vfs_syscalls.c revision 1.162 1 /* $NetBSD: vfs_syscalls.c,v 1.162 2000/09/19 22:01:41 fvdl 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 "opt_compat_netbsd.h"
44 #include "opt_compat_43.h"
45
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/namei.h>
49 #include <sys/filedesc.h>
50 #include <sys/kernel.h>
51 #include <sys/file.h>
52 #include <sys/stat.h>
53 #include <sys/vnode.h>
54 #include <sys/mount.h>
55 #include <sys/proc.h>
56 #include <sys/uio.h>
57 #include <sys/malloc.h>
58 #include <sys/dirent.h>
59 #include <sys/syscallargs.h>
60
61 #include <miscfs/genfs/genfs.h>
62 #include <miscfs/syncfs/syncfs.h>
63
64 #include <uvm/uvm_extern.h>
65
66 #include <sys/sysctl.h>
67
68 static int change_dir __P((struct nameidata *, struct proc *));
69 static int change_mode __P((struct vnode *, int, struct proc *p));
70 static int change_owner __P((struct vnode *, uid_t, gid_t, struct proc *,
71 int));
72 static int change_utimes __P((struct vnode *vp, const struct timeval *,
73 struct proc *p));
74 static int rename_files __P((const char *, const char *, struct proc *, int));
75
76 void checkdirs __P((struct vnode *));
77
78 int dovfsusermount = 0;
79
80 /*
81 * Virtual File System System Calls
82 */
83
84 /*
85 * Mount a file system.
86 */
87
88 /*
89 * This table is used to maintain compatibility with 4.3BSD
90 * and NetBSD 0.9 mount syscalls. Note, the order is important!
91 *
92 * Also note that not all of these had actual numbers in 4.3BSD
93 * or NetBSD 0.9!
94 */
95 const char *mountcompatnames[] = {
96 NULL, /* 0 = MOUNT_NONE */
97 MOUNT_FFS, /* 1 */
98 MOUNT_NFS, /* 2 */
99 MOUNT_MFS, /* 3 */
100 MOUNT_MSDOS, /* 4 */
101 MOUNT_LFS, /* 5 */
102 NULL, /* 6 = MOUNT_LOFS */
103 MOUNT_FDESC, /* 7 */
104 MOUNT_PORTAL, /* 8 */
105 MOUNT_NULL, /* 9 */
106 MOUNT_UMAP, /* 10 */
107 MOUNT_KERNFS, /* 11 */
108 MOUNT_PROCFS, /* 12 */
109 MOUNT_AFS, /* 13 */
110 MOUNT_CD9660, /* 14 = MOUNT_ISOFS */
111 MOUNT_UNION, /* 15 */
112 MOUNT_ADOSFS, /* 16 */
113 MOUNT_EXT2FS, /* 17 */
114 MOUNT_CODA, /* 18 */
115 MOUNT_FILECORE, /* 19 */
116 MOUNT_NTFS, /* 20 */
117 };
118 const int nmountcompatnames = sizeof(mountcompatnames) /
119 sizeof(mountcompatnames[0]);
120
121 /* ARGSUSED */
122 int
123 sys_mount(p, v, retval)
124 struct proc *p;
125 void *v;
126 register_t *retval;
127 {
128 struct sys_mount_args /* {
129 syscallarg(const char *) type;
130 syscallarg(const char *) path;
131 syscallarg(int) flags;
132 syscallarg(void *) data;
133 } */ *uap = v;
134 struct vnode *vp;
135 struct mount *mp;
136 int error, flag = 0;
137 char fstypename[MFSNAMELEN];
138 struct vattr va;
139 struct nameidata nd;
140 struct vfsops *vfs;
141
142 if (dovfsusermount == 0 && (error = suser(p->p_ucred, &p->p_acflag)))
143 return (error);
144 /*
145 * Get vnode to be covered
146 */
147 NDINIT(&nd, LOOKUP, FOLLOW , UIO_USERSPACE,
148 SCARG(uap, path), p);
149 if ((error = namei(&nd)) != 0)
150 return (error);
151 vp = nd.ni_vp;
152 /*
153 * A lookup in VFS_MOUNT might result in an attempt to
154 * lock this vnode again, so make the lock resursive.
155 */
156 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
157 if (SCARG(uap, flags) & MNT_UPDATE) {
158 if ((vp->v_flag & VROOT) == 0) {
159 vput(vp);
160 return (EINVAL);
161 }
162 mp = vp->v_mount;
163 flag = mp->mnt_flag;
164 vfs = mp->mnt_op;
165 /*
166 * We only allow the filesystem to be reloaded if it
167 * is currently mounted read-only.
168 */
169 if ((SCARG(uap, flags) & MNT_RELOAD) &&
170 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
171 vput(vp);
172 return (EOPNOTSUPP); /* Needs translation */
173 }
174 /*
175 * In "highly secure" mode, don't let the caller do anything
176 * but downgrade a filesystem from read-write to read-only.
177 * (see also below; MNT_UPDATE is required.)
178 */
179 if (securelevel >= 2 &&
180 (SCARG(uap, flags) !=
181 (mp->mnt_flag | MNT_RDONLY |
182 MNT_RELOAD | MNT_FORCE | MNT_UPDATE))) {
183 vput(vp);
184 return (EPERM);
185 }
186 mp->mnt_flag |=
187 SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
188 /*
189 * Only root, or the user that did the original mount is
190 * permitted to update it.
191 */
192 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
193 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
194 vput(vp);
195 return (error);
196 }
197 /*
198 * Do not allow NFS export by non-root users. For non-root
199 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
200 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
201 */
202 if (p->p_ucred->cr_uid != 0) {
203 if (SCARG(uap, flags) & MNT_EXPORTED) {
204 vput(vp);
205 return (EPERM);
206 }
207 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
208 if (flag & MNT_NOEXEC)
209 SCARG(uap, flags) |= MNT_NOEXEC;
210 }
211 if (vfs_busy(mp, LK_NOWAIT, 0)) {
212 vput(vp);
213 return (EPERM);
214 }
215 VOP_UNLOCK(vp, 0);
216 goto update;
217 } else {
218 if (securelevel >= 2) {
219 vput(vp);
220 return (EPERM);
221 }
222 }
223 /*
224 * If the user is not root, ensure that they own the directory
225 * onto which we are attempting to mount.
226 */
227 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
228 (va.va_uid != p->p_ucred->cr_uid &&
229 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
230 vput(vp);
231 return (error);
232 }
233 /*
234 * Do not allow NFS export by non-root users. For non-root users,
235 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
236 * mount point is already MNT_NOEXEC.
237 */
238 if (p->p_ucred->cr_uid != 0) {
239 if (SCARG(uap, flags) & MNT_EXPORTED) {
240 vput(vp);
241 return (EPERM);
242 }
243 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
244 if (vp->v_mount->mnt_flag & MNT_NOEXEC)
245 SCARG(uap, flags) |= MNT_NOEXEC;
246 }
247 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
248 return (error);
249 if (vp->v_type != VDIR) {
250 vput(vp);
251 return (ENOTDIR);
252 }
253 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
254 if (error) {
255 #if defined(COMPAT_09) || defined(COMPAT_43)
256 /*
257 * Historically filesystem types were identified by number.
258 * If we get an integer for the filesystem type instead of a
259 * string, we check to see if it matches one of the historic
260 * filesystem types.
261 */
262 u_long fsindex = (u_long)SCARG(uap, type);
263 if (fsindex >= nmountcompatnames ||
264 mountcompatnames[fsindex] == NULL) {
265 vput(vp);
266 return (ENODEV);
267 }
268 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
269 #else
270 vput(vp);
271 return (error);
272 #endif
273 }
274 #ifdef COMPAT_10
275 /* Accept `ufs' as an alias for `ffs'. */
276 if (!strncmp(fstypename, "ufs", MFSNAMELEN))
277 strncpy(fstypename, "ffs", MFSNAMELEN);
278 #endif
279 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
280 vput(vp);
281 return (ENODEV);
282 }
283 if (vp->v_mountedhere != NULL) {
284 vput(vp);
285 return (EBUSY);
286 }
287
288 /*
289 * Allocate and initialize the file system.
290 */
291 mp = (struct mount *)malloc((u_long)sizeof(struct mount),
292 M_MOUNT, M_WAITOK);
293 memset((char *)mp, 0, (u_long)sizeof(struct mount));
294 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
295 (void)vfs_busy(mp, LK_NOWAIT, 0);
296 mp->mnt_op = vfs;
297 vfs->vfs_refcount++;
298 mp->mnt_vnodecovered = vp;
299 mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
300 mp->mnt_unmounter = NULL;
301 update:
302 /*
303 * Set the mount level flags.
304 */
305 if (SCARG(uap, flags) & MNT_RDONLY)
306 mp->mnt_flag |= MNT_RDONLY;
307 else if (mp->mnt_flag & MNT_RDONLY)
308 mp->mnt_flag |= MNT_WANTRDWR;
309 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
310 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
311 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
312 mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
313 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
314 MNT_NOCOREDUMP | MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM |
315 MNT_SOFTDEP);
316 /*
317 * Mount the filesystem.
318 */
319 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
320 if (mp->mnt_flag & MNT_UPDATE) {
321 vrele(vp);
322 if (mp->mnt_flag & MNT_WANTRDWR)
323 mp->mnt_flag &= ~MNT_RDONLY;
324 mp->mnt_flag &=~
325 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR);
326 if (error)
327 mp->mnt_flag = flag;
328 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
329 if (mp->mnt_syncer == NULL)
330 error = vfs_allocate_syncvnode(mp);
331 } else {
332 if (mp->mnt_syncer != NULL)
333 vfs_deallocate_syncvnode(mp);
334 }
335 vfs_unbusy(mp);
336 return (error);
337 }
338 /*
339 * Put the new filesystem on the mount list after root.
340 */
341 cache_purge(vp);
342 if (!error) {
343 vp->v_mountedhere = mp;
344 simple_lock(&mountlist_slock);
345 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
346 simple_unlock(&mountlist_slock);
347 checkdirs(vp);
348 VOP_UNLOCK(vp, 0);
349 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
350 error = vfs_allocate_syncvnode(mp);
351 vfs_unbusy(mp);
352 (void) VFS_STATFS(mp, &mp->mnt_stat, p);
353 if ((error = VFS_START(mp, 0, p)))
354 vrele(vp);
355 } else {
356 vp->v_mountedhere = (struct mount *)0;
357 vfs->vfs_refcount--;
358 vfs_unbusy(mp);
359 free((caddr_t)mp, M_MOUNT);
360 vput(vp);
361 }
362 return (error);
363 }
364
365 /*
366 * Scan all active processes to see if any of them have a current
367 * or root directory onto which the new filesystem has just been
368 * mounted. If so, replace them with the new mount point.
369 */
370 void
371 checkdirs(olddp)
372 struct vnode *olddp;
373 {
374 struct cwdinfo *cwdi;
375 struct vnode *newdp;
376 struct proc *p;
377
378 if (olddp->v_usecount == 1)
379 return;
380 if (VFS_ROOT(olddp->v_mountedhere, &newdp))
381 panic("mount: lost mount");
382 proclist_lock_read();
383 for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
384 cwdi = p->p_cwdi;
385 if (cwdi->cwdi_cdir == olddp) {
386 vrele(cwdi->cwdi_cdir);
387 VREF(newdp);
388 cwdi->cwdi_cdir = newdp;
389 }
390 if (cwdi->cwdi_rdir == olddp) {
391 vrele(cwdi->cwdi_rdir);
392 VREF(newdp);
393 cwdi->cwdi_rdir = newdp;
394 }
395 }
396 proclist_unlock_read();
397 if (rootvnode == olddp) {
398 vrele(rootvnode);
399 VREF(newdp);
400 rootvnode = newdp;
401 }
402 vput(newdp);
403 }
404
405 /*
406 * Unmount a file system.
407 *
408 * Note: unmount takes a path to the vnode mounted on as argument,
409 * not special file (as before).
410 */
411 /* ARGSUSED */
412 int
413 sys_unmount(p, v, retval)
414 struct proc *p;
415 void *v;
416 register_t *retval;
417 {
418 struct sys_unmount_args /* {
419 syscallarg(const char *) path;
420 syscallarg(int) flags;
421 } */ *uap = v;
422 struct vnode *vp;
423 struct mount *mp;
424 int error;
425 struct nameidata nd;
426
427 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
428 SCARG(uap, path), p);
429 if ((error = namei(&nd)) != 0)
430 return (error);
431 vp = nd.ni_vp;
432 mp = vp->v_mount;
433
434 /*
435 * Only root, or the user that did the original mount is
436 * permitted to unmount this filesystem.
437 */
438 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
439 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
440 vput(vp);
441 return (error);
442 }
443
444 /*
445 * Don't allow unmounting the root file system.
446 */
447 if (mp->mnt_flag & MNT_ROOTFS) {
448 vput(vp);
449 return (EINVAL);
450 }
451
452 /*
453 * Must be the root of the filesystem
454 */
455 if ((vp->v_flag & VROOT) == 0) {
456 vput(vp);
457 return (EINVAL);
458 }
459 vput(vp);
460
461 if (vfs_busy(mp, 0, 0))
462 return (EBUSY);
463
464 return (dounmount(mp, SCARG(uap, flags), p));
465 }
466
467 /*
468 * Do the actual file system unmount. File system is assumed to have been
469 * marked busy by the caller.
470 */
471 int
472 dounmount(mp, flags, p)
473 struct mount *mp;
474 int flags;
475 struct proc *p;
476 {
477 struct vnode *coveredvp;
478 int error;
479 int async;
480 int used_syncer;
481
482 simple_lock(&mountlist_slock);
483 vfs_unbusy(mp);
484 used_syncer = (mp->mnt_syncer != NULL);
485
486 /*
487 * XXX Freeze syncer. This should really be done on a mountpoint
488 * basis, but especially the softdep code possibly called from
489 * the syncer doesn't exactly work on a per-mountpoint basis,
490 * so the softdep code would become a maze of vfs_busy calls.
491 */
492 if (used_syncer)
493 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
494
495 mp->mnt_flag |= MNT_UNMOUNT;
496 mp->mnt_unmounter = p;
497 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
498 if (mp->mnt_flag & MNT_EXPUBLIC)
499 vfs_setpublicfs(NULL, NULL, NULL);
500 async = mp->mnt_flag & MNT_ASYNC;
501 mp->mnt_flag &= ~MNT_ASYNC;
502 cache_purgevfs(mp); /* remove cache entries for this file sys */
503 if (mp->mnt_syncer != NULL)
504 vfs_deallocate_syncvnode(mp);
505 if (((mp->mnt_flag & MNT_RDONLY) ||
506 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
507 (flags & MNT_FORCE))
508 error = VFS_UNMOUNT(mp, flags, p);
509 simple_lock(&mountlist_slock);
510 if (error) {
511 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
512 (void) vfs_allocate_syncvnode(mp);
513 mp->mnt_flag &= ~MNT_UNMOUNT;
514 mp->mnt_unmounter = NULL;
515 mp->mnt_flag |= async;
516 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
517 &mountlist_slock);
518 if (used_syncer)
519 lockmgr(&syncer_lock, LK_RELEASE, NULL);
520 while (mp->mnt_wcnt > 0) {
521 wakeup((caddr_t)mp);
522 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0);
523 }
524 return (error);
525 }
526 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
527 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
528 coveredvp->v_mountedhere = NULL;
529 vrele(coveredvp);
530 }
531 mp->mnt_op->vfs_refcount--;
532 if (mp->mnt_vnodelist.lh_first != NULL)
533 panic("unmount: dangling vnode");
534 mp->mnt_flag |= MNT_GONE;
535 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
536 if (used_syncer)
537 lockmgr(&syncer_lock, LK_RELEASE, NULL);
538 while(mp->mnt_wcnt > 0) {
539 wakeup((caddr_t)mp);
540 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0);
541 }
542 free((caddr_t)mp, M_MOUNT);
543 return (0);
544 }
545
546 /*
547 * Sync each mounted filesystem.
548 */
549 #ifdef DEBUG
550 int syncprt = 0;
551 struct ctldebug debug0 = { "syncprt", &syncprt };
552 #endif
553
554 /* ARGSUSED */
555 int
556 sys_sync(p, v, retval)
557 struct proc *p;
558 void *v;
559 register_t *retval;
560 {
561 struct mount *mp, *nmp;
562 int asyncflag;
563
564 simple_lock(&mountlist_slock);
565 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
566 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
567 nmp = mp->mnt_list.cqe_prev;
568 continue;
569 }
570 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
571 asyncflag = mp->mnt_flag & MNT_ASYNC;
572 mp->mnt_flag &= ~MNT_ASYNC;
573 uvm_vnp_sync(mp);
574 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
575 if (asyncflag)
576 mp->mnt_flag |= MNT_ASYNC;
577 }
578 simple_lock(&mountlist_slock);
579 nmp = mp->mnt_list.cqe_prev;
580 vfs_unbusy(mp);
581
582 }
583 simple_unlock(&mountlist_slock);
584 #ifdef DEBUG
585 if (syncprt)
586 vfs_bufstats();
587 #endif /* DEBUG */
588 return (0);
589 }
590
591 /*
592 * Change filesystem quotas.
593 */
594 /* ARGSUSED */
595 int
596 sys_quotactl(p, v, retval)
597 struct proc *p;
598 void *v;
599 register_t *retval;
600 {
601 struct sys_quotactl_args /* {
602 syscallarg(const char *) path;
603 syscallarg(int) cmd;
604 syscallarg(int) uid;
605 syscallarg(caddr_t) arg;
606 } */ *uap = v;
607 struct mount *mp;
608 int error;
609 struct nameidata nd;
610
611 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
612 if ((error = namei(&nd)) != 0)
613 return (error);
614 mp = nd.ni_vp->v_mount;
615 vrele(nd.ni_vp);
616 return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
617 SCARG(uap, arg), p));
618 }
619
620 /*
621 * Get filesystem statistics.
622 */
623 /* ARGSUSED */
624 int
625 sys_statfs(p, v, retval)
626 struct proc *p;
627 void *v;
628 register_t *retval;
629 {
630 struct sys_statfs_args /* {
631 syscallarg(const char *) path;
632 syscallarg(struct statfs *) buf;
633 } */ *uap = v;
634 struct mount *mp;
635 struct statfs *sp;
636 int error;
637 struct nameidata nd;
638
639 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
640 if ((error = namei(&nd)) != 0)
641 return (error);
642 mp = nd.ni_vp->v_mount;
643 sp = &mp->mnt_stat;
644 vrele(nd.ni_vp);
645 if ((error = VFS_STATFS(mp, sp, p)) != 0)
646 return (error);
647 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
648 sp->f_oflags = sp->f_flags & 0xffff;
649 return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
650 }
651
652 /*
653 * Get filesystem statistics.
654 */
655 /* ARGSUSED */
656 int
657 sys_fstatfs(p, v, retval)
658 struct proc *p;
659 void *v;
660 register_t *retval;
661 {
662 struct sys_fstatfs_args /* {
663 syscallarg(int) fd;
664 syscallarg(struct statfs *) buf;
665 } */ *uap = v;
666 struct file *fp;
667 struct mount *mp;
668 struct statfs *sp;
669 int error;
670
671 /* getvnode() will use the descriptor for us */
672 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
673 return (error);
674 mp = ((struct vnode *)fp->f_data)->v_mount;
675 sp = &mp->mnt_stat;
676 if ((error = VFS_STATFS(mp, sp, p)) != 0)
677 goto out;
678 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
679 sp->f_oflags = sp->f_flags & 0xffff;
680 error = copyout(sp, SCARG(uap, buf), sizeof(*sp));
681 out:
682 FILE_UNUSE(fp, p);
683 return (error);
684 }
685
686 /*
687 * Get statistics on all filesystems.
688 */
689 int
690 sys_getfsstat(p, v, retval)
691 struct proc *p;
692 void *v;
693 register_t *retval;
694 {
695 struct sys_getfsstat_args /* {
696 syscallarg(struct statfs *) buf;
697 syscallarg(long) bufsize;
698 syscallarg(int) flags;
699 } */ *uap = v;
700 struct mount *mp, *nmp;
701 struct statfs *sp;
702 caddr_t sfsp;
703 long count, maxcount, error;
704
705 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
706 sfsp = (caddr_t)SCARG(uap, buf);
707 simple_lock(&mountlist_slock);
708 count = 0;
709 for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
710 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
711 nmp = mp->mnt_list.cqe_next;
712 continue;
713 }
714 if (sfsp && count < maxcount) {
715 sp = &mp->mnt_stat;
716 /*
717 * If MNT_NOWAIT or MNT_LAZY is specified, do not
718 * refresh the fsstat cache. MNT_WAIT or MNT_LAXY
719 * overrides MNT_NOWAIT.
720 */
721 if (SCARG(uap, flags) != MNT_NOWAIT &&
722 SCARG(uap, flags) != MNT_LAZY &&
723 (SCARG(uap, flags) == MNT_WAIT ||
724 SCARG(uap, flags) == 0) &&
725 (error = VFS_STATFS(mp, sp, p)) != 0) {
726 simple_lock(&mountlist_slock);
727 nmp = mp->mnt_list.cqe_next;
728 vfs_unbusy(mp);
729 continue;
730 }
731 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
732 sp->f_oflags = sp->f_flags & 0xffff;
733 error = copyout(sp, sfsp, sizeof(*sp));
734 if (error) {
735 vfs_unbusy(mp);
736 return (error);
737 }
738 sfsp += sizeof(*sp);
739 }
740 count++;
741 simple_lock(&mountlist_slock);
742 nmp = mp->mnt_list.cqe_next;
743 vfs_unbusy(mp);
744 }
745 simple_unlock(&mountlist_slock);
746 if (sfsp && count > maxcount)
747 *retval = maxcount;
748 else
749 *retval = count;
750 return (0);
751 }
752
753 /*
754 * Change current working directory to a given file descriptor.
755 */
756 /* ARGSUSED */
757 int
758 sys_fchdir(p, v, retval)
759 struct proc *p;
760 void *v;
761 register_t *retval;
762 {
763 struct sys_fchdir_args /* {
764 syscallarg(int) fd;
765 } */ *uap = v;
766 struct filedesc *fdp = p->p_fd;
767 struct cwdinfo *cwdi = p->p_cwdi;
768 struct vnode *vp, *tdp;
769 struct mount *mp;
770 struct file *fp;
771 int error;
772
773 /* getvnode() will use the descriptor for us */
774 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
775 return (error);
776 vp = (struct vnode *)fp->f_data;
777
778 VREF(vp);
779 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
780 if (vp->v_type != VDIR)
781 error = ENOTDIR;
782 else
783 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
784 while (!error && (mp = vp->v_mountedhere) != NULL) {
785 if (vfs_busy(mp, 0, 0))
786 continue;
787 error = VFS_ROOT(mp, &tdp);
788 vfs_unbusy(mp);
789 if (error)
790 break;
791 vput(vp);
792 vp = tdp;
793 }
794 if (error) {
795 vput(vp);
796 goto out;
797 }
798 VOP_UNLOCK(vp, 0);
799
800 /*
801 * Disallow changing to a directory not under the process's
802 * current root directory (if there is one).
803 */
804 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
805 vrele(vp);
806 error = EPERM; /* operation not permitted */
807 goto out;
808 }
809
810 vrele(cwdi->cwdi_cdir);
811 cwdi->cwdi_cdir = vp;
812 out:
813 FILE_UNUSE(fp, p);
814 return (error);
815 }
816
817 /*
818 * Change this process's notion of the root directory to a given file descriptor.
819 */
820
821 int
822 sys_fchroot(p, v, retval)
823 struct proc *p;
824 void *v;
825 register_t *retval;
826 {
827 struct sys_fchroot_args *uap = v;
828 struct filedesc *fdp = p->p_fd;
829 struct cwdinfo *cwdi = p->p_cwdi;
830 struct vnode *vp;
831 struct file *fp;
832 int error;
833
834 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
835 return error;
836 /* getvnode() will use the descriptor for us */
837 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
838 return error;
839 vp = (struct vnode *) fp->f_data;
840 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
841 if (vp->v_type != VDIR)
842 error = ENOTDIR;
843 else
844 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
845 VOP_UNLOCK(vp, 0);
846 if (error)
847 goto out;
848 VREF(vp);
849
850 /*
851 * Prevent escaping from chroot by putting the root under
852 * the working directory. Silently chdir to / if we aren't
853 * already there.
854 */
855 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
856 /*
857 * XXX would be more failsafe to change directory to a
858 * deadfs node here instead
859 */
860 vrele(cwdi->cwdi_cdir);
861 VREF(vp);
862 cwdi->cwdi_cdir = vp;
863 }
864
865 if (cwdi->cwdi_rdir != NULL)
866 vrele(cwdi->cwdi_rdir);
867 cwdi->cwdi_rdir = vp;
868 out:
869 FILE_UNUSE(fp, p);
870 return (error);
871 }
872
873
874
875 /*
876 * Change current working directory (``.'').
877 */
878 /* ARGSUSED */
879 int
880 sys_chdir(p, v, retval)
881 struct proc *p;
882 void *v;
883 register_t *retval;
884 {
885 struct sys_chdir_args /* {
886 syscallarg(const char *) path;
887 } */ *uap = v;
888 struct cwdinfo *cwdi = p->p_cwdi;
889 int error;
890 struct nameidata nd;
891
892 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
893 SCARG(uap, path), p);
894 if ((error = change_dir(&nd, p)) != 0)
895 return (error);
896 vrele(cwdi->cwdi_cdir);
897 cwdi->cwdi_cdir = nd.ni_vp;
898 return (0);
899 }
900
901 /*
902 * Change notion of root (``/'') directory.
903 */
904 /* ARGSUSED */
905 int
906 sys_chroot(p, v, retval)
907 struct proc *p;
908 void *v;
909 register_t *retval;
910 {
911 struct sys_chroot_args /* {
912 syscallarg(const char *) path;
913 } */ *uap = v;
914 struct cwdinfo *cwdi = p->p_cwdi;
915 struct vnode *vp;
916 int error;
917 struct nameidata nd;
918
919 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
920 return (error);
921 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
922 SCARG(uap, path), p);
923 if ((error = change_dir(&nd, p)) != 0)
924 return (error);
925 if (cwdi->cwdi_rdir != NULL)
926 vrele(cwdi->cwdi_rdir);
927 vp = nd.ni_vp;
928 cwdi->cwdi_rdir = vp;
929
930 /*
931 * Prevent escaping from chroot by putting the root under
932 * the working directory. Silently chdir to / if we aren't
933 * already there.
934 */
935 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
936 /*
937 * XXX would be more failsafe to change directory to a
938 * deadfs node here instead
939 */
940 vrele(cwdi->cwdi_cdir);
941 VREF(vp);
942 cwdi->cwdi_cdir = vp;
943 }
944
945 return (0);
946 }
947
948 /*
949 * Common routine for chroot and chdir.
950 */
951 static int
952 change_dir(ndp, p)
953 struct nameidata *ndp;
954 struct proc *p;
955 {
956 struct vnode *vp;
957 int error;
958
959 if ((error = namei(ndp)) != 0)
960 return (error);
961 vp = ndp->ni_vp;
962 if (vp->v_type != VDIR)
963 error = ENOTDIR;
964 else
965 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
966
967 if (error)
968 vput(vp);
969 else
970 VOP_UNLOCK(vp, 0);
971 return (error);
972 }
973
974 /*
975 * Check permissions, allocate an open file structure,
976 * and call the device open routine if any.
977 */
978 int
979 sys_open(p, v, retval)
980 struct proc *p;
981 void *v;
982 register_t *retval;
983 {
984 struct sys_open_args /* {
985 syscallarg(const char *) path;
986 syscallarg(int) flags;
987 syscallarg(int) mode;
988 } */ *uap = v;
989 struct cwdinfo *cwdi = p->p_cwdi;
990 struct filedesc *fdp = p->p_fd;
991 struct file *fp;
992 struct vnode *vp;
993 int flags, cmode;
994 int type, indx, error;
995 struct flock lf;
996 struct nameidata nd;
997
998 flags = FFLAGS(SCARG(uap, flags));
999 if ((flags & (FREAD | FWRITE)) == 0)
1000 return (EINVAL);
1001 /* falloc() will use the file descriptor for us */
1002 if ((error = falloc(p, &fp, &indx)) != 0)
1003 return (error);
1004 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1005 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1006 p->p_dupfd = -indx - 1; /* XXX check for fdopen */
1007 if ((error = vn_open(&nd, flags, cmode)) != 0) {
1008 FILE_UNUSE(fp, p);
1009 ffree(fp);
1010 if ((error == ENODEV || error == ENXIO) &&
1011 p->p_dupfd >= 0 && /* XXX from fdopen */
1012 (error =
1013 dupfdopen(p, indx, p->p_dupfd, flags, error)) == 0) {
1014 *retval = indx;
1015 return (0);
1016 }
1017 if (error == ERESTART)
1018 error = EINTR;
1019 fdremove(fdp, indx);
1020 return (error);
1021 }
1022 p->p_dupfd = 0;
1023 vp = nd.ni_vp;
1024 fp->f_flag = flags & FMASK;
1025 fp->f_type = DTYPE_VNODE;
1026 fp->f_ops = &vnops;
1027 fp->f_data = (caddr_t)vp;
1028 if (flags & (O_EXLOCK | O_SHLOCK)) {
1029 lf.l_whence = SEEK_SET;
1030 lf.l_start = 0;
1031 lf.l_len = 0;
1032 if (flags & O_EXLOCK)
1033 lf.l_type = F_WRLCK;
1034 else
1035 lf.l_type = F_RDLCK;
1036 type = F_FLOCK;
1037 if ((flags & FNONBLOCK) == 0)
1038 type |= F_WAIT;
1039 VOP_UNLOCK(vp, 0);
1040 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1041 if (error) {
1042 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1043 FILE_UNUSE(fp, p);
1044 ffree(fp);
1045 fdremove(fdp, indx);
1046 return (error);
1047 }
1048 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1049 fp->f_flag |= FHASLOCK;
1050 }
1051 VOP_UNLOCK(vp, 0);
1052 *retval = indx;
1053 FILE_UNUSE(fp, p);
1054 return (0);
1055 }
1056
1057 /*
1058 * Get file handle system call
1059 */
1060 int
1061 sys_getfh(p, v, retval)
1062 struct proc *p;
1063 void *v;
1064 register_t *retval;
1065 {
1066 struct sys_getfh_args /* {
1067 syscallarg(char *) fname;
1068 syscallarg(fhandle_t *) fhp;
1069 } */ *uap = v;
1070 struct vnode *vp;
1071 fhandle_t fh;
1072 int error;
1073 struct nameidata nd;
1074
1075 /*
1076 * Must be super user
1077 */
1078 error = suser(p->p_ucred, &p->p_acflag);
1079 if (error)
1080 return (error);
1081 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1082 SCARG(uap, fname), p);
1083 error = namei(&nd);
1084 if (error)
1085 return (error);
1086 vp = nd.ni_vp;
1087 memset((caddr_t)&fh, 0, sizeof(fh));
1088 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1089 error = VFS_VPTOFH(vp, &fh.fh_fid);
1090 vput(vp);
1091 if (error)
1092 return (error);
1093 error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1094 return (error);
1095 }
1096
1097 /*
1098 * Open a file given a file handle.
1099 *
1100 * Check permissions, allocate an open file structure,
1101 * and call the device open routine if any.
1102 */
1103 int
1104 sys_fhopen(p, v, retval)
1105 struct proc *p;
1106 void *v;
1107 register_t *retval;
1108 {
1109 struct sys_fhopen_args /* {
1110 syscallarg(const fhandle_t *) fhp;
1111 syscallarg(int) flags;
1112 } */ *uap = v;
1113 struct filedesc *fdp = p->p_fd;
1114 struct file *fp;
1115 struct vnode *vp = NULL;
1116 struct mount *mp;
1117 struct ucred *cred = p->p_ucred;
1118 int flags;
1119 struct file *nfp;
1120 int type, indx, error=0;
1121 struct flock lf;
1122 struct vattr va;
1123 fhandle_t fh;
1124
1125 /*
1126 * Must be super user
1127 */
1128 if ((error = suser(p->p_ucred, &p->p_acflag)))
1129 return (error);
1130
1131 flags = FFLAGS(SCARG(uap, flags));
1132 if ((flags & (FREAD | FWRITE)) == 0)
1133 return (EINVAL);
1134 if ((flags & O_CREAT))
1135 return (EINVAL);
1136 /* falloc() will use the file descriptor for us */
1137 if ((error = falloc(p, &nfp, &indx)) != 0)
1138 return (error);
1139 fp = nfp;
1140 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1141 goto bad;
1142
1143 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1144 error = ESTALE;
1145 goto bad;
1146 }
1147
1148 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1149 vp = NULL; /* most likely unnecessary sanity for bad: */
1150 goto bad;
1151 }
1152
1153 /* Now do an effective vn_open */
1154
1155 if (vp->v_type == VSOCK) {
1156 error = EOPNOTSUPP;
1157 goto bad;
1158 }
1159 if (flags & FREAD) {
1160 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1161 goto bad;
1162 }
1163 if (flags & (FWRITE | O_TRUNC)) {
1164 if (vp->v_type == VDIR) {
1165 error = EISDIR;
1166 goto bad;
1167 }
1168 if ((error = vn_writechk(vp)) != 0 ||
1169 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1170 goto bad;
1171 }
1172 if (flags & O_TRUNC) {
1173 VOP_UNLOCK(vp, 0); /* XXX */
1174 VOP_LEASE(vp, p, cred, LEASE_WRITE);
1175 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
1176 VATTR_NULL(&va);
1177 va.va_size = 0;
1178 if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1179 goto bad;
1180 }
1181 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1182 goto bad;
1183 if (flags & FWRITE)
1184 vp->v_writecount++;
1185
1186 /* done with modified vn_open, now finish what sys_open does. */
1187
1188 fp->f_flag = flags & FMASK;
1189 fp->f_type = DTYPE_VNODE;
1190 fp->f_ops = &vnops;
1191 fp->f_data = (caddr_t)vp;
1192 if (flags & (O_EXLOCK | O_SHLOCK)) {
1193 lf.l_whence = SEEK_SET;
1194 lf.l_start = 0;
1195 lf.l_len = 0;
1196 if (flags & O_EXLOCK)
1197 lf.l_type = F_WRLCK;
1198 else
1199 lf.l_type = F_RDLCK;
1200 type = F_FLOCK;
1201 if ((flags & FNONBLOCK) == 0)
1202 type |= F_WAIT;
1203 VOP_UNLOCK(vp, 0);
1204 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1205 if (error) {
1206 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1207 FILE_UNUSE(fp, p);
1208 ffree(fp);
1209 fdremove(fdp, indx);
1210 return (error);
1211 }
1212 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1213 fp->f_flag |= FHASLOCK;
1214 }
1215 VOP_UNLOCK(vp, 0);
1216 *retval = indx;
1217 FILE_UNUSE(fp, p);
1218 return (0);
1219
1220 bad:
1221 FILE_UNUSE(fp, p);
1222 ffree(fp);
1223 fdremove(fdp, indx);
1224 if (vp != NULL)
1225 vput(vp);
1226 return (error);
1227 }
1228
1229 /* ARGSUSED */
1230 int
1231 sys_fhstat(p, v, retval)
1232 struct proc *p;
1233 void *v;
1234 register_t *retval;
1235 {
1236 struct sys_fhstat_args /* {
1237 syscallarg(const fhandle_t *) fhp;
1238 syscallarg(struct stat *) sb;
1239 } */ *uap = v;
1240 struct stat sb;
1241 int error;
1242 fhandle_t fh;
1243 struct mount *mp;
1244 struct vnode *vp;
1245
1246 /*
1247 * Must be super user
1248 */
1249 if ((error = suser(p->p_ucred, &p->p_acflag)))
1250 return (error);
1251
1252 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1253 return (error);
1254
1255 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1256 return (ESTALE);
1257 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1258 return (error);
1259 error = vn_stat(vp, &sb, p);
1260 vput(vp);
1261 if (error)
1262 return (error);
1263 error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1264 return (error);
1265 }
1266
1267 /* ARGSUSED */
1268 int
1269 sys_fhstatfs(p, v, retval)
1270 struct proc *p;
1271 void *v;
1272 register_t *retval;
1273 {
1274 struct sys_fhstatfs_args /*
1275 syscallarg(const fhandle_t *) fhp;
1276 syscallarg(struct statfs *) buf;
1277 } */ *uap = v;
1278 struct statfs sp;
1279 fhandle_t fh;
1280 struct mount *mp;
1281 struct vnode *vp;
1282 int error;
1283
1284 /*
1285 * Must be super user
1286 */
1287 if ((error = suser(p->p_ucred, &p->p_acflag)))
1288 return (error);
1289
1290 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1291 return (error);
1292
1293 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1294 return (ESTALE);
1295 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1296 return (error);
1297 mp = vp->v_mount;
1298 vput(vp);
1299 if ((error = VFS_STATFS(mp, &sp, p)) != 0)
1300 return (error);
1301 sp.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1302 sp.f_oflags = sp.f_flags & 0xffff;
1303 return (copyout(&sp, SCARG(uap, buf), sizeof(sp)));
1304 }
1305
1306 /*
1307 * Create a special file.
1308 */
1309 /* ARGSUSED */
1310 int
1311 sys_mknod(p, v, retval)
1312 struct proc *p;
1313 void *v;
1314 register_t *retval;
1315 {
1316 struct sys_mknod_args /* {
1317 syscallarg(const char *) path;
1318 syscallarg(int) mode;
1319 syscallarg(int) dev;
1320 } */ *uap = v;
1321 struct vnode *vp;
1322 struct vattr vattr;
1323 int error;
1324 int whiteout = 0;
1325 struct nameidata nd;
1326
1327 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1328 return (error);
1329 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1330 if ((error = namei(&nd)) != 0)
1331 return (error);
1332 vp = nd.ni_vp;
1333 if (vp != NULL)
1334 error = EEXIST;
1335 else {
1336 VATTR_NULL(&vattr);
1337 vattr.va_mode =
1338 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1339 vattr.va_rdev = SCARG(uap, dev);
1340 whiteout = 0;
1341
1342 switch (SCARG(uap, mode) & S_IFMT) {
1343 case S_IFMT: /* used by badsect to flag bad sectors */
1344 vattr.va_type = VBAD;
1345 break;
1346 case S_IFCHR:
1347 vattr.va_type = VCHR;
1348 break;
1349 case S_IFBLK:
1350 vattr.va_type = VBLK;
1351 break;
1352 case S_IFWHT:
1353 whiteout = 1;
1354 break;
1355 default:
1356 error = EINVAL;
1357 break;
1358 }
1359 }
1360 if (!error) {
1361 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1362 if (whiteout) {
1363 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1364 if (error)
1365 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1366 vput(nd.ni_dvp);
1367 } else {
1368 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1369 &nd.ni_cnd, &vattr);
1370 }
1371 } else {
1372 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1373 if (nd.ni_dvp == vp)
1374 vrele(nd.ni_dvp);
1375 else
1376 vput(nd.ni_dvp);
1377 if (vp)
1378 vrele(vp);
1379 }
1380 return (error);
1381 }
1382
1383 /*
1384 * Create a named pipe.
1385 */
1386 /* ARGSUSED */
1387 int
1388 sys_mkfifo(p, v, retval)
1389 struct proc *p;
1390 void *v;
1391 register_t *retval;
1392 {
1393 struct sys_mkfifo_args /* {
1394 syscallarg(const char *) path;
1395 syscallarg(int) mode;
1396 } */ *uap = v;
1397 struct vattr vattr;
1398 int error;
1399 struct nameidata nd;
1400
1401 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1402 if ((error = namei(&nd)) != 0)
1403 return (error);
1404 if (nd.ni_vp != NULL) {
1405 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1406 if (nd.ni_dvp == nd.ni_vp)
1407 vrele(nd.ni_dvp);
1408 else
1409 vput(nd.ni_dvp);
1410 vrele(nd.ni_vp);
1411 return (EEXIST);
1412 }
1413 VATTR_NULL(&vattr);
1414 vattr.va_type = VFIFO;
1415 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1416 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1417 return (VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr));
1418 }
1419
1420 /*
1421 * Make a hard file link.
1422 */
1423 /* ARGSUSED */
1424 int
1425 sys_link(p, v, retval)
1426 struct proc *p;
1427 void *v;
1428 register_t *retval;
1429 {
1430 struct sys_link_args /* {
1431 syscallarg(const char *) path;
1432 syscallarg(const char *) link;
1433 } */ *uap = v;
1434 struct vnode *vp;
1435 struct nameidata nd;
1436 int error;
1437
1438 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1439 if ((error = namei(&nd)) != 0)
1440 return (error);
1441 vp = nd.ni_vp;
1442 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1443 if ((error = namei(&nd)) != 0)
1444 goto out;
1445 if (nd.ni_vp) {
1446 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1447 if (nd.ni_dvp == nd.ni_vp)
1448 vrele(nd.ni_dvp);
1449 else
1450 vput(nd.ni_dvp);
1451 vrele(nd.ni_vp);
1452 error = EEXIST;
1453 goto out;
1454 }
1455 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1456 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1457 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1458 out:
1459 vrele(vp);
1460 return (error);
1461 }
1462
1463 /*
1464 * Make a symbolic link.
1465 */
1466 /* ARGSUSED */
1467 int
1468 sys_symlink(p, v, retval)
1469 struct proc *p;
1470 void *v;
1471 register_t *retval;
1472 {
1473 struct sys_symlink_args /* {
1474 syscallarg(const char *) path;
1475 syscallarg(const char *) link;
1476 } */ *uap = v;
1477 struct vattr vattr;
1478 char *path;
1479 int error;
1480 struct nameidata nd;
1481
1482 path = PNBUF_GET();
1483 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1484 if (error)
1485 goto out;
1486 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1487 if ((error = namei(&nd)) != 0)
1488 goto out;
1489 if (nd.ni_vp) {
1490 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1491 if (nd.ni_dvp == nd.ni_vp)
1492 vrele(nd.ni_dvp);
1493 else
1494 vput(nd.ni_dvp);
1495 vrele(nd.ni_vp);
1496 error = EEXIST;
1497 goto out;
1498 }
1499 VATTR_NULL(&vattr);
1500 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1501 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1502 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1503 out:
1504 PNBUF_PUT(path);
1505 return (error);
1506 }
1507
1508 /*
1509 * Delete a whiteout from the filesystem.
1510 */
1511 /* ARGSUSED */
1512 int
1513 sys_undelete(p, v, retval)
1514 struct proc *p;
1515 void *v;
1516 register_t *retval;
1517 {
1518 struct sys_undelete_args /* {
1519 syscallarg(const char *) path;
1520 } */ *uap = v;
1521 int error;
1522 struct nameidata nd;
1523
1524 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1525 SCARG(uap, path), p);
1526 error = namei(&nd);
1527 if (error)
1528 return (error);
1529
1530 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
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 if (nd.ni_vp)
1537 vrele(nd.ni_vp);
1538 return (EEXIST);
1539 }
1540
1541 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1542 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1543 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1544 vput(nd.ni_dvp);
1545 return (error);
1546 }
1547
1548 /*
1549 * Delete a name from the filesystem.
1550 */
1551 /* ARGSUSED */
1552 int
1553 sys_unlink(p, v, retval)
1554 struct proc *p;
1555 void *v;
1556 register_t *retval;
1557 {
1558 struct sys_unlink_args /* {
1559 syscallarg(const char *) path;
1560 } */ *uap = v;
1561 struct vnode *vp;
1562 int error;
1563 struct nameidata nd;
1564
1565 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1566 SCARG(uap, path), p);
1567 if ((error = namei(&nd)) != 0)
1568 return (error);
1569 vp = nd.ni_vp;
1570
1571 /*
1572 * The root of a mounted filesystem cannot be deleted.
1573 */
1574 if (vp->v_flag & VROOT) {
1575 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1576 if (nd.ni_dvp == vp)
1577 vrele(nd.ni_dvp);
1578 else
1579 vput(nd.ni_dvp);
1580 vput(vp);
1581 error = EBUSY;
1582 goto out;
1583 }
1584
1585 (void)uvm_vnp_uncache(vp);
1586
1587 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1588 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1589 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1590 out:
1591 return (error);
1592 }
1593
1594 /*
1595 * Reposition read/write file offset.
1596 */
1597 int
1598 sys_lseek(p, v, retval)
1599 struct proc *p;
1600 void *v;
1601 register_t *retval;
1602 {
1603 struct sys_lseek_args /* {
1604 syscallarg(int) fd;
1605 syscallarg(int) pad;
1606 syscallarg(off_t) offset;
1607 syscallarg(int) whence;
1608 } */ *uap = v;
1609 struct ucred *cred = p->p_ucred;
1610 struct filedesc *fdp = p->p_fd;
1611 struct file *fp;
1612 struct vnode *vp;
1613 struct vattr vattr;
1614 off_t newoff;
1615 int error;
1616
1617 if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles ||
1618 (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL ||
1619 (fp->f_iflags & FIF_WANTCLOSE) != 0)
1620 return (EBADF);
1621
1622 FILE_USE(fp);
1623
1624 vp = (struct vnode *)fp->f_data;
1625 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1626 error = ESPIPE;
1627 goto out;
1628 }
1629
1630 switch (SCARG(uap, whence)) {
1631 case SEEK_CUR:
1632 newoff = fp->f_offset + SCARG(uap, offset);
1633 break;
1634 case SEEK_END:
1635 error = VOP_GETATTR(vp, &vattr, cred, p);
1636 if (error)
1637 goto out;
1638 newoff = SCARG(uap, offset) + vattr.va_size;
1639 break;
1640 case SEEK_SET:
1641 newoff = SCARG(uap, offset);
1642 break;
1643 default:
1644 error = EINVAL;
1645 goto out;
1646 }
1647 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1648 goto out;
1649
1650 *(off_t *)retval = fp->f_offset = newoff;
1651 out:
1652 FILE_UNUSE(fp, p);
1653 return (error);
1654 }
1655
1656 /*
1657 * Positional read system call.
1658 */
1659 int
1660 sys_pread(p, v, retval)
1661 struct proc *p;
1662 void *v;
1663 register_t *retval;
1664 {
1665 struct sys_pread_args /* {
1666 syscallarg(int) fd;
1667 syscallarg(void *) buf;
1668 syscallarg(size_t) nbyte;
1669 syscallarg(off_t) offset;
1670 } */ *uap = v;
1671 struct filedesc *fdp = p->p_fd;
1672 struct file *fp;
1673 struct vnode *vp;
1674 off_t offset;
1675 int error, fd = SCARG(uap, fd);
1676
1677 if ((u_int)fd >= fdp->fd_nfiles ||
1678 (fp = fdp->fd_ofiles[fd]) == NULL ||
1679 (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1680 (fp->f_flag & FREAD) == 0)
1681 return (EBADF);
1682
1683 FILE_USE(fp);
1684
1685 vp = (struct vnode *)fp->f_data;
1686 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1687 error = ESPIPE;
1688 goto out;
1689 }
1690
1691 offset = SCARG(uap, offset);
1692
1693 /*
1694 * XXX This works because no file systems actually
1695 * XXX take any action on the seek operation.
1696 */
1697 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1698 goto out;
1699
1700 /* dofileread() will unuse the descriptor for us */
1701 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1702 &offset, 0, retval));
1703
1704 out:
1705 FILE_UNUSE(fp, p);
1706 return (error);
1707 }
1708
1709 /*
1710 * Positional scatter read system call.
1711 */
1712 int
1713 sys_preadv(p, v, retval)
1714 struct proc *p;
1715 void *v;
1716 register_t *retval;
1717 {
1718 struct sys_preadv_args /* {
1719 syscallarg(int) fd;
1720 syscallarg(const struct iovec *) iovp;
1721 syscallarg(int) iovcnt;
1722 syscallarg(off_t) offset;
1723 } */ *uap = v;
1724 struct filedesc *fdp = p->p_fd;
1725 struct file *fp;
1726 struct vnode *vp;
1727 off_t offset;
1728 int error, fd = SCARG(uap, fd);
1729
1730 if ((u_int)fd >= fdp->fd_nfiles ||
1731 (fp = fdp->fd_ofiles[fd]) == NULL ||
1732 (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1733 (fp->f_flag & FREAD) == 0)
1734 return (EBADF);
1735
1736 FILE_USE(fp);
1737
1738 vp = (struct vnode *)fp->f_data;
1739 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1740 error = ESPIPE;
1741 goto out;
1742 }
1743
1744 offset = SCARG(uap, offset);
1745
1746 /*
1747 * XXX This works because no file systems actually
1748 * XXX take any action on the seek operation.
1749 */
1750 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1751 goto out;
1752
1753 /* dofilereadv() will unuse the descriptor for us */
1754 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1755 &offset, 0, retval));
1756
1757 out:
1758 FILE_UNUSE(fp, p);
1759 return (error);
1760 }
1761
1762 /*
1763 * Positional write system call.
1764 */
1765 int
1766 sys_pwrite(p, v, retval)
1767 struct proc *p;
1768 void *v;
1769 register_t *retval;
1770 {
1771 struct sys_pwrite_args /* {
1772 syscallarg(int) fd;
1773 syscallarg(const void *) buf;
1774 syscallarg(size_t) nbyte;
1775 syscallarg(off_t) offset;
1776 } */ *uap = v;
1777 struct filedesc *fdp = p->p_fd;
1778 struct file *fp;
1779 struct vnode *vp;
1780 off_t offset;
1781 int error, fd = SCARG(uap, fd);
1782
1783 if ((u_int)fd >= fdp->fd_nfiles ||
1784 (fp = fdp->fd_ofiles[fd]) == NULL ||
1785 (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1786 (fp->f_flag & FWRITE) == 0)
1787 return (EBADF);
1788
1789 FILE_USE(fp);
1790
1791 vp = (struct vnode *)fp->f_data;
1792 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1793 error = ESPIPE;
1794 goto out;
1795 }
1796
1797 offset = SCARG(uap, offset);
1798
1799 /*
1800 * XXX This works because no file systems actually
1801 * XXX take any action on the seek operation.
1802 */
1803 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1804 goto out;
1805
1806 /* dofilewrite() will unuse the descriptor for us */
1807 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1808 &offset, 0, retval));
1809
1810 out:
1811 FILE_UNUSE(fp, p);
1812 return (error);
1813 }
1814
1815 /*
1816 * Positional gather write system call.
1817 */
1818 int
1819 sys_pwritev(p, v, retval)
1820 struct proc *p;
1821 void *v;
1822 register_t *retval;
1823 {
1824 struct sys_pwritev_args /* {
1825 syscallarg(int) fd;
1826 syscallarg(const struct iovec *) iovp;
1827 syscallarg(int) iovcnt;
1828 syscallarg(off_t) offset;
1829 } */ *uap = v;
1830 struct filedesc *fdp = p->p_fd;
1831 struct file *fp;
1832 struct vnode *vp;
1833 off_t offset;
1834 int error, fd = SCARG(uap, fd);
1835
1836 if ((u_int)fd >= fdp->fd_nfiles ||
1837 (fp = fdp->fd_ofiles[fd]) == NULL ||
1838 (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1839 (fp->f_flag & FWRITE) == 0)
1840 return (EBADF);
1841
1842 FILE_USE(fp);
1843
1844 vp = (struct vnode *)fp->f_data;
1845 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1846 error = ESPIPE;
1847 goto out;
1848 }
1849
1850 offset = SCARG(uap, offset);
1851
1852 /*
1853 * XXX This works because no file systems actually
1854 * XXX take any action on the seek operation.
1855 */
1856 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1857 goto out;
1858
1859 /* dofilewritev() will unuse the descriptor for us */
1860 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1861 &offset, 0, retval));
1862
1863 out:
1864 FILE_UNUSE(fp, p);
1865 return (error);
1866 }
1867
1868 /*
1869 * Check access permissions.
1870 */
1871 int
1872 sys_access(p, v, retval)
1873 struct proc *p;
1874 void *v;
1875 register_t *retval;
1876 {
1877 struct sys_access_args /* {
1878 syscallarg(const char *) path;
1879 syscallarg(int) flags;
1880 } */ *uap = v;
1881 struct ucred *cred = p->p_ucred;
1882 struct vnode *vp;
1883 int error, flags, t_gid, t_uid;
1884 struct nameidata nd;
1885
1886 t_uid = cred->cr_uid;
1887 t_gid = cred->cr_gid;
1888 cred->cr_uid = p->p_cred->p_ruid;
1889 cred->cr_gid = p->p_cred->p_rgid;
1890 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1891 SCARG(uap, path), p);
1892 if ((error = namei(&nd)) != 0)
1893 goto out1;
1894 vp = nd.ni_vp;
1895
1896 /* Flags == 0 means only check for existence. */
1897 if (SCARG(uap, flags)) {
1898 flags = 0;
1899 if (SCARG(uap, flags) & R_OK)
1900 flags |= VREAD;
1901 if (SCARG(uap, flags) & W_OK)
1902 flags |= VWRITE;
1903 if (SCARG(uap, flags) & X_OK)
1904 flags |= VEXEC;
1905
1906 error = VOP_ACCESS(vp, flags, cred, p);
1907 if (!error && (flags & VWRITE))
1908 error = vn_writechk(vp);
1909 }
1910 vput(vp);
1911 out1:
1912 cred->cr_uid = t_uid;
1913 cred->cr_gid = t_gid;
1914 return (error);
1915 }
1916
1917 /*
1918 * Get file status; this version follows links.
1919 */
1920 /* ARGSUSED */
1921 int
1922 sys___stat13(p, v, retval)
1923 struct proc *p;
1924 void *v;
1925 register_t *retval;
1926 {
1927 struct sys___stat13_args /* {
1928 syscallarg(const char *) path;
1929 syscallarg(struct stat *) ub;
1930 } */ *uap = v;
1931 struct stat sb;
1932 int error;
1933 struct nameidata nd;
1934
1935 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1936 SCARG(uap, path), p);
1937 if ((error = namei(&nd)) != 0)
1938 return (error);
1939 error = vn_stat(nd.ni_vp, &sb, p);
1940 vput(nd.ni_vp);
1941 if (error)
1942 return (error);
1943 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1944 return (error);
1945 }
1946
1947 /*
1948 * Get file status; this version does not follow links.
1949 */
1950 /* ARGSUSED */
1951 int
1952 sys___lstat13(p, v, retval)
1953 struct proc *p;
1954 void *v;
1955 register_t *retval;
1956 {
1957 struct sys___lstat13_args /* {
1958 syscallarg(const char *) path;
1959 syscallarg(struct stat *) ub;
1960 } */ *uap = v;
1961 struct stat sb;
1962 int error;
1963 struct nameidata nd;
1964
1965 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1966 SCARG(uap, path), p);
1967 if ((error = namei(&nd)) != 0)
1968 return (error);
1969 error = vn_stat(nd.ni_vp, &sb, p);
1970 vput(nd.ni_vp);
1971 if (error)
1972 return (error);
1973 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1974 return (error);
1975 }
1976
1977 /*
1978 * Get configurable pathname variables.
1979 */
1980 /* ARGSUSED */
1981 int
1982 sys_pathconf(p, v, retval)
1983 struct proc *p;
1984 void *v;
1985 register_t *retval;
1986 {
1987 struct sys_pathconf_args /* {
1988 syscallarg(const char *) path;
1989 syscallarg(int) name;
1990 } */ *uap = v;
1991 int error;
1992 struct nameidata nd;
1993
1994 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1995 SCARG(uap, path), p);
1996 if ((error = namei(&nd)) != 0)
1997 return (error);
1998 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
1999 vput(nd.ni_vp);
2000 return (error);
2001 }
2002
2003 /*
2004 * Return target name of a symbolic link.
2005 */
2006 /* ARGSUSED */
2007 int
2008 sys_readlink(p, v, retval)
2009 struct proc *p;
2010 void *v;
2011 register_t *retval;
2012 {
2013 struct sys_readlink_args /* {
2014 syscallarg(const char *) path;
2015 syscallarg(char *) buf;
2016 syscallarg(size_t) count;
2017 } */ *uap = v;
2018 struct vnode *vp;
2019 struct iovec aiov;
2020 struct uio auio;
2021 int error;
2022 struct nameidata nd;
2023
2024 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2025 SCARG(uap, path), p);
2026 if ((error = namei(&nd)) != 0)
2027 return (error);
2028 vp = nd.ni_vp;
2029 if (vp->v_type != VLNK)
2030 error = EINVAL;
2031 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2032 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2033 aiov.iov_base = SCARG(uap, buf);
2034 aiov.iov_len = SCARG(uap, count);
2035 auio.uio_iov = &aiov;
2036 auio.uio_iovcnt = 1;
2037 auio.uio_offset = 0;
2038 auio.uio_rw = UIO_READ;
2039 auio.uio_segflg = UIO_USERSPACE;
2040 auio.uio_procp = p;
2041 auio.uio_resid = SCARG(uap, count);
2042 error = VOP_READLINK(vp, &auio, p->p_ucred);
2043 }
2044 vput(vp);
2045 *retval = SCARG(uap, count) - auio.uio_resid;
2046 return (error);
2047 }
2048
2049 /*
2050 * Change flags of a file given a path name.
2051 */
2052 /* ARGSUSED */
2053 int
2054 sys_chflags(p, v, retval)
2055 struct proc *p;
2056 void *v;
2057 register_t *retval;
2058 {
2059 struct sys_chflags_args /* {
2060 syscallarg(const char *) path;
2061 syscallarg(u_long) flags;
2062 } */ *uap = v;
2063 struct vnode *vp;
2064 struct vattr vattr;
2065 int error;
2066 struct nameidata nd;
2067
2068 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2069 if ((error = namei(&nd)) != 0)
2070 return (error);
2071 vp = nd.ni_vp;
2072 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2073 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2074 /* Non-superusers cannot change the flags on devices, even if they
2075 own them. */
2076 if (suser(p->p_ucred, &p->p_acflag)) {
2077 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2078 goto out;
2079 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2080 error = EINVAL;
2081 goto out;
2082 }
2083 }
2084 VATTR_NULL(&vattr);
2085 vattr.va_flags = SCARG(uap, flags);
2086 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2087 out:
2088 vput(vp);
2089 return (error);
2090 }
2091
2092 /*
2093 * Change flags of a file given a file descriptor.
2094 */
2095 /* ARGSUSED */
2096 int
2097 sys_fchflags(p, v, retval)
2098 struct proc *p;
2099 void *v;
2100 register_t *retval;
2101 {
2102 struct sys_fchflags_args /* {
2103 syscallarg(int) fd;
2104 syscallarg(u_long) flags;
2105 } */ *uap = v;
2106 struct vattr vattr;
2107 struct vnode *vp;
2108 struct file *fp;
2109 int error;
2110
2111 /* getvnode() will use the descriptor for us */
2112 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2113 return (error);
2114 vp = (struct vnode *)fp->f_data;
2115 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2116 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2117 /* Non-superusers cannot change the flags on devices, even if they
2118 own them. */
2119 if (suser(p->p_ucred, &p->p_acflag)) {
2120 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p))
2121 != 0)
2122 goto out;
2123 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2124 error = EINVAL;
2125 goto out;
2126 }
2127 }
2128 VATTR_NULL(&vattr);
2129 vattr.va_flags = SCARG(uap, flags);
2130 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2131 out:
2132 VOP_UNLOCK(vp, 0);
2133 FILE_UNUSE(fp, p);
2134 return (error);
2135 }
2136
2137 /*
2138 * Change flags of a file given a file descriptor; this version does
2139 * not follow links.
2140 */
2141 int
2142 sys_lchflags(p, v, retval)
2143 struct proc *p;
2144 void *v;
2145 register_t *retval;
2146 {
2147 register struct sys_chflags_args /* {
2148 syscallarg(const char *) path;
2149 syscallarg(u_long) flags;
2150 } */ *uap = v;
2151 register struct vnode *vp;
2152 struct vattr vattr;
2153 int error;
2154 struct nameidata nd;
2155
2156 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2157 if ((error = namei(&nd)) != 0)
2158 return (error);
2159 vp = nd.ni_vp;
2160 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2161 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2162 /* Non-superusers cannot change the flags on devices, even if they
2163 own them. */
2164 if (suser(p->p_ucred, &p->p_acflag)) {
2165 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2166 goto out;
2167 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2168 error = EINVAL;
2169 goto out;
2170 }
2171 }
2172 VATTR_NULL(&vattr);
2173 vattr.va_flags = SCARG(uap, flags);
2174 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2175 out:
2176 vput(vp);
2177 return (error);
2178 }
2179
2180 /*
2181 * Change mode of a file given path name; this version follows links.
2182 */
2183 /* ARGSUSED */
2184 int
2185 sys_chmod(p, v, retval)
2186 struct proc *p;
2187 void *v;
2188 register_t *retval;
2189 {
2190 struct sys_chmod_args /* {
2191 syscallarg(const char *) path;
2192 syscallarg(int) mode;
2193 } */ *uap = v;
2194 int error;
2195 struct nameidata nd;
2196
2197 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2198 if ((error = namei(&nd)) != 0)
2199 return (error);
2200
2201 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2202
2203 vrele(nd.ni_vp);
2204 return (error);
2205 }
2206
2207 /*
2208 * Change mode of a file given a file descriptor.
2209 */
2210 /* ARGSUSED */
2211 int
2212 sys_fchmod(p, v, retval)
2213 struct proc *p;
2214 void *v;
2215 register_t *retval;
2216 {
2217 struct sys_fchmod_args /* {
2218 syscallarg(int) fd;
2219 syscallarg(int) mode;
2220 } */ *uap = v;
2221 struct file *fp;
2222 int error;
2223
2224 /* getvnode() will use the descriptor for us */
2225 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2226 return (error);
2227
2228 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2229 FILE_UNUSE(fp, p);
2230 return (error);
2231 }
2232
2233 /*
2234 * Change mode of a file given path name; this version does not follow links.
2235 */
2236 /* ARGSUSED */
2237 int
2238 sys_lchmod(p, v, retval)
2239 struct proc *p;
2240 void *v;
2241 register_t *retval;
2242 {
2243 struct sys_lchmod_args /* {
2244 syscallarg(const char *) path;
2245 syscallarg(int) mode;
2246 } */ *uap = v;
2247 int error;
2248 struct nameidata nd;
2249
2250 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2251 if ((error = namei(&nd)) != 0)
2252 return (error);
2253
2254 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2255
2256 vrele(nd.ni_vp);
2257 return (error);
2258 }
2259
2260 /*
2261 * Common routine to set mode given a vnode.
2262 */
2263 static int
2264 change_mode(vp, mode, p)
2265 struct vnode *vp;
2266 int mode;
2267 struct proc *p;
2268 {
2269 struct vattr vattr;
2270 int error;
2271
2272 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2273 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2274 VATTR_NULL(&vattr);
2275 vattr.va_mode = mode & ALLPERMS;
2276 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2277 VOP_UNLOCK(vp, 0);
2278 return (error);
2279 }
2280
2281 /*
2282 * Set ownership given a path name; this version follows links.
2283 */
2284 /* ARGSUSED */
2285 int
2286 sys_chown(p, v, retval)
2287 struct proc *p;
2288 void *v;
2289 register_t *retval;
2290 {
2291 struct sys_chown_args /* {
2292 syscallarg(const char *) path;
2293 syscallarg(uid_t) uid;
2294 syscallarg(gid_t) gid;
2295 } */ *uap = v;
2296 int error;
2297 struct nameidata nd;
2298
2299 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2300 if ((error = namei(&nd)) != 0)
2301 return (error);
2302
2303 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2304
2305 vrele(nd.ni_vp);
2306 return (error);
2307 }
2308
2309 /*
2310 * Set ownership given a path name; this version follows links.
2311 * Provides POSIX semantics.
2312 */
2313 /* ARGSUSED */
2314 int
2315 sys___posix_chown(p, v, retval)
2316 struct proc *p;
2317 void *v;
2318 register_t *retval;
2319 {
2320 struct sys_chown_args /* {
2321 syscallarg(const char *) path;
2322 syscallarg(uid_t) uid;
2323 syscallarg(gid_t) gid;
2324 } */ *uap = v;
2325 int error;
2326 struct nameidata nd;
2327
2328 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2329 if ((error = namei(&nd)) != 0)
2330 return (error);
2331
2332 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2333
2334 vrele(nd.ni_vp);
2335 return (error);
2336 }
2337
2338 /*
2339 * Set ownership given a file descriptor.
2340 */
2341 /* ARGSUSED */
2342 int
2343 sys_fchown(p, v, retval)
2344 struct proc *p;
2345 void *v;
2346 register_t *retval;
2347 {
2348 struct sys_fchown_args /* {
2349 syscallarg(int) fd;
2350 syscallarg(uid_t) uid;
2351 syscallarg(gid_t) gid;
2352 } */ *uap = v;
2353 int error;
2354 struct file *fp;
2355
2356 /* getvnode() will use the descriptor for us */
2357 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2358 return (error);
2359
2360 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2361 SCARG(uap, gid), p, 0);
2362 FILE_UNUSE(fp, p);
2363 return (error);
2364 }
2365
2366 /*
2367 * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2368 */
2369 /* ARGSUSED */
2370 int
2371 sys___posix_fchown(p, v, retval)
2372 struct proc *p;
2373 void *v;
2374 register_t *retval;
2375 {
2376 struct sys_fchown_args /* {
2377 syscallarg(int) fd;
2378 syscallarg(uid_t) uid;
2379 syscallarg(gid_t) gid;
2380 } */ *uap = v;
2381 int error;
2382 struct file *fp;
2383
2384 /* getvnode() will use the descriptor for us */
2385 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2386 return (error);
2387
2388 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2389 SCARG(uap, gid), p, 1);
2390 FILE_UNUSE(fp, p);
2391 return (error);
2392 }
2393
2394 /*
2395 * Set ownership given a path name; this version does not follow links.
2396 */
2397 /* ARGSUSED */
2398 int
2399 sys_lchown(p, v, retval)
2400 struct proc *p;
2401 void *v;
2402 register_t *retval;
2403 {
2404 struct sys_lchown_args /* {
2405 syscallarg(const char *) path;
2406 syscallarg(uid_t) uid;
2407 syscallarg(gid_t) gid;
2408 } */ *uap = v;
2409 int error;
2410 struct nameidata nd;
2411
2412 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2413 if ((error = namei(&nd)) != 0)
2414 return (error);
2415
2416 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2417
2418 vrele(nd.ni_vp);
2419 return (error);
2420 }
2421
2422 /*
2423 * Set ownership given a path name; this version does not follow links.
2424 * Provides POSIX/XPG semantics.
2425 */
2426 /* ARGSUSED */
2427 int
2428 sys___posix_lchown(p, v, retval)
2429 struct proc *p;
2430 void *v;
2431 register_t *retval;
2432 {
2433 struct sys_lchown_args /* {
2434 syscallarg(const char *) path;
2435 syscallarg(uid_t) uid;
2436 syscallarg(gid_t) gid;
2437 } */ *uap = v;
2438 int error;
2439 struct nameidata nd;
2440
2441 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2442 if ((error = namei(&nd)) != 0)
2443 return (error);
2444
2445 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2446
2447 vrele(nd.ni_vp);
2448 return (error);
2449 }
2450
2451 /*
2452 * Common routine to set ownership given a vnode.
2453 */
2454 static int
2455 change_owner(vp, uid, gid, p, posix_semantics)
2456 struct vnode *vp;
2457 uid_t uid;
2458 gid_t gid;
2459 struct proc *p;
2460 int posix_semantics;
2461 {
2462 struct vattr vattr;
2463 mode_t newmode;
2464 int error;
2465
2466 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2467 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2468 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2469 goto out;
2470
2471 #define CHANGED(x) ((x) != -1)
2472 newmode = vattr.va_mode;
2473 if (posix_semantics) {
2474 /*
2475 * POSIX/XPG semantics: if the caller is not the super-user,
2476 * clear set-user-id and set-group-id bits. Both POSIX and
2477 * the XPG consider the behaviour for calls by the super-user
2478 * implementation-defined; we leave the set-user-id and set-
2479 * group-id settings intact in that case.
2480 */
2481 if (suser(p->p_ucred, NULL) != 0)
2482 newmode &= ~(S_ISUID | S_ISGID);
2483 } else {
2484 /*
2485 * NetBSD semantics: when changing owner and/or group,
2486 * clear the respective bit(s).
2487 */
2488 if (CHANGED(uid))
2489 newmode &= ~S_ISUID;
2490 if (CHANGED(gid))
2491 newmode &= ~S_ISGID;
2492 }
2493 /* Update va_mode iff altered. */
2494 if (vattr.va_mode == newmode)
2495 newmode = VNOVAL;
2496
2497 VATTR_NULL(&vattr);
2498 vattr.va_uid = CHANGED(uid) ? uid : VNOVAL;
2499 vattr.va_gid = CHANGED(gid) ? gid : VNOVAL;
2500 vattr.va_mode = newmode;
2501 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2502 #undef CHANGED
2503
2504 out:
2505 VOP_UNLOCK(vp, 0);
2506 return (error);
2507 }
2508
2509 /*
2510 * Set the access and modification times given a path name; this
2511 * version follows links.
2512 */
2513 /* ARGSUSED */
2514 int
2515 sys_utimes(p, v, retval)
2516 struct proc *p;
2517 void *v;
2518 register_t *retval;
2519 {
2520 struct sys_utimes_args /* {
2521 syscallarg(const char *) path;
2522 syscallarg(const struct timeval *) tptr;
2523 } */ *uap = v;
2524 int error;
2525 struct nameidata nd;
2526
2527 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2528 if ((error = namei(&nd)) != 0)
2529 return (error);
2530
2531 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2532
2533 vrele(nd.ni_vp);
2534 return (error);
2535 }
2536
2537 /*
2538 * Set the access and modification times given a file descriptor.
2539 */
2540 /* ARGSUSED */
2541 int
2542 sys_futimes(p, v, retval)
2543 struct proc *p;
2544 void *v;
2545 register_t *retval;
2546 {
2547 struct sys_futimes_args /* {
2548 syscallarg(int) fd;
2549 syscallarg(const struct timeval *) tptr;
2550 } */ *uap = v;
2551 int error;
2552 struct file *fp;
2553
2554 /* getvnode() will use the descriptor for us */
2555 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2556 return (error);
2557
2558 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2559 FILE_UNUSE(fp, p);
2560 return (error);
2561 }
2562
2563 /*
2564 * Set the access and modification times given a path name; this
2565 * version does not follow links.
2566 */
2567 /* ARGSUSED */
2568 int
2569 sys_lutimes(p, v, retval)
2570 struct proc *p;
2571 void *v;
2572 register_t *retval;
2573 {
2574 struct sys_lutimes_args /* {
2575 syscallarg(const char *) path;
2576 syscallarg(const struct timeval *) tptr;
2577 } */ *uap = v;
2578 int error;
2579 struct nameidata nd;
2580
2581 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2582 if ((error = namei(&nd)) != 0)
2583 return (error);
2584
2585 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2586
2587 vrele(nd.ni_vp);
2588 return (error);
2589 }
2590
2591 /*
2592 * Common routine to set access and modification times given a vnode.
2593 */
2594 static int
2595 change_utimes(vp, tptr, p)
2596 struct vnode *vp;
2597 const struct timeval *tptr;
2598 struct proc *p;
2599 {
2600 struct timeval tv[2];
2601 struct vattr vattr;
2602 int error;
2603
2604 VATTR_NULL(&vattr);
2605 if (tptr == NULL) {
2606 microtime(&tv[0]);
2607 tv[1] = tv[0];
2608 vattr.va_vaflags |= VA_UTIMES_NULL;
2609 } else {
2610 error = copyin(tptr, tv, sizeof(tv));
2611 if (error)
2612 return (error);
2613 }
2614 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2615 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2616 vattr.va_atime.tv_sec = tv[0].tv_sec;
2617 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2618 vattr.va_mtime.tv_sec = tv[1].tv_sec;
2619 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2620 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2621 VOP_UNLOCK(vp, 0);
2622 return (error);
2623 }
2624
2625 /*
2626 * Truncate a file given its path name.
2627 */
2628 /* ARGSUSED */
2629 int
2630 sys_truncate(p, v, retval)
2631 struct proc *p;
2632 void *v;
2633 register_t *retval;
2634 {
2635 struct sys_truncate_args /* {
2636 syscallarg(const char *) path;
2637 syscallarg(int) pad;
2638 syscallarg(off_t) length;
2639 } */ *uap = v;
2640 struct vnode *vp;
2641 struct vattr vattr;
2642 int error;
2643 struct nameidata nd;
2644
2645 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2646 if ((error = namei(&nd)) != 0)
2647 return (error);
2648 vp = nd.ni_vp;
2649 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2650 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2651 if (vp->v_type == VDIR)
2652 error = EISDIR;
2653 else if ((error = vn_writechk(vp)) == 0 &&
2654 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2655 VATTR_NULL(&vattr);
2656 vattr.va_size = SCARG(uap, length);
2657 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2658 }
2659 vput(vp);
2660 return (error);
2661 }
2662
2663 /*
2664 * Truncate a file given a file descriptor.
2665 */
2666 /* ARGSUSED */
2667 int
2668 sys_ftruncate(p, v, retval)
2669 struct proc *p;
2670 void *v;
2671 register_t *retval;
2672 {
2673 struct sys_ftruncate_args /* {
2674 syscallarg(int) fd;
2675 syscallarg(int) pad;
2676 syscallarg(off_t) length;
2677 } */ *uap = v;
2678 struct vattr vattr;
2679 struct vnode *vp;
2680 struct file *fp;
2681 int error;
2682
2683 /* getvnode() will use the descriptor for us */
2684 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2685 return (error);
2686 if ((fp->f_flag & FWRITE) == 0) {
2687 error = EINVAL;
2688 goto out;
2689 }
2690 vp = (struct vnode *)fp->f_data;
2691 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2692 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2693 if (vp->v_type == VDIR)
2694 error = EISDIR;
2695 else if ((error = vn_writechk(vp)) == 0) {
2696 VATTR_NULL(&vattr);
2697 vattr.va_size = SCARG(uap, length);
2698 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2699 }
2700 VOP_UNLOCK(vp, 0);
2701 out:
2702 FILE_UNUSE(fp, p);
2703 return (error);
2704 }
2705
2706 /*
2707 * Sync an open file.
2708 */
2709 /* ARGSUSED */
2710 int
2711 sys_fsync(p, v, retval)
2712 struct proc *p;
2713 void *v;
2714 register_t *retval;
2715 {
2716 struct sys_fsync_args /* {
2717 syscallarg(int) fd;
2718 } */ *uap = v;
2719 struct vnode *vp;
2720 struct file *fp;
2721 int error;
2722
2723 /* getvnode() will use the descriptor for us */
2724 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2725 return (error);
2726 vp = (struct vnode *)fp->f_data;
2727 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2728 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
2729 if (error == 0 && bioops.io_fsync != NULL &&
2730 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2731 (*bioops.io_fsync)(vp);
2732 VOP_UNLOCK(vp, 0);
2733 FILE_UNUSE(fp, p);
2734 return (error);
2735 }
2736
2737 /*
2738 * Sync the data of an open file.
2739 */
2740 /* ARGSUSED */
2741 int
2742 sys_fdatasync(p, v, retval)
2743 struct proc *p;
2744 void *v;
2745 register_t *retval;
2746 {
2747 struct sys_fdatasync_args /* {
2748 syscallarg(int) fd;
2749 } */ *uap = v;
2750 struct vnode *vp;
2751 struct file *fp;
2752 int error;
2753
2754 /* getvnode() will use the descriptor for us */
2755 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2756 return (error);
2757 vp = (struct vnode *)fp->f_data;
2758 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2759 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
2760 VOP_UNLOCK(vp, 0);
2761 FILE_UNUSE(fp, p);
2762 return (error);
2763 }
2764
2765 /*
2766 * Rename files, (standard) BSD semantics frontend.
2767 */
2768 /* ARGSUSED */
2769 int
2770 sys_rename(p, v, retval)
2771 struct proc *p;
2772 void *v;
2773 register_t *retval;
2774 {
2775 struct sys_rename_args /* {
2776 syscallarg(const char *) from;
2777 syscallarg(const char *) to;
2778 } */ *uap = v;
2779
2780 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
2781 }
2782
2783 /*
2784 * Rename files, POSIX semantics frontend.
2785 */
2786 /* ARGSUSED */
2787 int
2788 sys___posix_rename(p, v, retval)
2789 struct proc *p;
2790 void *v;
2791 register_t *retval;
2792 {
2793 struct sys___posix_rename_args /* {
2794 syscallarg(const char *) from;
2795 syscallarg(const char *) to;
2796 } */ *uap = v;
2797
2798 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
2799 }
2800
2801 /*
2802 * Rename files. Source and destination must either both be directories,
2803 * or both not be directories. If target is a directory, it must be empty.
2804 * If `from' and `to' refer to the same object, the value of the `retain'
2805 * argument is used to determine whether `from' will be
2806 *
2807 * (retain == 0) deleted unless `from' and `to' refer to the same
2808 * object in the file system's name space (BSD).
2809 * (retain == 1) always retained (POSIX).
2810 */
2811 static int
2812 rename_files(from, to, p, retain)
2813 const char *from, *to;
2814 struct proc *p;
2815 int retain;
2816 {
2817 struct vnode *tvp, *fvp, *tdvp;
2818 struct nameidata fromnd, tond;
2819 int error;
2820
2821 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2822 from, p);
2823 if ((error = namei(&fromnd)) != 0)
2824 return (error);
2825 fvp = fromnd.ni_vp;
2826 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
2827 UIO_USERSPACE, to, p);
2828 if ((error = namei(&tond)) != 0) {
2829 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2830 vrele(fromnd.ni_dvp);
2831 vrele(fvp);
2832 goto out1;
2833 }
2834 tdvp = tond.ni_dvp;
2835 tvp = tond.ni_vp;
2836
2837 if (tvp != NULL) {
2838 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2839 error = ENOTDIR;
2840 goto out;
2841 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2842 error = EISDIR;
2843 goto out;
2844 }
2845 }
2846
2847 if (fvp == tdvp)
2848 error = EINVAL;
2849
2850 /*
2851 * Source and destination refer to the same object.
2852 */
2853 if (fvp == tvp) {
2854 if (retain)
2855 error = -1;
2856 else if (fromnd.ni_dvp == tdvp &&
2857 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2858 !memcmp(fromnd.ni_cnd.cn_nameptr,
2859 tond.ni_cnd.cn_nameptr,
2860 fromnd.ni_cnd.cn_namelen))
2861 error = -1;
2862 }
2863
2864 out:
2865 if (!error) {
2866 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2867 if (fromnd.ni_dvp != tdvp)
2868 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2869 if (tvp) {
2870 (void)uvm_vnp_uncache(tvp);
2871 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2872 }
2873 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2874 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2875 } else {
2876 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2877 if (tdvp == tvp)
2878 vrele(tdvp);
2879 else
2880 vput(tdvp);
2881 if (tvp)
2882 vput(tvp);
2883 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2884 vrele(fromnd.ni_dvp);
2885 vrele(fvp);
2886 }
2887 vrele(tond.ni_startdir);
2888 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2889 out1:
2890 if (fromnd.ni_startdir)
2891 vrele(fromnd.ni_startdir);
2892 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2893 return (error == -1 ? 0 : error);
2894 }
2895
2896 /*
2897 * Make a directory file.
2898 */
2899 /* ARGSUSED */
2900 int
2901 sys_mkdir(p, v, retval)
2902 struct proc *p;
2903 void *v;
2904 register_t *retval;
2905 {
2906 struct sys_mkdir_args /* {
2907 syscallarg(const char *) path;
2908 syscallarg(int) mode;
2909 } */ *uap = v;
2910 struct vnode *vp;
2911 struct vattr vattr;
2912 int error;
2913 struct nameidata nd;
2914
2915 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
2916 if ((error = namei(&nd)) != 0)
2917 return (error);
2918 vp = nd.ni_vp;
2919 if (vp != NULL) {
2920 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2921 if (nd.ni_dvp == vp)
2922 vrele(nd.ni_dvp);
2923 else
2924 vput(nd.ni_dvp);
2925 vrele(vp);
2926 return (EEXIST);
2927 }
2928 VATTR_NULL(&vattr);
2929 vattr.va_type = VDIR;
2930 vattr.va_mode =
2931 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
2932 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2933 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2934 if (!error)
2935 vput(nd.ni_vp);
2936 return (error);
2937 }
2938
2939 /*
2940 * Remove a directory file.
2941 */
2942 /* ARGSUSED */
2943 int
2944 sys_rmdir(p, v, retval)
2945 struct proc *p;
2946 void *v;
2947 register_t *retval;
2948 {
2949 struct sys_rmdir_args /* {
2950 syscallarg(const char *) path;
2951 } */ *uap = v;
2952 struct vnode *vp;
2953 int error;
2954 struct nameidata nd;
2955
2956 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
2957 SCARG(uap, path), p);
2958 if ((error = namei(&nd)) != 0)
2959 return (error);
2960 vp = nd.ni_vp;
2961 if (vp->v_type != VDIR) {
2962 error = ENOTDIR;
2963 goto out;
2964 }
2965 /*
2966 * No rmdir "." please.
2967 */
2968 if (nd.ni_dvp == vp) {
2969 error = EINVAL;
2970 goto out;
2971 }
2972 /*
2973 * The root of a mounted filesystem cannot be deleted.
2974 */
2975 if (vp->v_flag & VROOT)
2976 error = EBUSY;
2977 out:
2978 if (!error) {
2979 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2980 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2981 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2982 } else {
2983 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2984 if (nd.ni_dvp == vp)
2985 vrele(nd.ni_dvp);
2986 else
2987 vput(nd.ni_dvp);
2988 vput(vp);
2989 }
2990 return (error);
2991 }
2992
2993 /*
2994 * Read a block of directory entries in a file system independent format.
2995 */
2996 int
2997 sys_getdents(p, v, retval)
2998 struct proc *p;
2999 void *v;
3000 register_t *retval;
3001 {
3002 struct sys_getdents_args /* {
3003 syscallarg(int) fd;
3004 syscallarg(char *) buf;
3005 syscallarg(size_t) count;
3006 } */ *uap = v;
3007 struct file *fp;
3008 int error, done;
3009
3010 /* getvnode() will use the descriptor for us */
3011 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3012 return (error);
3013 if ((fp->f_flag & FREAD) == 0) {
3014 error = EBADF;
3015 goto out;
3016 }
3017 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3018 SCARG(uap, count), &done, p, 0, 0);
3019 *retval = done;
3020 out:
3021 FILE_UNUSE(fp, p);
3022 return (error);
3023 }
3024
3025 /*
3026 * Set the mode mask for creation of filesystem nodes.
3027 */
3028 int
3029 sys_umask(p, v, retval)
3030 struct proc *p;
3031 void *v;
3032 register_t *retval;
3033 {
3034 struct sys_umask_args /* {
3035 syscallarg(mode_t) newmask;
3036 } */ *uap = v;
3037 struct cwdinfo *cwdi;
3038
3039 cwdi = p->p_cwdi;
3040 *retval = cwdi->cwdi_cmask;
3041 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3042 return (0);
3043 }
3044
3045 /*
3046 * Void all references to file by ripping underlying filesystem
3047 * away from vnode.
3048 */
3049 /* ARGSUSED */
3050 int
3051 sys_revoke(p, v, retval)
3052 struct proc *p;
3053 void *v;
3054 register_t *retval;
3055 {
3056 struct sys_revoke_args /* {
3057 syscallarg(const char *) path;
3058 } */ *uap = v;
3059 struct vnode *vp;
3060 struct vattr vattr;
3061 int error;
3062 struct nameidata nd;
3063
3064 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3065 if ((error = namei(&nd)) != 0)
3066 return (error);
3067 vp = nd.ni_vp;
3068 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3069 goto out;
3070 if (p->p_ucred->cr_uid != vattr.va_uid &&
3071 (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3072 goto out;
3073 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3074 VOP_REVOKE(vp, REVOKEALL);
3075 out:
3076 vrele(vp);
3077 return (error);
3078 }
3079
3080 /*
3081 * Convert a user file descriptor to a kernel file entry.
3082 */
3083 int
3084 getvnode(fdp, fd, fpp)
3085 struct filedesc *fdp;
3086 int fd;
3087 struct file **fpp;
3088 {
3089 struct vnode *vp;
3090 struct file *fp;
3091
3092 if ((u_int)fd >= fdp->fd_nfiles ||
3093 (fp = fdp->fd_ofiles[fd]) == NULL ||
3094 (fp->f_iflags & FIF_WANTCLOSE) != 0)
3095 return (EBADF);
3096
3097 FILE_USE(fp);
3098
3099 if (fp->f_type != DTYPE_VNODE) {
3100 FILE_UNUSE(fp, NULL);
3101 return (EINVAL);
3102 }
3103
3104 vp = (struct vnode *)fp->f_data;
3105 if (vp->v_type == VBAD) {
3106 FILE_UNUSE(fp, NULL);
3107 return (EBADF);
3108 }
3109
3110 *fpp = fp;
3111 return (0);
3112 }
3113