253 fdremove(fdp, indx);
1254 if (vp != NULL)
1255 vput(vp);
1256 return (error);
1257 }
1258
1259 /* ARGSUSED */
1260 int
1261 sys_fhstat(l, v, retval)
1262 struct lwp *l;
1263 void *v;
1264 register_t *retval;
1265 {
1266 struct sys_fhstat_args /* {
1267 syscallarg(const fhandle_t *) fhp;
1268 syscallarg(struct stat *) sb;
1269 } */ *uap = v;
1270 struct proc *p = l->l_proc;
1271 struct stat sb;
1272 int error;
1273 fhandle_t fh;
1274 struct mount *mp;
1275 struct vnode *vp;
1276
1277 /*
1278 * Must be super user
1279 */
1280 if ((error = suser(p->p_ucred, &p->p_acflag)))
1281 return (error);
1282
1283 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1284 return (error);
1285
1286 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1287 return (ESTALE);
1288 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1289 return (error);
1290 error = vn_stat(vp, &sb, p);
1291 vput(vp);
1292 if (error)
1293 return (error);
1294 error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1295 return (error);
1296 }
1297
1298 /* ARGSUSED */
1299 int
1300 sys_fhstatfs(l, v, retval)
1301 struct lwp *l;
1302 void *v;
1303 register_t *retval;
1304 {
1305 struct sys_fhstatfs_args /*
1306 syscallarg(const fhandle_t *) fhp;
1307 syscallarg(struct statfs *) buf;
1308 } */ *uap = v;
1309 struct proc *p = l->l_proc;
1310 struct statfs sp;
1311 fhandle_t fh;
1312 struct mount *mp;
1313 struct vnode *vp;
1314 int error;
1315
1316 /*
1317 * Must be super user
1318 */
1319 if ((error = suser(p->p_ucred, &p->p_acflag)))
1320 return (error);
1321
1322 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1323 return (error);
1324
1325 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1326 return (ESTALE);
1327 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1328 return (error);
1329 mp = vp->v_mount;
1330 vput(vp);
1331 if ((error = VFS_STATFS(mp, &sp, p)) != 0)
1332 return (error);
1333 sp.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1334 sp.f_oflags = sp.f_flags & 0xffff;
1335 return (copyout(&sp, SCARG(uap, buf), sizeof(sp)));
1336 }
1337
1338 /*
1339 * Create a special file.
1340 */
1341 /* ARGSUSED */
1342 int
1343 sys_mknod(l, v, retval)
1344 struct lwp *l;
1345 void *v;
1346 register_t *retval;
1347 {
1348 struct sys_mknod_args /* {
1349 syscallarg(const char *) path;
1350 syscallarg(int) mode;
1351 syscallarg(int) dev;
1352 } */ *uap = v;
1353 struct proc *p = l->l_proc;
1354 struct vnode *vp;
1355 struct vattr vattr;
1356 int error;
1357 int whiteout = 0;
1358 struct nameidata nd;
1359
1360 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1361 return (error);
1362 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1363 if ((error = namei(&nd)) != 0)
1364 return (error);
1365 vp = nd.ni_vp;
1366 if (vp != NULL)
1367 error = EEXIST;
1368 else {
1369 VATTR_NULL(&vattr);
1370 vattr.va_mode =
1371 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1372 vattr.va_rdev = SCARG(uap, dev);
1373 whiteout = 0;
1374
1375 switch (SCARG(uap, mode) & S_IFMT) {
1376 case S_IFMT: /* used by badsect to flag bad sectors */
1377 vattr.va_type = VBAD;
1378 break;
1379 case S_IFCHR:
1380 vattr.va_type = VCHR;
1381 break;
1382 case S_IFBLK:
1383 vattr.va_type = VBLK;
1384 break;
1385 case S_IFWHT:
1386 whiteout = 1;
1387 break;
1388 default:
1389 error = EINVAL;
1390 break;
1391 }
1392 }
1393 if (!error) {
1394 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1395 if (whiteout) {
1396 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1397 if (error)
1398 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1399 vput(nd.ni_dvp);
1400 } else {
1401 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1402 &nd.ni_cnd, &vattr);
1403 if (error == 0)
1404 vput(nd.ni_vp);
1405 }
1406 } else {
1407 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1408 if (nd.ni_dvp == vp)
1409 vrele(nd.ni_dvp);
1410 else
1411 vput(nd.ni_dvp);
1412 if (vp)
1413 vrele(vp);
1414 }
1415 return (error);
1416 }
1417
1418 /*
1419 * Create a named pipe.
1420 */
1421 /* ARGSUSED */
1422 int
1423 sys_mkfifo(l, v, retval)
1424 struct lwp *l;
1425 void *v;
1426 register_t *retval;
1427 {
1428 struct sys_mkfifo_args /* {
1429 syscallarg(const char *) path;
1430 syscallarg(int) mode;
1431 } */ *uap = v;
1432 struct proc *p = l->l_proc;
1433 struct vattr vattr;
1434 int error;
1435 struct nameidata nd;
1436
1437 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1438 if ((error = namei(&nd)) != 0)
1439 return (error);
1440 if (nd.ni_vp != NULL) {
1441 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1442 if (nd.ni_dvp == nd.ni_vp)
1443 vrele(nd.ni_dvp);
1444 else
1445 vput(nd.ni_dvp);
1446 vrele(nd.ni_vp);
1447 return (EEXIST);
1448 }
1449 VATTR_NULL(&vattr);
1450 vattr.va_type = VFIFO;
1451 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1452 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1453 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1454 if (error == 0)
1455 vput(nd.ni_vp);
1456 return (error);
1457 }
1458
1459 /*
1460 * Make a hard file link.
1461 */
1462 /* ARGSUSED */
1463 int
1464 sys_link(l, v, retval)
1465 struct lwp *l;
1466 void *v;
1467 register_t *retval;
1468 {
1469 struct sys_link_args /* {
1470 syscallarg(const char *) path;
1471 syscallarg(const char *) link;
1472 } */ *uap = v;
1473 struct proc *p = l->l_proc;
1474 struct vnode *vp;
1475 struct nameidata nd;
1476 int error;
1477
1478 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1479 if ((error = namei(&nd)) != 0)
1480 return (error);
1481 vp = nd.ni_vp;
1482 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1483 if ((error = namei(&nd)) != 0)
1484 goto out;
1485 if (nd.ni_vp) {
1486 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1487 if (nd.ni_dvp == nd.ni_vp)
1488 vrele(nd.ni_dvp);
1489 else
1490 vput(nd.ni_dvp);
1491 vrele(nd.ni_vp);
1492 error = EEXIST;
1493 goto out;
1494 }
1495 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1496 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1497 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1498 out:
1499 vrele(vp);
1500 return (error);
1501 }
1502
1503 /*
1504 * Make a symbolic link.
1505 */
1506 /* ARGSUSED */
1507 int
1508 sys_symlink(l, v, retval)
1509 struct lwp *l;
1510 void *v;
1511 register_t *retval;
1512 {
1513 struct sys_symlink_args /* {
1514 syscallarg(const char *) path;
1515 syscallarg(const char *) link;
1516 } */ *uap = v;
1517 struct proc *p = l->l_proc;
1518 struct vattr vattr;
1519 char *path;
1520 int error;
1521 struct nameidata nd;
1522
1523 path = PNBUF_GET();
1524 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1525 if (error)
1526 goto out;
1527 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1528 if ((error = namei(&nd)) != 0)
1529 goto out;
1530 if (nd.ni_vp) {
1531 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1532 if (nd.ni_dvp == nd.ni_vp)
1533 vrele(nd.ni_dvp);
1534 else
1535 vput(nd.ni_dvp);
1536 vrele(nd.ni_vp);
1537 error = EEXIST;
1538 goto out;
1539 }
1540 VATTR_NULL(&vattr);
1541 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1542 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1543 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1544 if (error == 0)
1545 vput(nd.ni_vp);
1546 out:
1547 PNBUF_PUT(path);
1548 return (error);
1549 }
1550
1551 /*
1552 * Delete a whiteout from the filesystem.
1553 */
1554 /* ARGSUSED */
1555 int
1556 sys_undelete(l, v, retval)
1557 struct lwp *l;
1558 void *v;
1559 register_t *retval;
1560 {
1561 struct sys_undelete_args /* {
1562 syscallarg(const char *) path;
1563 } */ *uap = v;
1564 struct proc *p = l->l_proc;
1565 int error;
1566 struct nameidata nd;
1567
1568 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1569 SCARG(uap, path), p);
1570 error = namei(&nd);
1571 if (error)
1572 return (error);
1573
1574 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1575 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1576 if (nd.ni_dvp == nd.ni_vp)
1577 vrele(nd.ni_dvp);
1578 else
1579 vput(nd.ni_dvp);
1580 if (nd.ni_vp)
1581 vrele(nd.ni_vp);
1582 return (EEXIST);
1583 }
1584
1585 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1586 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1587 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1588 vput(nd.ni_dvp);
1589 return (error);
1590 }
1591
1592 /*
1593 * Delete a name from the filesystem.
1594 */
1595 /* ARGSUSED */
1596 int
1597 sys_unlink(l, v, retval)
1598 struct lwp *l;
1599 void *v;
1600 register_t *retval;
1601 {
1602 struct sys_unlink_args /* {
1603 syscallarg(const char *) path;
1604 } */ *uap = v;
1605 struct proc *p = l->l_proc;
1606 struct vnode *vp;
1607 int error;
1608 struct nameidata nd;
1609
1610 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1611 SCARG(uap, path), p);
1612 if ((error = namei(&nd)) != 0)
1613 return (error);
1614 vp = nd.ni_vp;
1615
1616 /*
1617 * The root of a mounted filesystem cannot be deleted.
1618 */
1619 if (vp->v_flag & VROOT) {
1620 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1621 if (nd.ni_dvp == vp)
1622 vrele(nd.ni_dvp);
1623 else
1624 vput(nd.ni_dvp);
1625 vput(vp);
1626 error = EBUSY;
1627 goto out;
1628 }
1629
1630 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1631 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1632 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1633 out:
1634 return (error);
1635 }
1636
1637 /*
1638 * Reposition read/write file offset.
1639 */
1640 int
1641 sys_lseek(l, v, retval)
1642 struct lwp *l;
1643 void *v;
1644 register_t *retval;
1645 {
1646 struct sys_lseek_args /* {
1647 syscallarg(int) fd;
1648 syscallarg(int) pad;
1649 syscallarg(off_t) offset;
1650 syscallarg(int) whence;
1651 } */ *uap = v;
1652 struct proc *p = l->l_proc;
1653 struct ucred *cred = p->p_ucred;
1654 struct filedesc *fdp = p->p_fd;
1655 struct file *fp;
1656 struct vnode *vp;
1657 struct vattr vattr;
1658 off_t newoff;
1659 int error;
1660
1661 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1662 return (EBADF);
1663
1664 FILE_USE(fp);
1665
1666 vp = (struct vnode *)fp->f_data;
1667 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1668 error = ESPIPE;
1669 goto out;
1670 }
1671
1672 switch (SCARG(uap, whence)) {
1673 case SEEK_CUR:
1674 newoff = fp->f_offset + SCARG(uap, offset);
1675 break;
1676 case SEEK_END:
1677 error = VOP_GETATTR(vp, &vattr, cred, p);
1678 if (error)
1679 goto out;
1680 newoff = SCARG(uap, offset) + vattr.va_size;
1681 break;
1682 case SEEK_SET:
1683 newoff = SCARG(uap, offset);
1684 break;
1685 default:
1686 error = EINVAL;
1687 goto out;
1688 }
1689 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1690 goto out;
1691
1692 *(off_t *)retval = fp->f_offset = newoff;
1693 out:
1694 FILE_UNUSE(fp, p);
1695 return (error);
1696 }
1697
1698 /*
1699 * Positional read system call.
1700 */
1701 int
1702 sys_pread(l, v, retval)
1703 struct lwp *l;
1704 void *v;
1705 register_t *retval;
1706 {
1707 struct sys_pread_args /* {
1708 syscallarg(int) fd;
1709 syscallarg(void *) buf;
1710 syscallarg(size_t) nbyte;
1711 syscallarg(off_t) offset;
1712 } */ *uap = v;
1713 struct proc *p = l->l_proc;
1714 struct filedesc *fdp = p->p_fd;
1715 struct file *fp;
1716 struct vnode *vp;
1717 off_t offset;
1718 int error, fd = SCARG(uap, fd);
1719
1720 if ((fp = fd_getfile(fdp, fd)) == NULL)
1721 return (EBADF);
1722
1723 if ((fp->f_flag & FREAD) == 0)
1724 return (EBADF);
1725
1726 FILE_USE(fp);
1727
1728 vp = (struct vnode *)fp->f_data;
1729 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1730 error = ESPIPE;
1731 goto out;
1732 }
1733
1734 offset = SCARG(uap, offset);
1735
1736 /*
1737 * XXX This works because no file systems actually
1738 * XXX take any action on the seek operation.
1739 */
1740 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1741 goto out;
1742
1743 /* dofileread() will unuse the descriptor for us */
1744 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1745 &offset, 0, retval));
1746
1747 out:
1748 FILE_UNUSE(fp, p);
1749 return (error);
1750 }
1751
1752 /*
1753 * Positional scatter read system call.
1754 */
1755 int
1756 sys_preadv(l, v, retval)
1757 struct lwp *l;
1758 void *v;
1759 register_t *retval;
1760 {
1761 struct sys_preadv_args /* {
1762 syscallarg(int) fd;
1763 syscallarg(const struct iovec *) iovp;
1764 syscallarg(int) iovcnt;
1765 syscallarg(off_t) offset;
1766 } */ *uap = v;
1767 struct proc *p = l->l_proc;
1768 struct filedesc *fdp = p->p_fd;
1769 struct file *fp;
1770 struct vnode *vp;
1771 off_t offset;
1772 int error, fd = SCARG(uap, fd);
1773
1774 if ((fp = fd_getfile(fdp, fd)) == NULL)
1775 return (EBADF);
1776
1777 if ((fp->f_flag & FREAD) == 0)
1778 return (EBADF);
1779
1780 FILE_USE(fp);
1781
1782 vp = (struct vnode *)fp->f_data;
1783 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1784 error = ESPIPE;
1785 goto out;
1786 }
1787
1788 offset = SCARG(uap, offset);
1789
1790 /*
1791 * XXX This works because no file systems actually
1792 * XXX take any action on the seek operation.
1793 */
1794 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1795 goto out;
1796
1797 /* dofilereadv() will unuse the descriptor for us */
1798 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1799 &offset, 0, retval));
1800
1801 out:
1802 FILE_UNUSE(fp, p);
1803 return (error);
1804 }
1805
1806 /*
1807 * Positional write system call.
1808 */
1809 int
1810 sys_pwrite(l, v, retval)
1811 struct lwp *l;
1812 void *v;
1813 register_t *retval;
1814 {
1815 struct sys_pwrite_args /* {
1816 syscallarg(int) fd;
1817 syscallarg(const void *) buf;
1818 syscallarg(size_t) nbyte;
1819 syscallarg(off_t) offset;
1820 } */ *uap = v;
1821 struct proc *p = l->l_proc;
1822 struct filedesc *fdp = p->p_fd;
1823 struct file *fp;
1824 struct vnode *vp;
1825 off_t offset;
1826 int error, fd = SCARG(uap, fd);
1827
1828 if ((fp = fd_getfile(fdp, fd)) == NULL)
1829 return (EBADF);
1830
1831 if ((fp->f_flag & FWRITE) == 0)
1832 return (EBADF);
1833
1834 FILE_USE(fp);
1835
1836 vp = (struct vnode *)fp->f_data;
1837 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1838 error = ESPIPE;
1839 goto out;
1840 }
1841
1842 offset = SCARG(uap, offset);
1843
1844 /*
1845 * XXX This works because no file systems actually
1846 * XXX take any action on the seek operation.
1847 */
1848 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1849 goto out;
1850
1851 /* dofilewrite() will unuse the descriptor for us */
1852 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1853 &offset, 0, retval));
1854
1855 out:
1856 FILE_UNUSE(fp, p);
1857 return (error);
1858 }
1859
1860 /*
1861 * Positional gather write system call.
1862 */
1863 int
1864 sys_pwritev(l, v, retval)
1865 struct lwp *l;
1866 void *v;
1867 register_t *retval;
1868 {
1869 struct sys_pwritev_args /* {
1870 syscallarg(int) fd;
1871 syscallarg(const struct iovec *) iovp;
1872 syscallarg(int) iovcnt;
1873 syscallarg(off_t) offset;
1874 } */ *uap = v;
1875 struct proc *p = l->l_proc;
1876 struct filedesc *fdp = p->p_fd;
1877 struct file *fp;
1878 struct vnode *vp;
1879 off_t offset;
1880 int error, fd = SCARG(uap, fd);
1881
1882 if ((fp = fd_getfile(fdp, fd)) == NULL)
1883 return (EBADF);
1884
1885 if ((fp->f_flag & FWRITE) == 0)
1886 return (EBADF);
1887
1888 FILE_USE(fp);
1889
1890 vp = (struct vnode *)fp->f_data;
1891 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1892 error = ESPIPE;
1893 goto out;
1894 }
1895
1896 offset = SCARG(uap, offset);
1897
1898 /*
1899 * XXX This works because no file systems actually
1900 * XXX take any action on the seek operation.
1901 */
1902 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1903 goto out;
1904
1905 /* dofilewritev() will unuse the descriptor for us */
1906 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1907 &offset, 0, retval));
1908
1909 out:
1910 FILE_UNUSE(fp, p);
1911 return (error);
1912 }
1913
1914 /*
1915 * Check access permissions.
1916 */
1917 int
1918 sys_access(l, v, retval)
1919 struct lwp *l;
1920 void *v;
1921 register_t *retval;
1922 {
1923 struct sys_access_args /* {
1924 syscallarg(const char *) path;
1925 syscallarg(int) flags;
1926 } */ *uap = v;
1927 struct proc *p = l->l_proc;
1928 struct ucred *cred = crget();
1929 struct vnode *vp;
1930 int error, flags;
1931 struct nameidata nd;
1932
1933 (void)memcpy(cred, p->p_ucred, sizeof(*cred));
1934 cred->cr_ref = 1;
1935 cred->cr_uid = p->p_cred->p_ruid;
1936 cred->cr_gid = p->p_cred->p_rgid;
1937 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1938 SCARG(uap, path), p);
1939 /* Override default credentials */
1940 nd.ni_cnd.cn_cred = cred;
1941 if ((error = namei(&nd)) != 0)
1942 goto out;
1943 vp = nd.ni_vp;
1944
1945 /* Flags == 0 means only check for existence. */
1946 if (SCARG(uap, flags)) {
1947 flags = 0;
1948 if (SCARG(uap, flags) & R_OK)
1949 flags |= VREAD;
1950 if (SCARG(uap, flags) & W_OK)
1951 flags |= VWRITE;
1952 if (SCARG(uap, flags) & X_OK)
1953 flags |= VEXEC;
1954
1955 error = VOP_ACCESS(vp, flags, cred, p);
1956 if (!error && (flags & VWRITE))
1957 error = vn_writechk(vp);
1958 }
1959 vput(vp);
1960 out:
1961 crfree(cred);
1962 return (error);
1963 }
1964
1965 /*
1966 * Get file status; this version follows links.
1967 */
1968 /* ARGSUSED */
1969 int
1970 sys___stat13(l, v, retval)
1971 struct lwp *l;
1972 void *v;
1973 register_t *retval;
1974 {
1975 struct sys___stat13_args /* {
1976 syscallarg(const char *) path;
1977 syscallarg(struct stat *) ub;
1978 } */ *uap = v;
1979 struct proc *p = l->l_proc;
1980 struct stat sb;
1981 int error;
1982 struct nameidata nd;
1983
1984 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1985 SCARG(uap, path), p);
1986 if ((error = namei(&nd)) != 0)
1987 return (error);
1988 error = vn_stat(nd.ni_vp, &sb, p);
1989 vput(nd.ni_vp);
1990 if (error)
1991 return (error);
1992 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1993 return (error);
1994 }
1995
1996 /*
1997 * Get file status; this version does not follow links.
1998 */
1999 /* ARGSUSED */
2000 int
2001 sys___lstat13(l, v, retval)
2002 struct lwp *l;
2003 void *v;
2004 register_t *retval;
2005 {
2006 struct sys___lstat13_args /* {
2007 syscallarg(const char *) path;
2008 syscallarg(struct stat *) ub;
2009 } */ *uap = v;
2010 struct proc *p = l->l_proc;
2011 struct stat sb;
2012 int error;
2013 struct nameidata nd;
2014
2015 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2016 SCARG(uap, path), p);
2017 if ((error = namei(&nd)) != 0)
2018 return (error);
2019 error = vn_stat(nd.ni_vp, &sb, p);
2020 vput(nd.ni_vp);
2021 if (error)
2022 return (error);
2023 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2024 return (error);
2025 }
2026
2027 /*
2028 * Get configurable pathname variables.
2029 */
2030 /* ARGSUSED */
2031 int
2032 sys_pathconf(l, v, retval)
2033 struct lwp *l;
2034 void *v;
2035 register_t *retval;
2036 {
2037 struct sys_pathconf_args /* {
2038 syscallarg(const char *) path;
2039 syscallarg(int) name;
2040 } */ *uap = v;
2041 struct proc *p = l->l_proc;
2042 int error;
2043 struct nameidata nd;
2044
2045 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2046 SCARG(uap, path), p);
2047 if ((error = namei(&nd)) != 0)
2048 return (error);
2049 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2050 vput(nd.ni_vp);
2051 return (error);
2052 }
2053
2054 /*
2055 * Return target name of a symbolic link.
2056 */
2057 /* ARGSUSED */
2058 int
2059 sys_readlink(l, v, retval)
2060 struct lwp *l;
2061 void *v;
2062 register_t *retval;
2063 {
2064 struct sys_readlink_args /* {
2065 syscallarg(const char *) path;
2066 syscallarg(char *) buf;
2067 syscallarg(size_t) count;
2068 } */ *uap = v;
2069 struct proc *p = l->l_proc;
2070 struct vnode *vp;
2071 struct iovec aiov;
2072 struct uio auio;
2073 int error;
2074 struct nameidata nd;
2075
2076 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2077 SCARG(uap, path), p);
2078 if ((error = namei(&nd)) != 0)
2079 return (error);
2080 vp = nd.ni_vp;
2081 if (vp->v_type != VLNK)
2082 error = EINVAL;
2083 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2084 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2085 aiov.iov_base = SCARG(uap, buf);
2086 aiov.iov_len = SCARG(uap, count);
2087 auio.uio_iov = &aiov;
2088 auio.uio_iovcnt = 1;
2089 auio.uio_offset = 0;
2090 auio.uio_rw = UIO_READ;
2091 auio.uio_segflg = UIO_USERSPACE;
2092 auio.uio_procp = p;
2093 auio.uio_resid = SCARG(uap, count);
2094 error = VOP_READLINK(vp, &auio, p->p_ucred);
2095 }
2096 vput(vp);
2097 *retval = SCARG(uap, count) - auio.uio_resid;
2098 return (error);
2099 }
2100
2101 /*
2102 * Change flags of a file given a path name.
2103 */
2104 /* ARGSUSED */
2105 int
2106 sys_chflags(l, v, retval)
2107 struct lwp *l;
2108 void *v;
2109 register_t *retval;
2110 {
2111 struct sys_chflags_args /* {
2112 syscallarg(const char *) path;
2113 syscallarg(u_long) flags;
2114 } */ *uap = v;
2115 struct proc *p = l->l_proc;
2116 struct vnode *vp;
2117 int error;
2118 struct nameidata nd;
2119
2120 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2121 if ((error = namei(&nd)) != 0)
2122 return (error);
2123 vp = nd.ni_vp;
2124 error = change_flags(vp, SCARG(uap, flags), p);
2125 vput(vp);
2126 return (error);
2127 }
2128
2129 /*
2130 * Change flags of a file given a file descriptor.
2131 */
2132 /* ARGSUSED */
2133 int
2134 sys_fchflags(l, v, retval)
2135 struct lwp *l;
2136 void *v;
2137 register_t *retval;
2138 {
2139 struct sys_fchflags_args /* {
2140 syscallarg(int) fd;
2141 syscallarg(u_long) flags;
2142 } */ *uap = v;
2143 struct proc *p = l->l_proc;
2144 struct vnode *vp;
2145 struct file *fp;
2146 int error;
2147
2148 /* getvnode() will use the descriptor for us */
2149 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2150 return (error);
2151 vp = (struct vnode *)fp->f_data;
2152 error = change_flags(vp, SCARG(uap, flags), p);
2153 VOP_UNLOCK(vp, 0);
2154 FILE_UNUSE(fp, p);
2155 return (error);
2156 }
2157
2158 /*
2159 * Change flags of a file given a path name; this version does
2160 * not follow links.
2161 */
2162 int
2163 sys_lchflags(l, v, retval)
2164 struct lwp *l;
2165 void *v;
2166 register_t *retval;
2167 {
2168 struct sys_lchflags_args /* {
2169 syscallarg(const char *) path;
2170 syscallarg(u_long) flags;
2171 } */ *uap = v;
2172 struct proc *p = l->l_proc;
2173 struct vnode *vp;
2174 int error;
2175 struct nameidata nd;
2176
2177 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2178 if ((error = namei(&nd)) != 0)
2179 return (error);
2180 vp = nd.ni_vp;
2181 error = change_flags(vp, SCARG(uap, flags), p);
2182 vput(vp);
2183 return (error);
2184 }
2185
2186 /*
2187 * Common routine to change flags of a file.
2188 */
2189 int
2190 change_flags(vp, flags, p)
2191 struct vnode *vp;
2192 u_long flags;
2193 struct proc *p;
2194 {
2195 struct vattr vattr;
2196 int error;
2197
2198 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2199 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2200 /*
2201 * Non-superusers cannot change the flags on devices, even if they
2202 * own them.
2203 */
2204 if (suser(p->p_ucred, &p->p_acflag) != 0) {
2205 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2206 goto out;
2207 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2208 error = EINVAL;
2209 goto out;
2210 }
2211 }
2212 VATTR_NULL(&vattr);
2213 vattr.va_flags = flags;
2214 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2215 out:
2216 return (error);
2217 }
2218
2219 /*
2220 * Change mode of a file given path name; this version follows links.
2221 */
2222 /* ARGSUSED */
2223 int
2224 sys_chmod(l, v, retval)
2225 struct lwp *l;
2226 void *v;
2227 register_t *retval;
2228 {
2229 struct sys_chmod_args /* {
2230 syscallarg(const char *) path;
2231 syscallarg(int) mode;
2232 } */ *uap = v;
2233 struct proc *p = l->l_proc;
2234 int error;
2235 struct nameidata nd;
2236
2237 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2238 if ((error = namei(&nd)) != 0)
2239 return (error);
2240
2241 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2242
2243 vrele(nd.ni_vp);
2244 return (error);
2245 }
2246
2247 /*
2248 * Change mode of a file given a file descriptor.
2249 */
2250 /* ARGSUSED */
2251 int
2252 sys_fchmod(l, v, retval)
2253 struct lwp *l;
2254 void *v;
2255 register_t *retval;
2256 {
2257 struct sys_fchmod_args /* {
2258 syscallarg(int) fd;
2259 syscallarg(int) mode;
2260 } */ *uap = v;
2261 struct proc *p = l->l_proc;
2262 struct file *fp;
2263 int error;
2264
2265 /* getvnode() will use the descriptor for us */
2266 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2267 return (error);
2268
2269 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2270 FILE_UNUSE(fp, p);
2271 return (error);
2272 }
2273
2274 /*
2275 * Change mode of a file given path name; this version does not follow links.
2276 */
2277 /* ARGSUSED */
2278 int
2279 sys_lchmod(l, v, retval)
2280 struct lwp *l;
2281 void *v;
2282 register_t *retval;
2283 {
2284 struct sys_lchmod_args /* {
2285 syscallarg(const char *) path;
2286 syscallarg(int) mode;
2287 } */ *uap = v;
2288 struct proc *p = l->l_proc;
2289 int error;
2290 struct nameidata nd;
2291
2292 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2293 if ((error = namei(&nd)) != 0)
2294 return (error);
2295
2296 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2297
2298 vrele(nd.ni_vp);
2299 return (error);
2300 }
2301
2302 /*
2303 * Common routine to set mode given a vnode.
2304 */
2305 static int
2306 change_mode(vp, mode, p)
2307 struct vnode *vp;
2308 int mode;
2309 struct proc *p;
2310 {
2311 struct vattr vattr;
2312 int error;
2313
2314 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2315 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2316 VATTR_NULL(&vattr);
2317 vattr.va_mode = mode & ALLPERMS;
2318 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2319 VOP_UNLOCK(vp, 0);
2320 return (error);
2321 }
2322
2323 /*
2324 * Set ownership given a path name; this version follows links.
2325 */
2326 /* ARGSUSED */
2327 int
2328 sys_chown(l, v, retval)
2329 struct lwp *l;
2330 void *v;
2331 register_t *retval;
2332 {
2333 struct sys_chown_args /* {
2334 syscallarg(const char *) path;
2335 syscallarg(uid_t) uid;
2336 syscallarg(gid_t) gid;
2337 } */ *uap = v;
2338 struct proc *p = l->l_proc;
2339 int error;
2340 struct nameidata nd;
2341
2342 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2343 if ((error = namei(&nd)) != 0)
2344 return (error);
2345
2346 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2347
2348 vrele(nd.ni_vp);
2349 return (error);
2350 }
2351
2352 /*
2353 * Set ownership given a path name; this version follows links.
2354 * Provides POSIX semantics.
2355 */
2356 /* ARGSUSED */
2357 int
2358 sys___posix_chown(l, v, retval)
2359 struct lwp *l;
2360 void *v;
2361 register_t *retval;
2362 {
2363 struct sys_chown_args /* {
2364 syscallarg(const char *) path;
2365 syscallarg(uid_t) uid;
2366 syscallarg(gid_t) gid;
2367 } */ *uap = v;
2368 struct proc *p = l->l_proc;
2369 int error;
2370 struct nameidata nd;
2371
2372 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2373 if ((error = namei(&nd)) != 0)
2374 return (error);
2375
2376 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2377
2378 vrele(nd.ni_vp);
2379 return (error);
2380 }
2381
2382 /*
2383 * Set ownership given a file descriptor.
2384 */
2385 /* ARGSUSED */
2386 int
2387 sys_fchown(l, v, retval)
2388 struct lwp *l;
2389 void *v;
2390 register_t *retval;
2391 {
2392 struct sys_fchown_args /* {
2393 syscallarg(int) fd;
2394 syscallarg(uid_t) uid;
2395 syscallarg(gid_t) gid;
2396 } */ *uap = v;
2397 struct proc *p = l->l_proc;
2398 int error;
2399 struct file *fp;
2400
2401 /* getvnode() will use the descriptor for us */
2402 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2403 return (error);
2404
2405 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2406 SCARG(uap, gid), p, 0);
2407 FILE_UNUSE(fp, p);
2408 return (error);
2409 }
2410
2411 /*
2412 * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2413 */
2414 /* ARGSUSED */
2415 int
2416 sys___posix_fchown(l, v, retval)
2417 struct lwp *l;
2418 void *v;
2419 register_t *retval;
2420 {
2421 struct sys_fchown_args /* {
2422 syscallarg(int) fd;
2423 syscallarg(uid_t) uid;
2424 syscallarg(gid_t) gid;
2425 } */ *uap = v;
2426 struct proc *p = l->l_proc;
2427 int error;
2428 struct file *fp;
2429
2430 /* getvnode() will use the descriptor for us */
2431 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2432 return (error);
2433
2434 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2435 SCARG(uap, gid), p, 1);
2436 FILE_UNUSE(fp, p);
2437 return (error);
2438 }
2439
2440 /*
2441 * Set ownership given a path name; this version does not follow links.
2442 */
2443 /* ARGSUSED */
2444 int
2445 sys_lchown(l, v, retval)
2446 struct lwp *l;
2447 void *v;
2448 register_t *retval;
2449 {
2450 struct sys_lchown_args /* {
2451 syscallarg(const char *) path;
2452 syscallarg(uid_t) uid;
2453 syscallarg(gid_t) gid;
2454 } */ *uap = v;
2455 struct proc *p = l->l_proc;
2456 int error;
2457 struct nameidata nd;
2458
2459 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2460 if ((error = namei(&nd)) != 0)
2461 return (error);
2462
2463 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2464
2465 vrele(nd.ni_vp);
2466 return (error);
2467 }
2468
2469 /*
2470 * Set ownership given a path name; this version does not follow links.
2471 * Provides POSIX/XPG semantics.
2472 */
2473 /* ARGSUSED */
2474 int
2475 sys___posix_lchown(l, v, retval)
2476 struct lwp *l;
2477 void *v;
2478 register_t *retval;
2479 {
2480 struct sys_lchown_args /* {
2481 syscallarg(const char *) path;
2482 syscallarg(uid_t) uid;
2483 syscallarg(gid_t) gid;
2484 } */ *uap = v;
2485 struct proc *p = l->l_proc;
2486 int error;
2487 struct nameidata nd;
2488
2489 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2490 if ((error = namei(&nd)) != 0)
2491 return (error);
2492
2493 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2494
2495 vrele(nd.ni_vp);
2496 return (error);
2497 }
2498
2499 /*
2500 * Common routine to set ownership given a vnode.
2501 */
2502 static int
2503 change_owner(vp, uid, gid, p, posix_semantics)
2504 struct vnode *vp;
2505 uid_t uid;
2506 gid_t gid;
2507 struct proc *p;
2508 int posix_semantics;
2509 {
2510 struct vattr vattr;
2511 mode_t newmode;
2512 int error;
2513
2514 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2515 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2516 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2517 goto out;
2518
2519 #define CHANGED(x) ((x) != -1)
2520 newmode = vattr.va_mode;
2521 if (posix_semantics) {
2522 /*
2523 * POSIX/XPG semantics: if the caller is not the super-user,
2524 * clear set-user-id and set-group-id bits. Both POSIX and
2525 * the XPG consider the behaviour for calls by the super-user
2526 * implementation-defined; we leave the set-user-id and set-
2527 * group-id settings intact in that case.
2528 */
2529 if (suser(p->p_ucred, NULL) != 0)
2530 newmode &= ~(S_ISUID | S_ISGID);
2531 } else {
2532 /*
2533 * NetBSD semantics: when changing owner and/or group,
2534 * clear the respective bit(s).
2535 */
2536 if (CHANGED(uid))
2537 newmode &= ~S_ISUID;
2538 if (CHANGED(gid))
2539 newmode &= ~S_ISGID;
2540 }
2541 /* Update va_mode iff altered. */
2542 if (vattr.va_mode == newmode)
2543 newmode = VNOVAL;
2544
2545 VATTR_NULL(&vattr);
2546 vattr.va_uid = CHANGED(uid) ? uid : VNOVAL;
2547 vattr.va_gid = CHANGED(gid) ? gid : VNOVAL;
2548 vattr.va_mode = newmode;
2549 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2550 #undef CHANGED
2551
2552 out:
2553 VOP_UNLOCK(vp, 0);
2554 return (error);
2555 }
2556
2557 /*
2558 * Set the access and modification times given a path name; this
2559 * version follows links.
2560 */
2561 /* ARGSUSED */
2562 int
2563 sys_utimes(l, v, retval)
2564 struct lwp *l;
2565 void *v;
2566 register_t *retval;
2567 {
2568 struct sys_utimes_args /* {
2569 syscallarg(const char *) path;
2570 syscallarg(const struct timeval *) tptr;
2571 } */ *uap = v;
2572 struct proc *p = l->l_proc;
2573 int error;
2574 struct nameidata nd;
2575
2576 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2577 if ((error = namei(&nd)) != 0)
2578 return (error);
2579
2580 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2581
2582 vrele(nd.ni_vp);
2583 return (error);
2584 }
2585
2586 /*
2587 * Set the access and modification times given a file descriptor.
2588 */
2589 /* ARGSUSED */
2590 int
2591 sys_futimes(l, v, retval)
2592 struct lwp *l;
2593 void *v;
2594 register_t *retval;
2595 {
2596 struct sys_futimes_args /* {
2597 syscallarg(int) fd;
2598 syscallarg(const struct timeval *) tptr;
2599 } */ *uap = v;
2600 struct proc *p = l->l_proc;
2601 int error;
2602 struct file *fp;
2603
2604 /* getvnode() will use the descriptor for us */
2605 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2606 return (error);
2607
2608 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2609 FILE_UNUSE(fp, p);
2610 return (error);
2611 }
2612
2613 /*
2614 * Set the access and modification times given a path name; this
2615 * version does not follow links.
2616 */
2617 /* ARGSUSED */
2618 int
2619 sys_lutimes(l, v, retval)
2620 struct lwp *l;
2621 void *v;
2622 register_t *retval;
2623 {
2624 struct sys_lutimes_args /* {
2625 syscallarg(const char *) path;
2626 syscallarg(const struct timeval *) tptr;
2627 } */ *uap = v;
2628 struct proc *p = l->l_proc;
2629 int error;
2630 struct nameidata nd;
2631
2632 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2633 if ((error = namei(&nd)) != 0)
2634 return (error);
2635
2636 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2637
2638 vrele(nd.ni_vp);
2639 return (error);
2640 }
2641
2642 /*
2643 * Common routine to set access and modification times given a vnode.
2644 */
2645 static int
2646 change_utimes(vp, tptr, p)
2647 struct vnode *vp;
2648 const struct timeval *tptr;
2649 struct proc *p;
2650 {
2651 struct timeval tv[2];
2652 struct vattr vattr;
2653 int error;
2654
2655 VATTR_NULL(&vattr);
2656 if (tptr == NULL) {
2657 microtime(&tv[0]);
2658 tv[1] = tv[0];
2659 vattr.va_vaflags |= VA_UTIMES_NULL;
2660 } else {
2661 error = copyin(tptr, tv, sizeof(tv));
2662 if (error)
2663 return (error);
2664 }
2665 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2666 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2667 vattr.va_atime.tv_sec = tv[0].tv_sec;
2668 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2669 vattr.va_mtime.tv_sec = tv[1].tv_sec;
2670 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2671 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2672 VOP_UNLOCK(vp, 0);
2673 return (error);
2674 }
2675
2676 /*
2677 * Truncate a file given its path name.
2678 */
2679 /* ARGSUSED */
2680 int
2681 sys_truncate(l, v, retval)
2682 struct lwp *l;
2683 void *v;
2684 register_t *retval;
2685 {
2686 struct sys_truncate_args /* {
2687 syscallarg(const char *) path;
2688 syscallarg(int) pad;
2689 syscallarg(off_t) length;
2690 } */ *uap = v;
2691 struct proc *p = l->l_proc;
2692 struct vnode *vp;
2693 struct vattr vattr;
2694 int error;
2695 struct nameidata nd;
2696
2697 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2698 if ((error = namei(&nd)) != 0)
2699 return (error);
2700 vp = nd.ni_vp;
2701 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2702 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2703 if (vp->v_type == VDIR)
2704 error = EISDIR;
2705 else if ((error = vn_writechk(vp)) == 0 &&
2706 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2707 VATTR_NULL(&vattr);
2708 vattr.va_size = SCARG(uap, length);
2709 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2710 }
2711 vput(vp);
2712 return (error);
2713 }
2714
2715 /*
2716 * Truncate a file given a file descriptor.
2717 */
2718 /* ARGSUSED */
2719 int
2720 sys_ftruncate(l, v, retval)
2721 struct lwp *l;
2722 void *v;
2723 register_t *retval;
2724 {
2725 struct sys_ftruncate_args /* {
2726 syscallarg(int) fd;
2727 syscallarg(int) pad;
2728 syscallarg(off_t) length;
2729 } */ *uap = v;
2730 struct proc *p = l->l_proc;
2731 struct vattr vattr;
2732 struct vnode *vp;
2733 struct file *fp;
2734 int error;
2735
2736 /* getvnode() will use the descriptor for us */
2737 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2738 return (error);
2739 if ((fp->f_flag & FWRITE) == 0) {
2740 error = EINVAL;
2741 goto out;
2742 }
2743 vp = (struct vnode *)fp->f_data;
2744 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2745 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2746 if (vp->v_type == VDIR)
2747 error = EISDIR;
2748 else if ((error = vn_writechk(vp)) == 0) {
2749 VATTR_NULL(&vattr);
2750 vattr.va_size = SCARG(uap, length);
2751 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2752 }
2753 VOP_UNLOCK(vp, 0);
2754 out:
2755 FILE_UNUSE(fp, p);
2756 return (error);
2757 }
2758
2759 /*
2760 * Sync an open file.
2761 */
2762 /* ARGSUSED */
2763 int
2764 sys_fsync(l, v, retval)
2765 struct lwp *l;
2766 void *v;
2767 register_t *retval;
2768 {
2769 struct sys_fsync_args /* {
2770 syscallarg(int) fd;
2771 } */ *uap = v;
2772 struct proc *p = l->l_proc;
2773 struct vnode *vp;
2774 struct file *fp;
2775 int error;
2776
2777 /* getvnode() will use the descriptor for us */
2778 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2779 return (error);
2780 vp = (struct vnode *)fp->f_data;
2781 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2782 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
2783 if (error == 0 && bioops.io_fsync != NULL &&
2784 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2785 (*bioops.io_fsync)(vp);
2786 VOP_UNLOCK(vp, 0);
2787 FILE_UNUSE(fp, p);
2788 return (error);
2789 }
2790
2791 /*
2792 * Sync the data of an open file.
2793 */
2794 /* ARGSUSED */
2795 int
2796 sys_fdatasync(l, v, retval)
2797 struct lwp *l;
2798 void *v;
2799 register_t *retval;
2800 {
2801 struct sys_fdatasync_args /* {
2802 syscallarg(int) fd;
2803 } */ *uap = v;
2804 struct proc *p = l->l_proc;
2805 struct vnode *vp;
2806 struct file *fp;
2807 int error;
2808
2809 /* getvnode() will use the descriptor for us */
2810 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2811 return (error);
2812 vp = (struct vnode *)fp->f_data;
2813 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2814 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
2815 VOP_UNLOCK(vp, 0);
2816 FILE_UNUSE(fp, p);
2817 return (error);
2818 }
2819
2820 /*
2821 * Rename files, (standard) BSD semantics frontend.
2822 */
2823 /* ARGSUSED */
2824 int
2825 sys_rename(l, v, retval)
2826 struct lwp *l;
2827 void *v;
2828 register_t *retval;
2829 {
2830 struct sys_rename_args /* {
2831 syscallarg(const char *) from;
2832 syscallarg(const char *) to;
2833 } */ *uap = v;
2834 struct proc *p = l->l_proc;
2835
2836 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
2837 }
2838
2839 /*
2840 * Rename files, POSIX semantics frontend.
2841 */
2842 /* ARGSUSED */
2843 int
2844 sys___posix_rename(l, v, retval)
2845 struct lwp *l;
2846 void *v;
2847 register_t *retval;
2848 {
2849 struct sys___posix_rename_args /* {
2850 syscallarg(const char *) from;
2851 syscallarg(const char *) to;
2852 } */ *uap = v;
2853 struct proc *p = l->l_proc;
2854
2855 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
2856 }
2857
2858 /*
2859 * Rename files. Source and destination must either both be directories,
2860 * or both not be directories. If target is a directory, it must be empty.
2861 * If `from' and `to' refer to the same object, the value of the `retain'
2862 * argument is used to determine whether `from' will be
2863 *
2864 * (retain == 0) deleted unless `from' and `to' refer to the same
2865 * object in the file system's name space (BSD).
2866 * (retain == 1) always retained (POSIX).
2867 */
2868 static int
2869 rename_files(from, to, p, retain)
2870 const char *from, *to;
2871 struct proc *p;
2872 int retain;
2873 {
2874 struct vnode *tvp, *fvp, *tdvp;
2875 struct nameidata fromnd, tond;
2876 int error;
2877
2878 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2879 from, p);
2880 if ((error = namei(&fromnd)) != 0)
2881 return (error);
2882 fvp = fromnd.ni_vp;
2883 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
2884 UIO_USERSPACE, to, p);
2885 if ((error = namei(&tond)) != 0) {
2886 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2887 vrele(fromnd.ni_dvp);
2888 vrele(fvp);
2889 goto out1;
2890 }
2891 tdvp = tond.ni_dvp;
2892 tvp = tond.ni_vp;
2893
2894 if (tvp != NULL) {
2895 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2896 error = ENOTDIR;
2897 goto out;
2898 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2899 error = EISDIR;
2900 goto out;
2901 }
2902 }
2903
2904 if (fvp == tdvp)
2905 error = EINVAL;
2906
2907 /*
2908 * Source and destination refer to the same object.
2909 */
2910 if (fvp == tvp) {
2911 if (retain)
2912 error = -1;
2913 else if (fromnd.ni_dvp == tdvp &&
2914 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2915 !memcmp(fromnd.ni_cnd.cn_nameptr,
2916 tond.ni_cnd.cn_nameptr,
2917 fromnd.ni_cnd.cn_namelen))
2918 error = -1;
2919 }
2920
2921 out:
2922 if (!error) {
2923 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2924 if (fromnd.ni_dvp != tdvp)
2925 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2926 if (tvp) {
2927 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2928 }
2929 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2930 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2931 } else {
2932 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2933 if (tdvp == tvp)
2934 vrele(tdvp);
2935 else
2936 vput(tdvp);
2937 if (tvp)
2938 vput(tvp);
2939 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2940 vrele(fromnd.ni_dvp);
2941 vrele(fvp);
2942 }
2943 vrele(tond.ni_startdir);
2944 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2945 out1:
2946 if (fromnd.ni_startdir)
2947 vrele(fromnd.ni_startdir);
2948 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2949 return (error == -1 ? 0 : error);
2950 }
2951
2952 /*
2953 * Make a directory file.
2954 */
2955 /* ARGSUSED */
2956 int
2957 sys_mkdir(l, v, retval)
2958 struct lwp *l;
2959 void *v;
2960 register_t *retval;
2961 {
2962 struct sys_mkdir_args /* {
2963 syscallarg(const char *) path;
2964 syscallarg(int) mode;
2965 } */ *uap = v;
2966 struct proc *p = l->l_proc;
2967 struct vnode *vp;
2968 struct vattr vattr;
2969 int error;
2970 struct nameidata nd;
2971
2972 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
2973 if ((error = namei(&nd)) != 0)
2974 return (error);
2975 vp = nd.ni_vp;
2976 if (vp != NULL) {
2977 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2978 if (nd.ni_dvp == vp)
2979 vrele(nd.ni_dvp);
2980 else
2981 vput(nd.ni_dvp);
2982 vrele(vp);
2983 return (EEXIST);
2984 }
2985 VATTR_NULL(&vattr);
2986 vattr.va_type = VDIR;
2987 vattr.va_mode =
2988 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
2989 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2990 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2991 if (!error)
2992 vput(nd.ni_vp);
2993 return (error);
2994 }
2995
2996 /*
2997 * Remove a directory file.
2998 */
2999 /* ARGSUSED */
3000 int
3001 sys_rmdir(l, v, retval)
3002 struct lwp *l;
3003 void *v;
3004 register_t *retval;
3005 {
3006 struct sys_rmdir_args /* {
3007 syscallarg(const char *) path;
3008 } */ *uap = v;
3009 struct proc *p = l->l_proc;
3010 struct vnode *vp;
3011 int error;
3012 struct nameidata nd;
3013
3014 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3015 SCARG(uap, path), p);
3016 if ((error = namei(&nd)) != 0)
3017 return (error);
3018 vp = nd.ni_vp;
3019 if (vp->v_type != VDIR) {
3020 error = ENOTDIR;
3021 goto out;
3022 }
3023 /*
3024 * No rmdir "." please.
3025 */
3026 if (nd.ni_dvp == vp) {
3027 error = EINVAL;
3028 goto out;
3029 }
3030 /*
3031 * The root of a mounted filesystem cannot be deleted.
3032 */
3033 if (vp->v_flag & VROOT)
3034 error = EBUSY;
3035 out:
3036 if (!error) {
3037 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3038 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3039 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3040 } else {
3041 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3042 if (nd.ni_dvp == vp)
3043 vrele(nd.ni_dvp);
3044 else
3045 vput(nd.ni_dvp);
3046 vput(vp);
3047 }
3048 return (error);
3049 }
3050
3051 /*
3052 * Read a block of directory entries in a file system independent format.
3053 */
3054 int
3055 sys_getdents(l, v, retval)
3056 struct lwp *l;
3057 void *v;
3058 register_t *retval;
3059 {
3060 struct sys_getdents_args /* {
3061 syscallarg(int) fd;
3062 syscallarg(char *) buf;
3063 syscallarg(size_t) count;
3064 } */ *uap = v;
3065 struct proc *p = l->l_proc;
3066 struct file *fp;
3067 int error, done;
3068
3069 /* getvnode() will use the descriptor for us */
3070 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3071 return (error);
3072 if ((fp->f_flag & FREAD) == 0) {
3073 error = EBADF;
3074 goto out;
3075 }
3076 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3077 SCARG(uap, count), &done, p, 0, 0);
3078 *retval = done;
3079 out:
3080 FILE_UNUSE(fp, p);
3081 return (error);
3082 }
3083
3084 /*
3085 * Set the mode mask for creation of filesystem nodes.
3086 */
3087 int
3088 sys_umask(l, v, retval)
3089 struct lwp *l;
3090 void *v;
3091 register_t *retval;
3092 {
3093 struct sys_umask_args /* {
3094 syscallarg(mode_t) newmask;
3095 } */ *uap = v;
3096 struct proc *p = l->l_proc;
3097 struct cwdinfo *cwdi;
3098
3099 cwdi = p->p_cwdi;
3100 *retval = cwdi->cwdi_cmask;
3101 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3102 return (0);
3103 }
3104
3105 /*
3106 * Void all references to file by ripping underlying filesystem
3107 * away from vnode.
3108 */
3109 /* ARGSUSED */
3110 int
3111 sys_revoke(l, v, retval)
3112 struct lwp *l;
3113 void *v;
3114 register_t *retval;
3115 {
3116 struct sys_revoke_args /* {
3117 syscallarg(const char *) path;
3118 } */ *uap = v;
3119 struct proc *p = l->l_proc;
3120 struct vnode *vp;
3121 struct vattr vattr;
3122 int error;
3123 struct nameidata nd;
3124
3125 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3126 if ((error = namei(&nd)) != 0)
3127 return (error);
3128 vp = nd.ni_vp;
3129 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3130 goto out;
3131 if (p->p_ucred->cr_uid != vattr.va_uid &&
3132 (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3133 goto out;
3134 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3135 VOP_REVOKE(vp, REVOKEALL);
3136 out:
3137 vrele(vp);
3138 return (error);
3139 }
3140
3141 /*
3142 * Convert a user file descriptor to a kernel file entry.
3143 */
3144 int
3145 getvnode(fdp, fd, fpp)
3146 struct filedesc *fdp;
3147 int fd;
3148 struct file **fpp;
3149 {
3150 struct vnode *vp;
3151 struct file *fp;
3152
3153 if ((fp = fd_getfile(fdp, fd)) == NULL)
3154 return (EBADF);
3155
3156 FILE_USE(fp);
3157
3158 if (fp->f_type != DTYPE_VNODE) {
3159 FILE_UNUSE(fp, NULL);
3160 return (EINVAL);
3161 }
3162
3163 vp = (struct vnode *)fp->f_data;
3164 if (vp->v_type == VBAD) {
3165 FILE_UNUSE(fp, NULL);
3166 return (EBADF);
3167 }
3168
3169 *fpp = fp;
3170 return (0);
3171 }
3172