vfs_syscalls.c revision 1.185 1 /* $NetBSD: vfs_syscalls.c,v 1.185 2003/04/16 21:44:21 christos 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 1.185 2003/04/16 21:44:21 christos 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/proc.h>
59 #include <sys/uio.h>
60 #include <sys/malloc.h>
61 #include <sys/dirent.h>
62 #include <sys/sysctl.h>
63 #include <sys/sa.h>
64 #include <sys/syscallargs.h>
65
66 #include <miscfs/genfs/genfs.h>
67 #include <miscfs/syncfs/syncfs.h>
68
69 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
70
71 static int change_dir __P((struct nameidata *, struct proc *));
72 static int change_flags __P((struct vnode *, u_long, struct proc *));
73 static int change_mode __P((struct vnode *, int, struct proc *p));
74 static int change_owner __P((struct vnode *, uid_t, gid_t, struct proc *,
75 int));
76 static int change_utimes __P((struct vnode *vp, const struct timeval *,
77 struct proc *p));
78 static int rename_files __P((const char *, const char *, struct proc *, int));
79 static int dostatfs __P((struct mount *, struct statfs *, struct proc *, int,
80 int));
81
82 void checkdirs __P((struct vnode *));
83
84 int dovfsusermount = 0;
85
86 /*
87 * Virtual File System System Calls
88 */
89
90 /*
91 * Mount a file system.
92 */
93
94 #if defined(COMPAT_09) || defined(COMPAT_43)
95 /*
96 * This table is used to maintain compatibility with 4.3BSD
97 * and NetBSD 0.9 mount syscalls. Note, the order is important!
98 *
99 * Do not modify this table. It should only contain filesystems
100 * supported by NetBSD 0.9 and 4.3BSD.
101 */
102 const char * const mountcompatnames[] = {
103 NULL, /* 0 = MOUNT_NONE */
104 MOUNT_FFS, /* 1 = MOUNT_UFS */
105 MOUNT_NFS, /* 2 */
106 MOUNT_MFS, /* 3 */
107 MOUNT_MSDOS, /* 4 */
108 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */
109 MOUNT_FDESC, /* 6 */
110 MOUNT_KERNFS, /* 7 */
111 NULL, /* 8 = MOUNT_DEVFS */
112 MOUNT_AFS, /* 9 */
113 };
114 const int nmountcompatnames = sizeof(mountcompatnames) /
115 sizeof(mountcompatnames[0]);
116 #endif /* COMPAT_09 || COMPAT_43 */
117
118 /* ARGSUSED */
119 int
120 sys_mount(l, v, retval)
121 struct lwp *l;
122 void *v;
123 register_t *retval;
124 {
125 struct sys_mount_args /* {
126 syscallarg(const char *) type;
127 syscallarg(const char *) path;
128 syscallarg(int) flags;
129 syscallarg(void *) data;
130 } */ *uap = v;
131 struct proc *p = l->l_proc;
132 struct vnode *vp;
133 struct mount *mp;
134 int error, flag = 0;
135 char fstypename[MFSNAMELEN];
136 struct vattr va;
137 struct nameidata nd;
138 struct vfsops *vfs;
139
140 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 &&
141 (error = suser(p->p_ucred, &p->p_acflag)))
142 return (error);
143 /*
144 * Get vnode to be covered
145 */
146 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE,
147 SCARG(uap, path), p);
148 if ((error = namei(&nd)) != 0)
149 return (error);
150 vp = nd.ni_vp;
151 /*
152 * A lookup in VFS_MOUNT might result in an attempt to
153 * lock this vnode again, so make the lock recursive.
154 */
155 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
156 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) {
157 if ((vp->v_flag & VROOT) == 0) {
158 vput(vp);
159 return (EINVAL);
160 }
161 mp = vp->v_mount;
162 flag = mp->mnt_flag;
163 vfs = mp->mnt_op;
164 /*
165 * We only allow the filesystem to be reloaded if it
166 * is currently mounted read-only.
167 */
168 if ((SCARG(uap, flags) & MNT_RELOAD) &&
169 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
170 vput(vp);
171 return (EOPNOTSUPP); /* Needs translation */
172 }
173 /*
174 * In "highly secure" mode, don't let the caller do anything
175 * but downgrade a filesystem from read-write to read-only.
176 * (see also below; MNT_UPDATE or MNT_GETARGS is required.)
177 */
178 if (securelevel >= 2 &&
179 SCARG(uap, flags) != MNT_GETARGS &&
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 |= SCARG(uap, flags) &
187 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
188 /*
189 * Only root, or the user that did the original mount is
190 * permitted to update it.
191 */
192 if ((mp->mnt_flag & MNT_GETARGS) == 0 &&
193 mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
194 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
195 vput(vp);
196 return (error);
197 }
198 /*
199 * Do not allow NFS export by non-root users. For non-root
200 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
201 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
202 */
203 if (p->p_ucred->cr_uid != 0) {
204 if (SCARG(uap, flags) & MNT_EXPORTED) {
205 vput(vp);
206 return (EPERM);
207 }
208 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
209 if (flag & MNT_NOEXEC)
210 SCARG(uap, flags) |= MNT_NOEXEC;
211 }
212 if (vfs_busy(mp, LK_NOWAIT, 0)) {
213 vput(vp);
214 return (EPERM);
215 }
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_IGNORE | MNT_NOATIME | MNT_NODEVMTIME |
315 MNT_SYMPERM | 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 | MNT_GETARGS)) {
321 if (mp->mnt_flag & MNT_WANTRDWR)
322 mp->mnt_flag &= ~MNT_RDONLY;
323 if (error || (mp->mnt_flag & MNT_GETARGS))
324 mp->mnt_flag = flag;
325 mp->mnt_flag &=~
326 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR |
327 MNT_GETARGS);
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 VOP_UNLOCK(vp, 0);
337 vrele(vp);
338 return (error);
339 }
340 /*
341 * Put the new filesystem on the mount list after root.
342 */
343 cache_purge(vp);
344 if (!error) {
345 vp->v_mountedhere = mp;
346 simple_lock(&mountlist_slock);
347 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
348 simple_unlock(&mountlist_slock);
349 checkdirs(vp);
350 VOP_UNLOCK(vp, 0);
351 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
352 error = vfs_allocate_syncvnode(mp);
353 vfs_unbusy(mp);
354 (void) VFS_STATFS(mp, &mp->mnt_stat, p);
355 if ((error = VFS_START(mp, 0, p)))
356 vrele(vp);
357 } else {
358 vp->v_mountedhere = (struct mount *)0;
359 vfs->vfs_refcount--;
360 vfs_unbusy(mp);
361 free(mp, M_MOUNT);
362 vput(vp);
363 }
364 return (error);
365 }
366
367 /*
368 * Scan all active processes to see if any of them have a current
369 * or root directory onto which the new filesystem has just been
370 * mounted. If so, replace them with the new mount point.
371 */
372 void
373 checkdirs(olddp)
374 struct vnode *olddp;
375 {
376 struct cwdinfo *cwdi;
377 struct vnode *newdp;
378 struct proc *p;
379
380 if (olddp->v_usecount == 1)
381 return;
382 if (VFS_ROOT(olddp->v_mountedhere, &newdp))
383 panic("mount: lost mount");
384 proclist_lock_read();
385 LIST_FOREACH(p, &allproc, p_list) {
386 cwdi = p->p_cwdi;
387 if (cwdi->cwdi_cdir == olddp) {
388 vrele(cwdi->cwdi_cdir);
389 VREF(newdp);
390 cwdi->cwdi_cdir = newdp;
391 }
392 if (cwdi->cwdi_rdir == olddp) {
393 vrele(cwdi->cwdi_rdir);
394 VREF(newdp);
395 cwdi->cwdi_rdir = newdp;
396 }
397 }
398 proclist_unlock_read();
399 if (rootvnode == olddp) {
400 vrele(rootvnode);
401 VREF(newdp);
402 rootvnode = newdp;
403 }
404 vput(newdp);
405 }
406
407 /*
408 * Unmount a file system.
409 *
410 * Note: unmount takes a path to the vnode mounted on as argument,
411 * not special file (as before).
412 */
413 /* ARGSUSED */
414 int
415 sys_unmount(l, v, retval)
416 struct lwp *l;
417 void *v;
418 register_t *retval;
419 {
420 struct sys_unmount_args /* {
421 syscallarg(const char *) path;
422 syscallarg(int) flags;
423 } */ *uap = v;
424 struct proc *p = l->l_proc;
425 struct vnode *vp;
426 struct mount *mp;
427 int error;
428 struct nameidata nd;
429
430 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
431 SCARG(uap, path), p);
432 if ((error = namei(&nd)) != 0)
433 return (error);
434 vp = nd.ni_vp;
435 mp = vp->v_mount;
436
437 /*
438 * Only root, or the user that did the original mount is
439 * permitted to unmount this filesystem.
440 */
441 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
442 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
443 vput(vp);
444 return (error);
445 }
446
447 /*
448 * Don't allow unmounting the root file system.
449 */
450 if (mp->mnt_flag & MNT_ROOTFS) {
451 vput(vp);
452 return (EINVAL);
453 }
454
455 /*
456 * Must be the root of the filesystem
457 */
458 if ((vp->v_flag & VROOT) == 0) {
459 vput(vp);
460 return (EINVAL);
461 }
462 vput(vp);
463
464 /*
465 * XXX Freeze syncer. Must do this before locking the
466 * mount point. See dounmount() for details.
467 */
468 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
469
470 if (vfs_busy(mp, 0, 0)) {
471 lockmgr(&syncer_lock, LK_RELEASE, NULL);
472 return (EBUSY);
473 }
474
475 return (dounmount(mp, SCARG(uap, flags), p));
476 }
477
478 /*
479 * Do the actual file system unmount. File system is assumed to have been
480 * marked busy by the caller.
481 */
482 int
483 dounmount(mp, flags, p)
484 struct mount *mp;
485 int flags;
486 struct proc *p;
487 {
488 struct vnode *coveredvp;
489 int error;
490 int async;
491 int used_syncer;
492
493 simple_lock(&mountlist_slock);
494 vfs_unbusy(mp);
495 used_syncer = (mp->mnt_syncer != NULL);
496
497 /*
498 * XXX Syncer must be frozen when we get here. This should really
499 * be done on a per-mountpoint basis, but especially the softdep
500 * code possibly called from the syncer doens't exactly work on a
501 * per-mountpoint basis, so the softdep code would become a maze
502 * of vfs_busy() calls.
503 *
504 * The caller of dounmount() must acquire syncer_lock because
505 * the syncer itself acquires locks in syncer_lock -> vfs_busy
506 * order, and we must preserve that order to avoid deadlock.
507 *
508 * So, if the file system did not use the syncer, now is
509 * the time to release the syncer_lock.
510 */
511 if (used_syncer == 0)
512 lockmgr(&syncer_lock, LK_RELEASE, NULL);
513
514 mp->mnt_flag |= MNT_UNMOUNT;
515 mp->mnt_unmounter = p;
516 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
517 if (mp->mnt_flag & MNT_EXPUBLIC)
518 vfs_setpublicfs(NULL, NULL, NULL);
519 async = mp->mnt_flag & MNT_ASYNC;
520 mp->mnt_flag &= ~MNT_ASYNC;
521 cache_purgevfs(mp); /* remove cache entries for this file sys */
522 if (mp->mnt_syncer != NULL)
523 vfs_deallocate_syncvnode(mp);
524 if (((mp->mnt_flag & MNT_RDONLY) ||
525 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
526 (flags & MNT_FORCE))
527 error = VFS_UNMOUNT(mp, flags, p);
528 simple_lock(&mountlist_slock);
529 if (error) {
530 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
531 (void) vfs_allocate_syncvnode(mp);
532 mp->mnt_flag &= ~MNT_UNMOUNT;
533 mp->mnt_unmounter = NULL;
534 mp->mnt_flag |= async;
535 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
536 &mountlist_slock);
537 if (used_syncer)
538 lockmgr(&syncer_lock, LK_RELEASE, NULL);
539 while (mp->mnt_wcnt > 0) {
540 wakeup(mp);
541 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0);
542 }
543 return (error);
544 }
545 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
546 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
547 coveredvp->v_mountedhere = NULL;
548 vrele(coveredvp);
549 }
550 mp->mnt_op->vfs_refcount--;
551 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL)
552 panic("unmount: dangling vnode");
553 mp->mnt_flag |= MNT_GONE;
554 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
555 if (used_syncer)
556 lockmgr(&syncer_lock, LK_RELEASE, NULL);
557 while(mp->mnt_wcnt > 0) {
558 wakeup(mp);
559 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0);
560 }
561 free(mp, M_MOUNT);
562 return (0);
563 }
564
565 /*
566 * Sync each mounted filesystem.
567 */
568 #ifdef DEBUG
569 int syncprt = 0;
570 struct ctldebug debug0 = { "syncprt", &syncprt };
571 #endif
572
573 /* ARGSUSED */
574 int
575 sys_sync(l, v, retval)
576 struct lwp *l;
577 void *v;
578 register_t *retval;
579 {
580 struct mount *mp, *nmp;
581 int asyncflag;
582 struct proc *p = l == NULL ? &proc0 : l->l_proc;
583
584 simple_lock(&mountlist_slock);
585 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
586 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
587 nmp = mp->mnt_list.cqe_prev;
588 continue;
589 }
590 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
591 asyncflag = mp->mnt_flag & MNT_ASYNC;
592 mp->mnt_flag &= ~MNT_ASYNC;
593 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
594 if (asyncflag)
595 mp->mnt_flag |= MNT_ASYNC;
596 }
597 simple_lock(&mountlist_slock);
598 nmp = mp->mnt_list.cqe_prev;
599 vfs_unbusy(mp);
600
601 }
602 simple_unlock(&mountlist_slock);
603 #ifdef DEBUG
604 if (syncprt)
605 vfs_bufstats();
606 #endif /* DEBUG */
607 return (0);
608 }
609
610 /*
611 * Change filesystem quotas.
612 */
613 /* ARGSUSED */
614 int
615 sys_quotactl(l, v, retval)
616 struct lwp *l;
617 void *v;
618 register_t *retval;
619 {
620 struct sys_quotactl_args /* {
621 syscallarg(const char *) path;
622 syscallarg(int) cmd;
623 syscallarg(int) uid;
624 syscallarg(caddr_t) arg;
625 } */ *uap = v;
626 struct proc *p = l->l_proc;
627 struct mount *mp;
628 int error;
629 struct nameidata nd;
630
631 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
632 if ((error = namei(&nd)) != 0)
633 return (error);
634 mp = nd.ni_vp->v_mount;
635 vrele(nd.ni_vp);
636 return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
637 SCARG(uap, arg), p));
638 }
639
640 static int
641 dostatfs(struct mount *mp, struct statfs *sp, struct proc *p, int flags,
642 int root)
643 {
644 struct cwdinfo *cwdi = p->p_cwdi;
645 int error = 0;
646
647 /*
648 * If MNT_NOWAIT or MNT_LAZY is specified, do not
649 * refresh the fsstat cache. MNT_WAIT or MNT_LAXY
650 * overrides MNT_NOWAIT.
651 */
652 if (flags == MNT_NOWAIT || flags == MNT_LAZY ||
653 (flags != MNT_WAIT && flags != 0)) {
654 memcpy(sp, &mp->mnt_stat, sizeof(*sp));
655 goto done;
656 }
657
658 if ((error = VFS_STATFS(mp, sp, p)) != 0) {
659 return error;
660 }
661
662 if (cwdi->cwdi_rdir == NULL)
663 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
664 done:
665 if (cwdi->cwdi_rdir != NULL) {
666 size_t len;
667 char *bp;
668 char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
669 if (!path)
670 return ENOMEM;
671
672 bp = path + MAXPATHLEN;
673 *--bp = '\0';
674 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
675 MAXPATHLEN / 2, 0, p);
676 if (error) {
677 free(path, M_TEMP);
678 return error;
679 }
680 len = strlen(bp);
681 /*
682 * for mount points that are below our root, we can see
683 * them, so we fix up the pathname and return them. The
684 * rest we cannot see, so we don't allow viewing the
685 * data.
686 */
687 if (strncmp(bp, sp->f_mntonname, len) == 0) {
688 strcpy(sp->f_mntonname, &sp->f_mntonname[len]);
689 if (sp->f_mntonname[0] == '\0')
690 (void)strcpy(sp->f_mntonname, "/");
691 } else {
692 if (root)
693 (void)strcpy(sp->f_mntonname, "/");
694 else
695 error = EPERM;
696 }
697 free(path, M_TEMP);
698 }
699 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
700 sp->f_oflags = sp->f_flags & 0xffff;
701 return error;
702 }
703
704 /*
705 * Get filesystem statistics.
706 */
707 /* ARGSUSED */
708 int
709 sys_statfs(l, v, retval)
710 struct lwp *l;
711 void *v;
712 register_t *retval;
713 {
714 struct sys_statfs_args /* {
715 syscallarg(const char *) path;
716 syscallarg(struct statfs *) buf;
717 } */ *uap = v;
718 struct proc *p = l->l_proc;
719 struct mount *mp;
720 struct statfs sbuf;
721 int error;
722 struct nameidata nd;
723
724 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
725 if ((error = namei(&nd)) != 0)
726 return error;
727 mp = nd.ni_vp->v_mount;
728 vrele(nd.ni_vp);
729 if ((error = dostatfs(mp, &sbuf, p, 0, 1)) != 0)
730 return error;
731 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
732 }
733
734 /*
735 * Get filesystem statistics.
736 */
737 /* ARGSUSED */
738 int
739 sys_fstatfs(l, v, retval)
740 struct lwp *l;
741 void *v;
742 register_t *retval;
743 {
744 struct sys_fstatfs_args /* {
745 syscallarg(int) fd;
746 syscallarg(struct statfs *) buf;
747 } */ *uap = v;
748 struct proc *p = l->l_proc;
749 struct file *fp;
750 struct mount *mp;
751 struct statfs sbuf;
752 int error;
753
754 /* getvnode() will use the descriptor for us */
755 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
756 return (error);
757 mp = ((struct vnode *)fp->f_data)->v_mount;
758 if ((error = dostatfs(mp, &sbuf, p, 0, 1)) != 0)
759 goto out;
760 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
761 out:
762 FILE_UNUSE(fp, p);
763 return error;
764 }
765
766
767 /*
768 * Get statistics on all filesystems.
769 */
770 int
771 sys_getfsstat(l, v, retval)
772 struct lwp *l;
773 void *v;
774 register_t *retval;
775 {
776 struct sys_getfsstat_args /* {
777 syscallarg(struct statfs *) buf;
778 syscallarg(long) bufsize;
779 syscallarg(int) flags;
780 } */ *uap = v;
781 int root = 0;
782 struct proc *p = l->l_proc;
783 struct mount *mp, *nmp;
784 struct statfs sbuf;
785 caddr_t sfsp;
786 long count, maxcount, error = 0;
787
788 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
789 sfsp = (caddr_t)SCARG(uap, buf);
790 simple_lock(&mountlist_slock);
791 count = 0;
792 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
793 mp = nmp) {
794 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
795 nmp = CIRCLEQ_NEXT(mp, mnt_list);
796 continue;
797 }
798 if (sfsp && count < maxcount) {
799 error = dostatfs(mp, &sbuf, p, SCARG(uap, flags), 0);
800 if (error) {
801 nmp = CIRCLEQ_NEXT(mp, mnt_list);
802 vfs_unbusy(mp);
803 continue;
804 }
805 error = copyout(&sbuf, sfsp, sizeof(sbuf));
806 if (error) {
807 vfs_unbusy(mp);
808 return (error);
809 }
810 sfsp += sizeof(sbuf);
811 root |= strcmp(sbuf.f_mntonname, "/") == 0;
812 }
813 count++;
814 nmp = CIRCLEQ_NEXT(mp, mnt_list);
815 vfs_unbusy(mp);
816 }
817 simple_unlock(&mountlist_slock);
818 if (root == 0 && p->p_cwdi->cwdi_rdir) {
819 /*
820 * fake a root entry
821 */
822 if ((error = dostatfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p,
823 SCARG(uap, flags), 1)) != 0)
824 return error;
825 if (sfsp)
826 error = copyout(&sbuf, sfsp, sizeof(sbuf));
827 count++;
828 }
829 if (sfsp && count > maxcount)
830 *retval = maxcount;
831 else
832 *retval = count;
833 return error;
834 }
835
836 /*
837 * Change current working directory to a given file descriptor.
838 */
839 /* ARGSUSED */
840 int
841 sys_fchdir(l, v, retval)
842 struct lwp *l;
843 void *v;
844 register_t *retval;
845 {
846 struct sys_fchdir_args /* {
847 syscallarg(int) fd;
848 } */ *uap = v;
849 struct proc *p = l->l_proc;
850 struct filedesc *fdp = p->p_fd;
851 struct cwdinfo *cwdi = p->p_cwdi;
852 struct vnode *vp, *tdp;
853 struct mount *mp;
854 struct file *fp;
855 int error;
856
857 /* getvnode() will use the descriptor for us */
858 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
859 return (error);
860 vp = (struct vnode *)fp->f_data;
861
862 VREF(vp);
863 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
864 if (vp->v_type != VDIR)
865 error = ENOTDIR;
866 else
867 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
868 while (!error && (mp = vp->v_mountedhere) != NULL) {
869 if (vfs_busy(mp, 0, 0))
870 continue;
871 error = VFS_ROOT(mp, &tdp);
872 vfs_unbusy(mp);
873 if (error)
874 break;
875 vput(vp);
876 vp = tdp;
877 }
878 if (error) {
879 vput(vp);
880 goto out;
881 }
882 VOP_UNLOCK(vp, 0);
883
884 /*
885 * Disallow changing to a directory not under the process's
886 * current root directory (if there is one).
887 */
888 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
889 vrele(vp);
890 error = EPERM; /* operation not permitted */
891 goto out;
892 }
893
894 vrele(cwdi->cwdi_cdir);
895 cwdi->cwdi_cdir = vp;
896 out:
897 FILE_UNUSE(fp, p);
898 return (error);
899 }
900
901 /*
902 * Change this process's notion of the root directory to a given file descriptor.
903 */
904
905 int
906 sys_fchroot(l, v, retval)
907 struct lwp *l;
908 void *v;
909 register_t *retval;
910 {
911 struct sys_fchroot_args *uap = v;
912 struct proc *p = l->l_proc;
913 struct filedesc *fdp = p->p_fd;
914 struct cwdinfo *cwdi = p->p_cwdi;
915 struct vnode *vp;
916 struct file *fp;
917 int error;
918
919 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
920 return error;
921 /* getvnode() will use the descriptor for us */
922 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
923 return error;
924 vp = (struct vnode *) fp->f_data;
925 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
926 if (vp->v_type != VDIR)
927 error = ENOTDIR;
928 else
929 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
930 VOP_UNLOCK(vp, 0);
931 if (error)
932 goto out;
933 VREF(vp);
934
935 /*
936 * Prevent escaping from chroot by putting the root under
937 * the working directory. Silently chdir to / if we aren't
938 * already there.
939 */
940 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
941 /*
942 * XXX would be more failsafe to change directory to a
943 * deadfs node here instead
944 */
945 vrele(cwdi->cwdi_cdir);
946 VREF(vp);
947 cwdi->cwdi_cdir = vp;
948 }
949
950 if (cwdi->cwdi_rdir != NULL)
951 vrele(cwdi->cwdi_rdir);
952 cwdi->cwdi_rdir = vp;
953 out:
954 FILE_UNUSE(fp, p);
955 return (error);
956 }
957
958
959
960 /*
961 * Change current working directory (``.'').
962 */
963 /* ARGSUSED */
964 int
965 sys_chdir(l, v, retval)
966 struct lwp *l;
967 void *v;
968 register_t *retval;
969 {
970 struct sys_chdir_args /* {
971 syscallarg(const char *) path;
972 } */ *uap = v;
973 struct proc *p = l->l_proc;
974 struct cwdinfo *cwdi = p->p_cwdi;
975 int error;
976 struct nameidata nd;
977
978 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
979 SCARG(uap, path), p);
980 if ((error = change_dir(&nd, p)) != 0)
981 return (error);
982 vrele(cwdi->cwdi_cdir);
983 cwdi->cwdi_cdir = nd.ni_vp;
984 return (0);
985 }
986
987 /*
988 * Change notion of root (``/'') directory.
989 */
990 /* ARGSUSED */
991 int
992 sys_chroot(l, v, retval)
993 struct lwp *l;
994 void *v;
995 register_t *retval;
996 {
997 struct sys_chroot_args /* {
998 syscallarg(const char *) path;
999 } */ *uap = v;
1000 struct proc *p = l->l_proc;
1001 struct cwdinfo *cwdi = p->p_cwdi;
1002 struct vnode *vp;
1003 int error;
1004 struct nameidata nd;
1005
1006 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1007 return (error);
1008 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1009 SCARG(uap, path), p);
1010 if ((error = change_dir(&nd, p)) != 0)
1011 return (error);
1012 if (cwdi->cwdi_rdir != NULL)
1013 vrele(cwdi->cwdi_rdir);
1014 vp = nd.ni_vp;
1015 cwdi->cwdi_rdir = vp;
1016
1017 /*
1018 * Prevent escaping from chroot by putting the root under
1019 * the working directory. Silently chdir to / if we aren't
1020 * already there.
1021 */
1022 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
1023 /*
1024 * XXX would be more failsafe to change directory to a
1025 * deadfs node here instead
1026 */
1027 vrele(cwdi->cwdi_cdir);
1028 VREF(vp);
1029 cwdi->cwdi_cdir = vp;
1030 }
1031
1032 return (0);
1033 }
1034
1035 /*
1036 * Common routine for chroot and chdir.
1037 */
1038 static int
1039 change_dir(ndp, p)
1040 struct nameidata *ndp;
1041 struct proc *p;
1042 {
1043 struct vnode *vp;
1044 int error;
1045
1046 if ((error = namei(ndp)) != 0)
1047 return (error);
1048 vp = ndp->ni_vp;
1049 if (vp->v_type != VDIR)
1050 error = ENOTDIR;
1051 else
1052 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
1053
1054 if (error)
1055 vput(vp);
1056 else
1057 VOP_UNLOCK(vp, 0);
1058 return (error);
1059 }
1060
1061 /*
1062 * Check permissions, allocate an open file structure,
1063 * and call the device open routine if any.
1064 */
1065 int
1066 sys_open(l, v, retval)
1067 struct lwp *l;
1068 void *v;
1069 register_t *retval;
1070 {
1071 struct sys_open_args /* {
1072 syscallarg(const char *) path;
1073 syscallarg(int) flags;
1074 syscallarg(int) mode;
1075 } */ *uap = v;
1076 struct proc *p = l->l_proc;
1077 struct cwdinfo *cwdi = p->p_cwdi;
1078 struct filedesc *fdp = p->p_fd;
1079 struct file *fp;
1080 struct vnode *vp;
1081 int flags, cmode;
1082 int type, indx, error;
1083 struct flock lf;
1084 struct nameidata nd;
1085
1086 flags = FFLAGS(SCARG(uap, flags));
1087 if ((flags & (FREAD | FWRITE)) == 0)
1088 return (EINVAL);
1089 /* falloc() will use the file descriptor for us */
1090 if ((error = falloc(p, &fp, &indx)) != 0)
1091 return (error);
1092 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1093 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1094 p->p_dupfd = -indx - 1; /* XXX check for fdopen */
1095 if ((error = vn_open(&nd, flags, cmode)) != 0) {
1096 FILE_UNUSE(fp, p);
1097 ffree(fp);
1098 if ((error == ENODEV || error == ENXIO) &&
1099 p->p_dupfd >= 0 && /* XXX from fdopen */
1100 (error =
1101 dupfdopen(p, indx, p->p_dupfd, flags, error)) == 0) {
1102 *retval = indx;
1103 return (0);
1104 }
1105 if (error == ERESTART)
1106 error = EINTR;
1107 fdremove(fdp, indx);
1108 return (error);
1109 }
1110 p->p_dupfd = 0;
1111 vp = nd.ni_vp;
1112 fp->f_flag = flags & FMASK;
1113 fp->f_type = DTYPE_VNODE;
1114 fp->f_ops = &vnops;
1115 fp->f_data = vp;
1116 if (flags & (O_EXLOCK | O_SHLOCK)) {
1117 lf.l_whence = SEEK_SET;
1118 lf.l_start = 0;
1119 lf.l_len = 0;
1120 if (flags & O_EXLOCK)
1121 lf.l_type = F_WRLCK;
1122 else
1123 lf.l_type = F_RDLCK;
1124 type = F_FLOCK;
1125 if ((flags & FNONBLOCK) == 0)
1126 type |= F_WAIT;
1127 VOP_UNLOCK(vp, 0);
1128 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1129 if (error) {
1130 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1131 FILE_UNUSE(fp, p);
1132 ffree(fp);
1133 fdremove(fdp, indx);
1134 return (error);
1135 }
1136 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1137 fp->f_flag |= FHASLOCK;
1138 }
1139 VOP_UNLOCK(vp, 0);
1140 *retval = indx;
1141 FILE_SET_MATURE(fp);
1142 FILE_UNUSE(fp, p);
1143 return (0);
1144 }
1145
1146 /*
1147 * Get file handle system call
1148 */
1149 int
1150 sys_getfh(l, v, retval)
1151 struct lwp *l;
1152 void *v;
1153 register_t *retval;
1154 {
1155 struct sys_getfh_args /* {
1156 syscallarg(char *) fname;
1157 syscallarg(fhandle_t *) fhp;
1158 } */ *uap = v;
1159 struct proc *p = l->l_proc;
1160 struct vnode *vp;
1161 fhandle_t fh;
1162 int error;
1163 struct nameidata nd;
1164
1165 /*
1166 * Must be super user
1167 */
1168 error = suser(p->p_ucred, &p->p_acflag);
1169 if (error)
1170 return (error);
1171 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1172 SCARG(uap, fname), p);
1173 error = namei(&nd);
1174 if (error)
1175 return (error);
1176 vp = nd.ni_vp;
1177 memset(&fh, 0, sizeof(fh));
1178 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1179 error = VFS_VPTOFH(vp, &fh.fh_fid);
1180 vput(vp);
1181 if (error)
1182 return (error);
1183 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1184 return (error);
1185 }
1186
1187 /*
1188 * Open a file given a file handle.
1189 *
1190 * Check permissions, allocate an open file structure,
1191 * and call the device open routine if any.
1192 */
1193 int
1194 sys_fhopen(l, v, retval)
1195 struct lwp *l;
1196 void *v;
1197 register_t *retval;
1198 {
1199 struct sys_fhopen_args /* {
1200 syscallarg(const fhandle_t *) fhp;
1201 syscallarg(int) flags;
1202 } */ *uap = v;
1203 struct proc *p = l->l_proc;
1204 struct filedesc *fdp = p->p_fd;
1205 struct file *fp;
1206 struct vnode *vp = NULL;
1207 struct mount *mp;
1208 struct ucred *cred = p->p_ucred;
1209 int flags;
1210 struct file *nfp;
1211 int type, indx, error=0;
1212 struct flock lf;
1213 struct vattr va;
1214 fhandle_t fh;
1215
1216 /*
1217 * Must be super user
1218 */
1219 if ((error = suser(p->p_ucred, &p->p_acflag)))
1220 return (error);
1221
1222 flags = FFLAGS(SCARG(uap, flags));
1223 if ((flags & (FREAD | FWRITE)) == 0)
1224 return (EINVAL);
1225 if ((flags & O_CREAT))
1226 return (EINVAL);
1227 /* falloc() will use the file descriptor for us */
1228 if ((error = falloc(p, &nfp, &indx)) != 0)
1229 return (error);
1230 fp = nfp;
1231 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1232 goto bad;
1233
1234 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1235 error = ESTALE;
1236 goto bad;
1237 }
1238
1239 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1240 vp = NULL; /* most likely unnecessary sanity for bad: */
1241 goto bad;
1242 }
1243
1244 /* Now do an effective vn_open */
1245
1246 if (vp->v_type == VSOCK) {
1247 error = EOPNOTSUPP;
1248 goto bad;
1249 }
1250 if (flags & FREAD) {
1251 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1252 goto bad;
1253 }
1254 if (flags & (FWRITE | O_TRUNC)) {
1255 if (vp->v_type == VDIR) {
1256 error = EISDIR;
1257 goto bad;
1258 }
1259 if ((error = vn_writechk(vp)) != 0 ||
1260 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1261 goto bad;
1262 }
1263 if (flags & O_TRUNC) {
1264 VOP_UNLOCK(vp, 0); /* XXX */
1265 VOP_LEASE(vp, p, cred, LEASE_WRITE);
1266 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
1267 VATTR_NULL(&va);
1268 va.va_size = 0;
1269 if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1270 goto bad;
1271 }
1272 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1273 goto bad;
1274 if (vp->v_type == VREG &&
1275 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1276 error = EIO;
1277 goto bad;
1278 }
1279 if (flags & FWRITE)
1280 vp->v_writecount++;
1281
1282 /* done with modified vn_open, now finish what sys_open does. */
1283
1284 fp->f_flag = flags & FMASK;
1285 fp->f_type = DTYPE_VNODE;
1286 fp->f_ops = &vnops;
1287 fp->f_data = vp;
1288 if (flags & (O_EXLOCK | O_SHLOCK)) {
1289 lf.l_whence = SEEK_SET;
1290 lf.l_start = 0;
1291 lf.l_len = 0;
1292 if (flags & O_EXLOCK)
1293 lf.l_type = F_WRLCK;
1294 else
1295 lf.l_type = F_RDLCK;
1296 type = F_FLOCK;
1297 if ((flags & FNONBLOCK) == 0)
1298 type |= F_WAIT;
1299 VOP_UNLOCK(vp, 0);
1300 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1301 if (error) {
1302 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1303 FILE_UNUSE(fp, p);
1304 ffree(fp);
1305 fdremove(fdp, indx);
1306 return (error);
1307 }
1308 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1309 fp->f_flag |= FHASLOCK;
1310 }
1311 VOP_UNLOCK(vp, 0);
1312 *retval = indx;
1313 FILE_SET_MATURE(fp);
1314 FILE_UNUSE(fp, p);
1315 return (0);
1316
1317 bad:
1318 FILE_UNUSE(fp, p);
1319 ffree(fp);
1320 fdremove(fdp, indx);
1321 if (vp != NULL)
1322 vput(vp);
1323 return (error);
1324 }
1325
1326 /* ARGSUSED */
1327 int
1328 sys_fhstat(l, v, retval)
1329 struct lwp *l;
1330 void *v;
1331 register_t *retval;
1332 {
1333 struct sys_fhstat_args /* {
1334 syscallarg(const fhandle_t *) fhp;
1335 syscallarg(struct stat *) sb;
1336 } */ *uap = v;
1337 struct proc *p = l->l_proc;
1338 struct stat sb;
1339 int error;
1340 fhandle_t fh;
1341 struct mount *mp;
1342 struct vnode *vp;
1343
1344 /*
1345 * Must be super user
1346 */
1347 if ((error = suser(p->p_ucred, &p->p_acflag)))
1348 return (error);
1349
1350 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1351 return (error);
1352
1353 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1354 return (ESTALE);
1355 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1356 return (error);
1357 error = vn_stat(vp, &sb, p);
1358 vput(vp);
1359 if (error)
1360 return (error);
1361 error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1362 return (error);
1363 }
1364
1365 /* ARGSUSED */
1366 int
1367 sys_fhstatfs(l, v, retval)
1368 struct lwp *l;
1369 void *v;
1370 register_t *retval;
1371 {
1372 struct sys_fhstatfs_args /*
1373 syscallarg(const fhandle_t *) fhp;
1374 syscallarg(struct statfs *) buf;
1375 } */ *uap = v;
1376 struct proc *p = l->l_proc;
1377 struct statfs sbuf;
1378 fhandle_t fh;
1379 struct mount *mp;
1380 struct vnode *vp;
1381 int error;
1382
1383 /*
1384 * Must be super user
1385 */
1386 if ((error = suser(p->p_ucred, &p->p_acflag)))
1387 return (error);
1388
1389 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1390 return (error);
1391
1392 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1393 return (ESTALE);
1394 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1395 return (error);
1396 mp = vp->v_mount;
1397 vput(vp);
1398 if ((error = VFS_STATFS(mp, &sbuf, p)) != 0)
1399 return (error);
1400 return (copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)));
1401 }
1402
1403 /*
1404 * Create a special file.
1405 */
1406 /* ARGSUSED */
1407 int
1408 sys_mknod(l, v, retval)
1409 struct lwp *l;
1410 void *v;
1411 register_t *retval;
1412 {
1413 struct sys_mknod_args /* {
1414 syscallarg(const char *) path;
1415 syscallarg(int) mode;
1416 syscallarg(int) dev;
1417 } */ *uap = v;
1418 struct proc *p = l->l_proc;
1419 struct vnode *vp;
1420 struct vattr vattr;
1421 int error;
1422 int whiteout = 0;
1423 struct nameidata nd;
1424
1425 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1426 return (error);
1427 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1428 if ((error = namei(&nd)) != 0)
1429 return (error);
1430 vp = nd.ni_vp;
1431 if (vp != NULL)
1432 error = EEXIST;
1433 else {
1434 VATTR_NULL(&vattr);
1435 vattr.va_mode =
1436 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1437 vattr.va_rdev = SCARG(uap, dev);
1438 whiteout = 0;
1439
1440 switch (SCARG(uap, mode) & S_IFMT) {
1441 case S_IFMT: /* used by badsect to flag bad sectors */
1442 vattr.va_type = VBAD;
1443 break;
1444 case S_IFCHR:
1445 vattr.va_type = VCHR;
1446 break;
1447 case S_IFBLK:
1448 vattr.va_type = VBLK;
1449 break;
1450 case S_IFWHT:
1451 whiteout = 1;
1452 break;
1453 default:
1454 error = EINVAL;
1455 break;
1456 }
1457 }
1458 if (!error) {
1459 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1460 if (whiteout) {
1461 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1462 if (error)
1463 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1464 vput(nd.ni_dvp);
1465 } else {
1466 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1467 &nd.ni_cnd, &vattr);
1468 if (error == 0)
1469 vput(nd.ni_vp);
1470 }
1471 } else {
1472 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1473 if (nd.ni_dvp == vp)
1474 vrele(nd.ni_dvp);
1475 else
1476 vput(nd.ni_dvp);
1477 if (vp)
1478 vrele(vp);
1479 }
1480 return (error);
1481 }
1482
1483 /*
1484 * Create a named pipe.
1485 */
1486 /* ARGSUSED */
1487 int
1488 sys_mkfifo(l, v, retval)
1489 struct lwp *l;
1490 void *v;
1491 register_t *retval;
1492 {
1493 struct sys_mkfifo_args /* {
1494 syscallarg(const char *) path;
1495 syscallarg(int) mode;
1496 } */ *uap = v;
1497 struct proc *p = l->l_proc;
1498 struct vattr vattr;
1499 int error;
1500 struct nameidata nd;
1501
1502 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1503 if ((error = namei(&nd)) != 0)
1504 return (error);
1505 if (nd.ni_vp != NULL) {
1506 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1507 if (nd.ni_dvp == nd.ni_vp)
1508 vrele(nd.ni_dvp);
1509 else
1510 vput(nd.ni_dvp);
1511 vrele(nd.ni_vp);
1512 return (EEXIST);
1513 }
1514 VATTR_NULL(&vattr);
1515 vattr.va_type = VFIFO;
1516 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1517 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1518 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1519 if (error == 0)
1520 vput(nd.ni_vp);
1521 return (error);
1522 }
1523
1524 /*
1525 * Make a hard file link.
1526 */
1527 /* ARGSUSED */
1528 int
1529 sys_link(l, v, retval)
1530 struct lwp *l;
1531 void *v;
1532 register_t *retval;
1533 {
1534 struct sys_link_args /* {
1535 syscallarg(const char *) path;
1536 syscallarg(const char *) link;
1537 } */ *uap = v;
1538 struct proc *p = l->l_proc;
1539 struct vnode *vp;
1540 struct nameidata nd;
1541 int error;
1542
1543 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1544 if ((error = namei(&nd)) != 0)
1545 return (error);
1546 vp = nd.ni_vp;
1547 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1548 if ((error = namei(&nd)) != 0)
1549 goto out;
1550 if (nd.ni_vp) {
1551 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1552 if (nd.ni_dvp == nd.ni_vp)
1553 vrele(nd.ni_dvp);
1554 else
1555 vput(nd.ni_dvp);
1556 vrele(nd.ni_vp);
1557 error = EEXIST;
1558 goto out;
1559 }
1560 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1561 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1562 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1563 out:
1564 vrele(vp);
1565 return (error);
1566 }
1567
1568 /*
1569 * Make a symbolic link.
1570 */
1571 /* ARGSUSED */
1572 int
1573 sys_symlink(l, v, retval)
1574 struct lwp *l;
1575 void *v;
1576 register_t *retval;
1577 {
1578 struct sys_symlink_args /* {
1579 syscallarg(const char *) path;
1580 syscallarg(const char *) link;
1581 } */ *uap = v;
1582 struct proc *p = l->l_proc;
1583 struct vattr vattr;
1584 char *path;
1585 int error;
1586 struct nameidata nd;
1587
1588 path = PNBUF_GET();
1589 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1590 if (error)
1591 goto out;
1592 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1593 if ((error = namei(&nd)) != 0)
1594 goto out;
1595 if (nd.ni_vp) {
1596 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1597 if (nd.ni_dvp == nd.ni_vp)
1598 vrele(nd.ni_dvp);
1599 else
1600 vput(nd.ni_dvp);
1601 vrele(nd.ni_vp);
1602 error = EEXIST;
1603 goto out;
1604 }
1605 VATTR_NULL(&vattr);
1606 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1607 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1608 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1609 if (error == 0)
1610 vput(nd.ni_vp);
1611 out:
1612 PNBUF_PUT(path);
1613 return (error);
1614 }
1615
1616 /*
1617 * Delete a whiteout from the filesystem.
1618 */
1619 /* ARGSUSED */
1620 int
1621 sys_undelete(l, v, retval)
1622 struct lwp *l;
1623 void *v;
1624 register_t *retval;
1625 {
1626 struct sys_undelete_args /* {
1627 syscallarg(const char *) path;
1628 } */ *uap = v;
1629 struct proc *p = l->l_proc;
1630 int error;
1631 struct nameidata nd;
1632
1633 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1634 SCARG(uap, path), p);
1635 error = namei(&nd);
1636 if (error)
1637 return (error);
1638
1639 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1640 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1641 if (nd.ni_dvp == nd.ni_vp)
1642 vrele(nd.ni_dvp);
1643 else
1644 vput(nd.ni_dvp);
1645 if (nd.ni_vp)
1646 vrele(nd.ni_vp);
1647 return (EEXIST);
1648 }
1649
1650 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1651 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1652 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1653 vput(nd.ni_dvp);
1654 return (error);
1655 }
1656
1657 /*
1658 * Delete a name from the filesystem.
1659 */
1660 /* ARGSUSED */
1661 int
1662 sys_unlink(l, v, retval)
1663 struct lwp *l;
1664 void *v;
1665 register_t *retval;
1666 {
1667 struct sys_unlink_args /* {
1668 syscallarg(const char *) path;
1669 } */ *uap = v;
1670 struct proc *p = l->l_proc;
1671 struct vnode *vp;
1672 int error;
1673 struct nameidata nd;
1674
1675 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1676 SCARG(uap, path), p);
1677 if ((error = namei(&nd)) != 0)
1678 return (error);
1679 vp = nd.ni_vp;
1680
1681 /*
1682 * The root of a mounted filesystem cannot be deleted.
1683 */
1684 if (vp->v_flag & VROOT) {
1685 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1686 if (nd.ni_dvp == vp)
1687 vrele(nd.ni_dvp);
1688 else
1689 vput(nd.ni_dvp);
1690 vput(vp);
1691 error = EBUSY;
1692 goto out;
1693 }
1694
1695 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1696 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1697 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1698 out:
1699 return (error);
1700 }
1701
1702 /*
1703 * Reposition read/write file offset.
1704 */
1705 int
1706 sys_lseek(l, v, retval)
1707 struct lwp *l;
1708 void *v;
1709 register_t *retval;
1710 {
1711 struct sys_lseek_args /* {
1712 syscallarg(int) fd;
1713 syscallarg(int) pad;
1714 syscallarg(off_t) offset;
1715 syscallarg(int) whence;
1716 } */ *uap = v;
1717 struct proc *p = l->l_proc;
1718 struct ucred *cred = p->p_ucred;
1719 struct filedesc *fdp = p->p_fd;
1720 struct file *fp;
1721 struct vnode *vp;
1722 struct vattr vattr;
1723 off_t newoff;
1724 int error;
1725
1726 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1727 return (EBADF);
1728
1729 FILE_USE(fp);
1730
1731 vp = (struct vnode *)fp->f_data;
1732 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1733 error = ESPIPE;
1734 goto out;
1735 }
1736
1737 switch (SCARG(uap, whence)) {
1738 case SEEK_CUR:
1739 newoff = fp->f_offset + SCARG(uap, offset);
1740 break;
1741 case SEEK_END:
1742 error = VOP_GETATTR(vp, &vattr, cred, p);
1743 if (error)
1744 goto out;
1745 newoff = SCARG(uap, offset) + vattr.va_size;
1746 break;
1747 case SEEK_SET:
1748 newoff = SCARG(uap, offset);
1749 break;
1750 default:
1751 error = EINVAL;
1752 goto out;
1753 }
1754 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1755 goto out;
1756
1757 *(off_t *)retval = fp->f_offset = newoff;
1758 out:
1759 FILE_UNUSE(fp, p);
1760 return (error);
1761 }
1762
1763 /*
1764 * Positional read system call.
1765 */
1766 int
1767 sys_pread(l, v, retval)
1768 struct lwp *l;
1769 void *v;
1770 register_t *retval;
1771 {
1772 struct sys_pread_args /* {
1773 syscallarg(int) fd;
1774 syscallarg(void *) buf;
1775 syscallarg(size_t) nbyte;
1776 syscallarg(off_t) offset;
1777 } */ *uap = v;
1778 struct proc *p = l->l_proc;
1779 struct filedesc *fdp = p->p_fd;
1780 struct file *fp;
1781 struct vnode *vp;
1782 off_t offset;
1783 int error, fd = SCARG(uap, fd);
1784
1785 if ((fp = fd_getfile(fdp, fd)) == NULL)
1786 return (EBADF);
1787
1788 if ((fp->f_flag & FREAD) == 0) {
1789 simple_unlock(&fp->f_slock);
1790 return (EBADF);
1791 }
1792
1793 FILE_USE(fp);
1794
1795 vp = (struct vnode *)fp->f_data;
1796 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1797 error = ESPIPE;
1798 goto out;
1799 }
1800
1801 offset = SCARG(uap, offset);
1802
1803 /*
1804 * XXX This works because no file systems actually
1805 * XXX take any action on the seek operation.
1806 */
1807 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1808 goto out;
1809
1810 /* dofileread() will unuse the descriptor for us */
1811 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1812 &offset, 0, retval));
1813
1814 out:
1815 FILE_UNUSE(fp, p);
1816 return (error);
1817 }
1818
1819 /*
1820 * Positional scatter read system call.
1821 */
1822 int
1823 sys_preadv(l, v, retval)
1824 struct lwp *l;
1825 void *v;
1826 register_t *retval;
1827 {
1828 struct sys_preadv_args /* {
1829 syscallarg(int) fd;
1830 syscallarg(const struct iovec *) iovp;
1831 syscallarg(int) iovcnt;
1832 syscallarg(off_t) offset;
1833 } */ *uap = v;
1834 struct proc *p = l->l_proc;
1835 struct filedesc *fdp = p->p_fd;
1836 struct file *fp;
1837 struct vnode *vp;
1838 off_t offset;
1839 int error, fd = SCARG(uap, fd);
1840
1841 if ((fp = fd_getfile(fdp, fd)) == NULL)
1842 return (EBADF);
1843
1844 if ((fp->f_flag & FREAD) == 0) {
1845 simple_unlock(&fp->f_slock);
1846 return (EBADF);
1847 }
1848
1849 FILE_USE(fp);
1850
1851 vp = (struct vnode *)fp->f_data;
1852 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1853 error = ESPIPE;
1854 goto out;
1855 }
1856
1857 offset = SCARG(uap, offset);
1858
1859 /*
1860 * XXX This works because no file systems actually
1861 * XXX take any action on the seek operation.
1862 */
1863 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1864 goto out;
1865
1866 /* dofilereadv() will unuse the descriptor for us */
1867 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1868 &offset, 0, retval));
1869
1870 out:
1871 FILE_UNUSE(fp, p);
1872 return (error);
1873 }
1874
1875 /*
1876 * Positional write system call.
1877 */
1878 int
1879 sys_pwrite(l, v, retval)
1880 struct lwp *l;
1881 void *v;
1882 register_t *retval;
1883 {
1884 struct sys_pwrite_args /* {
1885 syscallarg(int) fd;
1886 syscallarg(const void *) buf;
1887 syscallarg(size_t) nbyte;
1888 syscallarg(off_t) offset;
1889 } */ *uap = v;
1890 struct proc *p = l->l_proc;
1891 struct filedesc *fdp = p->p_fd;
1892 struct file *fp;
1893 struct vnode *vp;
1894 off_t offset;
1895 int error, fd = SCARG(uap, fd);
1896
1897 if ((fp = fd_getfile(fdp, fd)) == NULL)
1898 return (EBADF);
1899
1900 if ((fp->f_flag & FWRITE) == 0) {
1901 simple_unlock(&fp->f_slock);
1902 return (EBADF);
1903 }
1904
1905 FILE_USE(fp);
1906
1907 vp = (struct vnode *)fp->f_data;
1908 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1909 error = ESPIPE;
1910 goto out;
1911 }
1912
1913 offset = SCARG(uap, offset);
1914
1915 /*
1916 * XXX This works because no file systems actually
1917 * XXX take any action on the seek operation.
1918 */
1919 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1920 goto out;
1921
1922 /* dofilewrite() will unuse the descriptor for us */
1923 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1924 &offset, 0, retval));
1925
1926 out:
1927 FILE_UNUSE(fp, p);
1928 return (error);
1929 }
1930
1931 /*
1932 * Positional gather write system call.
1933 */
1934 int
1935 sys_pwritev(l, v, retval)
1936 struct lwp *l;
1937 void *v;
1938 register_t *retval;
1939 {
1940 struct sys_pwritev_args /* {
1941 syscallarg(int) fd;
1942 syscallarg(const struct iovec *) iovp;
1943 syscallarg(int) iovcnt;
1944 syscallarg(off_t) offset;
1945 } */ *uap = v;
1946 struct proc *p = l->l_proc;
1947 struct filedesc *fdp = p->p_fd;
1948 struct file *fp;
1949 struct vnode *vp;
1950 off_t offset;
1951 int error, fd = SCARG(uap, fd);
1952
1953 if ((fp = fd_getfile(fdp, fd)) == NULL)
1954 return (EBADF);
1955
1956 if ((fp->f_flag & FWRITE) == 0) {
1957 simple_unlock(&fp->f_slock);
1958 return (EBADF);
1959 }
1960
1961 FILE_USE(fp);
1962
1963 vp = (struct vnode *)fp->f_data;
1964 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1965 error = ESPIPE;
1966 goto out;
1967 }
1968
1969 offset = SCARG(uap, offset);
1970
1971 /*
1972 * XXX This works because no file systems actually
1973 * XXX take any action on the seek operation.
1974 */
1975 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1976 goto out;
1977
1978 /* dofilewritev() will unuse the descriptor for us */
1979 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1980 &offset, 0, retval));
1981
1982 out:
1983 FILE_UNUSE(fp, p);
1984 return (error);
1985 }
1986
1987 /*
1988 * Check access permissions.
1989 */
1990 int
1991 sys_access(l, v, retval)
1992 struct lwp *l;
1993 void *v;
1994 register_t *retval;
1995 {
1996 struct sys_access_args /* {
1997 syscallarg(const char *) path;
1998 syscallarg(int) flags;
1999 } */ *uap = v;
2000 struct proc *p = l->l_proc;
2001 struct ucred *cred = crget();
2002 struct vnode *vp;
2003 int error, flags;
2004 struct nameidata nd;
2005
2006 (void)memcpy(cred, p->p_ucred, sizeof(*cred));
2007 cred->cr_ref = 1;
2008 cred->cr_uid = p->p_cred->p_ruid;
2009 cred->cr_gid = p->p_cred->p_rgid;
2010 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2011 SCARG(uap, path), p);
2012 /* Override default credentials */
2013 nd.ni_cnd.cn_cred = cred;
2014 if ((error = namei(&nd)) != 0)
2015 goto out;
2016 vp = nd.ni_vp;
2017
2018 /* Flags == 0 means only check for existence. */
2019 if (SCARG(uap, flags)) {
2020 flags = 0;
2021 if (SCARG(uap, flags) & R_OK)
2022 flags |= VREAD;
2023 if (SCARG(uap, flags) & W_OK)
2024 flags |= VWRITE;
2025 if (SCARG(uap, flags) & X_OK)
2026 flags |= VEXEC;
2027
2028 error = VOP_ACCESS(vp, flags, cred, p);
2029 if (!error && (flags & VWRITE))
2030 error = vn_writechk(vp);
2031 }
2032 vput(vp);
2033 out:
2034 crfree(cred);
2035 return (error);
2036 }
2037
2038 /*
2039 * Get file status; this version follows links.
2040 */
2041 /* ARGSUSED */
2042 int
2043 sys___stat13(l, v, retval)
2044 struct lwp *l;
2045 void *v;
2046 register_t *retval;
2047 {
2048 struct sys___stat13_args /* {
2049 syscallarg(const char *) path;
2050 syscallarg(struct stat *) ub;
2051 } */ *uap = v;
2052 struct proc *p = l->l_proc;
2053 struct stat sb;
2054 int error;
2055 struct nameidata nd;
2056
2057 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2058 SCARG(uap, path), p);
2059 if ((error = namei(&nd)) != 0)
2060 return (error);
2061 error = vn_stat(nd.ni_vp, &sb, p);
2062 vput(nd.ni_vp);
2063 if (error)
2064 return (error);
2065 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2066 return (error);
2067 }
2068
2069 /*
2070 * Get file status; this version does not follow links.
2071 */
2072 /* ARGSUSED */
2073 int
2074 sys___lstat13(l, v, retval)
2075 struct lwp *l;
2076 void *v;
2077 register_t *retval;
2078 {
2079 struct sys___lstat13_args /* {
2080 syscallarg(const char *) path;
2081 syscallarg(struct stat *) ub;
2082 } */ *uap = v;
2083 struct proc *p = l->l_proc;
2084 struct stat sb;
2085 int error;
2086 struct nameidata nd;
2087
2088 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2089 SCARG(uap, path), p);
2090 if ((error = namei(&nd)) != 0)
2091 return (error);
2092 error = vn_stat(nd.ni_vp, &sb, p);
2093 vput(nd.ni_vp);
2094 if (error)
2095 return (error);
2096 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2097 return (error);
2098 }
2099
2100 /*
2101 * Get configurable pathname variables.
2102 */
2103 /* ARGSUSED */
2104 int
2105 sys_pathconf(l, v, retval)
2106 struct lwp *l;
2107 void *v;
2108 register_t *retval;
2109 {
2110 struct sys_pathconf_args /* {
2111 syscallarg(const char *) path;
2112 syscallarg(int) name;
2113 } */ *uap = v;
2114 struct proc *p = l->l_proc;
2115 int error;
2116 struct nameidata nd;
2117
2118 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2119 SCARG(uap, path), p);
2120 if ((error = namei(&nd)) != 0)
2121 return (error);
2122 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2123 vput(nd.ni_vp);
2124 return (error);
2125 }
2126
2127 /*
2128 * Return target name of a symbolic link.
2129 */
2130 /* ARGSUSED */
2131 int
2132 sys_readlink(l, v, retval)
2133 struct lwp *l;
2134 void *v;
2135 register_t *retval;
2136 {
2137 struct sys_readlink_args /* {
2138 syscallarg(const char *) path;
2139 syscallarg(char *) buf;
2140 syscallarg(size_t) count;
2141 } */ *uap = v;
2142 struct proc *p = l->l_proc;
2143 struct vnode *vp;
2144 struct iovec aiov;
2145 struct uio auio;
2146 int error;
2147 struct nameidata nd;
2148
2149 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2150 SCARG(uap, path), p);
2151 if ((error = namei(&nd)) != 0)
2152 return (error);
2153 vp = nd.ni_vp;
2154 if (vp->v_type != VLNK)
2155 error = EINVAL;
2156 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2157 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2158 aiov.iov_base = SCARG(uap, buf);
2159 aiov.iov_len = SCARG(uap, count);
2160 auio.uio_iov = &aiov;
2161 auio.uio_iovcnt = 1;
2162 auio.uio_offset = 0;
2163 auio.uio_rw = UIO_READ;
2164 auio.uio_segflg = UIO_USERSPACE;
2165 auio.uio_procp = p;
2166 auio.uio_resid = SCARG(uap, count);
2167 error = VOP_READLINK(vp, &auio, p->p_ucred);
2168 }
2169 vput(vp);
2170 *retval = SCARG(uap, count) - auio.uio_resid;
2171 return (error);
2172 }
2173
2174 /*
2175 * Change flags of a file given a path name.
2176 */
2177 /* ARGSUSED */
2178 int
2179 sys_chflags(l, v, retval)
2180 struct lwp *l;
2181 void *v;
2182 register_t *retval;
2183 {
2184 struct sys_chflags_args /* {
2185 syscallarg(const char *) path;
2186 syscallarg(u_long) flags;
2187 } */ *uap = v;
2188 struct proc *p = l->l_proc;
2189 struct vnode *vp;
2190 int error;
2191 struct nameidata nd;
2192
2193 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2194 if ((error = namei(&nd)) != 0)
2195 return (error);
2196 vp = nd.ni_vp;
2197 error = change_flags(vp, SCARG(uap, flags), p);
2198 vput(vp);
2199 return (error);
2200 }
2201
2202 /*
2203 * Change flags of a file given a file descriptor.
2204 */
2205 /* ARGSUSED */
2206 int
2207 sys_fchflags(l, v, retval)
2208 struct lwp *l;
2209 void *v;
2210 register_t *retval;
2211 {
2212 struct sys_fchflags_args /* {
2213 syscallarg(int) fd;
2214 syscallarg(u_long) flags;
2215 } */ *uap = v;
2216 struct proc *p = l->l_proc;
2217 struct vnode *vp;
2218 struct file *fp;
2219 int error;
2220
2221 /* getvnode() will use the descriptor for us */
2222 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2223 return (error);
2224 vp = (struct vnode *)fp->f_data;
2225 error = change_flags(vp, SCARG(uap, flags), p);
2226 VOP_UNLOCK(vp, 0);
2227 FILE_UNUSE(fp, p);
2228 return (error);
2229 }
2230
2231 /*
2232 * Change flags of a file given a path name; this version does
2233 * not follow links.
2234 */
2235 int
2236 sys_lchflags(l, v, retval)
2237 struct lwp *l;
2238 void *v;
2239 register_t *retval;
2240 {
2241 struct sys_lchflags_args /* {
2242 syscallarg(const char *) path;
2243 syscallarg(u_long) flags;
2244 } */ *uap = v;
2245 struct proc *p = l->l_proc;
2246 struct vnode *vp;
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 vp = nd.ni_vp;
2254 error = change_flags(vp, SCARG(uap, flags), p);
2255 vput(vp);
2256 return (error);
2257 }
2258
2259 /*
2260 * Common routine to change flags of a file.
2261 */
2262 int
2263 change_flags(vp, flags, p)
2264 struct vnode *vp;
2265 u_long flags;
2266 struct proc *p;
2267 {
2268 struct vattr vattr;
2269 int error;
2270
2271 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2272 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2273 /*
2274 * Non-superusers cannot change the flags on devices, even if they
2275 * own them.
2276 */
2277 if (suser(p->p_ucred, &p->p_acflag) != 0) {
2278 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2279 goto out;
2280 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2281 error = EINVAL;
2282 goto out;
2283 }
2284 }
2285 VATTR_NULL(&vattr);
2286 vattr.va_flags = flags;
2287 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2288 out:
2289 return (error);
2290 }
2291
2292 /*
2293 * Change mode of a file given path name; this version follows links.
2294 */
2295 /* ARGSUSED */
2296 int
2297 sys_chmod(l, v, retval)
2298 struct lwp *l;
2299 void *v;
2300 register_t *retval;
2301 {
2302 struct sys_chmod_args /* {
2303 syscallarg(const char *) path;
2304 syscallarg(int) mode;
2305 } */ *uap = v;
2306 struct proc *p = l->l_proc;
2307 int error;
2308 struct nameidata nd;
2309
2310 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2311 if ((error = namei(&nd)) != 0)
2312 return (error);
2313
2314 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2315
2316 vrele(nd.ni_vp);
2317 return (error);
2318 }
2319
2320 /*
2321 * Change mode of a file given a file descriptor.
2322 */
2323 /* ARGSUSED */
2324 int
2325 sys_fchmod(l, v, retval)
2326 struct lwp *l;
2327 void *v;
2328 register_t *retval;
2329 {
2330 struct sys_fchmod_args /* {
2331 syscallarg(int) fd;
2332 syscallarg(int) mode;
2333 } */ *uap = v;
2334 struct proc *p = l->l_proc;
2335 struct file *fp;
2336 int error;
2337
2338 /* getvnode() will use the descriptor for us */
2339 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2340 return (error);
2341
2342 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2343 FILE_UNUSE(fp, p);
2344 return (error);
2345 }
2346
2347 /*
2348 * Change mode of a file given path name; this version does not follow links.
2349 */
2350 /* ARGSUSED */
2351 int
2352 sys_lchmod(l, v, retval)
2353 struct lwp *l;
2354 void *v;
2355 register_t *retval;
2356 {
2357 struct sys_lchmod_args /* {
2358 syscallarg(const char *) path;
2359 syscallarg(int) mode;
2360 } */ *uap = v;
2361 struct proc *p = l->l_proc;
2362 int error;
2363 struct nameidata nd;
2364
2365 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2366 if ((error = namei(&nd)) != 0)
2367 return (error);
2368
2369 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2370
2371 vrele(nd.ni_vp);
2372 return (error);
2373 }
2374
2375 /*
2376 * Common routine to set mode given a vnode.
2377 */
2378 static int
2379 change_mode(vp, mode, p)
2380 struct vnode *vp;
2381 int mode;
2382 struct proc *p;
2383 {
2384 struct vattr vattr;
2385 int error;
2386
2387 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2388 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2389 VATTR_NULL(&vattr);
2390 vattr.va_mode = mode & ALLPERMS;
2391 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2392 VOP_UNLOCK(vp, 0);
2393 return (error);
2394 }
2395
2396 /*
2397 * Set ownership given a path name; this version follows links.
2398 */
2399 /* ARGSUSED */
2400 int
2401 sys_chown(l, v, retval)
2402 struct lwp *l;
2403 void *v;
2404 register_t *retval;
2405 {
2406 struct sys_chown_args /* {
2407 syscallarg(const char *) path;
2408 syscallarg(uid_t) uid;
2409 syscallarg(gid_t) gid;
2410 } */ *uap = v;
2411 struct proc *p = l->l_proc;
2412 int error;
2413 struct nameidata nd;
2414
2415 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2416 if ((error = namei(&nd)) != 0)
2417 return (error);
2418
2419 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2420
2421 vrele(nd.ni_vp);
2422 return (error);
2423 }
2424
2425 /*
2426 * Set ownership given a path name; this version follows links.
2427 * Provides POSIX semantics.
2428 */
2429 /* ARGSUSED */
2430 int
2431 sys___posix_chown(l, v, retval)
2432 struct lwp *l;
2433 void *v;
2434 register_t *retval;
2435 {
2436 struct sys_chown_args /* {
2437 syscallarg(const char *) path;
2438 syscallarg(uid_t) uid;
2439 syscallarg(gid_t) gid;
2440 } */ *uap = v;
2441 struct proc *p = l->l_proc;
2442 int error;
2443 struct nameidata nd;
2444
2445 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2446 if ((error = namei(&nd)) != 0)
2447 return (error);
2448
2449 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2450
2451 vrele(nd.ni_vp);
2452 return (error);
2453 }
2454
2455 /*
2456 * Set ownership given a file descriptor.
2457 */
2458 /* ARGSUSED */
2459 int
2460 sys_fchown(l, v, retval)
2461 struct lwp *l;
2462 void *v;
2463 register_t *retval;
2464 {
2465 struct sys_fchown_args /* {
2466 syscallarg(int) fd;
2467 syscallarg(uid_t) uid;
2468 syscallarg(gid_t) gid;
2469 } */ *uap = v;
2470 struct proc *p = l->l_proc;
2471 int error;
2472 struct file *fp;
2473
2474 /* getvnode() will use the descriptor for us */
2475 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2476 return (error);
2477
2478 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2479 SCARG(uap, gid), p, 0);
2480 FILE_UNUSE(fp, p);
2481 return (error);
2482 }
2483
2484 /*
2485 * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2486 */
2487 /* ARGSUSED */
2488 int
2489 sys___posix_fchown(l, v, retval)
2490 struct lwp *l;
2491 void *v;
2492 register_t *retval;
2493 {
2494 struct sys_fchown_args /* {
2495 syscallarg(int) fd;
2496 syscallarg(uid_t) uid;
2497 syscallarg(gid_t) gid;
2498 } */ *uap = v;
2499 struct proc *p = l->l_proc;
2500 int error;
2501 struct file *fp;
2502
2503 /* getvnode() will use the descriptor for us */
2504 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2505 return (error);
2506
2507 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2508 SCARG(uap, gid), p, 1);
2509 FILE_UNUSE(fp, p);
2510 return (error);
2511 }
2512
2513 /*
2514 * Set ownership given a path name; this version does not follow links.
2515 */
2516 /* ARGSUSED */
2517 int
2518 sys_lchown(l, v, retval)
2519 struct lwp *l;
2520 void *v;
2521 register_t *retval;
2522 {
2523 struct sys_lchown_args /* {
2524 syscallarg(const char *) path;
2525 syscallarg(uid_t) uid;
2526 syscallarg(gid_t) gid;
2527 } */ *uap = v;
2528 struct proc *p = l->l_proc;
2529 int error;
2530 struct nameidata nd;
2531
2532 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2533 if ((error = namei(&nd)) != 0)
2534 return (error);
2535
2536 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2537
2538 vrele(nd.ni_vp);
2539 return (error);
2540 }
2541
2542 /*
2543 * Set ownership given a path name; this version does not follow links.
2544 * Provides POSIX/XPG semantics.
2545 */
2546 /* ARGSUSED */
2547 int
2548 sys___posix_lchown(l, v, retval)
2549 struct lwp *l;
2550 void *v;
2551 register_t *retval;
2552 {
2553 struct sys_lchown_args /* {
2554 syscallarg(const char *) path;
2555 syscallarg(uid_t) uid;
2556 syscallarg(gid_t) gid;
2557 } */ *uap = v;
2558 struct proc *p = l->l_proc;
2559 int error;
2560 struct nameidata nd;
2561
2562 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2563 if ((error = namei(&nd)) != 0)
2564 return (error);
2565
2566 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2567
2568 vrele(nd.ni_vp);
2569 return (error);
2570 }
2571
2572 /*
2573 * Common routine to set ownership given a vnode.
2574 */
2575 static int
2576 change_owner(vp, uid, gid, p, posix_semantics)
2577 struct vnode *vp;
2578 uid_t uid;
2579 gid_t gid;
2580 struct proc *p;
2581 int posix_semantics;
2582 {
2583 struct vattr vattr;
2584 mode_t newmode;
2585 int error;
2586
2587 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2588 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2589 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2590 goto out;
2591
2592 #define CHANGED(x) ((int)(x) != -1)
2593 newmode = vattr.va_mode;
2594 if (posix_semantics) {
2595 /*
2596 * POSIX/XPG semantics: if the caller is not the super-user,
2597 * clear set-user-id and set-group-id bits. Both POSIX and
2598 * the XPG consider the behaviour for calls by the super-user
2599 * implementation-defined; we leave the set-user-id and set-
2600 * group-id settings intact in that case.
2601 */
2602 if (suser(p->p_ucred, NULL) != 0)
2603 newmode &= ~(S_ISUID | S_ISGID);
2604 } else {
2605 /*
2606 * NetBSD semantics: when changing owner and/or group,
2607 * clear the respective bit(s).
2608 */
2609 if (CHANGED(uid))
2610 newmode &= ~S_ISUID;
2611 if (CHANGED(gid))
2612 newmode &= ~S_ISGID;
2613 }
2614 /* Update va_mode iff altered. */
2615 if (vattr.va_mode == newmode)
2616 newmode = VNOVAL;
2617
2618 VATTR_NULL(&vattr);
2619 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2620 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2621 vattr.va_mode = newmode;
2622 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2623 #undef CHANGED
2624
2625 out:
2626 VOP_UNLOCK(vp, 0);
2627 return (error);
2628 }
2629
2630 /*
2631 * Set the access and modification times given a path name; this
2632 * version follows links.
2633 */
2634 /* ARGSUSED */
2635 int
2636 sys_utimes(l, v, retval)
2637 struct lwp *l;
2638 void *v;
2639 register_t *retval;
2640 {
2641 struct sys_utimes_args /* {
2642 syscallarg(const char *) path;
2643 syscallarg(const struct timeval *) tptr;
2644 } */ *uap = v;
2645 struct proc *p = l->l_proc;
2646 int error;
2647 struct nameidata nd;
2648
2649 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2650 if ((error = namei(&nd)) != 0)
2651 return (error);
2652
2653 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2654
2655 vrele(nd.ni_vp);
2656 return (error);
2657 }
2658
2659 /*
2660 * Set the access and modification times given a file descriptor.
2661 */
2662 /* ARGSUSED */
2663 int
2664 sys_futimes(l, v, retval)
2665 struct lwp *l;
2666 void *v;
2667 register_t *retval;
2668 {
2669 struct sys_futimes_args /* {
2670 syscallarg(int) fd;
2671 syscallarg(const struct timeval *) tptr;
2672 } */ *uap = v;
2673 struct proc *p = l->l_proc;
2674 int error;
2675 struct file *fp;
2676
2677 /* getvnode() will use the descriptor for us */
2678 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2679 return (error);
2680
2681 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2682 FILE_UNUSE(fp, p);
2683 return (error);
2684 }
2685
2686 /*
2687 * Set the access and modification times given a path name; this
2688 * version does not follow links.
2689 */
2690 /* ARGSUSED */
2691 int
2692 sys_lutimes(l, v, retval)
2693 struct lwp *l;
2694 void *v;
2695 register_t *retval;
2696 {
2697 struct sys_lutimes_args /* {
2698 syscallarg(const char *) path;
2699 syscallarg(const struct timeval *) tptr;
2700 } */ *uap = v;
2701 struct proc *p = l->l_proc;
2702 int error;
2703 struct nameidata nd;
2704
2705 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2706 if ((error = namei(&nd)) != 0)
2707 return (error);
2708
2709 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2710
2711 vrele(nd.ni_vp);
2712 return (error);
2713 }
2714
2715 /*
2716 * Common routine to set access and modification times given a vnode.
2717 */
2718 static int
2719 change_utimes(vp, tptr, p)
2720 struct vnode *vp;
2721 const struct timeval *tptr;
2722 struct proc *p;
2723 {
2724 struct timeval tv[2];
2725 struct vattr vattr;
2726 int error;
2727
2728 VATTR_NULL(&vattr);
2729 if (tptr == NULL) {
2730 microtime(&tv[0]);
2731 tv[1] = tv[0];
2732 vattr.va_vaflags |= VA_UTIMES_NULL;
2733 } else {
2734 error = copyin(tptr, tv, sizeof(tv));
2735 if (error)
2736 return (error);
2737 }
2738 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2739 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2740 vattr.va_atime.tv_sec = tv[0].tv_sec;
2741 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2742 vattr.va_mtime.tv_sec = tv[1].tv_sec;
2743 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2744 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2745 VOP_UNLOCK(vp, 0);
2746 return (error);
2747 }
2748
2749 /*
2750 * Truncate a file given its path name.
2751 */
2752 /* ARGSUSED */
2753 int
2754 sys_truncate(l, v, retval)
2755 struct lwp *l;
2756 void *v;
2757 register_t *retval;
2758 {
2759 struct sys_truncate_args /* {
2760 syscallarg(const char *) path;
2761 syscallarg(int) pad;
2762 syscallarg(off_t) length;
2763 } */ *uap = v;
2764 struct proc *p = l->l_proc;
2765 struct vnode *vp;
2766 struct vattr vattr;
2767 int error;
2768 struct nameidata nd;
2769
2770 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2771 if ((error = namei(&nd)) != 0)
2772 return (error);
2773 vp = nd.ni_vp;
2774 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2775 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2776 if (vp->v_type == VDIR)
2777 error = EISDIR;
2778 else if ((error = vn_writechk(vp)) == 0 &&
2779 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2780 VATTR_NULL(&vattr);
2781 vattr.va_size = SCARG(uap, length);
2782 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2783 }
2784 vput(vp);
2785 return (error);
2786 }
2787
2788 /*
2789 * Truncate a file given a file descriptor.
2790 */
2791 /* ARGSUSED */
2792 int
2793 sys_ftruncate(l, v, retval)
2794 struct lwp *l;
2795 void *v;
2796 register_t *retval;
2797 {
2798 struct sys_ftruncate_args /* {
2799 syscallarg(int) fd;
2800 syscallarg(int) pad;
2801 syscallarg(off_t) length;
2802 } */ *uap = v;
2803 struct proc *p = l->l_proc;
2804 struct vattr vattr;
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 if ((fp->f_flag & FWRITE) == 0) {
2813 error = EINVAL;
2814 goto out;
2815 }
2816 vp = (struct vnode *)fp->f_data;
2817 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2818 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2819 if (vp->v_type == VDIR)
2820 error = EISDIR;
2821 else if ((error = vn_writechk(vp)) == 0) {
2822 VATTR_NULL(&vattr);
2823 vattr.va_size = SCARG(uap, length);
2824 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2825 }
2826 VOP_UNLOCK(vp, 0);
2827 out:
2828 FILE_UNUSE(fp, p);
2829 return (error);
2830 }
2831
2832 /*
2833 * Sync an open file.
2834 */
2835 /* ARGSUSED */
2836 int
2837 sys_fsync(l, v, retval)
2838 struct lwp *l;
2839 void *v;
2840 register_t *retval;
2841 {
2842 struct sys_fsync_args /* {
2843 syscallarg(int) fd;
2844 } */ *uap = v;
2845 struct proc *p = l->l_proc;
2846 struct vnode *vp;
2847 struct file *fp;
2848 int error;
2849
2850 /* getvnode() will use the descriptor for us */
2851 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2852 return (error);
2853 vp = (struct vnode *)fp->f_data;
2854 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2855 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
2856 if (error == 0 && bioops.io_fsync != NULL &&
2857 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2858 (*bioops.io_fsync)(vp);
2859 VOP_UNLOCK(vp, 0);
2860 FILE_UNUSE(fp, p);
2861 return (error);
2862 }
2863
2864 /*
2865 * Sync the data of an open file.
2866 */
2867 /* ARGSUSED */
2868 int
2869 sys_fdatasync(l, v, retval)
2870 struct lwp *l;
2871 void *v;
2872 register_t *retval;
2873 {
2874 struct sys_fdatasync_args /* {
2875 syscallarg(int) fd;
2876 } */ *uap = v;
2877 struct proc *p = l->l_proc;
2878 struct vnode *vp;
2879 struct file *fp;
2880 int error;
2881
2882 /* getvnode() will use the descriptor for us */
2883 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2884 return (error);
2885 vp = (struct vnode *)fp->f_data;
2886 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2887 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
2888 VOP_UNLOCK(vp, 0);
2889 FILE_UNUSE(fp, p);
2890 return (error);
2891 }
2892
2893 /*
2894 * Rename files, (standard) BSD semantics frontend.
2895 */
2896 /* ARGSUSED */
2897 int
2898 sys_rename(l, v, retval)
2899 struct lwp *l;
2900 void *v;
2901 register_t *retval;
2902 {
2903 struct sys_rename_args /* {
2904 syscallarg(const char *) from;
2905 syscallarg(const char *) to;
2906 } */ *uap = v;
2907 struct proc *p = l->l_proc;
2908
2909 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
2910 }
2911
2912 /*
2913 * Rename files, POSIX semantics frontend.
2914 */
2915 /* ARGSUSED */
2916 int
2917 sys___posix_rename(l, v, retval)
2918 struct lwp *l;
2919 void *v;
2920 register_t *retval;
2921 {
2922 struct sys___posix_rename_args /* {
2923 syscallarg(const char *) from;
2924 syscallarg(const char *) to;
2925 } */ *uap = v;
2926 struct proc *p = l->l_proc;
2927
2928 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
2929 }
2930
2931 /*
2932 * Rename files. Source and destination must either both be directories,
2933 * or both not be directories. If target is a directory, it must be empty.
2934 * If `from' and `to' refer to the same object, the value of the `retain'
2935 * argument is used to determine whether `from' will be
2936 *
2937 * (retain == 0) deleted unless `from' and `to' refer to the same
2938 * object in the file system's name space (BSD).
2939 * (retain == 1) always retained (POSIX).
2940 */
2941 static int
2942 rename_files(from, to, p, retain)
2943 const char *from, *to;
2944 struct proc *p;
2945 int retain;
2946 {
2947 struct vnode *tvp, *fvp, *tdvp;
2948 struct nameidata fromnd, tond;
2949 int error;
2950
2951 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2952 from, p);
2953 if ((error = namei(&fromnd)) != 0)
2954 return (error);
2955 fvp = fromnd.ni_vp;
2956 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
2957 UIO_USERSPACE, to, p);
2958 if ((error = namei(&tond)) != 0) {
2959 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2960 vrele(fromnd.ni_dvp);
2961 vrele(fvp);
2962 goto out1;
2963 }
2964 tdvp = tond.ni_dvp;
2965 tvp = tond.ni_vp;
2966
2967 if (tvp != NULL) {
2968 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2969 error = ENOTDIR;
2970 goto out;
2971 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2972 error = EISDIR;
2973 goto out;
2974 }
2975 }
2976
2977 if (fvp == tdvp)
2978 error = EINVAL;
2979
2980 /*
2981 * Source and destination refer to the same object.
2982 */
2983 if (fvp == tvp) {
2984 if (retain)
2985 error = -1;
2986 else if (fromnd.ni_dvp == tdvp &&
2987 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2988 !memcmp(fromnd.ni_cnd.cn_nameptr,
2989 tond.ni_cnd.cn_nameptr,
2990 fromnd.ni_cnd.cn_namelen))
2991 error = -1;
2992 }
2993
2994 out:
2995 if (!error) {
2996 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2997 if (fromnd.ni_dvp != tdvp)
2998 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2999 if (tvp) {
3000 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
3001 }
3002 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3003 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3004 } else {
3005 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3006 if (tdvp == tvp)
3007 vrele(tdvp);
3008 else
3009 vput(tdvp);
3010 if (tvp)
3011 vput(tvp);
3012 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3013 vrele(fromnd.ni_dvp);
3014 vrele(fvp);
3015 }
3016 vrele(tond.ni_startdir);
3017 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3018 out1:
3019 if (fromnd.ni_startdir)
3020 vrele(fromnd.ni_startdir);
3021 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3022 return (error == -1 ? 0 : error);
3023 }
3024
3025 /*
3026 * Make a directory file.
3027 */
3028 /* ARGSUSED */
3029 int
3030 sys_mkdir(l, v, retval)
3031 struct lwp *l;
3032 void *v;
3033 register_t *retval;
3034 {
3035 struct sys_mkdir_args /* {
3036 syscallarg(const char *) path;
3037 syscallarg(int) mode;
3038 } */ *uap = v;
3039 struct proc *p = l->l_proc;
3040 struct vnode *vp;
3041 struct vattr vattr;
3042 int error;
3043 struct nameidata nd;
3044
3045 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
3046 if ((error = namei(&nd)) != 0)
3047 return (error);
3048 vp = nd.ni_vp;
3049 if (vp != NULL) {
3050 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3051 if (nd.ni_dvp == vp)
3052 vrele(nd.ni_dvp);
3053 else
3054 vput(nd.ni_dvp);
3055 vrele(vp);
3056 return (EEXIST);
3057 }
3058 VATTR_NULL(&vattr);
3059 vattr.va_type = VDIR;
3060 vattr.va_mode =
3061 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3062 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3063 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3064 if (!error)
3065 vput(nd.ni_vp);
3066 return (error);
3067 }
3068
3069 /*
3070 * Remove a directory file.
3071 */
3072 /* ARGSUSED */
3073 int
3074 sys_rmdir(l, v, retval)
3075 struct lwp *l;
3076 void *v;
3077 register_t *retval;
3078 {
3079 struct sys_rmdir_args /* {
3080 syscallarg(const char *) path;
3081 } */ *uap = v;
3082 struct proc *p = l->l_proc;
3083 struct vnode *vp;
3084 int error;
3085 struct nameidata nd;
3086
3087 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3088 SCARG(uap, path), p);
3089 if ((error = namei(&nd)) != 0)
3090 return (error);
3091 vp = nd.ni_vp;
3092 if (vp->v_type != VDIR) {
3093 error = ENOTDIR;
3094 goto out;
3095 }
3096 /*
3097 * No rmdir "." please.
3098 */
3099 if (nd.ni_dvp == vp) {
3100 error = EINVAL;
3101 goto out;
3102 }
3103 /*
3104 * The root of a mounted filesystem cannot be deleted.
3105 */
3106 if (vp->v_flag & VROOT)
3107 error = EBUSY;
3108 out:
3109 if (!error) {
3110 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3111 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3112 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3113 } else {
3114 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3115 if (nd.ni_dvp == vp)
3116 vrele(nd.ni_dvp);
3117 else
3118 vput(nd.ni_dvp);
3119 vput(vp);
3120 }
3121 return (error);
3122 }
3123
3124 /*
3125 * Read a block of directory entries in a file system independent format.
3126 */
3127 int
3128 sys_getdents(l, v, retval)
3129 struct lwp *l;
3130 void *v;
3131 register_t *retval;
3132 {
3133 struct sys_getdents_args /* {
3134 syscallarg(int) fd;
3135 syscallarg(char *) buf;
3136 syscallarg(size_t) count;
3137 } */ *uap = v;
3138 struct proc *p = l->l_proc;
3139 struct file *fp;
3140 int error, done;
3141
3142 /* getvnode() will use the descriptor for us */
3143 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3144 return (error);
3145 if ((fp->f_flag & FREAD) == 0) {
3146 error = EBADF;
3147 goto out;
3148 }
3149 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3150 SCARG(uap, count), &done, p, 0, 0);
3151 *retval = done;
3152 out:
3153 FILE_UNUSE(fp, p);
3154 return (error);
3155 }
3156
3157 /*
3158 * Set the mode mask for creation of filesystem nodes.
3159 */
3160 int
3161 sys_umask(l, v, retval)
3162 struct lwp *l;
3163 void *v;
3164 register_t *retval;
3165 {
3166 struct sys_umask_args /* {
3167 syscallarg(mode_t) newmask;
3168 } */ *uap = v;
3169 struct proc *p = l->l_proc;
3170 struct cwdinfo *cwdi;
3171
3172 cwdi = p->p_cwdi;
3173 *retval = cwdi->cwdi_cmask;
3174 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3175 return (0);
3176 }
3177
3178 /*
3179 * Void all references to file by ripping underlying filesystem
3180 * away from vnode.
3181 */
3182 /* ARGSUSED */
3183 int
3184 sys_revoke(l, v, retval)
3185 struct lwp *l;
3186 void *v;
3187 register_t *retval;
3188 {
3189 struct sys_revoke_args /* {
3190 syscallarg(const char *) path;
3191 } */ *uap = v;
3192 struct proc *p = l->l_proc;
3193 struct vnode *vp;
3194 struct vattr vattr;
3195 int error;
3196 struct nameidata nd;
3197
3198 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3199 if ((error = namei(&nd)) != 0)
3200 return (error);
3201 vp = nd.ni_vp;
3202 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3203 goto out;
3204 if (p->p_ucred->cr_uid != vattr.va_uid &&
3205 (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3206 goto out;
3207 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3208 VOP_REVOKE(vp, REVOKEALL);
3209 out:
3210 vrele(vp);
3211 return (error);
3212 }
3213
3214 /*
3215 * Convert a user file descriptor to a kernel file entry.
3216 */
3217 int
3218 getvnode(fdp, fd, fpp)
3219 struct filedesc *fdp;
3220 int fd;
3221 struct file **fpp;
3222 {
3223 struct vnode *vp;
3224 struct file *fp;
3225
3226 if ((fp = fd_getfile(fdp, fd)) == NULL)
3227 return (EBADF);
3228
3229 FILE_USE(fp);
3230
3231 if (fp->f_type != DTYPE_VNODE) {
3232 FILE_UNUSE(fp, NULL);
3233 return (EINVAL);
3234 }
3235
3236 vp = (struct vnode *)fp->f_data;
3237 if (vp->v_type == VBAD) {
3238 FILE_UNUSE(fp, NULL);
3239 return (EBADF);
3240 }
3241
3242 *fpp = fp;
3243 return (0);
3244 }
3245