vfs_syscalls.c revision 1.181 1 /* $NetBSD: vfs_syscalls.c,v 1.181 2003/02/01 06:23:45 thorpej 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 1.181 2003/02/01 06:23:45 thorpej 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
80 void checkdirs __P((struct vnode *));
81
82 int dovfsusermount = 0;
83
84 /*
85 * Virtual File System System Calls
86 */
87
88 /*
89 * Mount a file system.
90 */
91
92 #if defined(COMPAT_09) || defined(COMPAT_43)
93 /*
94 * This table is used to maintain compatibility with 4.3BSD
95 * and NetBSD 0.9 mount syscalls. Note, the order is important!
96 *
97 * Do not modify this table. It should only contain filesystems
98 * supported by NetBSD 0.9 and 4.3BSD.
99 */
100 const char * const mountcompatnames[] = {
101 NULL, /* 0 = MOUNT_NONE */
102 MOUNT_FFS, /* 1 = MOUNT_UFS */
103 MOUNT_NFS, /* 2 */
104 MOUNT_MFS, /* 3 */
105 MOUNT_MSDOS, /* 4 */
106 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */
107 MOUNT_FDESC, /* 6 */
108 MOUNT_KERNFS, /* 7 */
109 NULL, /* 8 = MOUNT_DEVFS */
110 MOUNT_AFS, /* 9 */
111 };
112 const int nmountcompatnames = sizeof(mountcompatnames) /
113 sizeof(mountcompatnames[0]);
114 #endif /* COMPAT_09 || COMPAT_43 */
115
116 /* ARGSUSED */
117 int
118 sys_mount(l, v, retval)
119 struct lwp *l;
120 void *v;
121 register_t *retval;
122 {
123 struct sys_mount_args /* {
124 syscallarg(const char *) type;
125 syscallarg(const char *) path;
126 syscallarg(int) flags;
127 syscallarg(void *) data;
128 } */ *uap = v;
129 struct proc *p = l->l_proc;
130 struct vnode *vp;
131 struct mount *mp;
132 int error, flag = 0;
133 char fstypename[MFSNAMELEN];
134 struct vattr va;
135 struct nameidata nd;
136 struct vfsops *vfs;
137
138 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 &&
139 (error = suser(p->p_ucred, &p->p_acflag)))
140 return (error);
141 /*
142 * Get vnode to be covered
143 */
144 NDINIT(&nd, LOOKUP, FOLLOW , UIO_USERSPACE,
145 SCARG(uap, path), p);
146 if ((error = namei(&nd)) != 0)
147 return (error);
148 vp = nd.ni_vp;
149 /*
150 * A lookup in VFS_MOUNT might result in an attempt to
151 * lock this vnode again, so make the lock resursive.
152 */
153 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
154 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) {
155 if ((vp->v_flag & VROOT) == 0) {
156 vput(vp);
157 return (EINVAL);
158 }
159 mp = vp->v_mount;
160 flag = mp->mnt_flag;
161 vfs = mp->mnt_op;
162 /*
163 * We only allow the filesystem to be reloaded if it
164 * is currently mounted read-only.
165 */
166 if ((SCARG(uap, flags) & MNT_RELOAD) &&
167 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
168 vput(vp);
169 return (EOPNOTSUPP); /* Needs translation */
170 }
171 /*
172 * In "highly secure" mode, don't let the caller do anything
173 * but downgrade a filesystem from read-write to read-only.
174 * (see also below; MNT_UPDATE or MNT_GETARGS is required.)
175 */
176 if (securelevel >= 2 &&
177 SCARG(uap, flags) != MNT_GETARGS &&
178 SCARG(uap, flags) !=
179 (mp->mnt_flag | MNT_RDONLY |
180 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) {
181 vput(vp);
182 return (EPERM);
183 }
184 mp->mnt_flag |= SCARG(uap, flags) &
185 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
186 /*
187 * Only root, or the user that did the original mount is
188 * permitted to update it.
189 */
190 if ((mp->mnt_flag & MNT_GETARGS) == 0 &&
191 mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
192 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
193 vput(vp);
194 return (error);
195 }
196 /*
197 * Do not allow NFS export by non-root users. For non-root
198 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
199 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
200 */
201 if (p->p_ucred->cr_uid != 0) {
202 if (SCARG(uap, flags) & MNT_EXPORTED) {
203 vput(vp);
204 return (EPERM);
205 }
206 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
207 if (flag & MNT_NOEXEC)
208 SCARG(uap, flags) |= MNT_NOEXEC;
209 }
210 if (vfs_busy(mp, LK_NOWAIT, 0)) {
211 vput(vp);
212 return (EPERM);
213 }
214 goto update;
215 } else {
216 if (securelevel >= 2) {
217 vput(vp);
218 return (EPERM);
219 }
220 }
221 /*
222 * If the user is not root, ensure that they own the directory
223 * onto which we are attempting to mount.
224 */
225 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
226 (va.va_uid != p->p_ucred->cr_uid &&
227 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
228 vput(vp);
229 return (error);
230 }
231 /*
232 * Do not allow NFS export by non-root users. For non-root users,
233 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
234 * mount point is already MNT_NOEXEC.
235 */
236 if (p->p_ucred->cr_uid != 0) {
237 if (SCARG(uap, flags) & MNT_EXPORTED) {
238 vput(vp);
239 return (EPERM);
240 }
241 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
242 if (vp->v_mount->mnt_flag & MNT_NOEXEC)
243 SCARG(uap, flags) |= MNT_NOEXEC;
244 }
245 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
246 return (error);
247 if (vp->v_type != VDIR) {
248 vput(vp);
249 return (ENOTDIR);
250 }
251 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
252 if (error) {
253 #if defined(COMPAT_09) || defined(COMPAT_43)
254 /*
255 * Historically filesystem types were identified by number.
256 * If we get an integer for the filesystem type instead of a
257 * string, we check to see if it matches one of the historic
258 * filesystem types.
259 */
260 u_long fsindex = (u_long)SCARG(uap, type);
261 if (fsindex >= nmountcompatnames ||
262 mountcompatnames[fsindex] == NULL) {
263 vput(vp);
264 return (ENODEV);
265 }
266 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
267 #else
268 vput(vp);
269 return (error);
270 #endif
271 }
272 #ifdef COMPAT_10
273 /* Accept `ufs' as an alias for `ffs'. */
274 if (!strncmp(fstypename, "ufs", MFSNAMELEN))
275 strncpy(fstypename, "ffs", MFSNAMELEN);
276 #endif
277 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
278 vput(vp);
279 return (ENODEV);
280 }
281 if (vp->v_mountedhere != NULL) {
282 vput(vp);
283 return (EBUSY);
284 }
285
286 /*
287 * Allocate and initialize the file system.
288 */
289 mp = (struct mount *)malloc((u_long)sizeof(struct mount),
290 M_MOUNT, M_WAITOK);
291 memset((char *)mp, 0, (u_long)sizeof(struct mount));
292 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
293 (void)vfs_busy(mp, LK_NOWAIT, 0);
294 mp->mnt_op = vfs;
295 vfs->vfs_refcount++;
296 mp->mnt_vnodecovered = vp;
297 mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
298 mp->mnt_unmounter = NULL;
299 update:
300 /*
301 * Set the mount level flags.
302 */
303 if (SCARG(uap, flags) & MNT_RDONLY)
304 mp->mnt_flag |= MNT_RDONLY;
305 else if (mp->mnt_flag & MNT_RDONLY)
306 mp->mnt_flag |= MNT_WANTRDWR;
307 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
308 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
309 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
310 mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
311 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
312 MNT_NOCOREDUMP | MNT_IGNORE | MNT_NOATIME | MNT_NODEVMTIME |
313 MNT_SYMPERM | MNT_SOFTDEP);
314 /*
315 * Mount the filesystem.
316 */
317 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
318 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) {
319 if (mp->mnt_flag & MNT_WANTRDWR)
320 mp->mnt_flag &= ~MNT_RDONLY;
321 if (error || (mp->mnt_flag & MNT_GETARGS))
322 mp->mnt_flag = flag;
323 mp->mnt_flag &=~
324 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR |
325 MNT_GETARGS);
326 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
327 if (mp->mnt_syncer == NULL)
328 error = vfs_allocate_syncvnode(mp);
329 } else {
330 if (mp->mnt_syncer != NULL)
331 vfs_deallocate_syncvnode(mp);
332 }
333 vfs_unbusy(mp);
334 VOP_UNLOCK(vp, 0);
335 vrele(vp);
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 LIST_FOREACH(p, &allproc, p_list) {
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(l, v, retval)
414 struct lwp *l;
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 proc *p = l->l_proc;
423 struct vnode *vp;
424 struct mount *mp;
425 int error;
426 struct nameidata nd;
427
428 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
429 SCARG(uap, path), p);
430 if ((error = namei(&nd)) != 0)
431 return (error);
432 vp = nd.ni_vp;
433 mp = vp->v_mount;
434
435 /*
436 * Only root, or the user that did the original mount is
437 * permitted to unmount this filesystem.
438 */
439 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
440 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
441 vput(vp);
442 return (error);
443 }
444
445 /*
446 * Don't allow unmounting the root file system.
447 */
448 if (mp->mnt_flag & MNT_ROOTFS) {
449 vput(vp);
450 return (EINVAL);
451 }
452
453 /*
454 * Must be the root of the filesystem
455 */
456 if ((vp->v_flag & VROOT) == 0) {
457 vput(vp);
458 return (EINVAL);
459 }
460 vput(vp);
461
462 /*
463 * XXX Freeze syncer. Must do this before locking the
464 * mount point. See dounmount() for details.
465 */
466 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
467
468 if (vfs_busy(mp, 0, 0)) {
469 lockmgr(&syncer_lock, LK_RELEASE, NULL);
470 return (EBUSY);
471 }
472
473 return (dounmount(mp, SCARG(uap, flags), p));
474 }
475
476 /*
477 * Do the actual file system unmount. File system is assumed to have been
478 * marked busy by the caller.
479 */
480 int
481 dounmount(mp, flags, p)
482 struct mount *mp;
483 int flags;
484 struct proc *p;
485 {
486 struct vnode *coveredvp;
487 int error;
488 int async;
489 int used_syncer;
490
491 simple_lock(&mountlist_slock);
492 vfs_unbusy(mp);
493 used_syncer = (mp->mnt_syncer != NULL);
494
495 /*
496 * XXX Syncer must be frozen when we get here. This should really
497 * be done on a per-mountpoint basis, but especially the softdep
498 * code possibly called from the syncer doens't exactly work on a
499 * per-mountpoint basis, so the softdep code would become a maze
500 * of vfs_busy() calls.
501 *
502 * The caller of dounmount() must acquire syncer_lock because
503 * the syncer itself acquires locks in syncer_lock -> vfs_busy
504 * order, and we must preserve that order to avoid deadlock.
505 *
506 * So, if the file system did not use the syncer, now is
507 * the time to release the syncer_lock.
508 */
509 if (used_syncer == 0)
510 lockmgr(&syncer_lock, LK_RELEASE, NULL);
511
512 mp->mnt_flag |= MNT_UNMOUNT;
513 mp->mnt_unmounter = p;
514 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
515 if (mp->mnt_flag & MNT_EXPUBLIC)
516 vfs_setpublicfs(NULL, NULL, NULL);
517 async = mp->mnt_flag & MNT_ASYNC;
518 mp->mnt_flag &= ~MNT_ASYNC;
519 cache_purgevfs(mp); /* remove cache entries for this file sys */
520 if (mp->mnt_syncer != NULL)
521 vfs_deallocate_syncvnode(mp);
522 if (((mp->mnt_flag & MNT_RDONLY) ||
523 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
524 (flags & MNT_FORCE))
525 error = VFS_UNMOUNT(mp, flags, p);
526 simple_lock(&mountlist_slock);
527 if (error) {
528 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
529 (void) vfs_allocate_syncvnode(mp);
530 mp->mnt_flag &= ~MNT_UNMOUNT;
531 mp->mnt_unmounter = NULL;
532 mp->mnt_flag |= async;
533 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
534 &mountlist_slock);
535 if (used_syncer)
536 lockmgr(&syncer_lock, LK_RELEASE, NULL);
537 while (mp->mnt_wcnt > 0) {
538 wakeup((caddr_t)mp);
539 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0);
540 }
541 return (error);
542 }
543 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
544 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
545 coveredvp->v_mountedhere = NULL;
546 vrele(coveredvp);
547 }
548 mp->mnt_op->vfs_refcount--;
549 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL)
550 panic("unmount: dangling vnode");
551 mp->mnt_flag |= MNT_GONE;
552 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
553 if (used_syncer)
554 lockmgr(&syncer_lock, LK_RELEASE, NULL);
555 while(mp->mnt_wcnt > 0) {
556 wakeup((caddr_t)mp);
557 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0);
558 }
559 free((caddr_t)mp, M_MOUNT);
560 return (0);
561 }
562
563 /*
564 * Sync each mounted filesystem.
565 */
566 #ifdef DEBUG
567 int syncprt = 0;
568 struct ctldebug debug0 = { "syncprt", &syncprt };
569 #endif
570
571 /* ARGSUSED */
572 int
573 sys_sync(l, v, retval)
574 struct lwp *l;
575 void *v;
576 register_t *retval;
577 {
578 struct mount *mp, *nmp;
579 int asyncflag;
580 struct proc *p = l == NULL ? &proc0 : l->l_proc;
581
582 simple_lock(&mountlist_slock);
583 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
584 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
585 nmp = mp->mnt_list.cqe_prev;
586 continue;
587 }
588 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
589 asyncflag = mp->mnt_flag & MNT_ASYNC;
590 mp->mnt_flag &= ~MNT_ASYNC;
591 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
592 if (asyncflag)
593 mp->mnt_flag |= MNT_ASYNC;
594 }
595 simple_lock(&mountlist_slock);
596 nmp = mp->mnt_list.cqe_prev;
597 vfs_unbusy(mp);
598
599 }
600 simple_unlock(&mountlist_slock);
601 #ifdef DEBUG
602 if (syncprt)
603 vfs_bufstats();
604 #endif /* DEBUG */
605 return (0);
606 }
607
608 /*
609 * Change filesystem quotas.
610 */
611 /* ARGSUSED */
612 int
613 sys_quotactl(l, v, retval)
614 struct lwp *l;
615 void *v;
616 register_t *retval;
617 {
618 struct sys_quotactl_args /* {
619 syscallarg(const char *) path;
620 syscallarg(int) cmd;
621 syscallarg(int) uid;
622 syscallarg(caddr_t) arg;
623 } */ *uap = v;
624 struct proc *p = l->l_proc;
625 struct mount *mp;
626 int error;
627 struct nameidata nd;
628
629 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
630 if ((error = namei(&nd)) != 0)
631 return (error);
632 mp = nd.ni_vp->v_mount;
633 vrele(nd.ni_vp);
634 return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
635 SCARG(uap, arg), p));
636 }
637
638 /*
639 * Get filesystem statistics.
640 */
641 /* ARGSUSED */
642 int
643 sys_statfs(l, v, retval)
644 struct lwp *l;
645 void *v;
646 register_t *retval;
647 {
648 struct sys_statfs_args /* {
649 syscallarg(const char *) path;
650 syscallarg(struct statfs *) buf;
651 } */ *uap = v;
652 struct proc *p = l->l_proc;
653 struct mount *mp;
654 struct statfs *sp;
655 int error;
656 struct nameidata nd;
657
658 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
659 if ((error = namei(&nd)) != 0)
660 return (error);
661 mp = nd.ni_vp->v_mount;
662 sp = &mp->mnt_stat;
663 vrele(nd.ni_vp);
664 if ((error = VFS_STATFS(mp, sp, p)) != 0)
665 return (error);
666 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
667 sp->f_oflags = sp->f_flags & 0xffff;
668 return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
669 }
670
671 /*
672 * Get filesystem statistics.
673 */
674 /* ARGSUSED */
675 int
676 sys_fstatfs(l, v, retval)
677 struct lwp *l;
678 void *v;
679 register_t *retval;
680 {
681 struct sys_fstatfs_args /* {
682 syscallarg(int) fd;
683 syscallarg(struct statfs *) buf;
684 } */ *uap = v;
685 struct proc *p = l->l_proc;
686 struct file *fp;
687 struct mount *mp;
688 struct statfs *sp;
689 int error;
690
691 /* getvnode() will use the descriptor for us */
692 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
693 return (error);
694 mp = ((struct vnode *)fp->f_data)->v_mount;
695 sp = &mp->mnt_stat;
696 if ((error = VFS_STATFS(mp, sp, p)) != 0)
697 goto out;
698 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
699 sp->f_oflags = sp->f_flags & 0xffff;
700 error = copyout(sp, SCARG(uap, buf), sizeof(*sp));
701 out:
702 FILE_UNUSE(fp, p);
703 return (error);
704 }
705
706 /*
707 * Get statistics on all filesystems.
708 */
709 int
710 sys_getfsstat(l, v, retval)
711 struct lwp *l;
712 void *v;
713 register_t *retval;
714 {
715 struct sys_getfsstat_args /* {
716 syscallarg(struct statfs *) buf;
717 syscallarg(long) bufsize;
718 syscallarg(int) flags;
719 } */ *uap = v;
720 struct proc *p = l->l_proc;
721 struct mount *mp, *nmp;
722 struct statfs *sp;
723 caddr_t sfsp;
724 long count, maxcount, error;
725
726 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
727 sfsp = (caddr_t)SCARG(uap, buf);
728 simple_lock(&mountlist_slock);
729 count = 0;
730 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
731 mp = nmp) {
732 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
733 nmp = CIRCLEQ_NEXT(mp, mnt_list);
734 continue;
735 }
736 if (sfsp && count < maxcount) {
737 sp = &mp->mnt_stat;
738 /*
739 * If MNT_NOWAIT or MNT_LAZY is specified, do not
740 * refresh the fsstat cache. MNT_WAIT or MNT_LAXY
741 * overrides MNT_NOWAIT.
742 */
743 if (SCARG(uap, flags) != MNT_NOWAIT &&
744 SCARG(uap, flags) != MNT_LAZY &&
745 (SCARG(uap, flags) == MNT_WAIT ||
746 SCARG(uap, flags) == 0) &&
747 (error = VFS_STATFS(mp, sp, p)) != 0) {
748 simple_lock(&mountlist_slock);
749 nmp = CIRCLEQ_NEXT(mp, mnt_list);
750 vfs_unbusy(mp);
751 continue;
752 }
753 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
754 sp->f_oflags = sp->f_flags & 0xffff;
755 error = copyout(sp, sfsp, sizeof(*sp));
756 if (error) {
757 vfs_unbusy(mp);
758 return (error);
759 }
760 sfsp += sizeof(*sp);
761 }
762 count++;
763 simple_lock(&mountlist_slock);
764 nmp = CIRCLEQ_NEXT(mp, mnt_list);
765 vfs_unbusy(mp);
766 }
767 simple_unlock(&mountlist_slock);
768 if (sfsp && count > maxcount)
769 *retval = maxcount;
770 else
771 *retval = count;
772 return (0);
773 }
774
775 /*
776 * Change current working directory to a given file descriptor.
777 */
778 /* ARGSUSED */
779 int
780 sys_fchdir(l, v, retval)
781 struct lwp *l;
782 void *v;
783 register_t *retval;
784 {
785 struct sys_fchdir_args /* {
786 syscallarg(int) fd;
787 } */ *uap = v;
788 struct proc *p = l->l_proc;
789 struct filedesc *fdp = p->p_fd;
790 struct cwdinfo *cwdi = p->p_cwdi;
791 struct vnode *vp, *tdp;
792 struct mount *mp;
793 struct file *fp;
794 int error;
795
796 /* getvnode() will use the descriptor for us */
797 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
798 return (error);
799 vp = (struct vnode *)fp->f_data;
800
801 VREF(vp);
802 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
803 if (vp->v_type != VDIR)
804 error = ENOTDIR;
805 else
806 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
807 while (!error && (mp = vp->v_mountedhere) != NULL) {
808 if (vfs_busy(mp, 0, 0))
809 continue;
810 error = VFS_ROOT(mp, &tdp);
811 vfs_unbusy(mp);
812 if (error)
813 break;
814 vput(vp);
815 vp = tdp;
816 }
817 if (error) {
818 vput(vp);
819 goto out;
820 }
821 VOP_UNLOCK(vp, 0);
822
823 /*
824 * Disallow changing to a directory not under the process's
825 * current root directory (if there is one).
826 */
827 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
828 vrele(vp);
829 error = EPERM; /* operation not permitted */
830 goto out;
831 }
832
833 vrele(cwdi->cwdi_cdir);
834 cwdi->cwdi_cdir = vp;
835 out:
836 FILE_UNUSE(fp, p);
837 return (error);
838 }
839
840 /*
841 * Change this process's notion of the root directory to a given file descriptor.
842 */
843
844 int
845 sys_fchroot(l, v, retval)
846 struct lwp *l;
847 void *v;
848 register_t *retval;
849 {
850 struct sys_fchroot_args *uap = v;
851 struct proc *p = l->l_proc;
852 struct filedesc *fdp = p->p_fd;
853 struct cwdinfo *cwdi = p->p_cwdi;
854 struct vnode *vp;
855 struct file *fp;
856 int error;
857
858 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
859 return error;
860 /* getvnode() will use the descriptor for us */
861 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
862 return error;
863 vp = (struct vnode *) fp->f_data;
864 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
865 if (vp->v_type != VDIR)
866 error = ENOTDIR;
867 else
868 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
869 VOP_UNLOCK(vp, 0);
870 if (error)
871 goto out;
872 VREF(vp);
873
874 /*
875 * Prevent escaping from chroot by putting the root under
876 * the working directory. Silently chdir to / if we aren't
877 * already there.
878 */
879 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
880 /*
881 * XXX would be more failsafe to change directory to a
882 * deadfs node here instead
883 */
884 vrele(cwdi->cwdi_cdir);
885 VREF(vp);
886 cwdi->cwdi_cdir = vp;
887 }
888
889 if (cwdi->cwdi_rdir != NULL)
890 vrele(cwdi->cwdi_rdir);
891 cwdi->cwdi_rdir = vp;
892 out:
893 FILE_UNUSE(fp, p);
894 return (error);
895 }
896
897
898
899 /*
900 * Change current working directory (``.'').
901 */
902 /* ARGSUSED */
903 int
904 sys_chdir(l, v, retval)
905 struct lwp *l;
906 void *v;
907 register_t *retval;
908 {
909 struct sys_chdir_args /* {
910 syscallarg(const char *) path;
911 } */ *uap = v;
912 struct proc *p = l->l_proc;
913 struct cwdinfo *cwdi = p->p_cwdi;
914 int error;
915 struct nameidata nd;
916
917 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
918 SCARG(uap, path), p);
919 if ((error = change_dir(&nd, p)) != 0)
920 return (error);
921 vrele(cwdi->cwdi_cdir);
922 cwdi->cwdi_cdir = nd.ni_vp;
923 return (0);
924 }
925
926 /*
927 * Change notion of root (``/'') directory.
928 */
929 /* ARGSUSED */
930 int
931 sys_chroot(l, v, retval)
932 struct lwp *l;
933 void *v;
934 register_t *retval;
935 {
936 struct sys_chroot_args /* {
937 syscallarg(const char *) path;
938 } */ *uap = v;
939 struct proc *p = l->l_proc;
940 struct cwdinfo *cwdi = p->p_cwdi;
941 struct vnode *vp;
942 int error;
943 struct nameidata nd;
944
945 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
946 return (error);
947 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
948 SCARG(uap, path), p);
949 if ((error = change_dir(&nd, p)) != 0)
950 return (error);
951 if (cwdi->cwdi_rdir != NULL)
952 vrele(cwdi->cwdi_rdir);
953 vp = nd.ni_vp;
954 cwdi->cwdi_rdir = vp;
955
956 /*
957 * Prevent escaping from chroot by putting the root under
958 * the working directory. Silently chdir to / if we aren't
959 * already there.
960 */
961 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
962 /*
963 * XXX would be more failsafe to change directory to a
964 * deadfs node here instead
965 */
966 vrele(cwdi->cwdi_cdir);
967 VREF(vp);
968 cwdi->cwdi_cdir = vp;
969 }
970
971 return (0);
972 }
973
974 /*
975 * Common routine for chroot and chdir.
976 */
977 static int
978 change_dir(ndp, p)
979 struct nameidata *ndp;
980 struct proc *p;
981 {
982 struct vnode *vp;
983 int error;
984
985 if ((error = namei(ndp)) != 0)
986 return (error);
987 vp = ndp->ni_vp;
988 if (vp->v_type != VDIR)
989 error = ENOTDIR;
990 else
991 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
992
993 if (error)
994 vput(vp);
995 else
996 VOP_UNLOCK(vp, 0);
997 return (error);
998 }
999
1000 /*
1001 * Check permissions, allocate an open file structure,
1002 * and call the device open routine if any.
1003 */
1004 int
1005 sys_open(l, v, retval)
1006 struct lwp *l;
1007 void *v;
1008 register_t *retval;
1009 {
1010 struct sys_open_args /* {
1011 syscallarg(const char *) path;
1012 syscallarg(int) flags;
1013 syscallarg(int) mode;
1014 } */ *uap = v;
1015 struct proc *p = l->l_proc;
1016 struct cwdinfo *cwdi = p->p_cwdi;
1017 struct filedesc *fdp = p->p_fd;
1018 struct file *fp;
1019 struct vnode *vp;
1020 int flags, cmode;
1021 int type, indx, error;
1022 struct flock lf;
1023 struct nameidata nd;
1024
1025 flags = FFLAGS(SCARG(uap, flags));
1026 if ((flags & (FREAD | FWRITE)) == 0)
1027 return (EINVAL);
1028 /* falloc() will use the file descriptor for us */
1029 if ((error = falloc(p, &fp, &indx)) != 0)
1030 return (error);
1031 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1032 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1033 p->p_dupfd = -indx - 1; /* XXX check for fdopen */
1034 if ((error = vn_open(&nd, flags, cmode)) != 0) {
1035 FILE_UNUSE(fp, p);
1036 ffree(fp);
1037 if ((error == ENODEV || error == ENXIO) &&
1038 p->p_dupfd >= 0 && /* XXX from fdopen */
1039 (error =
1040 dupfdopen(p, indx, p->p_dupfd, flags, error)) == 0) {
1041 *retval = indx;
1042 return (0);
1043 }
1044 if (error == ERESTART)
1045 error = EINTR;
1046 fdremove(fdp, indx);
1047 return (error);
1048 }
1049 p->p_dupfd = 0;
1050 vp = nd.ni_vp;
1051 fp->f_flag = flags & FMASK;
1052 fp->f_type = DTYPE_VNODE;
1053 fp->f_ops = &vnops;
1054 fp->f_data = (caddr_t)vp;
1055 if (flags & (O_EXLOCK | O_SHLOCK)) {
1056 lf.l_whence = SEEK_SET;
1057 lf.l_start = 0;
1058 lf.l_len = 0;
1059 if (flags & O_EXLOCK)
1060 lf.l_type = F_WRLCK;
1061 else
1062 lf.l_type = F_RDLCK;
1063 type = F_FLOCK;
1064 if ((flags & FNONBLOCK) == 0)
1065 type |= F_WAIT;
1066 VOP_UNLOCK(vp, 0);
1067 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1068 if (error) {
1069 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1070 FILE_UNUSE(fp, p);
1071 ffree(fp);
1072 fdremove(fdp, indx);
1073 return (error);
1074 }
1075 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1076 fp->f_flag |= FHASLOCK;
1077 }
1078 VOP_UNLOCK(vp, 0);
1079 *retval = indx;
1080 FILE_SET_MATURE(fp);
1081 FILE_UNUSE(fp, p);
1082 return (0);
1083 }
1084
1085 /*
1086 * Get file handle system call
1087 */
1088 int
1089 sys_getfh(l, v, retval)
1090 struct lwp *l;
1091 void *v;
1092 register_t *retval;
1093 {
1094 struct sys_getfh_args /* {
1095 syscallarg(char *) fname;
1096 syscallarg(fhandle_t *) fhp;
1097 } */ *uap = v;
1098 struct proc *p = l->l_proc;
1099 struct vnode *vp;
1100 fhandle_t fh;
1101 int error;
1102 struct nameidata nd;
1103
1104 /*
1105 * Must be super user
1106 */
1107 error = suser(p->p_ucred, &p->p_acflag);
1108 if (error)
1109 return (error);
1110 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1111 SCARG(uap, fname), p);
1112 error = namei(&nd);
1113 if (error)
1114 return (error);
1115 vp = nd.ni_vp;
1116 memset((caddr_t)&fh, 0, sizeof(fh));
1117 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1118 error = VFS_VPTOFH(vp, &fh.fh_fid);
1119 vput(vp);
1120 if (error)
1121 return (error);
1122 error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1123 return (error);
1124 }
1125
1126 /*
1127 * Open a file given a file handle.
1128 *
1129 * Check permissions, allocate an open file structure,
1130 * and call the device open routine if any.
1131 */
1132 int
1133 sys_fhopen(l, v, retval)
1134 struct lwp *l;
1135 void *v;
1136 register_t *retval;
1137 {
1138 struct sys_fhopen_args /* {
1139 syscallarg(const fhandle_t *) fhp;
1140 syscallarg(int) flags;
1141 } */ *uap = v;
1142 struct proc *p = l->l_proc;
1143 struct filedesc *fdp = p->p_fd;
1144 struct file *fp;
1145 struct vnode *vp = NULL;
1146 struct mount *mp;
1147 struct ucred *cred = p->p_ucred;
1148 int flags;
1149 struct file *nfp;
1150 int type, indx, error=0;
1151 struct flock lf;
1152 struct vattr va;
1153 fhandle_t fh;
1154
1155 /*
1156 * Must be super user
1157 */
1158 if ((error = suser(p->p_ucred, &p->p_acflag)))
1159 return (error);
1160
1161 flags = FFLAGS(SCARG(uap, flags));
1162 if ((flags & (FREAD | FWRITE)) == 0)
1163 return (EINVAL);
1164 if ((flags & O_CREAT))
1165 return (EINVAL);
1166 /* falloc() will use the file descriptor for us */
1167 if ((error = falloc(p, &nfp, &indx)) != 0)
1168 return (error);
1169 fp = nfp;
1170 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1171 goto bad;
1172
1173 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1174 error = ESTALE;
1175 goto bad;
1176 }
1177
1178 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1179 vp = NULL; /* most likely unnecessary sanity for bad: */
1180 goto bad;
1181 }
1182
1183 /* Now do an effective vn_open */
1184
1185 if (vp->v_type == VSOCK) {
1186 error = EOPNOTSUPP;
1187 goto bad;
1188 }
1189 if (flags & FREAD) {
1190 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1191 goto bad;
1192 }
1193 if (flags & (FWRITE | O_TRUNC)) {
1194 if (vp->v_type == VDIR) {
1195 error = EISDIR;
1196 goto bad;
1197 }
1198 if ((error = vn_writechk(vp)) != 0 ||
1199 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1200 goto bad;
1201 }
1202 if (flags & O_TRUNC) {
1203 VOP_UNLOCK(vp, 0); /* XXX */
1204 VOP_LEASE(vp, p, cred, LEASE_WRITE);
1205 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
1206 VATTR_NULL(&va);
1207 va.va_size = 0;
1208 if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1209 goto bad;
1210 }
1211 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1212 goto bad;
1213 if (vp->v_type == VREG &&
1214 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1215 error = EIO;
1216 goto bad;
1217 }
1218 if (flags & FWRITE)
1219 vp->v_writecount++;
1220
1221 /* done with modified vn_open, now finish what sys_open does. */
1222
1223 fp->f_flag = flags & FMASK;
1224 fp->f_type = DTYPE_VNODE;
1225 fp->f_ops = &vnops;
1226 fp->f_data = (caddr_t)vp;
1227 if (flags & (O_EXLOCK | O_SHLOCK)) {
1228 lf.l_whence = SEEK_SET;
1229 lf.l_start = 0;
1230 lf.l_len = 0;
1231 if (flags & O_EXLOCK)
1232 lf.l_type = F_WRLCK;
1233 else
1234 lf.l_type = F_RDLCK;
1235 type = F_FLOCK;
1236 if ((flags & FNONBLOCK) == 0)
1237 type |= F_WAIT;
1238 VOP_UNLOCK(vp, 0);
1239 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1240 if (error) {
1241 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1242 FILE_UNUSE(fp, p);
1243 ffree(fp);
1244 fdremove(fdp, indx);
1245 return (error);
1246 }
1247 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1248 fp->f_flag |= FHASLOCK;
1249 }
1250 VOP_UNLOCK(vp, 0);
1251 *retval = indx;
1252 FILE_SET_MATURE(fp);
1253 FILE_UNUSE(fp, p);
1254 return (0);
1255
1256 bad:
1257 FILE_UNUSE(fp, p);
1258 ffree(fp);
1259 fdremove(fdp, indx);
1260 if (vp != NULL)
1261 vput(vp);
1262 return (error);
1263 }
1264
1265 /* ARGSUSED */
1266 int
1267 sys_fhstat(l, v, retval)
1268 struct lwp *l;
1269 void *v;
1270 register_t *retval;
1271 {
1272 struct sys_fhstat_args /* {
1273 syscallarg(const fhandle_t *) fhp;
1274 syscallarg(struct stat *) sb;
1275 } */ *uap = v;
1276 struct proc *p = l->l_proc;
1277 struct stat sb;
1278 int error;
1279 fhandle_t fh;
1280 struct mount *mp;
1281 struct vnode *vp;
1282
1283 /*
1284 * Must be super user
1285 */
1286 if ((error = suser(p->p_ucred, &p->p_acflag)))
1287 return (error);
1288
1289 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1290 return (error);
1291
1292 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1293 return (ESTALE);
1294 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1295 return (error);
1296 error = vn_stat(vp, &sb, p);
1297 vput(vp);
1298 if (error)
1299 return (error);
1300 error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1301 return (error);
1302 }
1303
1304 /* ARGSUSED */
1305 int
1306 sys_fhstatfs(l, v, retval)
1307 struct lwp *l;
1308 void *v;
1309 register_t *retval;
1310 {
1311 struct sys_fhstatfs_args /*
1312 syscallarg(const fhandle_t *) fhp;
1313 syscallarg(struct statfs *) buf;
1314 } */ *uap = v;
1315 struct proc *p = l->l_proc;
1316 struct statfs sp;
1317 fhandle_t fh;
1318 struct mount *mp;
1319 struct vnode *vp;
1320 int error;
1321
1322 /*
1323 * Must be super user
1324 */
1325 if ((error = suser(p->p_ucred, &p->p_acflag)))
1326 return (error);
1327
1328 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1329 return (error);
1330
1331 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1332 return (ESTALE);
1333 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1334 return (error);
1335 mp = vp->v_mount;
1336 vput(vp);
1337 if ((error = VFS_STATFS(mp, &sp, p)) != 0)
1338 return (error);
1339 sp.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1340 sp.f_oflags = sp.f_flags & 0xffff;
1341 return (copyout(&sp, SCARG(uap, buf), sizeof(sp)));
1342 }
1343
1344 /*
1345 * Create a special file.
1346 */
1347 /* ARGSUSED */
1348 int
1349 sys_mknod(l, v, retval)
1350 struct lwp *l;
1351 void *v;
1352 register_t *retval;
1353 {
1354 struct sys_mknod_args /* {
1355 syscallarg(const char *) path;
1356 syscallarg(int) mode;
1357 syscallarg(int) dev;
1358 } */ *uap = v;
1359 struct proc *p = l->l_proc;
1360 struct vnode *vp;
1361 struct vattr vattr;
1362 int error;
1363 int whiteout = 0;
1364 struct nameidata nd;
1365
1366 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1367 return (error);
1368 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1369 if ((error = namei(&nd)) != 0)
1370 return (error);
1371 vp = nd.ni_vp;
1372 if (vp != NULL)
1373 error = EEXIST;
1374 else {
1375 VATTR_NULL(&vattr);
1376 vattr.va_mode =
1377 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1378 vattr.va_rdev = SCARG(uap, dev);
1379 whiteout = 0;
1380
1381 switch (SCARG(uap, mode) & S_IFMT) {
1382 case S_IFMT: /* used by badsect to flag bad sectors */
1383 vattr.va_type = VBAD;
1384 break;
1385 case S_IFCHR:
1386 vattr.va_type = VCHR;
1387 break;
1388 case S_IFBLK:
1389 vattr.va_type = VBLK;
1390 break;
1391 case S_IFWHT:
1392 whiteout = 1;
1393 break;
1394 default:
1395 error = EINVAL;
1396 break;
1397 }
1398 }
1399 if (!error) {
1400 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1401 if (whiteout) {
1402 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1403 if (error)
1404 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1405 vput(nd.ni_dvp);
1406 } else {
1407 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1408 &nd.ni_cnd, &vattr);
1409 if (error == 0)
1410 vput(nd.ni_vp);
1411 }
1412 } else {
1413 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1414 if (nd.ni_dvp == vp)
1415 vrele(nd.ni_dvp);
1416 else
1417 vput(nd.ni_dvp);
1418 if (vp)
1419 vrele(vp);
1420 }
1421 return (error);
1422 }
1423
1424 /*
1425 * Create a named pipe.
1426 */
1427 /* ARGSUSED */
1428 int
1429 sys_mkfifo(l, v, retval)
1430 struct lwp *l;
1431 void *v;
1432 register_t *retval;
1433 {
1434 struct sys_mkfifo_args /* {
1435 syscallarg(const char *) path;
1436 syscallarg(int) mode;
1437 } */ *uap = v;
1438 struct proc *p = l->l_proc;
1439 struct vattr vattr;
1440 int error;
1441 struct nameidata nd;
1442
1443 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1444 if ((error = namei(&nd)) != 0)
1445 return (error);
1446 if (nd.ni_vp != NULL) {
1447 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1448 if (nd.ni_dvp == nd.ni_vp)
1449 vrele(nd.ni_dvp);
1450 else
1451 vput(nd.ni_dvp);
1452 vrele(nd.ni_vp);
1453 return (EEXIST);
1454 }
1455 VATTR_NULL(&vattr);
1456 vattr.va_type = VFIFO;
1457 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1458 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1459 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1460 if (error == 0)
1461 vput(nd.ni_vp);
1462 return (error);
1463 }
1464
1465 /*
1466 * Make a hard file link.
1467 */
1468 /* ARGSUSED */
1469 int
1470 sys_link(l, v, retval)
1471 struct lwp *l;
1472 void *v;
1473 register_t *retval;
1474 {
1475 struct sys_link_args /* {
1476 syscallarg(const char *) path;
1477 syscallarg(const char *) link;
1478 } */ *uap = v;
1479 struct proc *p = l->l_proc;
1480 struct vnode *vp;
1481 struct nameidata nd;
1482 int error;
1483
1484 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1485 if ((error = namei(&nd)) != 0)
1486 return (error);
1487 vp = nd.ni_vp;
1488 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1489 if ((error = namei(&nd)) != 0)
1490 goto out;
1491 if (nd.ni_vp) {
1492 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1493 if (nd.ni_dvp == nd.ni_vp)
1494 vrele(nd.ni_dvp);
1495 else
1496 vput(nd.ni_dvp);
1497 vrele(nd.ni_vp);
1498 error = EEXIST;
1499 goto out;
1500 }
1501 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1502 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1503 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1504 out:
1505 vrele(vp);
1506 return (error);
1507 }
1508
1509 /*
1510 * Make a symbolic link.
1511 */
1512 /* ARGSUSED */
1513 int
1514 sys_symlink(l, v, retval)
1515 struct lwp *l;
1516 void *v;
1517 register_t *retval;
1518 {
1519 struct sys_symlink_args /* {
1520 syscallarg(const char *) path;
1521 syscallarg(const char *) link;
1522 } */ *uap = v;
1523 struct proc *p = l->l_proc;
1524 struct vattr vattr;
1525 char *path;
1526 int error;
1527 struct nameidata nd;
1528
1529 path = PNBUF_GET();
1530 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1531 if (error)
1532 goto out;
1533 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1534 if ((error = namei(&nd)) != 0)
1535 goto out;
1536 if (nd.ni_vp) {
1537 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1538 if (nd.ni_dvp == nd.ni_vp)
1539 vrele(nd.ni_dvp);
1540 else
1541 vput(nd.ni_dvp);
1542 vrele(nd.ni_vp);
1543 error = EEXIST;
1544 goto out;
1545 }
1546 VATTR_NULL(&vattr);
1547 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1548 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1549 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1550 if (error == 0)
1551 vput(nd.ni_vp);
1552 out:
1553 PNBUF_PUT(path);
1554 return (error);
1555 }
1556
1557 /*
1558 * Delete a whiteout from the filesystem.
1559 */
1560 /* ARGSUSED */
1561 int
1562 sys_undelete(l, v, retval)
1563 struct lwp *l;
1564 void *v;
1565 register_t *retval;
1566 {
1567 struct sys_undelete_args /* {
1568 syscallarg(const char *) path;
1569 } */ *uap = v;
1570 struct proc *p = l->l_proc;
1571 int error;
1572 struct nameidata nd;
1573
1574 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1575 SCARG(uap, path), p);
1576 error = namei(&nd);
1577 if (error)
1578 return (error);
1579
1580 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1581 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1582 if (nd.ni_dvp == nd.ni_vp)
1583 vrele(nd.ni_dvp);
1584 else
1585 vput(nd.ni_dvp);
1586 if (nd.ni_vp)
1587 vrele(nd.ni_vp);
1588 return (EEXIST);
1589 }
1590
1591 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1592 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1593 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1594 vput(nd.ni_dvp);
1595 return (error);
1596 }
1597
1598 /*
1599 * Delete a name from the filesystem.
1600 */
1601 /* ARGSUSED */
1602 int
1603 sys_unlink(l, v, retval)
1604 struct lwp *l;
1605 void *v;
1606 register_t *retval;
1607 {
1608 struct sys_unlink_args /* {
1609 syscallarg(const char *) path;
1610 } */ *uap = v;
1611 struct proc *p = l->l_proc;
1612 struct vnode *vp;
1613 int error;
1614 struct nameidata nd;
1615
1616 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1617 SCARG(uap, path), p);
1618 if ((error = namei(&nd)) != 0)
1619 return (error);
1620 vp = nd.ni_vp;
1621
1622 /*
1623 * The root of a mounted filesystem cannot be deleted.
1624 */
1625 if (vp->v_flag & VROOT) {
1626 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1627 if (nd.ni_dvp == vp)
1628 vrele(nd.ni_dvp);
1629 else
1630 vput(nd.ni_dvp);
1631 vput(vp);
1632 error = EBUSY;
1633 goto out;
1634 }
1635
1636 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1637 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1638 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1639 out:
1640 return (error);
1641 }
1642
1643 /*
1644 * Reposition read/write file offset.
1645 */
1646 int
1647 sys_lseek(l, v, retval)
1648 struct lwp *l;
1649 void *v;
1650 register_t *retval;
1651 {
1652 struct sys_lseek_args /* {
1653 syscallarg(int) fd;
1654 syscallarg(int) pad;
1655 syscallarg(off_t) offset;
1656 syscallarg(int) whence;
1657 } */ *uap = v;
1658 struct proc *p = l->l_proc;
1659 struct ucred *cred = p->p_ucred;
1660 struct filedesc *fdp = p->p_fd;
1661 struct file *fp;
1662 struct vnode *vp;
1663 struct vattr vattr;
1664 off_t newoff;
1665 int error;
1666
1667 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1668 return (EBADF);
1669
1670 FILE_USE(fp);
1671
1672 vp = (struct vnode *)fp->f_data;
1673 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1674 error = ESPIPE;
1675 goto out;
1676 }
1677
1678 switch (SCARG(uap, whence)) {
1679 case SEEK_CUR:
1680 newoff = fp->f_offset + SCARG(uap, offset);
1681 break;
1682 case SEEK_END:
1683 error = VOP_GETATTR(vp, &vattr, cred, p);
1684 if (error)
1685 goto out;
1686 newoff = SCARG(uap, offset) + vattr.va_size;
1687 break;
1688 case SEEK_SET:
1689 newoff = SCARG(uap, offset);
1690 break;
1691 default:
1692 error = EINVAL;
1693 goto out;
1694 }
1695 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1696 goto out;
1697
1698 *(off_t *)retval = fp->f_offset = newoff;
1699 out:
1700 FILE_UNUSE(fp, p);
1701 return (error);
1702 }
1703
1704 /*
1705 * Positional read system call.
1706 */
1707 int
1708 sys_pread(l, v, retval)
1709 struct lwp *l;
1710 void *v;
1711 register_t *retval;
1712 {
1713 struct sys_pread_args /* {
1714 syscallarg(int) fd;
1715 syscallarg(void *) buf;
1716 syscallarg(size_t) nbyte;
1717 syscallarg(off_t) offset;
1718 } */ *uap = v;
1719 struct proc *p = l->l_proc;
1720 struct filedesc *fdp = p->p_fd;
1721 struct file *fp;
1722 struct vnode *vp;
1723 off_t offset;
1724 int error, fd = SCARG(uap, fd);
1725
1726 if ((fp = fd_getfile(fdp, fd)) == NULL)
1727 return (EBADF);
1728
1729 if ((fp->f_flag & FREAD) == 0)
1730 return (EBADF);
1731
1732 FILE_USE(fp);
1733
1734 vp = (struct vnode *)fp->f_data;
1735 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1736 error = ESPIPE;
1737 goto out;
1738 }
1739
1740 offset = SCARG(uap, offset);
1741
1742 /*
1743 * XXX This works because no file systems actually
1744 * XXX take any action on the seek operation.
1745 */
1746 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1747 goto out;
1748
1749 /* dofileread() will unuse the descriptor for us */
1750 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1751 &offset, 0, retval));
1752
1753 out:
1754 FILE_UNUSE(fp, p);
1755 return (error);
1756 }
1757
1758 /*
1759 * Positional scatter read system call.
1760 */
1761 int
1762 sys_preadv(l, v, retval)
1763 struct lwp *l;
1764 void *v;
1765 register_t *retval;
1766 {
1767 struct sys_preadv_args /* {
1768 syscallarg(int) fd;
1769 syscallarg(const struct iovec *) iovp;
1770 syscallarg(int) iovcnt;
1771 syscallarg(off_t) offset;
1772 } */ *uap = v;
1773 struct proc *p = l->l_proc;
1774 struct filedesc *fdp = p->p_fd;
1775 struct file *fp;
1776 struct vnode *vp;
1777 off_t offset;
1778 int error, fd = SCARG(uap, fd);
1779
1780 if ((fp = fd_getfile(fdp, fd)) == NULL)
1781 return (EBADF);
1782
1783 if ((fp->f_flag & FREAD) == 0)
1784 return (EBADF);
1785
1786 FILE_USE(fp);
1787
1788 vp = (struct vnode *)fp->f_data;
1789 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1790 error = ESPIPE;
1791 goto out;
1792 }
1793
1794 offset = SCARG(uap, offset);
1795
1796 /*
1797 * XXX This works because no file systems actually
1798 * XXX take any action on the seek operation.
1799 */
1800 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1801 goto out;
1802
1803 /* dofilereadv() will unuse the descriptor for us */
1804 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1805 &offset, 0, retval));
1806
1807 out:
1808 FILE_UNUSE(fp, p);
1809 return (error);
1810 }
1811
1812 /*
1813 * Positional write system call.
1814 */
1815 int
1816 sys_pwrite(l, v, retval)
1817 struct lwp *l;
1818 void *v;
1819 register_t *retval;
1820 {
1821 struct sys_pwrite_args /* {
1822 syscallarg(int) fd;
1823 syscallarg(const void *) buf;
1824 syscallarg(size_t) nbyte;
1825 syscallarg(off_t) offset;
1826 } */ *uap = v;
1827 struct proc *p = l->l_proc;
1828 struct filedesc *fdp = p->p_fd;
1829 struct file *fp;
1830 struct vnode *vp;
1831 off_t offset;
1832 int error, fd = SCARG(uap, fd);
1833
1834 if ((fp = fd_getfile(fdp, fd)) == NULL)
1835 return (EBADF);
1836
1837 if ((fp->f_flag & FWRITE) == 0)
1838 return (EBADF);
1839
1840 FILE_USE(fp);
1841
1842 vp = (struct vnode *)fp->f_data;
1843 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1844 error = ESPIPE;
1845 goto out;
1846 }
1847
1848 offset = SCARG(uap, offset);
1849
1850 /*
1851 * XXX This works because no file systems actually
1852 * XXX take any action on the seek operation.
1853 */
1854 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1855 goto out;
1856
1857 /* dofilewrite() will unuse the descriptor for us */
1858 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1859 &offset, 0, retval));
1860
1861 out:
1862 FILE_UNUSE(fp, p);
1863 return (error);
1864 }
1865
1866 /*
1867 * Positional gather write system call.
1868 */
1869 int
1870 sys_pwritev(l, v, retval)
1871 struct lwp *l;
1872 void *v;
1873 register_t *retval;
1874 {
1875 struct sys_pwritev_args /* {
1876 syscallarg(int) fd;
1877 syscallarg(const struct iovec *) iovp;
1878 syscallarg(int) iovcnt;
1879 syscallarg(off_t) offset;
1880 } */ *uap = v;
1881 struct proc *p = l->l_proc;
1882 struct filedesc *fdp = p->p_fd;
1883 struct file *fp;
1884 struct vnode *vp;
1885 off_t offset;
1886 int error, fd = SCARG(uap, fd);
1887
1888 if ((fp = fd_getfile(fdp, fd)) == NULL)
1889 return (EBADF);
1890
1891 if ((fp->f_flag & FWRITE) == 0)
1892 return (EBADF);
1893
1894 FILE_USE(fp);
1895
1896 vp = (struct vnode *)fp->f_data;
1897 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1898 error = ESPIPE;
1899 goto out;
1900 }
1901
1902 offset = SCARG(uap, offset);
1903
1904 /*
1905 * XXX This works because no file systems actually
1906 * XXX take any action on the seek operation.
1907 */
1908 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1909 goto out;
1910
1911 /* dofilewritev() will unuse the descriptor for us */
1912 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1913 &offset, 0, retval));
1914
1915 out:
1916 FILE_UNUSE(fp, p);
1917 return (error);
1918 }
1919
1920 /*
1921 * Check access permissions.
1922 */
1923 int
1924 sys_access(l, v, retval)
1925 struct lwp *l;
1926 void *v;
1927 register_t *retval;
1928 {
1929 struct sys_access_args /* {
1930 syscallarg(const char *) path;
1931 syscallarg(int) flags;
1932 } */ *uap = v;
1933 struct proc *p = l->l_proc;
1934 struct ucred *cred = crget();
1935 struct vnode *vp;
1936 int error, flags;
1937 struct nameidata nd;
1938
1939 (void)memcpy(cred, p->p_ucred, sizeof(*cred));
1940 cred->cr_ref = 1;
1941 cred->cr_uid = p->p_cred->p_ruid;
1942 cred->cr_gid = p->p_cred->p_rgid;
1943 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1944 SCARG(uap, path), p);
1945 /* Override default credentials */
1946 nd.ni_cnd.cn_cred = cred;
1947 if ((error = namei(&nd)) != 0)
1948 goto out;
1949 vp = nd.ni_vp;
1950
1951 /* Flags == 0 means only check for existence. */
1952 if (SCARG(uap, flags)) {
1953 flags = 0;
1954 if (SCARG(uap, flags) & R_OK)
1955 flags |= VREAD;
1956 if (SCARG(uap, flags) & W_OK)
1957 flags |= VWRITE;
1958 if (SCARG(uap, flags) & X_OK)
1959 flags |= VEXEC;
1960
1961 error = VOP_ACCESS(vp, flags, cred, p);
1962 if (!error && (flags & VWRITE))
1963 error = vn_writechk(vp);
1964 }
1965 vput(vp);
1966 out:
1967 crfree(cred);
1968 return (error);
1969 }
1970
1971 /*
1972 * Get file status; this version follows links.
1973 */
1974 /* ARGSUSED */
1975 int
1976 sys___stat13(l, v, retval)
1977 struct lwp *l;
1978 void *v;
1979 register_t *retval;
1980 {
1981 struct sys___stat13_args /* {
1982 syscallarg(const char *) path;
1983 syscallarg(struct stat *) ub;
1984 } */ *uap = v;
1985 struct proc *p = l->l_proc;
1986 struct stat sb;
1987 int error;
1988 struct nameidata nd;
1989
1990 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1991 SCARG(uap, path), p);
1992 if ((error = namei(&nd)) != 0)
1993 return (error);
1994 error = vn_stat(nd.ni_vp, &sb, p);
1995 vput(nd.ni_vp);
1996 if (error)
1997 return (error);
1998 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1999 return (error);
2000 }
2001
2002 /*
2003 * Get file status; this version does not follow links.
2004 */
2005 /* ARGSUSED */
2006 int
2007 sys___lstat13(l, v, retval)
2008 struct lwp *l;
2009 void *v;
2010 register_t *retval;
2011 {
2012 struct sys___lstat13_args /* {
2013 syscallarg(const char *) path;
2014 syscallarg(struct stat *) ub;
2015 } */ *uap = v;
2016 struct proc *p = l->l_proc;
2017 struct stat sb;
2018 int error;
2019 struct nameidata nd;
2020
2021 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2022 SCARG(uap, path), p);
2023 if ((error = namei(&nd)) != 0)
2024 return (error);
2025 error = vn_stat(nd.ni_vp, &sb, p);
2026 vput(nd.ni_vp);
2027 if (error)
2028 return (error);
2029 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2030 return (error);
2031 }
2032
2033 /*
2034 * Get configurable pathname variables.
2035 */
2036 /* ARGSUSED */
2037 int
2038 sys_pathconf(l, v, retval)
2039 struct lwp *l;
2040 void *v;
2041 register_t *retval;
2042 {
2043 struct sys_pathconf_args /* {
2044 syscallarg(const char *) path;
2045 syscallarg(int) name;
2046 } */ *uap = v;
2047 struct proc *p = l->l_proc;
2048 int error;
2049 struct nameidata nd;
2050
2051 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2052 SCARG(uap, path), p);
2053 if ((error = namei(&nd)) != 0)
2054 return (error);
2055 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2056 vput(nd.ni_vp);
2057 return (error);
2058 }
2059
2060 /*
2061 * Return target name of a symbolic link.
2062 */
2063 /* ARGSUSED */
2064 int
2065 sys_readlink(l, v, retval)
2066 struct lwp *l;
2067 void *v;
2068 register_t *retval;
2069 {
2070 struct sys_readlink_args /* {
2071 syscallarg(const char *) path;
2072 syscallarg(char *) buf;
2073 syscallarg(size_t) count;
2074 } */ *uap = v;
2075 struct proc *p = l->l_proc;
2076 struct vnode *vp;
2077 struct iovec aiov;
2078 struct uio auio;
2079 int error;
2080 struct nameidata nd;
2081
2082 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2083 SCARG(uap, path), p);
2084 if ((error = namei(&nd)) != 0)
2085 return (error);
2086 vp = nd.ni_vp;
2087 if (vp->v_type != VLNK)
2088 error = EINVAL;
2089 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2090 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2091 aiov.iov_base = SCARG(uap, buf);
2092 aiov.iov_len = SCARG(uap, count);
2093 auio.uio_iov = &aiov;
2094 auio.uio_iovcnt = 1;
2095 auio.uio_offset = 0;
2096 auio.uio_rw = UIO_READ;
2097 auio.uio_segflg = UIO_USERSPACE;
2098 auio.uio_procp = p;
2099 auio.uio_resid = SCARG(uap, count);
2100 error = VOP_READLINK(vp, &auio, p->p_ucred);
2101 }
2102 vput(vp);
2103 *retval = SCARG(uap, count) - auio.uio_resid;
2104 return (error);
2105 }
2106
2107 /*
2108 * Change flags of a file given a path name.
2109 */
2110 /* ARGSUSED */
2111 int
2112 sys_chflags(l, v, retval)
2113 struct lwp *l;
2114 void *v;
2115 register_t *retval;
2116 {
2117 struct sys_chflags_args /* {
2118 syscallarg(const char *) path;
2119 syscallarg(u_long) flags;
2120 } */ *uap = v;
2121 struct proc *p = l->l_proc;
2122 struct vnode *vp;
2123 int error;
2124 struct nameidata nd;
2125
2126 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2127 if ((error = namei(&nd)) != 0)
2128 return (error);
2129 vp = nd.ni_vp;
2130 error = change_flags(vp, SCARG(uap, flags), p);
2131 vput(vp);
2132 return (error);
2133 }
2134
2135 /*
2136 * Change flags of a file given a file descriptor.
2137 */
2138 /* ARGSUSED */
2139 int
2140 sys_fchflags(l, v, retval)
2141 struct lwp *l;
2142 void *v;
2143 register_t *retval;
2144 {
2145 struct sys_fchflags_args /* {
2146 syscallarg(int) fd;
2147 syscallarg(u_long) flags;
2148 } */ *uap = v;
2149 struct proc *p = l->l_proc;
2150 struct vnode *vp;
2151 struct file *fp;
2152 int error;
2153
2154 /* getvnode() will use the descriptor for us */
2155 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2156 return (error);
2157 vp = (struct vnode *)fp->f_data;
2158 error = change_flags(vp, SCARG(uap, flags), p);
2159 VOP_UNLOCK(vp, 0);
2160 FILE_UNUSE(fp, p);
2161 return (error);
2162 }
2163
2164 /*
2165 * Change flags of a file given a path name; this version does
2166 * not follow links.
2167 */
2168 int
2169 sys_lchflags(l, v, retval)
2170 struct lwp *l;
2171 void *v;
2172 register_t *retval;
2173 {
2174 struct sys_lchflags_args /* {
2175 syscallarg(const char *) path;
2176 syscallarg(u_long) flags;
2177 } */ *uap = v;
2178 struct proc *p = l->l_proc;
2179 struct vnode *vp;
2180 int error;
2181 struct nameidata nd;
2182
2183 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2184 if ((error = namei(&nd)) != 0)
2185 return (error);
2186 vp = nd.ni_vp;
2187 error = change_flags(vp, SCARG(uap, flags), p);
2188 vput(vp);
2189 return (error);
2190 }
2191
2192 /*
2193 * Common routine to change flags of a file.
2194 */
2195 int
2196 change_flags(vp, flags, p)
2197 struct vnode *vp;
2198 u_long flags;
2199 struct proc *p;
2200 {
2201 struct vattr vattr;
2202 int error;
2203
2204 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2205 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2206 /*
2207 * Non-superusers cannot change the flags on devices, even if they
2208 * own them.
2209 */
2210 if (suser(p->p_ucred, &p->p_acflag) != 0) {
2211 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2212 goto out;
2213 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2214 error = EINVAL;
2215 goto out;
2216 }
2217 }
2218 VATTR_NULL(&vattr);
2219 vattr.va_flags = flags;
2220 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2221 out:
2222 return (error);
2223 }
2224
2225 /*
2226 * Change mode of a file given path name; this version follows links.
2227 */
2228 /* ARGSUSED */
2229 int
2230 sys_chmod(l, v, retval)
2231 struct lwp *l;
2232 void *v;
2233 register_t *retval;
2234 {
2235 struct sys_chmod_args /* {
2236 syscallarg(const char *) path;
2237 syscallarg(int) mode;
2238 } */ *uap = v;
2239 struct proc *p = l->l_proc;
2240 int error;
2241 struct nameidata nd;
2242
2243 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2244 if ((error = namei(&nd)) != 0)
2245 return (error);
2246
2247 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2248
2249 vrele(nd.ni_vp);
2250 return (error);
2251 }
2252
2253 /*
2254 * Change mode of a file given a file descriptor.
2255 */
2256 /* ARGSUSED */
2257 int
2258 sys_fchmod(l, v, retval)
2259 struct lwp *l;
2260 void *v;
2261 register_t *retval;
2262 {
2263 struct sys_fchmod_args /* {
2264 syscallarg(int) fd;
2265 syscallarg(int) mode;
2266 } */ *uap = v;
2267 struct proc *p = l->l_proc;
2268 struct file *fp;
2269 int error;
2270
2271 /* getvnode() will use the descriptor for us */
2272 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2273 return (error);
2274
2275 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2276 FILE_UNUSE(fp, p);
2277 return (error);
2278 }
2279
2280 /*
2281 * Change mode of a file given path name; this version does not follow links.
2282 */
2283 /* ARGSUSED */
2284 int
2285 sys_lchmod(l, v, retval)
2286 struct lwp *l;
2287 void *v;
2288 register_t *retval;
2289 {
2290 struct sys_lchmod_args /* {
2291 syscallarg(const char *) path;
2292 syscallarg(int) mode;
2293 } */ *uap = v;
2294 struct proc *p = l->l_proc;
2295 int error;
2296 struct nameidata nd;
2297
2298 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2299 if ((error = namei(&nd)) != 0)
2300 return (error);
2301
2302 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2303
2304 vrele(nd.ni_vp);
2305 return (error);
2306 }
2307
2308 /*
2309 * Common routine to set mode given a vnode.
2310 */
2311 static int
2312 change_mode(vp, mode, p)
2313 struct vnode *vp;
2314 int mode;
2315 struct proc *p;
2316 {
2317 struct vattr vattr;
2318 int error;
2319
2320 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2321 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2322 VATTR_NULL(&vattr);
2323 vattr.va_mode = mode & ALLPERMS;
2324 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2325 VOP_UNLOCK(vp, 0);
2326 return (error);
2327 }
2328
2329 /*
2330 * Set ownership given a path name; this version follows links.
2331 */
2332 /* ARGSUSED */
2333 int
2334 sys_chown(l, v, retval)
2335 struct lwp *l;
2336 void *v;
2337 register_t *retval;
2338 {
2339 struct sys_chown_args /* {
2340 syscallarg(const char *) path;
2341 syscallarg(uid_t) uid;
2342 syscallarg(gid_t) gid;
2343 } */ *uap = v;
2344 struct proc *p = l->l_proc;
2345 int error;
2346 struct nameidata nd;
2347
2348 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2349 if ((error = namei(&nd)) != 0)
2350 return (error);
2351
2352 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2353
2354 vrele(nd.ni_vp);
2355 return (error);
2356 }
2357
2358 /*
2359 * Set ownership given a path name; this version follows links.
2360 * Provides POSIX semantics.
2361 */
2362 /* ARGSUSED */
2363 int
2364 sys___posix_chown(l, v, retval)
2365 struct lwp *l;
2366 void *v;
2367 register_t *retval;
2368 {
2369 struct sys_chown_args /* {
2370 syscallarg(const char *) path;
2371 syscallarg(uid_t) uid;
2372 syscallarg(gid_t) gid;
2373 } */ *uap = v;
2374 struct proc *p = l->l_proc;
2375 int error;
2376 struct nameidata nd;
2377
2378 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2379 if ((error = namei(&nd)) != 0)
2380 return (error);
2381
2382 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2383
2384 vrele(nd.ni_vp);
2385 return (error);
2386 }
2387
2388 /*
2389 * Set ownership given a file descriptor.
2390 */
2391 /* ARGSUSED */
2392 int
2393 sys_fchown(l, v, retval)
2394 struct lwp *l;
2395 void *v;
2396 register_t *retval;
2397 {
2398 struct sys_fchown_args /* {
2399 syscallarg(int) fd;
2400 syscallarg(uid_t) uid;
2401 syscallarg(gid_t) gid;
2402 } */ *uap = v;
2403 struct proc *p = l->l_proc;
2404 int error;
2405 struct file *fp;
2406
2407 /* getvnode() will use the descriptor for us */
2408 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2409 return (error);
2410
2411 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2412 SCARG(uap, gid), p, 0);
2413 FILE_UNUSE(fp, p);
2414 return (error);
2415 }
2416
2417 /*
2418 * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2419 */
2420 /* ARGSUSED */
2421 int
2422 sys___posix_fchown(l, v, retval)
2423 struct lwp *l;
2424 void *v;
2425 register_t *retval;
2426 {
2427 struct sys_fchown_args /* {
2428 syscallarg(int) fd;
2429 syscallarg(uid_t) uid;
2430 syscallarg(gid_t) gid;
2431 } */ *uap = v;
2432 struct proc *p = l->l_proc;
2433 int error;
2434 struct file *fp;
2435
2436 /* getvnode() will use the descriptor for us */
2437 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2438 return (error);
2439
2440 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2441 SCARG(uap, gid), p, 1);
2442 FILE_UNUSE(fp, p);
2443 return (error);
2444 }
2445
2446 /*
2447 * Set ownership given a path name; this version does not follow links.
2448 */
2449 /* ARGSUSED */
2450 int
2451 sys_lchown(l, v, retval)
2452 struct lwp *l;
2453 void *v;
2454 register_t *retval;
2455 {
2456 struct sys_lchown_args /* {
2457 syscallarg(const char *) path;
2458 syscallarg(uid_t) uid;
2459 syscallarg(gid_t) gid;
2460 } */ *uap = v;
2461 struct proc *p = l->l_proc;
2462 int error;
2463 struct nameidata nd;
2464
2465 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2466 if ((error = namei(&nd)) != 0)
2467 return (error);
2468
2469 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2470
2471 vrele(nd.ni_vp);
2472 return (error);
2473 }
2474
2475 /*
2476 * Set ownership given a path name; this version does not follow links.
2477 * Provides POSIX/XPG semantics.
2478 */
2479 /* ARGSUSED */
2480 int
2481 sys___posix_lchown(l, v, retval)
2482 struct lwp *l;
2483 void *v;
2484 register_t *retval;
2485 {
2486 struct sys_lchown_args /* {
2487 syscallarg(const char *) path;
2488 syscallarg(uid_t) uid;
2489 syscallarg(gid_t) gid;
2490 } */ *uap = v;
2491 struct proc *p = l->l_proc;
2492 int error;
2493 struct nameidata nd;
2494
2495 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2496 if ((error = namei(&nd)) != 0)
2497 return (error);
2498
2499 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2500
2501 vrele(nd.ni_vp);
2502 return (error);
2503 }
2504
2505 /*
2506 * Common routine to set ownership given a vnode.
2507 */
2508 static int
2509 change_owner(vp, uid, gid, p, posix_semantics)
2510 struct vnode *vp;
2511 uid_t uid;
2512 gid_t gid;
2513 struct proc *p;
2514 int posix_semantics;
2515 {
2516 struct vattr vattr;
2517 mode_t newmode;
2518 int error;
2519
2520 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2521 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2522 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2523 goto out;
2524
2525 #define CHANGED(x) ((int)(x) != -1)
2526 newmode = vattr.va_mode;
2527 if (posix_semantics) {
2528 /*
2529 * POSIX/XPG semantics: if the caller is not the super-user,
2530 * clear set-user-id and set-group-id bits. Both POSIX and
2531 * the XPG consider the behaviour for calls by the super-user
2532 * implementation-defined; we leave the set-user-id and set-
2533 * group-id settings intact in that case.
2534 */
2535 if (suser(p->p_ucred, NULL) != 0)
2536 newmode &= ~(S_ISUID | S_ISGID);
2537 } else {
2538 /*
2539 * NetBSD semantics: when changing owner and/or group,
2540 * clear the respective bit(s).
2541 */
2542 if (CHANGED(uid))
2543 newmode &= ~S_ISUID;
2544 if (CHANGED(gid))
2545 newmode &= ~S_ISGID;
2546 }
2547 /* Update va_mode iff altered. */
2548 if (vattr.va_mode == newmode)
2549 newmode = VNOVAL;
2550
2551 VATTR_NULL(&vattr);
2552 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2553 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2554 vattr.va_mode = newmode;
2555 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2556 #undef CHANGED
2557
2558 out:
2559 VOP_UNLOCK(vp, 0);
2560 return (error);
2561 }
2562
2563 /*
2564 * Set the access and modification times given a path name; this
2565 * version follows links.
2566 */
2567 /* ARGSUSED */
2568 int
2569 sys_utimes(l, v, retval)
2570 struct lwp *l;
2571 void *v;
2572 register_t *retval;
2573 {
2574 struct sys_utimes_args /* {
2575 syscallarg(const char *) path;
2576 syscallarg(const struct timeval *) tptr;
2577 } */ *uap = v;
2578 struct proc *p = l->l_proc;
2579 int error;
2580 struct nameidata nd;
2581
2582 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2583 if ((error = namei(&nd)) != 0)
2584 return (error);
2585
2586 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2587
2588 vrele(nd.ni_vp);
2589 return (error);
2590 }
2591
2592 /*
2593 * Set the access and modification times given a file descriptor.
2594 */
2595 /* ARGSUSED */
2596 int
2597 sys_futimes(l, v, retval)
2598 struct lwp *l;
2599 void *v;
2600 register_t *retval;
2601 {
2602 struct sys_futimes_args /* {
2603 syscallarg(int) fd;
2604 syscallarg(const struct timeval *) tptr;
2605 } */ *uap = v;
2606 struct proc *p = l->l_proc;
2607 int error;
2608 struct file *fp;
2609
2610 /* getvnode() will use the descriptor for us */
2611 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2612 return (error);
2613
2614 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2615 FILE_UNUSE(fp, p);
2616 return (error);
2617 }
2618
2619 /*
2620 * Set the access and modification times given a path name; this
2621 * version does not follow links.
2622 */
2623 /* ARGSUSED */
2624 int
2625 sys_lutimes(l, v, retval)
2626 struct lwp *l;
2627 void *v;
2628 register_t *retval;
2629 {
2630 struct sys_lutimes_args /* {
2631 syscallarg(const char *) path;
2632 syscallarg(const struct timeval *) tptr;
2633 } */ *uap = v;
2634 struct proc *p = l->l_proc;
2635 int error;
2636 struct nameidata nd;
2637
2638 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2639 if ((error = namei(&nd)) != 0)
2640 return (error);
2641
2642 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2643
2644 vrele(nd.ni_vp);
2645 return (error);
2646 }
2647
2648 /*
2649 * Common routine to set access and modification times given a vnode.
2650 */
2651 static int
2652 change_utimes(vp, tptr, p)
2653 struct vnode *vp;
2654 const struct timeval *tptr;
2655 struct proc *p;
2656 {
2657 struct timeval tv[2];
2658 struct vattr vattr;
2659 int error;
2660
2661 VATTR_NULL(&vattr);
2662 if (tptr == NULL) {
2663 microtime(&tv[0]);
2664 tv[1] = tv[0];
2665 vattr.va_vaflags |= VA_UTIMES_NULL;
2666 } else {
2667 error = copyin(tptr, tv, sizeof(tv));
2668 if (error)
2669 return (error);
2670 }
2671 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2672 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2673 vattr.va_atime.tv_sec = tv[0].tv_sec;
2674 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2675 vattr.va_mtime.tv_sec = tv[1].tv_sec;
2676 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2677 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2678 VOP_UNLOCK(vp, 0);
2679 return (error);
2680 }
2681
2682 /*
2683 * Truncate a file given its path name.
2684 */
2685 /* ARGSUSED */
2686 int
2687 sys_truncate(l, v, retval)
2688 struct lwp *l;
2689 void *v;
2690 register_t *retval;
2691 {
2692 struct sys_truncate_args /* {
2693 syscallarg(const char *) path;
2694 syscallarg(int) pad;
2695 syscallarg(off_t) length;
2696 } */ *uap = v;
2697 struct proc *p = l->l_proc;
2698 struct vnode *vp;
2699 struct vattr vattr;
2700 int error;
2701 struct nameidata nd;
2702
2703 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2704 if ((error = namei(&nd)) != 0)
2705 return (error);
2706 vp = nd.ni_vp;
2707 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2708 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2709 if (vp->v_type == VDIR)
2710 error = EISDIR;
2711 else if ((error = vn_writechk(vp)) == 0 &&
2712 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2713 VATTR_NULL(&vattr);
2714 vattr.va_size = SCARG(uap, length);
2715 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2716 }
2717 vput(vp);
2718 return (error);
2719 }
2720
2721 /*
2722 * Truncate a file given a file descriptor.
2723 */
2724 /* ARGSUSED */
2725 int
2726 sys_ftruncate(l, v, retval)
2727 struct lwp *l;
2728 void *v;
2729 register_t *retval;
2730 {
2731 struct sys_ftruncate_args /* {
2732 syscallarg(int) fd;
2733 syscallarg(int) pad;
2734 syscallarg(off_t) length;
2735 } */ *uap = v;
2736 struct proc *p = l->l_proc;
2737 struct vattr vattr;
2738 struct vnode *vp;
2739 struct file *fp;
2740 int error;
2741
2742 /* getvnode() will use the descriptor for us */
2743 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2744 return (error);
2745 if ((fp->f_flag & FWRITE) == 0) {
2746 error = EINVAL;
2747 goto out;
2748 }
2749 vp = (struct vnode *)fp->f_data;
2750 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2751 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2752 if (vp->v_type == VDIR)
2753 error = EISDIR;
2754 else if ((error = vn_writechk(vp)) == 0) {
2755 VATTR_NULL(&vattr);
2756 vattr.va_size = SCARG(uap, length);
2757 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2758 }
2759 VOP_UNLOCK(vp, 0);
2760 out:
2761 FILE_UNUSE(fp, p);
2762 return (error);
2763 }
2764
2765 /*
2766 * Sync an open file.
2767 */
2768 /* ARGSUSED */
2769 int
2770 sys_fsync(l, v, retval)
2771 struct lwp *l;
2772 void *v;
2773 register_t *retval;
2774 {
2775 struct sys_fsync_args /* {
2776 syscallarg(int) fd;
2777 } */ *uap = v;
2778 struct proc *p = l->l_proc;
2779 struct vnode *vp;
2780 struct file *fp;
2781 int error;
2782
2783 /* getvnode() will use the descriptor for us */
2784 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2785 return (error);
2786 vp = (struct vnode *)fp->f_data;
2787 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2788 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
2789 if (error == 0 && bioops.io_fsync != NULL &&
2790 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2791 (*bioops.io_fsync)(vp);
2792 VOP_UNLOCK(vp, 0);
2793 FILE_UNUSE(fp, p);
2794 return (error);
2795 }
2796
2797 /*
2798 * Sync the data of an open file.
2799 */
2800 /* ARGSUSED */
2801 int
2802 sys_fdatasync(l, v, retval)
2803 struct lwp *l;
2804 void *v;
2805 register_t *retval;
2806 {
2807 struct sys_fdatasync_args /* {
2808 syscallarg(int) fd;
2809 } */ *uap = v;
2810 struct proc *p = l->l_proc;
2811 struct vnode *vp;
2812 struct file *fp;
2813 int error;
2814
2815 /* getvnode() will use the descriptor for us */
2816 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2817 return (error);
2818 vp = (struct vnode *)fp->f_data;
2819 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2820 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
2821 VOP_UNLOCK(vp, 0);
2822 FILE_UNUSE(fp, p);
2823 return (error);
2824 }
2825
2826 /*
2827 * Rename files, (standard) BSD semantics frontend.
2828 */
2829 /* ARGSUSED */
2830 int
2831 sys_rename(l, v, retval)
2832 struct lwp *l;
2833 void *v;
2834 register_t *retval;
2835 {
2836 struct sys_rename_args /* {
2837 syscallarg(const char *) from;
2838 syscallarg(const char *) to;
2839 } */ *uap = v;
2840 struct proc *p = l->l_proc;
2841
2842 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
2843 }
2844
2845 /*
2846 * Rename files, POSIX semantics frontend.
2847 */
2848 /* ARGSUSED */
2849 int
2850 sys___posix_rename(l, v, retval)
2851 struct lwp *l;
2852 void *v;
2853 register_t *retval;
2854 {
2855 struct sys___posix_rename_args /* {
2856 syscallarg(const char *) from;
2857 syscallarg(const char *) to;
2858 } */ *uap = v;
2859 struct proc *p = l->l_proc;
2860
2861 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
2862 }
2863
2864 /*
2865 * Rename files. Source and destination must either both be directories,
2866 * or both not be directories. If target is a directory, it must be empty.
2867 * If `from' and `to' refer to the same object, the value of the `retain'
2868 * argument is used to determine whether `from' will be
2869 *
2870 * (retain == 0) deleted unless `from' and `to' refer to the same
2871 * object in the file system's name space (BSD).
2872 * (retain == 1) always retained (POSIX).
2873 */
2874 static int
2875 rename_files(from, to, p, retain)
2876 const char *from, *to;
2877 struct proc *p;
2878 int retain;
2879 {
2880 struct vnode *tvp, *fvp, *tdvp;
2881 struct nameidata fromnd, tond;
2882 int error;
2883
2884 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2885 from, p);
2886 if ((error = namei(&fromnd)) != 0)
2887 return (error);
2888 fvp = fromnd.ni_vp;
2889 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
2890 UIO_USERSPACE, to, p);
2891 if ((error = namei(&tond)) != 0) {
2892 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2893 vrele(fromnd.ni_dvp);
2894 vrele(fvp);
2895 goto out1;
2896 }
2897 tdvp = tond.ni_dvp;
2898 tvp = tond.ni_vp;
2899
2900 if (tvp != NULL) {
2901 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2902 error = ENOTDIR;
2903 goto out;
2904 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2905 error = EISDIR;
2906 goto out;
2907 }
2908 }
2909
2910 if (fvp == tdvp)
2911 error = EINVAL;
2912
2913 /*
2914 * Source and destination refer to the same object.
2915 */
2916 if (fvp == tvp) {
2917 if (retain)
2918 error = -1;
2919 else if (fromnd.ni_dvp == tdvp &&
2920 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2921 !memcmp(fromnd.ni_cnd.cn_nameptr,
2922 tond.ni_cnd.cn_nameptr,
2923 fromnd.ni_cnd.cn_namelen))
2924 error = -1;
2925 }
2926
2927 out:
2928 if (!error) {
2929 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2930 if (fromnd.ni_dvp != tdvp)
2931 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2932 if (tvp) {
2933 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2934 }
2935 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2936 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2937 } else {
2938 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2939 if (tdvp == tvp)
2940 vrele(tdvp);
2941 else
2942 vput(tdvp);
2943 if (tvp)
2944 vput(tvp);
2945 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2946 vrele(fromnd.ni_dvp);
2947 vrele(fvp);
2948 }
2949 vrele(tond.ni_startdir);
2950 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2951 out1:
2952 if (fromnd.ni_startdir)
2953 vrele(fromnd.ni_startdir);
2954 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2955 return (error == -1 ? 0 : error);
2956 }
2957
2958 /*
2959 * Make a directory file.
2960 */
2961 /* ARGSUSED */
2962 int
2963 sys_mkdir(l, v, retval)
2964 struct lwp *l;
2965 void *v;
2966 register_t *retval;
2967 {
2968 struct sys_mkdir_args /* {
2969 syscallarg(const char *) path;
2970 syscallarg(int) mode;
2971 } */ *uap = v;
2972 struct proc *p = l->l_proc;
2973 struct vnode *vp;
2974 struct vattr vattr;
2975 int error;
2976 struct nameidata nd;
2977
2978 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
2979 if ((error = namei(&nd)) != 0)
2980 return (error);
2981 vp = nd.ni_vp;
2982 if (vp != NULL) {
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 vrele(vp);
2989 return (EEXIST);
2990 }
2991 VATTR_NULL(&vattr);
2992 vattr.va_type = VDIR;
2993 vattr.va_mode =
2994 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
2995 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2996 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2997 if (!error)
2998 vput(nd.ni_vp);
2999 return (error);
3000 }
3001
3002 /*
3003 * Remove a directory file.
3004 */
3005 /* ARGSUSED */
3006 int
3007 sys_rmdir(l, v, retval)
3008 struct lwp *l;
3009 void *v;
3010 register_t *retval;
3011 {
3012 struct sys_rmdir_args /* {
3013 syscallarg(const char *) path;
3014 } */ *uap = v;
3015 struct proc *p = l->l_proc;
3016 struct vnode *vp;
3017 int error;
3018 struct nameidata nd;
3019
3020 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3021 SCARG(uap, path), p);
3022 if ((error = namei(&nd)) != 0)
3023 return (error);
3024 vp = nd.ni_vp;
3025 if (vp->v_type != VDIR) {
3026 error = ENOTDIR;
3027 goto out;
3028 }
3029 /*
3030 * No rmdir "." please.
3031 */
3032 if (nd.ni_dvp == vp) {
3033 error = EINVAL;
3034 goto out;
3035 }
3036 /*
3037 * The root of a mounted filesystem cannot be deleted.
3038 */
3039 if (vp->v_flag & VROOT)
3040 error = EBUSY;
3041 out:
3042 if (!error) {
3043 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3044 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3045 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3046 } else {
3047 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3048 if (nd.ni_dvp == vp)
3049 vrele(nd.ni_dvp);
3050 else
3051 vput(nd.ni_dvp);
3052 vput(vp);
3053 }
3054 return (error);
3055 }
3056
3057 /*
3058 * Read a block of directory entries in a file system independent format.
3059 */
3060 int
3061 sys_getdents(l, v, retval)
3062 struct lwp *l;
3063 void *v;
3064 register_t *retval;
3065 {
3066 struct sys_getdents_args /* {
3067 syscallarg(int) fd;
3068 syscallarg(char *) buf;
3069 syscallarg(size_t) count;
3070 } */ *uap = v;
3071 struct proc *p = l->l_proc;
3072 struct file *fp;
3073 int error, done;
3074
3075 /* getvnode() will use the descriptor for us */
3076 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3077 return (error);
3078 if ((fp->f_flag & FREAD) == 0) {
3079 error = EBADF;
3080 goto out;
3081 }
3082 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3083 SCARG(uap, count), &done, p, 0, 0);
3084 *retval = done;
3085 out:
3086 FILE_UNUSE(fp, p);
3087 return (error);
3088 }
3089
3090 /*
3091 * Set the mode mask for creation of filesystem nodes.
3092 */
3093 int
3094 sys_umask(l, v, retval)
3095 struct lwp *l;
3096 void *v;
3097 register_t *retval;
3098 {
3099 struct sys_umask_args /* {
3100 syscallarg(mode_t) newmask;
3101 } */ *uap = v;
3102 struct proc *p = l->l_proc;
3103 struct cwdinfo *cwdi;
3104
3105 cwdi = p->p_cwdi;
3106 *retval = cwdi->cwdi_cmask;
3107 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3108 return (0);
3109 }
3110
3111 /*
3112 * Void all references to file by ripping underlying filesystem
3113 * away from vnode.
3114 */
3115 /* ARGSUSED */
3116 int
3117 sys_revoke(l, v, retval)
3118 struct lwp *l;
3119 void *v;
3120 register_t *retval;
3121 {
3122 struct sys_revoke_args /* {
3123 syscallarg(const char *) path;
3124 } */ *uap = v;
3125 struct proc *p = l->l_proc;
3126 struct vnode *vp;
3127 struct vattr vattr;
3128 int error;
3129 struct nameidata nd;
3130
3131 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3132 if ((error = namei(&nd)) != 0)
3133 return (error);
3134 vp = nd.ni_vp;
3135 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3136 goto out;
3137 if (p->p_ucred->cr_uid != vattr.va_uid &&
3138 (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3139 goto out;
3140 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3141 VOP_REVOKE(vp, REVOKEALL);
3142 out:
3143 vrele(vp);
3144 return (error);
3145 }
3146
3147 /*
3148 * Convert a user file descriptor to a kernel file entry.
3149 */
3150 int
3151 getvnode(fdp, fd, fpp)
3152 struct filedesc *fdp;
3153 int fd;
3154 struct file **fpp;
3155 {
3156 struct vnode *vp;
3157 struct file *fp;
3158
3159 if ((fp = fd_getfile(fdp, fd)) == NULL)
3160 return (EBADF);
3161
3162 FILE_USE(fp);
3163
3164 if (fp->f_type != DTYPE_VNODE) {
3165 FILE_UNUSE(fp, NULL);
3166 return (EINVAL);
3167 }
3168
3169 vp = (struct vnode *)fp->f_data;
3170 if (vp->v_type == VBAD) {
3171 FILE_UNUSE(fp, NULL);
3172 return (EBADF);
3173 }
3174
3175 *fpp = fp;
3176 return (0);
3177 }
3178