vfs_syscalls.c revision 1.158.2.3 1 /* $NetBSD: vfs_syscalls.c,v 1.158.2.3 2002/06/06 19:53:07 he Exp $ */
2
3 /*
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. All advertising materials mentioning features or use of this software
21 * must display the following acknowledgement:
22 * This product includes software developed by the University of
23 * California, Berkeley and its contributors.
24 * 4. Neither the name of the University nor the names of its contributors
25 * may be used to endorse or promote products derived from this software
26 * without specific prior written permission.
27 *
28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * SUCH DAMAGE.
39 *
40 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95
41 */
42
43 #include "opt_compat_netbsd.h"
44 #include "opt_compat_43.h"
45
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/namei.h>
49 #include <sys/filedesc.h>
50 #include <sys/kernel.h>
51 #include <sys/file.h>
52 #include <sys/stat.h>
53 #include <sys/vnode.h>
54 #include <sys/mount.h>
55 #include <sys/proc.h>
56 #include <sys/uio.h>
57 #include <sys/malloc.h>
58 #include <sys/dirent.h>
59
60 #include <sys/syscallargs.h>
61
62 #include <vm/vm.h>
63 #include <sys/sysctl.h>
64
65 #include <miscfs/genfs/genfs.h>
66 #include <miscfs/syncfs/syncfs.h>
67
68 #include <uvm/uvm_extern.h>
69
70 static int change_dir __P((struct nameidata *, struct proc *));
71 static int change_mode __P((struct vnode *, int, struct proc *p));
72 static int change_owner __P((struct vnode *, uid_t, gid_t, struct proc *,
73 int));
74 static int change_utimes __P((struct vnode *vp, const struct timeval *,
75 struct proc *p));
76 static int rename_files __P((const char *, const char *, struct proc *, int));
77
78 void checkdirs __P((struct vnode *));
79
80 int dovfsusermount = 0;
81
82 /*
83 * Virtual File System System Calls
84 */
85
86 /*
87 * Mount a file system.
88 */
89
90 /*
91 * This table is used to maintain compatibility with 4.3BSD
92 * and NetBSD 0.9 mount syscalls. Note, the order is important!
93 *
94 * Also note that not all of these had actual numbers in 4.3BSD
95 * or NetBSD 0.9!
96 */
97 const char *mountcompatnames[] = {
98 NULL, /* 0 = MOUNT_NONE */
99 MOUNT_FFS, /* 1 */
100 MOUNT_NFS, /* 2 */
101 MOUNT_MFS, /* 3 */
102 MOUNT_MSDOS, /* 4 */
103 MOUNT_LFS, /* 5 */
104 NULL, /* 6 = MOUNT_LOFS */
105 MOUNT_FDESC, /* 7 */
106 MOUNT_PORTAL, /* 8 */
107 MOUNT_NULL, /* 9 */
108 MOUNT_UMAP, /* 10 */
109 MOUNT_KERNFS, /* 11 */
110 MOUNT_PROCFS, /* 12 */
111 MOUNT_AFS, /* 13 */
112 MOUNT_CD9660, /* 14 = MOUNT_ISOFS */
113 MOUNT_UNION, /* 15 */
114 MOUNT_ADOSFS, /* 16 */
115 MOUNT_EXT2FS, /* 17 */
116 MOUNT_CODA, /* 18 */
117 MOUNT_FILECORE, /* 19 */
118 MOUNT_NTFS, /* 20 */
119 };
120 const int nmountcompatnames = sizeof(mountcompatnames) /
121 sizeof(mountcompatnames[0]);
122
123 /* ARGSUSED */
124 int
125 sys_mount(p, v, retval)
126 struct proc *p;
127 void *v;
128 register_t *retval;
129 {
130 struct sys_mount_args /* {
131 syscallarg(const char *) type;
132 syscallarg(const char *) path;
133 syscallarg(int) flags;
134 syscallarg(void *) data;
135 } */ *uap = v;
136 struct vnode *vp;
137 struct mount *mp;
138 int error, flag = 0;
139 char fstypename[MFSNAMELEN];
140 struct vattr va;
141 struct nameidata nd;
142 struct vfsops *vfs;
143
144 if (dovfsusermount == 0 && (error = suser(p->p_ucred, &p->p_acflag)))
145 return (error);
146 /*
147 * Get vnode to be covered
148 */
149 NDINIT(&nd, LOOKUP, FOLLOW , UIO_USERSPACE,
150 SCARG(uap, path), p);
151 if ((error = namei(&nd)) != 0)
152 return (error);
153 vp = nd.ni_vp;
154 /*
155 * A lookup in VFS_MOUNT might result in an attempt to
156 * lock this vnode again, so make the lock resursive.
157 */
158 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
159 if (SCARG(uap, flags) & MNT_UPDATE) {
160 if ((vp->v_flag & VROOT) == 0) {
161 vput(vp);
162 return (EINVAL);
163 }
164 mp = vp->v_mount;
165 flag = mp->mnt_flag;
166 vfs = mp->mnt_op;
167 /*
168 * We only allow the filesystem to be reloaded if it
169 * is currently mounted read-only.
170 */
171 if ((SCARG(uap, flags) & MNT_RELOAD) &&
172 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
173 vput(vp);
174 return (EOPNOTSUPP); /* Needs translation */
175 }
176 /*
177 * In "highly secure" mode, don't let the caller do anything
178 * but downgrade a filesystem from read-write to read-only.
179 * (see also below; MNT_UPDATE is required.)
180 */
181 if (securelevel >= 2 &&
182 (SCARG(uap, flags) !=
183 (mp->mnt_flag | MNT_RDONLY |
184 MNT_RELOAD | MNT_FORCE | MNT_UPDATE))) {
185 vput(vp);
186 return (EPERM);
187 }
188 mp->mnt_flag |=
189 SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
190 /*
191 * Only root, or the user that did the original mount is
192 * permitted to update it.
193 */
194 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
195 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
196 vput(vp);
197 return (error);
198 }
199 /*
200 * Do not allow NFS export by non-root users. For non-root
201 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
202 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
203 */
204 if (p->p_ucred->cr_uid != 0) {
205 if (SCARG(uap, flags) & MNT_EXPORTED) {
206 vput(vp);
207 return (EPERM);
208 }
209 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
210 if (flag & MNT_NOEXEC)
211 SCARG(uap, flags) |= MNT_NOEXEC;
212 }
213 if (vfs_busy(mp, LK_NOWAIT, 0)) {
214 vput(vp);
215 return (EPERM);
216 }
217 goto update;
218 } else {
219 if (securelevel >= 2) {
220 vput(vp);
221 return (EPERM);
222 }
223 }
224 /*
225 * If the user is not root, ensure that they own the directory
226 * onto which we are attempting to mount.
227 */
228 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
229 (va.va_uid != p->p_ucred->cr_uid &&
230 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
231 vput(vp);
232 return (error);
233 }
234 /*
235 * Do not allow NFS export by non-root users. For non-root users,
236 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
237 * mount point is already MNT_NOEXEC.
238 */
239 if (p->p_ucred->cr_uid != 0) {
240 if (SCARG(uap, flags) & MNT_EXPORTED) {
241 vput(vp);
242 return (EPERM);
243 }
244 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
245 if (vp->v_mount->mnt_flag & MNT_NOEXEC)
246 SCARG(uap, flags) |= MNT_NOEXEC;
247 }
248 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
249 return (error);
250 if (vp->v_type != VDIR) {
251 vput(vp);
252 return (ENOTDIR);
253 }
254 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
255 if (error) {
256 #if defined(COMPAT_09) || defined(COMPAT_43)
257 /*
258 * Historically filesystem types were identified by number.
259 * If we get an integer for the filesystem type instead of a
260 * string, we check to see if it matches one of the historic
261 * filesystem types.
262 */
263 u_long fsindex = (u_long)SCARG(uap, type);
264 if (fsindex >= nmountcompatnames ||
265 mountcompatnames[fsindex] == NULL) {
266 vput(vp);
267 return (ENODEV);
268 }
269 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
270 #else
271 vput(vp);
272 return (error);
273 #endif
274 }
275 #ifdef COMPAT_10
276 /* Accept `ufs' as an alias for `ffs'. */
277 if (!strncmp(fstypename, "ufs", MFSNAMELEN))
278 strncpy(fstypename, "ffs", MFSNAMELEN);
279 #endif
280 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
281 vput(vp);
282 return (ENODEV);
283 }
284 if (vp->v_mountedhere != NULL) {
285 vput(vp);
286 return (EBUSY);
287 }
288
289 /*
290 * Allocate and initialize the file system.
291 */
292 mp = (struct mount *)malloc((u_long)sizeof(struct mount),
293 M_MOUNT, M_WAITOK);
294 memset((char *)mp, 0, (u_long)sizeof(struct mount));
295 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
296 (void)vfs_busy(mp, LK_NOWAIT, 0);
297 mp->mnt_op = vfs;
298 vfs->vfs_refcount++;
299 mp->mnt_vnodecovered = vp;
300 mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
301 mp->mnt_unmounter = NULL;
302 update:
303 /*
304 * Set the mount level flags.
305 */
306 if (SCARG(uap, flags) & MNT_RDONLY)
307 mp->mnt_flag |= MNT_RDONLY;
308 else if (mp->mnt_flag & MNT_RDONLY)
309 mp->mnt_flag |= MNT_WANTRDWR;
310 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
311 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
312 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
313 mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
314 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
315 MNT_NOCOREDUMP | MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM |
316 MNT_SOFTDEP);
317 /*
318 * Mount the filesystem.
319 */
320 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
321 if (mp->mnt_flag & MNT_UPDATE) {
322 if (mp->mnt_flag & MNT_WANTRDWR)
323 mp->mnt_flag &= ~MNT_RDONLY;
324 mp->mnt_flag &=~
325 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR);
326 if (error)
327 mp->mnt_flag = flag;
328 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
329 if (mp->mnt_syncer == NULL)
330 error = vfs_allocate_syncvnode(mp);
331 } else {
332 if (mp->mnt_syncer != NULL)
333 vfs_deallocate_syncvnode(mp);
334 }
335 vfs_unbusy(mp);
336 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((caddr_t)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 for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
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(p, v, retval)
416 struct proc *p;
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 vnode *vp;
425 struct mount *mp;
426 int error;
427 struct nameidata nd;
428
429 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
430 SCARG(uap, path), p);
431 if ((error = namei(&nd)) != 0)
432 return (error);
433 vp = nd.ni_vp;
434 mp = vp->v_mount;
435
436 /*
437 * Only root, or the user that did the original mount is
438 * permitted to unmount this filesystem.
439 */
440 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
441 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
442 vput(vp);
443 return (error);
444 }
445
446 /*
447 * Don't allow unmounting the root file system.
448 */
449 if (mp->mnt_flag & MNT_ROOTFS) {
450 vput(vp);
451 return (EINVAL);
452 }
453
454 /*
455 * Must be the root of the filesystem
456 */
457 if ((vp->v_flag & VROOT) == 0) {
458 vput(vp);
459 return (EINVAL);
460 }
461 vput(vp);
462
463 if (vfs_busy(mp, 0, 0))
464 return (EBUSY);
465
466 return (dounmount(mp, SCARG(uap, flags), p));
467 }
468
469 /*
470 * Do the actual file system unmount. File system is assumed to have been
471 * marked busy by the caller.
472 */
473 int
474 dounmount(mp, flags, p)
475 struct mount *mp;
476 int flags;
477 struct proc *p;
478 {
479 struct vnode *coveredvp;
480 int error;
481 int async;
482 int used_syncer;
483
484 simple_lock(&mountlist_slock);
485 vfs_unbusy(mp);
486 used_syncer = (mp->mnt_syncer != NULL);
487
488 /*
489 * XXX Freeze syncer. This should really be done on a mountpoint
490 * basis, but especially the softdep code possibly called from
491 * the syncer doesn't exactly work on a per-mountpoint basis,
492 * so the softdep code would become a maze of vfs_busy calls.
493 */
494 if (used_syncer)
495 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
496
497 mp->mnt_flag |= MNT_UNMOUNT;
498 mp->mnt_unmounter = p;
499 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
500 if (mp->mnt_flag & MNT_EXPUBLIC)
501 vfs_setpublicfs(NULL, NULL, NULL);
502 async = mp->mnt_flag & MNT_ASYNC;
503 mp->mnt_flag &= ~MNT_ASYNC;
504 cache_purgevfs(mp); /* remove cache entries for this file sys */
505 if (mp->mnt_syncer != NULL)
506 vfs_deallocate_syncvnode(mp);
507 if (((mp->mnt_flag & MNT_RDONLY) ||
508 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
509 (flags & MNT_FORCE))
510 error = VFS_UNMOUNT(mp, flags, p);
511 simple_lock(&mountlist_slock);
512 if (error) {
513 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
514 (void) vfs_allocate_syncvnode(mp);
515 mp->mnt_flag &= ~MNT_UNMOUNT;
516 mp->mnt_unmounter = NULL;
517 mp->mnt_flag |= async;
518 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
519 &mountlist_slock);
520 if (used_syncer)
521 lockmgr(&syncer_lock, LK_RELEASE, NULL);
522 while (mp->mnt_wcnt > 0) {
523 wakeup((caddr_t)mp);
524 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0);
525 }
526 return (error);
527 }
528 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
529 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
530 coveredvp->v_mountedhere = NULL;
531 vrele(coveredvp);
532 }
533 mp->mnt_op->vfs_refcount--;
534 if (mp->mnt_vnodelist.lh_first != NULL)
535 panic("unmount: dangling vnode");
536 mp->mnt_flag |= MNT_GONE;
537 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
538 if (used_syncer)
539 lockmgr(&syncer_lock, LK_RELEASE, NULL);
540 while(mp->mnt_wcnt > 0) {
541 wakeup((caddr_t)mp);
542 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0);
543 }
544 free((caddr_t)mp, M_MOUNT);
545 return (0);
546 }
547
548 /*
549 * Sync each mounted filesystem.
550 */
551 #ifdef DEBUG
552 int syncprt = 0;
553 struct ctldebug debug0 = { "syncprt", &syncprt };
554 #endif
555
556 /* ARGSUSED */
557 int
558 sys_sync(p, v, retval)
559 struct proc *p;
560 void *v;
561 register_t *retval;
562 {
563 struct mount *mp, *nmp;
564 int asyncflag;
565
566 simple_lock(&mountlist_slock);
567 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
568 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
569 nmp = mp->mnt_list.cqe_prev;
570 continue;
571 }
572 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
573 asyncflag = mp->mnt_flag & MNT_ASYNC;
574 mp->mnt_flag &= ~MNT_ASYNC;
575 uvm_vnp_sync(mp);
576 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
577 if (asyncflag)
578 mp->mnt_flag |= MNT_ASYNC;
579 }
580 simple_lock(&mountlist_slock);
581 nmp = mp->mnt_list.cqe_prev;
582 vfs_unbusy(mp);
583
584 }
585 simple_unlock(&mountlist_slock);
586 #ifdef DEBUG
587 if (syncprt)
588 vfs_bufstats();
589 #endif /* DEBUG */
590 return (0);
591 }
592
593 /*
594 * Change filesystem quotas.
595 */
596 /* ARGSUSED */
597 int
598 sys_quotactl(p, v, retval)
599 struct proc *p;
600 void *v;
601 register_t *retval;
602 {
603 struct sys_quotactl_args /* {
604 syscallarg(const char *) path;
605 syscallarg(int) cmd;
606 syscallarg(int) uid;
607 syscallarg(caddr_t) arg;
608 } */ *uap = v;
609 struct mount *mp;
610 int error;
611 struct nameidata nd;
612
613 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
614 if ((error = namei(&nd)) != 0)
615 return (error);
616 mp = nd.ni_vp->v_mount;
617 vrele(nd.ni_vp);
618 return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
619 SCARG(uap, arg), p));
620 }
621
622 /*
623 * Get filesystem statistics.
624 */
625 /* ARGSUSED */
626 int
627 sys_statfs(p, v, retval)
628 struct proc *p;
629 void *v;
630 register_t *retval;
631 {
632 struct sys_statfs_args /* {
633 syscallarg(const char *) path;
634 syscallarg(struct statfs *) buf;
635 } */ *uap = v;
636 struct mount *mp;
637 struct statfs *sp;
638 int error;
639 struct nameidata nd;
640
641 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
642 if ((error = namei(&nd)) != 0)
643 return (error);
644 mp = nd.ni_vp->v_mount;
645 sp = &mp->mnt_stat;
646 vrele(nd.ni_vp);
647 if ((error = VFS_STATFS(mp, sp, p)) != 0)
648 return (error);
649 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
650 sp->f_oflags = sp->f_flags & 0xffff;
651 return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
652 }
653
654 /*
655 * Get filesystem statistics.
656 */
657 /* ARGSUSED */
658 int
659 sys_fstatfs(p, v, retval)
660 struct proc *p;
661 void *v;
662 register_t *retval;
663 {
664 struct sys_fstatfs_args /* {
665 syscallarg(int) fd;
666 syscallarg(struct statfs *) buf;
667 } */ *uap = v;
668 struct file *fp;
669 struct mount *mp;
670 struct statfs *sp;
671 int error;
672
673 /* getvnode() will use the descriptor for us */
674 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
675 return (error);
676 mp = ((struct vnode *)fp->f_data)->v_mount;
677 sp = &mp->mnt_stat;
678 if ((error = VFS_STATFS(mp, sp, p)) != 0)
679 goto out;
680 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
681 sp->f_oflags = sp->f_flags & 0xffff;
682 error = copyout(sp, SCARG(uap, buf), sizeof(*sp));
683 out:
684 FILE_UNUSE(fp, p);
685 return (error);
686 }
687
688 /*
689 * Get statistics on all filesystems.
690 */
691 int
692 sys_getfsstat(p, v, retval)
693 struct proc *p;
694 void *v;
695 register_t *retval;
696 {
697 struct sys_getfsstat_args /* {
698 syscallarg(struct statfs *) buf;
699 syscallarg(long) bufsize;
700 syscallarg(int) flags;
701 } */ *uap = v;
702 struct mount *mp, *nmp;
703 struct statfs *sp;
704 caddr_t sfsp;
705 long count, maxcount, error;
706
707 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
708 sfsp = (caddr_t)SCARG(uap, buf);
709 simple_lock(&mountlist_slock);
710 count = 0;
711 for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
712 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
713 nmp = mp->mnt_list.cqe_next;
714 continue;
715 }
716 if (sfsp && count < maxcount) {
717 sp = &mp->mnt_stat;
718 /*
719 * If MNT_NOWAIT or MNT_LAZY is specified, do not
720 * refresh the fsstat cache. MNT_WAIT or MNT_LAXY
721 * overrides MNT_NOWAIT.
722 */
723 if (SCARG(uap, flags) != MNT_NOWAIT &&
724 SCARG(uap, flags) != MNT_LAZY &&
725 (SCARG(uap, flags) == MNT_WAIT ||
726 SCARG(uap, flags) == 0) &&
727 (error = VFS_STATFS(mp, sp, p)) != 0) {
728 simple_lock(&mountlist_slock);
729 nmp = mp->mnt_list.cqe_next;
730 vfs_unbusy(mp);
731 continue;
732 }
733 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
734 sp->f_oflags = sp->f_flags & 0xffff;
735 error = copyout(sp, sfsp, sizeof(*sp));
736 if (error) {
737 vfs_unbusy(mp);
738 return (error);
739 }
740 sfsp += sizeof(*sp);
741 }
742 count++;
743 simple_lock(&mountlist_slock);
744 nmp = mp->mnt_list.cqe_next;
745 vfs_unbusy(mp);
746 }
747 simple_unlock(&mountlist_slock);
748 if (sfsp && count > maxcount)
749 *retval = maxcount;
750 else
751 *retval = count;
752 return (0);
753 }
754
755 /*
756 * Change current working directory to a given file descriptor.
757 */
758 /* ARGSUSED */
759 int
760 sys_fchdir(p, v, retval)
761 struct proc *p;
762 void *v;
763 register_t *retval;
764 {
765 struct sys_fchdir_args /* {
766 syscallarg(int) fd;
767 } */ *uap = v;
768 struct filedesc *fdp = p->p_fd;
769 struct cwdinfo *cwdi = p->p_cwdi;
770 struct vnode *vp, *tdp;
771 struct mount *mp;
772 struct file *fp;
773 int error;
774
775 /* getvnode() will use the descriptor for us */
776 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
777 return (error);
778 vp = (struct vnode *)fp->f_data;
779
780 VREF(vp);
781 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
782 if (vp->v_type != VDIR)
783 error = ENOTDIR;
784 else
785 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
786 while (!error && (mp = vp->v_mountedhere) != NULL) {
787 if (vfs_busy(mp, 0, 0))
788 continue;
789 error = VFS_ROOT(mp, &tdp);
790 vfs_unbusy(mp);
791 if (error)
792 break;
793 vput(vp);
794 vp = tdp;
795 }
796 if (error) {
797 vput(vp);
798 goto out;
799 }
800 VOP_UNLOCK(vp, 0);
801
802 /*
803 * Disallow changing to a directory not under the process's
804 * current root directory (if there is one).
805 */
806 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
807 vrele(vp);
808 error = EPERM; /* operation not permitted */
809 goto out;
810 }
811
812 vrele(cwdi->cwdi_cdir);
813 cwdi->cwdi_cdir = vp;
814 out:
815 FILE_UNUSE(fp, p);
816 return (error);
817 }
818
819 /*
820 * Change this process's notion of the root directory to a given file descriptor.
821 */
822
823 int
824 sys_fchroot(p, v, retval)
825 struct proc *p;
826 void *v;
827 register_t *retval;
828 {
829 struct sys_fchroot_args *uap = v;
830 struct filedesc *fdp = p->p_fd;
831 struct cwdinfo *cwdi = p->p_cwdi;
832 struct vnode *vp;
833 struct file *fp;
834 int error;
835
836 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
837 return error;
838 /* getvnode() will use the descriptor for us */
839 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
840 return error;
841 vp = (struct vnode *) fp->f_data;
842 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
843 if (vp->v_type != VDIR)
844 error = ENOTDIR;
845 else
846 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
847 VOP_UNLOCK(vp, 0);
848 if (error)
849 goto out;
850 VREF(vp);
851
852 /*
853 * Prevent escaping from chroot by putting the root under
854 * the working directory. Silently chdir to / if we aren't
855 * already there.
856 */
857 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
858 /*
859 * XXX would be more failsafe to change directory to a
860 * deadfs node here instead
861 */
862 vrele(cwdi->cwdi_cdir);
863 VREF(vp);
864 cwdi->cwdi_cdir = vp;
865 }
866
867 if (cwdi->cwdi_rdir != NULL)
868 vrele(cwdi->cwdi_rdir);
869 cwdi->cwdi_rdir = vp;
870 out:
871 FILE_UNUSE(fp, p);
872 return (error);
873 }
874
875
876
877 /*
878 * Change current working directory (``.'').
879 */
880 /* ARGSUSED */
881 int
882 sys_chdir(p, v, retval)
883 struct proc *p;
884 void *v;
885 register_t *retval;
886 {
887 struct sys_chdir_args /* {
888 syscallarg(const char *) path;
889 } */ *uap = v;
890 struct cwdinfo *cwdi = p->p_cwdi;
891 int error;
892 struct nameidata nd;
893
894 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
895 SCARG(uap, path), p);
896 if ((error = change_dir(&nd, p)) != 0)
897 return (error);
898 vrele(cwdi->cwdi_cdir);
899 cwdi->cwdi_cdir = nd.ni_vp;
900 return (0);
901 }
902
903 /*
904 * Change notion of root (``/'') directory.
905 */
906 /* ARGSUSED */
907 int
908 sys_chroot(p, v, retval)
909 struct proc *p;
910 void *v;
911 register_t *retval;
912 {
913 struct sys_chroot_args /* {
914 syscallarg(const char *) path;
915 } */ *uap = v;
916 struct cwdinfo *cwdi = p->p_cwdi;
917 struct vnode *vp;
918 int error;
919 struct nameidata nd;
920
921 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
922 return (error);
923 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
924 SCARG(uap, path), p);
925 if ((error = change_dir(&nd, p)) != 0)
926 return (error);
927 if (cwdi->cwdi_rdir != NULL)
928 vrele(cwdi->cwdi_rdir);
929 vp = nd.ni_vp;
930 cwdi->cwdi_rdir = vp;
931
932 /*
933 * Prevent escaping from chroot by putting the root under
934 * the working directory. Silently chdir to / if we aren't
935 * already there.
936 */
937 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
938 /*
939 * XXX would be more failsafe to change directory to a
940 * deadfs node here instead
941 */
942 vrele(cwdi->cwdi_cdir);
943 VREF(vp);
944 cwdi->cwdi_cdir = vp;
945 }
946
947 return (0);
948 }
949
950 /*
951 * Common routine for chroot and chdir.
952 */
953 static int
954 change_dir(ndp, p)
955 struct nameidata *ndp;
956 struct proc *p;
957 {
958 struct vnode *vp;
959 int error;
960
961 if ((error = namei(ndp)) != 0)
962 return (error);
963 vp = ndp->ni_vp;
964 if (vp->v_type != VDIR)
965 error = ENOTDIR;
966 else
967 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
968
969 if (error)
970 vput(vp);
971 else
972 VOP_UNLOCK(vp, 0);
973 return (error);
974 }
975
976 /*
977 * Check permissions, allocate an open file structure,
978 * and call the device open routine if any.
979 */
980 int
981 sys_open(p, v, retval)
982 struct proc *p;
983 void *v;
984 register_t *retval;
985 {
986 struct sys_open_args /* {
987 syscallarg(const char *) path;
988 syscallarg(int) flags;
989 syscallarg(int) mode;
990 } */ *uap = v;
991 struct cwdinfo *cwdi = p->p_cwdi;
992 struct filedesc *fdp = p->p_fd;
993 struct file *fp;
994 struct vnode *vp;
995 int flags, cmode;
996 int type, indx, error;
997 struct flock lf;
998 struct nameidata nd;
999
1000 flags = FFLAGS(SCARG(uap, flags));
1001 if ((flags & (FREAD | FWRITE)) == 0)
1002 return (EINVAL);
1003 /* falloc() will use the file descriptor for us */
1004 if ((error = falloc(p, &fp, &indx)) != 0)
1005 return (error);
1006 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1007 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1008 p->p_dupfd = -indx - 1; /* XXX check for fdopen */
1009 if ((error = vn_open(&nd, flags, cmode)) != 0) {
1010 FILE_UNUSE(fp, p);
1011 ffree(fp);
1012 if ((error == ENODEV || error == ENXIO) &&
1013 p->p_dupfd >= 0 && /* XXX from fdopen */
1014 (error =
1015 dupfdopen(p, indx, p->p_dupfd, flags, error)) == 0) {
1016 *retval = indx;
1017 return (0);
1018 }
1019 if (error == ERESTART)
1020 error = EINTR;
1021 fdremove(fdp, indx);
1022 return (error);
1023 }
1024 p->p_dupfd = 0;
1025 vp = nd.ni_vp;
1026 fp->f_flag = flags & FMASK;
1027 fp->f_type = DTYPE_VNODE;
1028 fp->f_ops = &vnops;
1029 fp->f_data = (caddr_t)vp;
1030 if (flags & (O_EXLOCK | O_SHLOCK)) {
1031 lf.l_whence = SEEK_SET;
1032 lf.l_start = 0;
1033 lf.l_len = 0;
1034 if (flags & O_EXLOCK)
1035 lf.l_type = F_WRLCK;
1036 else
1037 lf.l_type = F_RDLCK;
1038 type = F_FLOCK;
1039 if ((flags & FNONBLOCK) == 0)
1040 type |= F_WAIT;
1041 VOP_UNLOCK(vp, 0);
1042 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1043 if (error) {
1044 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1045 FILE_UNUSE(fp, p);
1046 ffree(fp);
1047 fdremove(fdp, indx);
1048 return (error);
1049 }
1050 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1051 fp->f_flag |= FHASLOCK;
1052 }
1053 VOP_UNLOCK(vp, 0);
1054 *retval = indx;
1055 FILE_UNUSE(fp, p);
1056 return (0);
1057 }
1058
1059 /*
1060 * Get file handle system call
1061 */
1062 int
1063 sys_getfh(p, v, retval)
1064 struct proc *p;
1065 void *v;
1066 register_t *retval;
1067 {
1068 struct sys_getfh_args /* {
1069 syscallarg(char *) fname;
1070 syscallarg(fhandle_t *) fhp;
1071 } */ *uap = v;
1072 struct vnode *vp;
1073 fhandle_t fh;
1074 int error;
1075 struct nameidata nd;
1076
1077 /*
1078 * Must be super user
1079 */
1080 error = suser(p->p_ucred, &p->p_acflag);
1081 if (error)
1082 return (error);
1083 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1084 SCARG(uap, fname), p);
1085 error = namei(&nd);
1086 if (error)
1087 return (error);
1088 vp = nd.ni_vp;
1089 memset((caddr_t)&fh, 0, sizeof(fh));
1090 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1091 error = VFS_VPTOFH(vp, &fh.fh_fid);
1092 vput(vp);
1093 if (error)
1094 return (error);
1095 error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1096 return (error);
1097 }
1098
1099 /*
1100 * Open a file given a file handle.
1101 *
1102 * Check permissions, allocate an open file structure,
1103 * and call the device open routine if any.
1104 */
1105 int
1106 sys_fhopen(p, v, retval)
1107 struct proc *p;
1108 void *v;
1109 register_t *retval;
1110 {
1111 struct sys_fhopen_args /* {
1112 syscallarg(const fhandle_t *) fhp;
1113 syscallarg(int) flags;
1114 } */ *uap = v;
1115 struct filedesc *fdp = p->p_fd;
1116 struct file *fp;
1117 struct vnode *vp = NULL;
1118 struct mount *mp;
1119 struct ucred *cred = p->p_ucred;
1120 int flags;
1121 struct file *nfp;
1122 int type, indx, error=0;
1123 struct flock lf;
1124 struct vattr va;
1125 fhandle_t fh;
1126
1127 /*
1128 * Must be super user
1129 */
1130 if ((error = suser(p->p_ucred, &p->p_acflag)))
1131 return (error);
1132
1133 flags = FFLAGS(SCARG(uap, flags));
1134 if ((flags & (FREAD | FWRITE)) == 0)
1135 return (EINVAL);
1136 if ((flags & O_CREAT))
1137 return (EINVAL);
1138 /* falloc() will use the file descriptor for us */
1139 if ((error = falloc(p, &nfp, &indx)) != 0)
1140 return (error);
1141 fp = nfp;
1142 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1143 goto bad;
1144
1145 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1146 error = ESTALE;
1147 goto bad;
1148 }
1149
1150 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1151 vp = NULL; /* most likely unnecessary sanity for bad: */
1152 goto bad;
1153 }
1154
1155 /* Now do an effective vn_open */
1156
1157 if (vp->v_type == VSOCK) {
1158 error = EOPNOTSUPP;
1159 goto bad;
1160 }
1161 if (flags & FREAD) {
1162 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1163 goto bad;
1164 }
1165 if (flags & (FWRITE | O_TRUNC)) {
1166 if (vp->v_type == VDIR) {
1167 error = EISDIR;
1168 goto bad;
1169 }
1170 if ((error = vn_writechk(vp)) != 0 ||
1171 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1172 goto bad;
1173 }
1174 if (flags & O_TRUNC) {
1175 VOP_UNLOCK(vp, 0); /* XXX */
1176 VOP_LEASE(vp, p, cred, LEASE_WRITE);
1177 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
1178 VATTR_NULL(&va);
1179 va.va_size = 0;
1180 if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1181 goto bad;
1182 }
1183 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1184 goto bad;
1185 if (flags & FWRITE)
1186 vp->v_writecount++;
1187
1188 /* done with modified vn_open, now finish what sys_open does. */
1189
1190 fp->f_flag = flags & FMASK;
1191 fp->f_type = DTYPE_VNODE;
1192 fp->f_ops = &vnops;
1193 fp->f_data = (caddr_t)vp;
1194 if (flags & (O_EXLOCK | O_SHLOCK)) {
1195 lf.l_whence = SEEK_SET;
1196 lf.l_start = 0;
1197 lf.l_len = 0;
1198 if (flags & O_EXLOCK)
1199 lf.l_type = F_WRLCK;
1200 else
1201 lf.l_type = F_RDLCK;
1202 type = F_FLOCK;
1203 if ((flags & FNONBLOCK) == 0)
1204 type |= F_WAIT;
1205 VOP_UNLOCK(vp, 0);
1206 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1207 if (error) {
1208 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1209 FILE_UNUSE(fp, p);
1210 ffree(fp);
1211 fdremove(fdp, indx);
1212 return (error);
1213 }
1214 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1215 fp->f_flag |= FHASLOCK;
1216 }
1217 VOP_UNLOCK(vp, 0);
1218 *retval = indx;
1219 FILE_UNUSE(fp, p);
1220 return (0);
1221
1222 bad:
1223 FILE_UNUSE(fp, p);
1224 ffree(fp);
1225 fdremove(fdp, indx);
1226 if (vp != NULL)
1227 vput(vp);
1228 return (error);
1229 }
1230
1231 /* ARGSUSED */
1232 int
1233 sys_fhstat(p, v, retval)
1234 struct proc *p;
1235 void *v;
1236 register_t *retval;
1237 {
1238 struct sys_fhstat_args /* {
1239 syscallarg(const fhandle_t *) fhp;
1240 syscallarg(struct stat *) sb;
1241 } */ *uap = v;
1242 struct stat sb;
1243 int error;
1244 fhandle_t fh;
1245 struct mount *mp;
1246 struct vnode *vp;
1247
1248 /*
1249 * Must be super user
1250 */
1251 if ((error = suser(p->p_ucred, &p->p_acflag)))
1252 return (error);
1253
1254 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1255 return (error);
1256
1257 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1258 return (ESTALE);
1259 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1260 return (error);
1261 error = vn_stat(vp, &sb, p);
1262 vput(vp);
1263 if (error)
1264 return (error);
1265 error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1266 return (error);
1267 }
1268
1269 /* ARGSUSED */
1270 int
1271 sys_fhstatfs(p, v, retval)
1272 struct proc *p;
1273 void *v;
1274 register_t *retval;
1275 {
1276 struct sys_fhstatfs_args /*
1277 syscallarg(const fhandle_t *) fhp;
1278 syscallarg(struct statfs *) buf;
1279 } */ *uap = v;
1280 struct statfs sp;
1281 fhandle_t fh;
1282 struct mount *mp;
1283 struct vnode *vp;
1284 int error;
1285
1286 /*
1287 * Must be super user
1288 */
1289 if ((error = suser(p->p_ucred, &p->p_acflag)))
1290 return (error);
1291
1292 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1293 return (error);
1294
1295 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1296 return (ESTALE);
1297 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1298 return (error);
1299 mp = vp->v_mount;
1300 vput(vp);
1301 if ((error = VFS_STATFS(mp, &sp, p)) != 0)
1302 return (error);
1303 sp.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1304 sp.f_oflags = sp.f_flags & 0xffff;
1305 return (copyout(&sp, SCARG(uap, buf), sizeof(sp)));
1306 }
1307
1308 /*
1309 * Create a special file.
1310 */
1311 /* ARGSUSED */
1312 int
1313 sys_mknod(p, v, retval)
1314 struct proc *p;
1315 void *v;
1316 register_t *retval;
1317 {
1318 struct sys_mknod_args /* {
1319 syscallarg(const char *) path;
1320 syscallarg(int) mode;
1321 syscallarg(int) dev;
1322 } */ *uap = v;
1323 struct vnode *vp;
1324 struct vattr vattr;
1325 int error;
1326 int whiteout = 0;
1327 struct nameidata nd;
1328
1329 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1330 return (error);
1331 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1332 if ((error = namei(&nd)) != 0)
1333 return (error);
1334 vp = nd.ni_vp;
1335 if (vp != NULL)
1336 error = EEXIST;
1337 else {
1338 VATTR_NULL(&vattr);
1339 vattr.va_mode =
1340 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1341 vattr.va_rdev = SCARG(uap, dev);
1342 whiteout = 0;
1343
1344 switch (SCARG(uap, mode) & S_IFMT) {
1345 case S_IFMT: /* used by badsect to flag bad sectors */
1346 vattr.va_type = VBAD;
1347 break;
1348 case S_IFCHR:
1349 vattr.va_type = VCHR;
1350 break;
1351 case S_IFBLK:
1352 vattr.va_type = VBLK;
1353 break;
1354 case S_IFWHT:
1355 whiteout = 1;
1356 break;
1357 default:
1358 error = EINVAL;
1359 break;
1360 }
1361 }
1362 if (!error) {
1363 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1364 if (whiteout) {
1365 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1366 if (error)
1367 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1368 vput(nd.ni_dvp);
1369 } else {
1370 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1371 &nd.ni_cnd, &vattr);
1372 }
1373 } else {
1374 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1375 if (nd.ni_dvp == vp)
1376 vrele(nd.ni_dvp);
1377 else
1378 vput(nd.ni_dvp);
1379 if (vp)
1380 vrele(vp);
1381 }
1382 return (error);
1383 }
1384
1385 /*
1386 * Create a named pipe.
1387 */
1388 /* ARGSUSED */
1389 int
1390 sys_mkfifo(p, v, retval)
1391 struct proc *p;
1392 void *v;
1393 register_t *retval;
1394 {
1395 struct sys_mkfifo_args /* {
1396 syscallarg(const char *) path;
1397 syscallarg(int) mode;
1398 } */ *uap = v;
1399 struct vattr vattr;
1400 int error;
1401 struct nameidata nd;
1402
1403 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1404 if ((error = namei(&nd)) != 0)
1405 return (error);
1406 if (nd.ni_vp != NULL) {
1407 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1408 if (nd.ni_dvp == nd.ni_vp)
1409 vrele(nd.ni_dvp);
1410 else
1411 vput(nd.ni_dvp);
1412 vrele(nd.ni_vp);
1413 return (EEXIST);
1414 }
1415 VATTR_NULL(&vattr);
1416 vattr.va_type = VFIFO;
1417 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1418 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1419 return (VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr));
1420 }
1421
1422 /*
1423 * Make a hard file link.
1424 */
1425 /* ARGSUSED */
1426 int
1427 sys_link(p, v, retval)
1428 struct proc *p;
1429 void *v;
1430 register_t *retval;
1431 {
1432 struct sys_link_args /* {
1433 syscallarg(const char *) path;
1434 syscallarg(const char *) link;
1435 } */ *uap = v;
1436 struct vnode *vp;
1437 struct nameidata nd;
1438 int error;
1439
1440 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1441 if ((error = namei(&nd)) != 0)
1442 return (error);
1443 vp = nd.ni_vp;
1444 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1445 if ((error = namei(&nd)) != 0)
1446 goto out;
1447 if (nd.ni_vp) {
1448 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1449 if (nd.ni_dvp == nd.ni_vp)
1450 vrele(nd.ni_dvp);
1451 else
1452 vput(nd.ni_dvp);
1453 vrele(nd.ni_vp);
1454 error = EEXIST;
1455 goto out;
1456 }
1457 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1458 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1459 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1460 out:
1461 vrele(vp);
1462 return (error);
1463 }
1464
1465 /*
1466 * Make a symbolic link.
1467 */
1468 /* ARGSUSED */
1469 int
1470 sys_symlink(p, v, retval)
1471 struct proc *p;
1472 void *v;
1473 register_t *retval;
1474 {
1475 struct sys_symlink_args /* {
1476 syscallarg(const char *) path;
1477 syscallarg(const char *) link;
1478 } */ *uap = v;
1479 struct vattr vattr;
1480 char *path;
1481 int error;
1482 struct nameidata nd;
1483
1484 MALLOC(path, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
1485 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1486 if (error)
1487 goto out;
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 VATTR_NULL(&vattr);
1502 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1503 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1504 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1505 out:
1506 FREE(path, M_NAMEI);
1507 return (error);
1508 }
1509
1510 /*
1511 * Delete a whiteout from the filesystem.
1512 */
1513 /* ARGSUSED */
1514 int
1515 sys_undelete(p, v, retval)
1516 struct proc *p;
1517 void *v;
1518 register_t *retval;
1519 {
1520 struct sys_undelete_args /* {
1521 syscallarg(const char *) path;
1522 } */ *uap = v;
1523 int error;
1524 struct nameidata nd;
1525
1526 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1527 SCARG(uap, path), p);
1528 error = namei(&nd);
1529 if (error)
1530 return (error);
1531
1532 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1533 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1534 if (nd.ni_dvp == nd.ni_vp)
1535 vrele(nd.ni_dvp);
1536 else
1537 vput(nd.ni_dvp);
1538 if (nd.ni_vp)
1539 vrele(nd.ni_vp);
1540 return (EEXIST);
1541 }
1542
1543 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1544 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1545 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1546 vput(nd.ni_dvp);
1547 return (error);
1548 }
1549
1550 /*
1551 * Delete a name from the filesystem.
1552 */
1553 /* ARGSUSED */
1554 int
1555 sys_unlink(p, v, retval)
1556 struct proc *p;
1557 void *v;
1558 register_t *retval;
1559 {
1560 struct sys_unlink_args /* {
1561 syscallarg(const char *) path;
1562 } */ *uap = v;
1563 struct vnode *vp;
1564 int error;
1565 struct nameidata nd;
1566
1567 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1568 SCARG(uap, path), p);
1569 if ((error = namei(&nd)) != 0)
1570 return (error);
1571 vp = nd.ni_vp;
1572
1573 /*
1574 * The root of a mounted filesystem cannot be deleted.
1575 */
1576 if (vp->v_flag & VROOT) {
1577 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1578 if (nd.ni_dvp == vp)
1579 vrele(nd.ni_dvp);
1580 else
1581 vput(nd.ni_dvp);
1582 vput(vp);
1583 error = EBUSY;
1584 goto out;
1585 }
1586
1587 (void)uvm_vnp_uncache(vp);
1588
1589 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1590 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1591 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1592 out:
1593 return (error);
1594 }
1595
1596 /*
1597 * Reposition read/write file offset.
1598 */
1599 int
1600 sys_lseek(p, v, retval)
1601 struct proc *p;
1602 void *v;
1603 register_t *retval;
1604 {
1605 struct sys_lseek_args /* {
1606 syscallarg(int) fd;
1607 syscallarg(int) pad;
1608 syscallarg(off_t) offset;
1609 syscallarg(int) whence;
1610 } */ *uap = v;
1611 struct ucred *cred = p->p_ucred;
1612 struct filedesc *fdp = p->p_fd;
1613 struct file *fp;
1614 struct vnode *vp;
1615 struct vattr vattr;
1616 off_t newoff;
1617 int error;
1618
1619 if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles ||
1620 (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL ||
1621 (fp->f_iflags & FIF_WANTCLOSE) != 0)
1622 return (EBADF);
1623
1624 FILE_USE(fp);
1625
1626 vp = (struct vnode *)fp->f_data;
1627 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1628 error = ESPIPE;
1629 goto out;
1630 }
1631
1632 switch (SCARG(uap, whence)) {
1633 case SEEK_CUR:
1634 newoff = fp->f_offset + SCARG(uap, offset);
1635 break;
1636 case SEEK_END:
1637 error = VOP_GETATTR(vp, &vattr, cred, p);
1638 if (error)
1639 goto out;
1640 newoff = SCARG(uap, offset) + vattr.va_size;
1641 break;
1642 case SEEK_SET:
1643 newoff = SCARG(uap, offset);
1644 break;
1645 default:
1646 error = EINVAL;
1647 goto out;
1648 }
1649 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1650 goto out;
1651
1652 *(off_t *)retval = fp->f_offset = newoff;
1653 out:
1654 FILE_UNUSE(fp, p);
1655 return (error);
1656 }
1657
1658 /*
1659 * Positional read system call.
1660 */
1661 int
1662 sys_pread(p, v, retval)
1663 struct proc *p;
1664 void *v;
1665 register_t *retval;
1666 {
1667 struct sys_pread_args /* {
1668 syscallarg(int) fd;
1669 syscallarg(void *) buf;
1670 syscallarg(size_t) nbyte;
1671 syscallarg(off_t) offset;
1672 } */ *uap = v;
1673 struct filedesc *fdp = p->p_fd;
1674 struct file *fp;
1675 struct vnode *vp;
1676 off_t offset;
1677 int error, fd = SCARG(uap, fd);
1678
1679 if ((u_int)fd >= fdp->fd_nfiles ||
1680 (fp = fdp->fd_ofiles[fd]) == NULL ||
1681 (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1682 (fp->f_flag & FREAD) == 0)
1683 return (EBADF);
1684
1685 FILE_USE(fp);
1686
1687 vp = (struct vnode *)fp->f_data;
1688 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1689 error = ESPIPE;
1690 goto out;
1691 }
1692
1693 offset = SCARG(uap, offset);
1694
1695 /*
1696 * XXX This works because no file systems actually
1697 * XXX take any action on the seek operation.
1698 */
1699 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1700 goto out;
1701
1702 /* dofileread() will unuse the descriptor for us */
1703 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1704 &offset, 0, retval));
1705
1706 out:
1707 FILE_UNUSE(fp, p);
1708 return (error);
1709 }
1710
1711 /*
1712 * Positional scatter read system call.
1713 */
1714 int
1715 sys_preadv(p, v, retval)
1716 struct proc *p;
1717 void *v;
1718 register_t *retval;
1719 {
1720 struct sys_preadv_args /* {
1721 syscallarg(int) fd;
1722 syscallarg(const struct iovec *) iovp;
1723 syscallarg(int) iovcnt;
1724 syscallarg(off_t) offset;
1725 } */ *uap = v;
1726 struct filedesc *fdp = p->p_fd;
1727 struct file *fp;
1728 struct vnode *vp;
1729 off_t offset;
1730 int error, fd = SCARG(uap, fd);
1731
1732 if ((u_int)fd >= fdp->fd_nfiles ||
1733 (fp = fdp->fd_ofiles[fd]) == NULL ||
1734 (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1735 (fp->f_flag & FREAD) == 0)
1736 return (EBADF);
1737
1738 FILE_USE(fp);
1739
1740 vp = (struct vnode *)fp->f_data;
1741 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1742 error = ESPIPE;
1743 goto out;
1744 }
1745
1746 offset = SCARG(uap, offset);
1747
1748 /*
1749 * XXX This works because no file systems actually
1750 * XXX take any action on the seek operation.
1751 */
1752 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1753 goto out;
1754
1755 /* dofilereadv() will unuse the descriptor for us */
1756 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1757 &offset, 0, retval));
1758
1759 out:
1760 FILE_UNUSE(fp, p);
1761 return (error);
1762 }
1763
1764 /*
1765 * Positional write system call.
1766 */
1767 int
1768 sys_pwrite(p, v, retval)
1769 struct proc *p;
1770 void *v;
1771 register_t *retval;
1772 {
1773 struct sys_pwrite_args /* {
1774 syscallarg(int) fd;
1775 syscallarg(const void *) buf;
1776 syscallarg(size_t) nbyte;
1777 syscallarg(off_t) offset;
1778 } */ *uap = v;
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 ((u_int)fd >= fdp->fd_nfiles ||
1786 (fp = fdp->fd_ofiles[fd]) == NULL ||
1787 (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1788 (fp->f_flag & FWRITE) == 0)
1789 return (EBADF);
1790
1791 FILE_USE(fp);
1792
1793 vp = (struct vnode *)fp->f_data;
1794 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1795 error = ESPIPE;
1796 goto out;
1797 }
1798
1799 offset = SCARG(uap, offset);
1800
1801 /*
1802 * XXX This works because no file systems actually
1803 * XXX take any action on the seek operation.
1804 */
1805 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1806 goto out;
1807
1808 /* dofilewrite() will unuse the descriptor for us */
1809 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1810 &offset, 0, retval));
1811
1812 out:
1813 FILE_UNUSE(fp, p);
1814 return (error);
1815 }
1816
1817 /*
1818 * Positional gather write system call.
1819 */
1820 int
1821 sys_pwritev(p, v, retval)
1822 struct proc *p;
1823 void *v;
1824 register_t *retval;
1825 {
1826 struct sys_pwritev_args /* {
1827 syscallarg(int) fd;
1828 syscallarg(const struct iovec *) iovp;
1829 syscallarg(int) iovcnt;
1830 syscallarg(off_t) offset;
1831 } */ *uap = v;
1832 struct filedesc *fdp = p->p_fd;
1833 struct file *fp;
1834 struct vnode *vp;
1835 off_t offset;
1836 int error, fd = SCARG(uap, fd);
1837
1838 if ((u_int)fd >= fdp->fd_nfiles ||
1839 (fp = fdp->fd_ofiles[fd]) == NULL ||
1840 (fp->f_iflags & FIF_WANTCLOSE) != 0 ||
1841 (fp->f_flag & FWRITE) == 0)
1842 return (EBADF);
1843
1844 FILE_USE(fp);
1845
1846 vp = (struct vnode *)fp->f_data;
1847 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1848 error = ESPIPE;
1849 goto out;
1850 }
1851
1852 offset = SCARG(uap, offset);
1853
1854 /*
1855 * XXX This works because no file systems actually
1856 * XXX take any action on the seek operation.
1857 */
1858 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1859 goto out;
1860
1861 /* dofilewritev() will unuse the descriptor for us */
1862 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1863 &offset, 0, retval));
1864
1865 out:
1866 FILE_UNUSE(fp, p);
1867 return (error);
1868 }
1869
1870 /*
1871 * Check access permissions.
1872 */
1873 int
1874 sys_access(p, v, retval)
1875 struct proc *p;
1876 void *v;
1877 register_t *retval;
1878 {
1879 struct sys_access_args /* {
1880 syscallarg(const char *) path;
1881 syscallarg(int) flags;
1882 } */ *uap = v;
1883 struct ucred *cred = p->p_ucred;
1884 struct vnode *vp;
1885 int error, flags, t_gid, t_uid;
1886 struct nameidata nd;
1887
1888 t_uid = cred->cr_uid;
1889 t_gid = cred->cr_gid;
1890 cred->cr_uid = p->p_cred->p_ruid;
1891 cred->cr_gid = p->p_cred->p_rgid;
1892 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1893 SCARG(uap, path), p);
1894 if ((error = namei(&nd)) != 0)
1895 goto out1;
1896 vp = nd.ni_vp;
1897
1898 /* Flags == 0 means only check for existence. */
1899 if (SCARG(uap, flags)) {
1900 flags = 0;
1901 if (SCARG(uap, flags) & R_OK)
1902 flags |= VREAD;
1903 if (SCARG(uap, flags) & W_OK)
1904 flags |= VWRITE;
1905 if (SCARG(uap, flags) & X_OK)
1906 flags |= VEXEC;
1907
1908 error = VOP_ACCESS(vp, flags, cred, p);
1909 if (!error && (flags & VWRITE))
1910 error = vn_writechk(vp);
1911 }
1912 vput(vp);
1913 out1:
1914 cred->cr_uid = t_uid;
1915 cred->cr_gid = t_gid;
1916 return (error);
1917 }
1918
1919 /*
1920 * Get file status; this version follows links.
1921 */
1922 /* ARGSUSED */
1923 int
1924 sys___stat13(p, v, retval)
1925 struct proc *p;
1926 void *v;
1927 register_t *retval;
1928 {
1929 struct sys___stat13_args /* {
1930 syscallarg(const char *) path;
1931 syscallarg(struct stat *) ub;
1932 } */ *uap = v;
1933 struct stat sb;
1934 int error;
1935 struct nameidata nd;
1936
1937 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1938 SCARG(uap, path), p);
1939 if ((error = namei(&nd)) != 0)
1940 return (error);
1941 error = vn_stat(nd.ni_vp, &sb, p);
1942 vput(nd.ni_vp);
1943 if (error)
1944 return (error);
1945 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1946 return (error);
1947 }
1948
1949 /*
1950 * Get file status; this version does not follow links.
1951 */
1952 /* ARGSUSED */
1953 int
1954 sys___lstat13(p, v, retval)
1955 struct proc *p;
1956 void *v;
1957 register_t *retval;
1958 {
1959 struct sys___lstat13_args /* {
1960 syscallarg(const char *) path;
1961 syscallarg(struct stat *) ub;
1962 } */ *uap = v;
1963 struct stat sb;
1964 int error;
1965 struct nameidata nd;
1966
1967 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1968 SCARG(uap, path), p);
1969 if ((error = namei(&nd)) != 0)
1970 return (error);
1971 error = vn_stat(nd.ni_vp, &sb, p);
1972 vput(nd.ni_vp);
1973 if (error)
1974 return (error);
1975 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1976 return (error);
1977 }
1978
1979 /*
1980 * Get configurable pathname variables.
1981 */
1982 /* ARGSUSED */
1983 int
1984 sys_pathconf(p, v, retval)
1985 struct proc *p;
1986 void *v;
1987 register_t *retval;
1988 {
1989 struct sys_pathconf_args /* {
1990 syscallarg(const char *) path;
1991 syscallarg(int) name;
1992 } */ *uap = v;
1993 int error;
1994 struct nameidata nd;
1995
1996 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1997 SCARG(uap, path), p);
1998 if ((error = namei(&nd)) != 0)
1999 return (error);
2000 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2001 vput(nd.ni_vp);
2002 return (error);
2003 }
2004
2005 /*
2006 * Return target name of a symbolic link.
2007 */
2008 /* ARGSUSED */
2009 int
2010 sys_readlink(p, v, retval)
2011 struct proc *p;
2012 void *v;
2013 register_t *retval;
2014 {
2015 struct sys_readlink_args /* {
2016 syscallarg(const char *) path;
2017 syscallarg(char *) buf;
2018 syscallarg(size_t) count;
2019 } */ *uap = v;
2020 struct vnode *vp;
2021 struct iovec aiov;
2022 struct uio auio;
2023 int error;
2024 struct nameidata nd;
2025
2026 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2027 SCARG(uap, path), p);
2028 if ((error = namei(&nd)) != 0)
2029 return (error);
2030 vp = nd.ni_vp;
2031 if (vp->v_type != VLNK)
2032 error = EINVAL;
2033 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2034 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2035 aiov.iov_base = SCARG(uap, buf);
2036 aiov.iov_len = SCARG(uap, count);
2037 auio.uio_iov = &aiov;
2038 auio.uio_iovcnt = 1;
2039 auio.uio_offset = 0;
2040 auio.uio_rw = UIO_READ;
2041 auio.uio_segflg = UIO_USERSPACE;
2042 auio.uio_procp = p;
2043 auio.uio_resid = SCARG(uap, count);
2044 error = VOP_READLINK(vp, &auio, p->p_ucred);
2045 }
2046 vput(vp);
2047 *retval = SCARG(uap, count) - auio.uio_resid;
2048 return (error);
2049 }
2050
2051 /*
2052 * Change flags of a file given a path name.
2053 */
2054 /* ARGSUSED */
2055 int
2056 sys_chflags(p, v, retval)
2057 struct proc *p;
2058 void *v;
2059 register_t *retval;
2060 {
2061 struct sys_chflags_args /* {
2062 syscallarg(const char *) path;
2063 syscallarg(u_long) flags;
2064 } */ *uap = v;
2065 struct vnode *vp;
2066 struct vattr vattr;
2067 int error;
2068 struct nameidata nd;
2069
2070 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2071 if ((error = namei(&nd)) != 0)
2072 return (error);
2073 vp = nd.ni_vp;
2074 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2075 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2076 /* Non-superusers cannot change the flags on devices, even if they
2077 own them. */
2078 if (suser(p->p_ucred, &p->p_acflag)) {
2079 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2080 goto out;
2081 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2082 error = EINVAL;
2083 goto out;
2084 }
2085 }
2086 VATTR_NULL(&vattr);
2087 vattr.va_flags = SCARG(uap, flags);
2088 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2089 out:
2090 vput(vp);
2091 return (error);
2092 }
2093
2094 /*
2095 * Change flags of a file given a file descriptor.
2096 */
2097 /* ARGSUSED */
2098 int
2099 sys_fchflags(p, v, retval)
2100 struct proc *p;
2101 void *v;
2102 register_t *retval;
2103 {
2104 struct sys_fchflags_args /* {
2105 syscallarg(int) fd;
2106 syscallarg(u_long) flags;
2107 } */ *uap = v;
2108 struct vattr vattr;
2109 struct vnode *vp;
2110 struct file *fp;
2111 int error;
2112
2113 /* getvnode() will use the descriptor for us */
2114 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2115 return (error);
2116 vp = (struct vnode *)fp->f_data;
2117 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2118 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2119 /* Non-superusers cannot change the flags on devices, even if they
2120 own them. */
2121 if (suser(p->p_ucred, &p->p_acflag)) {
2122 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p))
2123 != 0)
2124 goto out;
2125 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2126 error = EINVAL;
2127 goto out;
2128 }
2129 }
2130 VATTR_NULL(&vattr);
2131 vattr.va_flags = SCARG(uap, flags);
2132 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2133 out:
2134 VOP_UNLOCK(vp, 0);
2135 FILE_UNUSE(fp, p);
2136 return (error);
2137 }
2138
2139 /*
2140 * Change flags of a file given a file descriptor; this version does
2141 * not follow links.
2142 */
2143 int
2144 sys_lchflags(p, v, retval)
2145 struct proc *p;
2146 void *v;
2147 register_t *retval;
2148 {
2149 register struct sys_chflags_args /* {
2150 syscallarg(const char *) path;
2151 syscallarg(u_long) flags;
2152 } */ *uap = v;
2153 register struct vnode *vp;
2154 struct vattr vattr;
2155 int error;
2156 struct nameidata nd;
2157
2158 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2159 if ((error = namei(&nd)) != 0)
2160 return (error);
2161 vp = nd.ni_vp;
2162 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2163 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2164 /* Non-superusers cannot change the flags on devices, even if they
2165 own them. */
2166 if (suser(p->p_ucred, &p->p_acflag)) {
2167 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2168 goto out;
2169 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2170 error = EINVAL;
2171 goto out;
2172 }
2173 }
2174 VATTR_NULL(&vattr);
2175 vattr.va_flags = SCARG(uap, flags);
2176 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2177 out:
2178 vput(vp);
2179 return (error);
2180 }
2181
2182 /*
2183 * Change mode of a file given path name; this version follows links.
2184 */
2185 /* ARGSUSED */
2186 int
2187 sys_chmod(p, v, retval)
2188 struct proc *p;
2189 void *v;
2190 register_t *retval;
2191 {
2192 struct sys_chmod_args /* {
2193 syscallarg(const char *) path;
2194 syscallarg(int) mode;
2195 } */ *uap = v;
2196 int error;
2197 struct nameidata nd;
2198
2199 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2200 if ((error = namei(&nd)) != 0)
2201 return (error);
2202
2203 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2204
2205 vrele(nd.ni_vp);
2206 return (error);
2207 }
2208
2209 /*
2210 * Change mode of a file given a file descriptor.
2211 */
2212 /* ARGSUSED */
2213 int
2214 sys_fchmod(p, v, retval)
2215 struct proc *p;
2216 void *v;
2217 register_t *retval;
2218 {
2219 struct sys_fchmod_args /* {
2220 syscallarg(int) fd;
2221 syscallarg(int) mode;
2222 } */ *uap = v;
2223 struct file *fp;
2224 int error;
2225
2226 /* getvnode() will use the descriptor for us */
2227 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2228 return (error);
2229
2230 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2231 FILE_UNUSE(fp, p);
2232 return (error);
2233 }
2234
2235 /*
2236 * Change mode of a file given path name; this version does not follow links.
2237 */
2238 /* ARGSUSED */
2239 int
2240 sys_lchmod(p, v, retval)
2241 struct proc *p;
2242 void *v;
2243 register_t *retval;
2244 {
2245 struct sys_lchmod_args /* {
2246 syscallarg(const char *) path;
2247 syscallarg(int) mode;
2248 } */ *uap = v;
2249 int error;
2250 struct nameidata nd;
2251
2252 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2253 if ((error = namei(&nd)) != 0)
2254 return (error);
2255
2256 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2257
2258 vrele(nd.ni_vp);
2259 return (error);
2260 }
2261
2262 /*
2263 * Common routine to set mode given a vnode.
2264 */
2265 static int
2266 change_mode(vp, mode, p)
2267 struct vnode *vp;
2268 int mode;
2269 struct proc *p;
2270 {
2271 struct vattr vattr;
2272 int error;
2273
2274 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2275 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2276 VATTR_NULL(&vattr);
2277 vattr.va_mode = mode & ALLPERMS;
2278 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2279 VOP_UNLOCK(vp, 0);
2280 return (error);
2281 }
2282
2283 /*
2284 * Set ownership given a path name; this version follows links.
2285 */
2286 /* ARGSUSED */
2287 int
2288 sys_chown(p, v, retval)
2289 struct proc *p;
2290 void *v;
2291 register_t *retval;
2292 {
2293 struct sys_chown_args /* {
2294 syscallarg(const char *) path;
2295 syscallarg(uid_t) uid;
2296 syscallarg(gid_t) gid;
2297 } */ *uap = v;
2298 int error;
2299 struct nameidata nd;
2300
2301 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2302 if ((error = namei(&nd)) != 0)
2303 return (error);
2304
2305 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2306
2307 vrele(nd.ni_vp);
2308 return (error);
2309 }
2310
2311 /*
2312 * Set ownership given a path name; this version follows links.
2313 * Provides POSIX semantics.
2314 */
2315 /* ARGSUSED */
2316 int
2317 sys___posix_chown(p, v, retval)
2318 struct proc *p;
2319 void *v;
2320 register_t *retval;
2321 {
2322 struct sys_chown_args /* {
2323 syscallarg(const char *) path;
2324 syscallarg(uid_t) uid;
2325 syscallarg(gid_t) gid;
2326 } */ *uap = v;
2327 int error;
2328 struct nameidata nd;
2329
2330 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2331 if ((error = namei(&nd)) != 0)
2332 return (error);
2333
2334 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2335
2336 vrele(nd.ni_vp);
2337 return (error);
2338 }
2339
2340 /*
2341 * Set ownership given a file descriptor.
2342 */
2343 /* ARGSUSED */
2344 int
2345 sys_fchown(p, v, retval)
2346 struct proc *p;
2347 void *v;
2348 register_t *retval;
2349 {
2350 struct sys_fchown_args /* {
2351 syscallarg(int) fd;
2352 syscallarg(uid_t) uid;
2353 syscallarg(gid_t) gid;
2354 } */ *uap = v;
2355 int error;
2356 struct file *fp;
2357
2358 /* getvnode() will use the descriptor for us */
2359 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2360 return (error);
2361
2362 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2363 SCARG(uap, gid), p, 0);
2364 FILE_UNUSE(fp, p);
2365 return (error);
2366 }
2367
2368 /*
2369 * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2370 */
2371 /* ARGSUSED */
2372 int
2373 sys___posix_fchown(p, v, retval)
2374 struct proc *p;
2375 void *v;
2376 register_t *retval;
2377 {
2378 struct sys_fchown_args /* {
2379 syscallarg(int) fd;
2380 syscallarg(uid_t) uid;
2381 syscallarg(gid_t) gid;
2382 } */ *uap = v;
2383 int error;
2384 struct file *fp;
2385
2386 /* getvnode() will use the descriptor for us */
2387 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2388 return (error);
2389
2390 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2391 SCARG(uap, gid), p, 1);
2392 FILE_UNUSE(fp, p);
2393 return (error);
2394 }
2395
2396 /*
2397 * Set ownership given a path name; this version does not follow links.
2398 */
2399 /* ARGSUSED */
2400 int
2401 sys_lchown(p, v, retval)
2402 struct proc *p;
2403 void *v;
2404 register_t *retval;
2405 {
2406 struct sys_lchown_args /* {
2407 syscallarg(const char *) path;
2408 syscallarg(uid_t) uid;
2409 syscallarg(gid_t) gid;
2410 } */ *uap = v;
2411 int error;
2412 struct nameidata nd;
2413
2414 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2415 if ((error = namei(&nd)) != 0)
2416 return (error);
2417
2418 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2419
2420 vrele(nd.ni_vp);
2421 return (error);
2422 }
2423
2424 /*
2425 * Set ownership given a path name; this version does not follow links.
2426 * Provides POSIX/XPG semantics.
2427 */
2428 /* ARGSUSED */
2429 int
2430 sys___posix_lchown(p, v, retval)
2431 struct proc *p;
2432 void *v;
2433 register_t *retval;
2434 {
2435 struct sys_lchown_args /* {
2436 syscallarg(const char *) path;
2437 syscallarg(uid_t) uid;
2438 syscallarg(gid_t) gid;
2439 } */ *uap = v;
2440 int error;
2441 struct nameidata nd;
2442
2443 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2444 if ((error = namei(&nd)) != 0)
2445 return (error);
2446
2447 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2448
2449 vrele(nd.ni_vp);
2450 return (error);
2451 }
2452
2453 /*
2454 * Common routine to set ownership given a vnode.
2455 */
2456 static int
2457 change_owner(vp, uid, gid, p, posix_semantics)
2458 struct vnode *vp;
2459 uid_t uid;
2460 gid_t gid;
2461 struct proc *p;
2462 int posix_semantics;
2463 {
2464 struct vattr vattr;
2465 mode_t newmode;
2466 int error;
2467
2468 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2469 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2470 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2471 goto out;
2472
2473 #define CHANGED(x) ((x) != -1)
2474 newmode = vattr.va_mode;
2475 if (posix_semantics) {
2476 /*
2477 * POSIX/XPG semantics: if the caller is not the super-user,
2478 * clear set-user-id and set-group-id bits. Both POSIX and
2479 * the XPG consider the behaviour for calls by the super-user
2480 * implementation-defined; we leave the set-user-id and set-
2481 * group-id settings intact in that case.
2482 */
2483 if (suser(p->p_ucred, NULL) != 0)
2484 newmode &= ~(S_ISUID | S_ISGID);
2485 } else {
2486 /*
2487 * NetBSD semantics: when changing owner and/or group,
2488 * clear the respective bit(s).
2489 */
2490 if (CHANGED(uid))
2491 newmode &= ~S_ISUID;
2492 if (CHANGED(gid))
2493 newmode &= ~S_ISGID;
2494 }
2495 /* Update va_mode iff altered. */
2496 if (vattr.va_mode == newmode)
2497 newmode = VNOVAL;
2498
2499 VATTR_NULL(&vattr);
2500 vattr.va_uid = CHANGED(uid) ? uid : VNOVAL;
2501 vattr.va_gid = CHANGED(gid) ? gid : VNOVAL;
2502 vattr.va_mode = newmode;
2503 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2504 #undef CHANGED
2505
2506 out:
2507 VOP_UNLOCK(vp, 0);
2508 return (error);
2509 }
2510
2511 /*
2512 * Set the access and modification times given a path name; this
2513 * version follows links.
2514 */
2515 /* ARGSUSED */
2516 int
2517 sys_utimes(p, v, retval)
2518 struct proc *p;
2519 void *v;
2520 register_t *retval;
2521 {
2522 struct sys_utimes_args /* {
2523 syscallarg(const char *) path;
2524 syscallarg(const struct timeval *) tptr;
2525 } */ *uap = v;
2526 int error;
2527 struct nameidata nd;
2528
2529 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2530 if ((error = namei(&nd)) != 0)
2531 return (error);
2532
2533 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2534
2535 vrele(nd.ni_vp);
2536 return (error);
2537 }
2538
2539 /*
2540 * Set the access and modification times given a file descriptor.
2541 */
2542 /* ARGSUSED */
2543 int
2544 sys_futimes(p, v, retval)
2545 struct proc *p;
2546 void *v;
2547 register_t *retval;
2548 {
2549 struct sys_futimes_args /* {
2550 syscallarg(int) fd;
2551 syscallarg(const struct timeval *) tptr;
2552 } */ *uap = v;
2553 int error;
2554 struct file *fp;
2555
2556 /* getvnode() will use the descriptor for us */
2557 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2558 return (error);
2559
2560 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2561 FILE_UNUSE(fp, p);
2562 return (error);
2563 }
2564
2565 /*
2566 * Set the access and modification times given a path name; this
2567 * version does not follow links.
2568 */
2569 /* ARGSUSED */
2570 int
2571 sys_lutimes(p, v, retval)
2572 struct proc *p;
2573 void *v;
2574 register_t *retval;
2575 {
2576 struct sys_lutimes_args /* {
2577 syscallarg(const char *) path;
2578 syscallarg(const struct timeval *) tptr;
2579 } */ *uap = v;
2580 int error;
2581 struct nameidata nd;
2582
2583 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2584 if ((error = namei(&nd)) != 0)
2585 return (error);
2586
2587 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2588
2589 vrele(nd.ni_vp);
2590 return (error);
2591 }
2592
2593 /*
2594 * Common routine to set access and modification times given a vnode.
2595 */
2596 static int
2597 change_utimes(vp, tptr, p)
2598 struct vnode *vp;
2599 const struct timeval *tptr;
2600 struct proc *p;
2601 {
2602 struct timeval tv[2];
2603 struct vattr vattr;
2604 int error;
2605
2606 VATTR_NULL(&vattr);
2607 if (tptr == NULL) {
2608 microtime(&tv[0]);
2609 tv[1] = tv[0];
2610 vattr.va_vaflags |= VA_UTIMES_NULL;
2611 } else {
2612 error = copyin(tptr, tv, sizeof(tv));
2613 if (error)
2614 return (error);
2615 }
2616 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2617 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2618 vattr.va_atime.tv_sec = tv[0].tv_sec;
2619 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2620 vattr.va_mtime.tv_sec = tv[1].tv_sec;
2621 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2622 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2623 VOP_UNLOCK(vp, 0);
2624 return (error);
2625 }
2626
2627 /*
2628 * Truncate a file given its path name.
2629 */
2630 /* ARGSUSED */
2631 int
2632 sys_truncate(p, v, retval)
2633 struct proc *p;
2634 void *v;
2635 register_t *retval;
2636 {
2637 struct sys_truncate_args /* {
2638 syscallarg(const char *) path;
2639 syscallarg(int) pad;
2640 syscallarg(off_t) length;
2641 } */ *uap = v;
2642 struct vnode *vp;
2643 struct vattr vattr;
2644 int error;
2645 struct nameidata nd;
2646
2647 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2648 if ((error = namei(&nd)) != 0)
2649 return (error);
2650 vp = nd.ni_vp;
2651 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2652 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2653 if (vp->v_type == VDIR)
2654 error = EISDIR;
2655 else if ((error = vn_writechk(vp)) == 0 &&
2656 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2657 VATTR_NULL(&vattr);
2658 vattr.va_size = SCARG(uap, length);
2659 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2660 }
2661 vput(vp);
2662 return (error);
2663 }
2664
2665 /*
2666 * Truncate a file given a file descriptor.
2667 */
2668 /* ARGSUSED */
2669 int
2670 sys_ftruncate(p, v, retval)
2671 struct proc *p;
2672 void *v;
2673 register_t *retval;
2674 {
2675 struct sys_ftruncate_args /* {
2676 syscallarg(int) fd;
2677 syscallarg(int) pad;
2678 syscallarg(off_t) length;
2679 } */ *uap = v;
2680 struct vattr vattr;
2681 struct vnode *vp;
2682 struct file *fp;
2683 int error;
2684
2685 /* getvnode() will use the descriptor for us */
2686 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2687 return (error);
2688 if ((fp->f_flag & FWRITE) == 0) {
2689 error = EINVAL;
2690 goto out;
2691 }
2692 vp = (struct vnode *)fp->f_data;
2693 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2694 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2695 if (vp->v_type == VDIR)
2696 error = EISDIR;
2697 else if ((error = vn_writechk(vp)) == 0) {
2698 VATTR_NULL(&vattr);
2699 vattr.va_size = SCARG(uap, length);
2700 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2701 }
2702 VOP_UNLOCK(vp, 0);
2703 out:
2704 FILE_UNUSE(fp, p);
2705 return (error);
2706 }
2707
2708 /*
2709 * Sync an open file.
2710 */
2711 /* ARGSUSED */
2712 int
2713 sys_fsync(p, v, retval)
2714 struct proc *p;
2715 void *v;
2716 register_t *retval;
2717 {
2718 struct sys_fsync_args /* {
2719 syscallarg(int) fd;
2720 } */ *uap = v;
2721 struct vnode *vp;
2722 struct file *fp;
2723 int error;
2724
2725 /* getvnode() will use the descriptor for us */
2726 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2727 return (error);
2728 vp = (struct vnode *)fp->f_data;
2729 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2730 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
2731 if (error == 0 && bioops.io_fsync != NULL &&
2732 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2733 (*bioops.io_fsync)(vp);
2734 VOP_UNLOCK(vp, 0);
2735 FILE_UNUSE(fp, p);
2736 return (error);
2737 }
2738
2739 /*
2740 * Sync the data of an open file.
2741 */
2742 /* ARGSUSED */
2743 int
2744 sys_fdatasync(p, v, retval)
2745 struct proc *p;
2746 void *v;
2747 register_t *retval;
2748 {
2749 struct sys_fdatasync_args /* {
2750 syscallarg(int) fd;
2751 } */ *uap = v;
2752 struct vnode *vp;
2753 struct file *fp;
2754 int error;
2755
2756 /* getvnode() will use the descriptor for us */
2757 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2758 return (error);
2759 vp = (struct vnode *)fp->f_data;
2760 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2761 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
2762 VOP_UNLOCK(vp, 0);
2763 FILE_UNUSE(fp, p);
2764 return (error);
2765 }
2766
2767 /*
2768 * Rename files, (standard) BSD semantics frontend.
2769 */
2770 /* ARGSUSED */
2771 int
2772 sys_rename(p, v, retval)
2773 struct proc *p;
2774 void *v;
2775 register_t *retval;
2776 {
2777 struct sys_rename_args /* {
2778 syscallarg(const char *) from;
2779 syscallarg(const char *) to;
2780 } */ *uap = v;
2781
2782 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
2783 }
2784
2785 /*
2786 * Rename files, POSIX semantics frontend.
2787 */
2788 /* ARGSUSED */
2789 int
2790 sys___posix_rename(p, v, retval)
2791 struct proc *p;
2792 void *v;
2793 register_t *retval;
2794 {
2795 struct sys___posix_rename_args /* {
2796 syscallarg(const char *) from;
2797 syscallarg(const char *) to;
2798 } */ *uap = v;
2799
2800 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
2801 }
2802
2803 /*
2804 * Rename files. Source and destination must either both be directories,
2805 * or both not be directories. If target is a directory, it must be empty.
2806 * If `from' and `to' refer to the same object, the value of the `retain'
2807 * argument is used to determine whether `from' will be
2808 *
2809 * (retain == 0) deleted unless `from' and `to' refer to the same
2810 * object in the file system's name space (BSD).
2811 * (retain == 1) always retained (POSIX).
2812 */
2813 static int
2814 rename_files(from, to, p, retain)
2815 const char *from, *to;
2816 struct proc *p;
2817 int retain;
2818 {
2819 struct vnode *tvp, *fvp, *tdvp;
2820 struct nameidata fromnd, tond;
2821 int error;
2822
2823 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2824 from, p);
2825 if ((error = namei(&fromnd)) != 0)
2826 return (error);
2827 fvp = fromnd.ni_vp;
2828 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
2829 UIO_USERSPACE, to, p);
2830 if ((error = namei(&tond)) != 0) {
2831 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2832 vrele(fromnd.ni_dvp);
2833 vrele(fvp);
2834 goto out1;
2835 }
2836 tdvp = tond.ni_dvp;
2837 tvp = tond.ni_vp;
2838
2839 if (tvp != NULL) {
2840 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2841 error = ENOTDIR;
2842 goto out;
2843 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2844 error = EISDIR;
2845 goto out;
2846 }
2847 }
2848
2849 if (fvp == tdvp)
2850 error = EINVAL;
2851
2852 /*
2853 * Source and destination refer to the same object.
2854 */
2855 if (fvp == tvp) {
2856 if (retain)
2857 error = -1;
2858 else if (fromnd.ni_dvp == tdvp &&
2859 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2860 !memcmp(fromnd.ni_cnd.cn_nameptr,
2861 tond.ni_cnd.cn_nameptr,
2862 fromnd.ni_cnd.cn_namelen))
2863 error = -1;
2864 }
2865
2866 out:
2867 if (!error) {
2868 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2869 if (fromnd.ni_dvp != tdvp)
2870 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2871 if (tvp) {
2872 (void)uvm_vnp_uncache(tvp);
2873 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2874 }
2875 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2876 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2877 } else {
2878 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2879 if (tdvp == tvp)
2880 vrele(tdvp);
2881 else
2882 vput(tdvp);
2883 if (tvp)
2884 vput(tvp);
2885 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2886 vrele(fromnd.ni_dvp);
2887 vrele(fvp);
2888 }
2889 vrele(tond.ni_startdir);
2890 FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
2891 out1:
2892 if (fromnd.ni_startdir)
2893 vrele(fromnd.ni_startdir);
2894 FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
2895 return (error == -1 ? 0 : error);
2896 }
2897
2898 /*
2899 * Make a directory file.
2900 */
2901 /* ARGSUSED */
2902 int
2903 sys_mkdir(p, v, retval)
2904 struct proc *p;
2905 void *v;
2906 register_t *retval;
2907 {
2908 struct sys_mkdir_args /* {
2909 syscallarg(const char *) path;
2910 syscallarg(int) mode;
2911 } */ *uap = v;
2912 struct vnode *vp;
2913 struct vattr vattr;
2914 int error;
2915 struct nameidata nd;
2916
2917 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
2918 if ((error = namei(&nd)) != 0)
2919 return (error);
2920 vp = nd.ni_vp;
2921 if (vp != NULL) {
2922 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2923 if (nd.ni_dvp == vp)
2924 vrele(nd.ni_dvp);
2925 else
2926 vput(nd.ni_dvp);
2927 vrele(vp);
2928 return (EEXIST);
2929 }
2930 VATTR_NULL(&vattr);
2931 vattr.va_type = VDIR;
2932 vattr.va_mode =
2933 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
2934 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2935 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2936 if (!error)
2937 vput(nd.ni_vp);
2938 return (error);
2939 }
2940
2941 /*
2942 * Remove a directory file.
2943 */
2944 /* ARGSUSED */
2945 int
2946 sys_rmdir(p, v, retval)
2947 struct proc *p;
2948 void *v;
2949 register_t *retval;
2950 {
2951 struct sys_rmdir_args /* {
2952 syscallarg(const char *) path;
2953 } */ *uap = v;
2954 struct vnode *vp;
2955 int error;
2956 struct nameidata nd;
2957
2958 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
2959 SCARG(uap, path), p);
2960 if ((error = namei(&nd)) != 0)
2961 return (error);
2962 vp = nd.ni_vp;
2963 if (vp->v_type != VDIR) {
2964 error = ENOTDIR;
2965 goto out;
2966 }
2967 /*
2968 * No rmdir "." please.
2969 */
2970 if (nd.ni_dvp == vp) {
2971 error = EINVAL;
2972 goto out;
2973 }
2974 /*
2975 * The root of a mounted filesystem cannot be deleted.
2976 */
2977 if (vp->v_flag & VROOT)
2978 error = EBUSY;
2979 out:
2980 if (!error) {
2981 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2982 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2983 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2984 } else {
2985 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2986 if (nd.ni_dvp == vp)
2987 vrele(nd.ni_dvp);
2988 else
2989 vput(nd.ni_dvp);
2990 vput(vp);
2991 }
2992 return (error);
2993 }
2994
2995 /*
2996 * Read a block of directory entries in a file system independent format.
2997 */
2998 int
2999 sys_getdents(p, v, retval)
3000 struct proc *p;
3001 void *v;
3002 register_t *retval;
3003 {
3004 struct sys_getdents_args /* {
3005 syscallarg(int) fd;
3006 syscallarg(char *) buf;
3007 syscallarg(size_t) count;
3008 } */ *uap = v;
3009 struct file *fp;
3010 int error, done;
3011
3012 /* getvnode() will use the descriptor for us */
3013 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3014 return (error);
3015 if ((fp->f_flag & FREAD) == 0) {
3016 error = EBADF;
3017 goto out;
3018 }
3019 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3020 SCARG(uap, count), &done, p, 0, 0);
3021 *retval = done;
3022 out:
3023 FILE_UNUSE(fp, p);
3024 return (error);
3025 }
3026
3027 /*
3028 * Set the mode mask for creation of filesystem nodes.
3029 */
3030 int
3031 sys_umask(p, v, retval)
3032 struct proc *p;
3033 void *v;
3034 register_t *retval;
3035 {
3036 struct sys_umask_args /* {
3037 syscallarg(mode_t) newmask;
3038 } */ *uap = v;
3039 struct cwdinfo *cwdi;
3040
3041 cwdi = p->p_cwdi;
3042 *retval = cwdi->cwdi_cmask;
3043 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3044 return (0);
3045 }
3046
3047 /*
3048 * Void all references to file by ripping underlying filesystem
3049 * away from vnode.
3050 */
3051 /* ARGSUSED */
3052 int
3053 sys_revoke(p, v, retval)
3054 struct proc *p;
3055 void *v;
3056 register_t *retval;
3057 {
3058 struct sys_revoke_args /* {
3059 syscallarg(const char *) path;
3060 } */ *uap = v;
3061 struct vnode *vp;
3062 struct vattr vattr;
3063 int error;
3064 struct nameidata nd;
3065
3066 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3067 if ((error = namei(&nd)) != 0)
3068 return (error);
3069 vp = nd.ni_vp;
3070 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3071 goto out;
3072 if (p->p_ucred->cr_uid != vattr.va_uid &&
3073 (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3074 goto out;
3075 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3076 VOP_REVOKE(vp, REVOKEALL);
3077 out:
3078 vrele(vp);
3079 return (error);
3080 }
3081
3082 /*
3083 * Convert a user file descriptor to a kernel file entry.
3084 */
3085 int
3086 getvnode(fdp, fd, fpp)
3087 struct filedesc *fdp;
3088 int fd;
3089 struct file **fpp;
3090 {
3091 struct vnode *vp;
3092 struct file *fp;
3093
3094 if ((u_int)fd >= fdp->fd_nfiles ||
3095 (fp = fdp->fd_ofiles[fd]) == NULL ||
3096 (fp->f_iflags & FIF_WANTCLOSE) != 0)
3097 return (EBADF);
3098
3099 FILE_USE(fp);
3100
3101 if (fp->f_type != DTYPE_VNODE) {
3102 FILE_UNUSE(fp, NULL);
3103 return (EINVAL);
3104 }
3105
3106 vp = (struct vnode *)fp->f_data;
3107 if (vp->v_type == VBAD) {
3108 FILE_UNUSE(fp, NULL);
3109 return (EBADF);
3110 }
3111
3112 *fpp = fp;
3113 return (0);
3114 }
3115