vfs_syscalls.c revision 1.164.2.4 1 /* $NetBSD: vfs_syscalls.c,v 1.164.2.4 2001/09/21 22:36:28 nathanw Exp $ */
2
3 /*
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. All advertising materials mentioning features or use of this software
21 * must display the following acknowledgement:
22 * This product includes software developed by the University of
23 * California, Berkeley and its contributors.
24 * 4. Neither the name of the University nor the names of its contributors
25 * may be used to endorse or promote products derived from this software
26 * without specific prior written permission.
27 *
28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * SUCH DAMAGE.
39 *
40 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95
41 */
42
43 #include "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/lwp.h>
56 #include <sys/proc.h>
57 #include <sys/uio.h>
58 #include <sys/malloc.h>
59 #include <sys/dirent.h>
60 #include <sys/syscallargs.h>
61
62 #include <miscfs/genfs/genfs.h>
63 #include <miscfs/syncfs/syncfs.h>
64
65 #include <uvm/uvm_extern.h>
66
67 #include <sys/sysctl.h>
68
69 static int change_dir __P((struct nameidata *, struct proc *));
70 static int change_flags __P((struct vnode *, u_long, 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 #if defined(COMPAT_09) || defined(COMPAT_43)
91 /*
92 * This table is used to maintain compatibility with 4.3BSD
93 * and NetBSD 0.9 mount syscalls. Note, the order is important!
94 *
95 * Do not modify this table. It should only contain filesystems
96 * supported by NetBSD 0.9 and 4.3BSD.
97 */
98 const char * const mountcompatnames[] = {
99 NULL, /* 0 = MOUNT_NONE */
100 MOUNT_FFS, /* 1 = MOUNT_UFS */
101 MOUNT_NFS, /* 2 */
102 MOUNT_MFS, /* 3 */
103 MOUNT_MSDOS, /* 4 */
104 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */
105 MOUNT_FDESC, /* 6 */
106 MOUNT_KERNFS, /* 7 */
107 NULL, /* 8 = MOUNT_DEVFS */
108 MOUNT_AFS, /* 9 */
109 };
110 const int nmountcompatnames = sizeof(mountcompatnames) /
111 sizeof(mountcompatnames[0]);
112 #endif /* COMPAT_09 || COMPAT_43 */
113
114 /* ARGSUSED */
115 int
116 sys_mount(l, v, retval)
117 struct lwp *l;
118 void *v;
119 register_t *retval;
120 {
121 struct sys_mount_args /* {
122 syscallarg(const char *) type;
123 syscallarg(const char *) path;
124 syscallarg(int) flags;
125 syscallarg(void *) data;
126 } */ *uap = v;
127 struct proc *p = l->l_proc;
128 struct vnode *vp;
129 struct mount *mp;
130 int error, flag = 0;
131 char fstypename[MFSNAMELEN];
132 struct vattr va;
133 struct nameidata nd;
134 struct vfsops *vfs;
135
136 if (dovfsusermount == 0 && (error = suser(p->p_ucred, &p->p_acflag)))
137 return (error);
138 /*
139 * Get vnode to be covered
140 */
141 NDINIT(&nd, LOOKUP, FOLLOW , UIO_USERSPACE,
142 SCARG(uap, path), p);
143 if ((error = namei(&nd)) != 0)
144 return (error);
145 vp = nd.ni_vp;
146 /*
147 * A lookup in VFS_MOUNT might result in an attempt to
148 * lock this vnode again, so make the lock resursive.
149 */
150 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
151 if (SCARG(uap, flags) & MNT_UPDATE) {
152 if ((vp->v_flag & VROOT) == 0) {
153 vput(vp);
154 return (EINVAL);
155 }
156 mp = vp->v_mount;
157 flag = mp->mnt_flag;
158 vfs = mp->mnt_op;
159 /*
160 * We only allow the filesystem to be reloaded if it
161 * is currently mounted read-only.
162 */
163 if ((SCARG(uap, flags) & MNT_RELOAD) &&
164 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
165 vput(vp);
166 return (EOPNOTSUPP); /* Needs translation */
167 }
168 /*
169 * In "highly secure" mode, don't let the caller do anything
170 * but downgrade a filesystem from read-write to read-only.
171 * (see also below; MNT_UPDATE is required.)
172 */
173 if (securelevel >= 2 &&
174 (SCARG(uap, flags) !=
175 (mp->mnt_flag | MNT_RDONLY |
176 MNT_RELOAD | MNT_FORCE | MNT_UPDATE))) {
177 vput(vp);
178 return (EPERM);
179 }
180 mp->mnt_flag |=
181 SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
182 /*
183 * Only root, or the user that did the original mount is
184 * permitted to update it.
185 */
186 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
187 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
188 vput(vp);
189 return (error);
190 }
191 /*
192 * Do not allow NFS export by non-root users. For non-root
193 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
194 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
195 */
196 if (p->p_ucred->cr_uid != 0) {
197 if (SCARG(uap, flags) & MNT_EXPORTED) {
198 vput(vp);
199 return (EPERM);
200 }
201 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
202 if (flag & MNT_NOEXEC)
203 SCARG(uap, flags) |= MNT_NOEXEC;
204 }
205 if (vfs_busy(mp, LK_NOWAIT, 0)) {
206 vput(vp);
207 return (EPERM);
208 }
209 VOP_UNLOCK(vp, 0);
210 goto update;
211 } else {
212 if (securelevel >= 2) {
213 vput(vp);
214 return (EPERM);
215 }
216 }
217 /*
218 * If the user is not root, ensure that they own the directory
219 * onto which we are attempting to mount.
220 */
221 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
222 (va.va_uid != p->p_ucred->cr_uid &&
223 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
224 vput(vp);
225 return (error);
226 }
227 /*
228 * Do not allow NFS export by non-root users. For non-root users,
229 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
230 * mount point is already MNT_NOEXEC.
231 */
232 if (p->p_ucred->cr_uid != 0) {
233 if (SCARG(uap, flags) & MNT_EXPORTED) {
234 vput(vp);
235 return (EPERM);
236 }
237 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
238 if (vp->v_mount->mnt_flag & MNT_NOEXEC)
239 SCARG(uap, flags) |= MNT_NOEXEC;
240 }
241 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
242 return (error);
243 if (vp->v_type != VDIR) {
244 vput(vp);
245 return (ENOTDIR);
246 }
247 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
248 if (error) {
249 #if defined(COMPAT_09) || defined(COMPAT_43)
250 /*
251 * Historically filesystem types were identified by number.
252 * If we get an integer for the filesystem type instead of a
253 * string, we check to see if it matches one of the historic
254 * filesystem types.
255 */
256 u_long fsindex = (u_long)SCARG(uap, type);
257 if (fsindex >= nmountcompatnames ||
258 mountcompatnames[fsindex] == NULL) {
259 vput(vp);
260 return (ENODEV);
261 }
262 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
263 #else
264 vput(vp);
265 return (error);
266 #endif
267 }
268 #ifdef COMPAT_10
269 /* Accept `ufs' as an alias for `ffs'. */
270 if (!strncmp(fstypename, "ufs", MFSNAMELEN))
271 strncpy(fstypename, "ffs", MFSNAMELEN);
272 #endif
273 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
274 vput(vp);
275 return (ENODEV);
276 }
277 if (vp->v_mountedhere != NULL) {
278 vput(vp);
279 return (EBUSY);
280 }
281
282 /*
283 * Allocate and initialize the file system.
284 */
285 mp = (struct mount *)malloc((u_long)sizeof(struct mount),
286 M_MOUNT, M_WAITOK);
287 memset((char *)mp, 0, (u_long)sizeof(struct mount));
288 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
289 (void)vfs_busy(mp, LK_NOWAIT, 0);
290 mp->mnt_op = vfs;
291 vfs->vfs_refcount++;
292 mp->mnt_vnodecovered = vp;
293 mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
294 mp->mnt_unmounter = NULL;
295 update:
296 /*
297 * Set the mount level flags.
298 */
299 if (SCARG(uap, flags) & MNT_RDONLY)
300 mp->mnt_flag |= MNT_RDONLY;
301 else if (mp->mnt_flag & MNT_RDONLY)
302 mp->mnt_flag |= MNT_WANTRDWR;
303 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
304 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
305 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
306 mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
307 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
308 MNT_NOCOREDUMP | MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM |
309 MNT_SOFTDEP);
310 /*
311 * Mount the filesystem.
312 */
313 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
314 if (mp->mnt_flag & MNT_UPDATE) {
315 vrele(vp);
316 if (mp->mnt_flag & MNT_WANTRDWR)
317 mp->mnt_flag &= ~MNT_RDONLY;
318 mp->mnt_flag &=~
319 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR);
320 if (error)
321 mp->mnt_flag = flag;
322 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
323 if (mp->mnt_syncer == NULL)
324 error = vfs_allocate_syncvnode(mp);
325 } else {
326 if (mp->mnt_syncer != NULL)
327 vfs_deallocate_syncvnode(mp);
328 }
329 vfs_unbusy(mp);
330 return (error);
331 }
332 /*
333 * Put the new filesystem on the mount list after root.
334 */
335 cache_purge(vp);
336 if (!error) {
337 vp->v_mountedhere = mp;
338 simple_lock(&mountlist_slock);
339 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
340 simple_unlock(&mountlist_slock);
341 checkdirs(vp);
342 VOP_UNLOCK(vp, 0);
343 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
344 error = vfs_allocate_syncvnode(mp);
345 vfs_unbusy(mp);
346 (void) VFS_STATFS(mp, &mp->mnt_stat, p);
347 if ((error = VFS_START(mp, 0, p)))
348 vrele(vp);
349 } else {
350 vp->v_mountedhere = (struct mount *)0;
351 vfs->vfs_refcount--;
352 vfs_unbusy(mp);
353 free((caddr_t)mp, M_MOUNT);
354 vput(vp);
355 }
356 return (error);
357 }
358
359 /*
360 * Scan all active processes to see if any of them have a current
361 * or root directory onto which the new filesystem has just been
362 * mounted. If so, replace them with the new mount point.
363 */
364 void
365 checkdirs(olddp)
366 struct vnode *olddp;
367 {
368 struct cwdinfo *cwdi;
369 struct vnode *newdp;
370 struct proc *p;
371
372 if (olddp->v_usecount == 1)
373 return;
374 if (VFS_ROOT(olddp->v_mountedhere, &newdp))
375 panic("mount: lost mount");
376 proclist_lock_read();
377 for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
378 cwdi = p->p_cwdi;
379 if (cwdi->cwdi_cdir == olddp) {
380 vrele(cwdi->cwdi_cdir);
381 VREF(newdp);
382 cwdi->cwdi_cdir = newdp;
383 }
384 if (cwdi->cwdi_rdir == olddp) {
385 vrele(cwdi->cwdi_rdir);
386 VREF(newdp);
387 cwdi->cwdi_rdir = newdp;
388 }
389 }
390 proclist_unlock_read();
391 if (rootvnode == olddp) {
392 vrele(rootvnode);
393 VREF(newdp);
394 rootvnode = newdp;
395 }
396 vput(newdp);
397 }
398
399 /*
400 * Unmount a file system.
401 *
402 * Note: unmount takes a path to the vnode mounted on as argument,
403 * not special file (as before).
404 */
405 /* ARGSUSED */
406 int
407 sys_unmount(l, v, retval)
408 struct lwp *l;
409 void *v;
410 register_t *retval;
411 {
412 struct sys_unmount_args /* {
413 syscallarg(const char *) path;
414 syscallarg(int) flags;
415 } */ *uap = v;
416 struct proc *p = l->l_proc;
417 struct vnode *vp;
418 struct mount *mp;
419 int error;
420 struct nameidata nd;
421
422 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
423 SCARG(uap, path), p);
424 if ((error = namei(&nd)) != 0)
425 return (error);
426 vp = nd.ni_vp;
427 mp = vp->v_mount;
428
429 /*
430 * Only root, or the user that did the original mount is
431 * permitted to unmount this filesystem.
432 */
433 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
434 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
435 vput(vp);
436 return (error);
437 }
438
439 /*
440 * Don't allow unmounting the root file system.
441 */
442 if (mp->mnt_flag & MNT_ROOTFS) {
443 vput(vp);
444 return (EINVAL);
445 }
446
447 /*
448 * Must be the root of the filesystem
449 */
450 if ((vp->v_flag & VROOT) == 0) {
451 vput(vp);
452 return (EINVAL);
453 }
454 vput(vp);
455
456 /*
457 * XXX Freeze syncer. Must do this before locking the
458 * mount point. See dounmount() for details.
459 */
460 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
461
462 if (vfs_busy(mp, 0, 0)) {
463 lockmgr(&syncer_lock, LK_RELEASE, NULL);
464 return (EBUSY);
465 }
466
467 return (dounmount(mp, SCARG(uap, flags), p));
468 }
469
470 /*
471 * Do the actual file system unmount. File system is assumed to have been
472 * marked busy by the caller.
473 */
474 int
475 dounmount(mp, flags, p)
476 struct mount *mp;
477 int flags;
478 struct proc *p;
479 {
480 struct vnode *coveredvp;
481 int error;
482 int async;
483 int used_syncer;
484
485 simple_lock(&mountlist_slock);
486 vfs_unbusy(mp);
487 used_syncer = (mp->mnt_syncer != NULL);
488
489 /*
490 * XXX Syncer must be frozen when we get here. This should really
491 * be done on a per-mountpoint basis, but especially the softdep
492 * code possibly called from the syncer doens't exactly work on a
493 * per-mountpoint basis, so the softdep code would become a maze
494 * of vfs_busy() calls.
495 *
496 * The caller of dounmount() must acquire syncer_lock because
497 * the syncer itself acquires locks in syncer_lock -> vfs_busy
498 * order, and we must preserve that order to avoid deadlock.
499 *
500 * So, if the file system did not use the syncer, now is
501 * the time to release the syncer_lock.
502 */
503 if (used_syncer == 0)
504 lockmgr(&syncer_lock, LK_RELEASE, NULL);
505
506 mp->mnt_flag |= MNT_UNMOUNT;
507 mp->mnt_unmounter = p;
508 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
509 if (mp->mnt_flag & MNT_EXPUBLIC)
510 vfs_setpublicfs(NULL, NULL, NULL);
511 async = mp->mnt_flag & MNT_ASYNC;
512 mp->mnt_flag &= ~MNT_ASYNC;
513 cache_purgevfs(mp); /* remove cache entries for this file sys */
514 if (mp->mnt_syncer != NULL)
515 vfs_deallocate_syncvnode(mp);
516 if (((mp->mnt_flag & MNT_RDONLY) ||
517 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
518 (flags & MNT_FORCE))
519 error = VFS_UNMOUNT(mp, flags, p);
520 simple_lock(&mountlist_slock);
521 if (error) {
522 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
523 (void) vfs_allocate_syncvnode(mp);
524 mp->mnt_flag &= ~MNT_UNMOUNT;
525 mp->mnt_unmounter = NULL;
526 mp->mnt_flag |= async;
527 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
528 &mountlist_slock);
529 if (used_syncer)
530 lockmgr(&syncer_lock, LK_RELEASE, NULL);
531 while (mp->mnt_wcnt > 0) {
532 wakeup((caddr_t)mp);
533 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0);
534 }
535 return (error);
536 }
537 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
538 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
539 coveredvp->v_mountedhere = NULL;
540 vrele(coveredvp);
541 }
542 mp->mnt_op->vfs_refcount--;
543 if (mp->mnt_vnodelist.lh_first != NULL)
544 panic("unmount: dangling vnode");
545 mp->mnt_flag |= MNT_GONE;
546 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
547 if (used_syncer)
548 lockmgr(&syncer_lock, LK_RELEASE, NULL);
549 while(mp->mnt_wcnt > 0) {
550 wakeup((caddr_t)mp);
551 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0);
552 }
553 free((caddr_t)mp, M_MOUNT);
554 return (0);
555 }
556
557 /*
558 * Sync each mounted filesystem.
559 */
560 #ifdef DEBUG
561 int syncprt = 0;
562 struct ctldebug debug0 = { "syncprt", &syncprt };
563 #endif
564
565 /* ARGSUSED */
566 int
567 sys_sync(l, v, retval)
568 struct lwp *l;
569 void *v;
570 register_t *retval;
571 {
572 struct mount *mp, *nmp;
573 int asyncflag;
574 struct proc *p = l->l_proc;
575
576 simple_lock(&mountlist_slock);
577 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
578 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
579 nmp = mp->mnt_list.cqe_prev;
580 continue;
581 }
582 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
583 asyncflag = mp->mnt_flag & MNT_ASYNC;
584 mp->mnt_flag &= ~MNT_ASYNC;
585 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
586 if (asyncflag)
587 mp->mnt_flag |= MNT_ASYNC;
588 }
589 simple_lock(&mountlist_slock);
590 nmp = mp->mnt_list.cqe_prev;
591 vfs_unbusy(mp);
592
593 }
594 simple_unlock(&mountlist_slock);
595 #ifdef DEBUG
596 if (syncprt)
597 vfs_bufstats();
598 #endif /* DEBUG */
599 return (0);
600 }
601
602 /*
603 * Change filesystem quotas.
604 */
605 /* ARGSUSED */
606 int
607 sys_quotactl(l, v, retval)
608 struct lwp *l;
609 void *v;
610 register_t *retval;
611 {
612 struct sys_quotactl_args /* {
613 syscallarg(const char *) path;
614 syscallarg(int) cmd;
615 syscallarg(int) uid;
616 syscallarg(caddr_t) arg;
617 } */ *uap = v;
618 struct proc *p = l->l_proc;
619 struct mount *mp;
620 int error;
621 struct nameidata nd;
622
623 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
624 if ((error = namei(&nd)) != 0)
625 return (error);
626 mp = nd.ni_vp->v_mount;
627 vrele(nd.ni_vp);
628 return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
629 SCARG(uap, arg), p));
630 }
631
632 /*
633 * Get filesystem statistics.
634 */
635 /* ARGSUSED */
636 int
637 sys_statfs(l, v, retval)
638 struct lwp *l;
639 void *v;
640 register_t *retval;
641 {
642 struct sys_statfs_args /* {
643 syscallarg(const char *) path;
644 syscallarg(struct statfs *) buf;
645 } */ *uap = v;
646 struct proc *p = l->l_proc;
647 struct mount *mp;
648 struct statfs *sp;
649 int error;
650 struct nameidata nd;
651
652 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
653 if ((error = namei(&nd)) != 0)
654 return (error);
655 mp = nd.ni_vp->v_mount;
656 sp = &mp->mnt_stat;
657 vrele(nd.ni_vp);
658 if ((error = VFS_STATFS(mp, sp, p)) != 0)
659 return (error);
660 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
661 sp->f_oflags = sp->f_flags & 0xffff;
662 return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
663 }
664
665 /*
666 * Get filesystem statistics.
667 */
668 /* ARGSUSED */
669 int
670 sys_fstatfs(l, v, retval)
671 struct lwp *l;
672 void *v;
673 register_t *retval;
674 {
675 struct sys_fstatfs_args /* {
676 syscallarg(int) fd;
677 syscallarg(struct statfs *) buf;
678 } */ *uap = v;
679 struct proc *p = l->l_proc;
680 struct file *fp;
681 struct mount *mp;
682 struct statfs *sp;
683 int error;
684
685 /* getvnode() will use the descriptor for us */
686 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
687 return (error);
688 mp = ((struct vnode *)fp->f_data)->v_mount;
689 sp = &mp->mnt_stat;
690 if ((error = VFS_STATFS(mp, sp, p)) != 0)
691 goto out;
692 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
693 sp->f_oflags = sp->f_flags & 0xffff;
694 error = copyout(sp, SCARG(uap, buf), sizeof(*sp));
695 out:
696 FILE_UNUSE(fp, p);
697 return (error);
698 }
699
700 /*
701 * Get statistics on all filesystems.
702 */
703 int
704 sys_getfsstat(l, v, retval)
705 struct lwp *l;
706 void *v;
707 register_t *retval;
708 {
709 struct sys_getfsstat_args /* {
710 syscallarg(struct statfs *) buf;
711 syscallarg(long) bufsize;
712 syscallarg(int) flags;
713 } */ *uap = v;
714 struct proc *p = l->l_proc;
715 struct mount *mp, *nmp;
716 struct statfs *sp;
717 caddr_t sfsp;
718 long count, maxcount, error;
719
720 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
721 sfsp = (caddr_t)SCARG(uap, buf);
722 simple_lock(&mountlist_slock);
723 count = 0;
724 for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
725 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
726 nmp = mp->mnt_list.cqe_next;
727 continue;
728 }
729 if (sfsp && count < maxcount) {
730 sp = &mp->mnt_stat;
731 /*
732 * If MNT_NOWAIT or MNT_LAZY is specified, do not
733 * refresh the fsstat cache. MNT_WAIT or MNT_LAXY
734 * overrides MNT_NOWAIT.
735 */
736 if (SCARG(uap, flags) != MNT_NOWAIT &&
737 SCARG(uap, flags) != MNT_LAZY &&
738 (SCARG(uap, flags) == MNT_WAIT ||
739 SCARG(uap, flags) == 0) &&
740 (error = VFS_STATFS(mp, sp, p)) != 0) {
741 simple_lock(&mountlist_slock);
742 nmp = mp->mnt_list.cqe_next;
743 vfs_unbusy(mp);
744 continue;
745 }
746 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
747 sp->f_oflags = sp->f_flags & 0xffff;
748 error = copyout(sp, sfsp, sizeof(*sp));
749 if (error) {
750 vfs_unbusy(mp);
751 return (error);
752 }
753 sfsp += sizeof(*sp);
754 }
755 count++;
756 simple_lock(&mountlist_slock);
757 nmp = mp->mnt_list.cqe_next;
758 vfs_unbusy(mp);
759 }
760 simple_unlock(&mountlist_slock);
761 if (sfsp && count > maxcount)
762 *retval = maxcount;
763 else
764 *retval = count;
765 return (0);
766 }
767
768 /*
769 * Change current working directory to a given file descriptor.
770 */
771 /* ARGSUSED */
772 int
773 sys_fchdir(l, v, retval)
774 struct lwp *l;
775 void *v;
776 register_t *retval;
777 {
778 struct sys_fchdir_args /* {
779 syscallarg(int) fd;
780 } */ *uap = v;
781 struct proc *p = l->l_proc;
782 struct filedesc *fdp = p->p_fd;
783 struct cwdinfo *cwdi = p->p_cwdi;
784 struct vnode *vp, *tdp;
785 struct mount *mp;
786 struct file *fp;
787 int error;
788
789 /* getvnode() will use the descriptor for us */
790 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
791 return (error);
792 vp = (struct vnode *)fp->f_data;
793
794 VREF(vp);
795 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
796 if (vp->v_type != VDIR)
797 error = ENOTDIR;
798 else
799 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
800 while (!error && (mp = vp->v_mountedhere) != NULL) {
801 if (vfs_busy(mp, 0, 0))
802 continue;
803 error = VFS_ROOT(mp, &tdp);
804 vfs_unbusy(mp);
805 if (error)
806 break;
807 vput(vp);
808 vp = tdp;
809 }
810 if (error) {
811 vput(vp);
812 goto out;
813 }
814 VOP_UNLOCK(vp, 0);
815
816 /*
817 * Disallow changing to a directory not under the process's
818 * current root directory (if there is one).
819 */
820 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
821 vrele(vp);
822 error = EPERM; /* operation not permitted */
823 goto out;
824 }
825
826 vrele(cwdi->cwdi_cdir);
827 cwdi->cwdi_cdir = vp;
828 out:
829 FILE_UNUSE(fp, p);
830 return (error);
831 }
832
833 /*
834 * Change this process's notion of the root directory to a given file descriptor.
835 */
836
837 int
838 sys_fchroot(l, v, retval)
839 struct lwp *l;
840 void *v;
841 register_t *retval;
842 {
843 struct sys_fchroot_args *uap = v;
844 struct proc *p = l->l_proc;
845 struct filedesc *fdp = p->p_fd;
846 struct cwdinfo *cwdi = p->p_cwdi;
847 struct vnode *vp;
848 struct file *fp;
849 int error;
850
851 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
852 return error;
853 /* getvnode() will use the descriptor for us */
854 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
855 return error;
856 vp = (struct vnode *) fp->f_data;
857 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
858 if (vp->v_type != VDIR)
859 error = ENOTDIR;
860 else
861 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
862 VOP_UNLOCK(vp, 0);
863 if (error)
864 goto out;
865 VREF(vp);
866
867 /*
868 * Prevent escaping from chroot by putting the root under
869 * the working directory. Silently chdir to / if we aren't
870 * already there.
871 */
872 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
873 /*
874 * XXX would be more failsafe to change directory to a
875 * deadfs node here instead
876 */
877 vrele(cwdi->cwdi_cdir);
878 VREF(vp);
879 cwdi->cwdi_cdir = vp;
880 }
881
882 if (cwdi->cwdi_rdir != NULL)
883 vrele(cwdi->cwdi_rdir);
884 cwdi->cwdi_rdir = vp;
885 out:
886 FILE_UNUSE(fp, p);
887 return (error);
888 }
889
890
891
892 /*
893 * Change current working directory (``.'').
894 */
895 /* ARGSUSED */
896 int
897 sys_chdir(l, v, retval)
898 struct lwp *l;
899 void *v;
900 register_t *retval;
901 {
902 struct sys_chdir_args /* {
903 syscallarg(const char *) path;
904 } */ *uap = v;
905 struct proc *p = l->l_proc;
906 struct cwdinfo *cwdi = p->p_cwdi;
907 int error;
908 struct nameidata nd;
909
910 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
911 SCARG(uap, path), p);
912 if ((error = change_dir(&nd, p)) != 0)
913 return (error);
914 vrele(cwdi->cwdi_cdir);
915 cwdi->cwdi_cdir = nd.ni_vp;
916 return (0);
917 }
918
919 /*
920 * Change notion of root (``/'') directory.
921 */
922 /* ARGSUSED */
923 int
924 sys_chroot(l, v, retval)
925 struct lwp *l;
926 void *v;
927 register_t *retval;
928 {
929 struct sys_chroot_args /* {
930 syscallarg(const char *) path;
931 } */ *uap = v;
932 struct proc *p = l->l_proc;
933 struct cwdinfo *cwdi = p->p_cwdi;
934 struct vnode *vp;
935 int error;
936 struct nameidata nd;
937
938 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
939 return (error);
940 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
941 SCARG(uap, path), p);
942 if ((error = change_dir(&nd, p)) != 0)
943 return (error);
944 if (cwdi->cwdi_rdir != NULL)
945 vrele(cwdi->cwdi_rdir);
946 vp = nd.ni_vp;
947 cwdi->cwdi_rdir = vp;
948
949 /*
950 * Prevent escaping from chroot by putting the root under
951 * the working directory. Silently chdir to / if we aren't
952 * already there.
953 */
954 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
955 /*
956 * XXX would be more failsafe to change directory to a
957 * deadfs node here instead
958 */
959 vrele(cwdi->cwdi_cdir);
960 VREF(vp);
961 cwdi->cwdi_cdir = vp;
962 }
963
964 return (0);
965 }
966
967 /*
968 * Common routine for chroot and chdir.
969 */
970 static int
971 change_dir(ndp, p)
972 struct nameidata *ndp;
973 struct proc *p;
974 {
975 struct vnode *vp;
976 int error;
977
978 if ((error = namei(ndp)) != 0)
979 return (error);
980 vp = ndp->ni_vp;
981 if (vp->v_type != VDIR)
982 error = ENOTDIR;
983 else
984 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
985
986 if (error)
987 vput(vp);
988 else
989 VOP_UNLOCK(vp, 0);
990 return (error);
991 }
992
993 /*
994 * Check permissions, allocate an open file structure,
995 * and call the device open routine if any.
996 */
997 int
998 sys_open(l, v, retval)
999 struct lwp *l;
1000 void *v;
1001 register_t *retval;
1002 {
1003 struct sys_open_args /* {
1004 syscallarg(const char *) path;
1005 syscallarg(int) flags;
1006 syscallarg(int) mode;
1007 } */ *uap = v;
1008 struct proc *p = l->l_proc;
1009 struct cwdinfo *cwdi = p->p_cwdi;
1010 struct filedesc *fdp = p->p_fd;
1011 struct file *fp;
1012 struct vnode *vp;
1013 int flags, cmode;
1014 int type, indx, error;
1015 struct flock lf;
1016 struct nameidata nd;
1017
1018 flags = FFLAGS(SCARG(uap, flags));
1019 if ((flags & (FREAD | FWRITE)) == 0)
1020 return (EINVAL);
1021 /* falloc() will use the file descriptor for us */
1022 if ((error = falloc(p, &fp, &indx)) != 0)
1023 return (error);
1024 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1025 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1026 p->p_dupfd = -indx - 1; /* XXX check for fdopen */
1027 if ((error = vn_open(&nd, flags, cmode)) != 0) {
1028 FILE_UNUSE(fp, p);
1029 ffree(fp);
1030 if ((error == ENODEV || error == ENXIO) &&
1031 p->p_dupfd >= 0 && /* XXX from fdopen */
1032 (error =
1033 dupfdopen(p, indx, p->p_dupfd, flags, error)) == 0) {
1034 *retval = indx;
1035 return (0);
1036 }
1037 if (error == ERESTART)
1038 error = EINTR;
1039 fdremove(fdp, indx);
1040 return (error);
1041 }
1042 p->p_dupfd = 0;
1043 vp = nd.ni_vp;
1044 fp->f_flag = flags & FMASK;
1045 fp->f_type = DTYPE_VNODE;
1046 fp->f_ops = &vnops;
1047 fp->f_data = (caddr_t)vp;
1048 if (flags & (O_EXLOCK | O_SHLOCK)) {
1049 lf.l_whence = SEEK_SET;
1050 lf.l_start = 0;
1051 lf.l_len = 0;
1052 if (flags & O_EXLOCK)
1053 lf.l_type = F_WRLCK;
1054 else
1055 lf.l_type = F_RDLCK;
1056 type = F_FLOCK;
1057 if ((flags & FNONBLOCK) == 0)
1058 type |= F_WAIT;
1059 VOP_UNLOCK(vp, 0);
1060 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1061 if (error) {
1062 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1063 FILE_UNUSE(fp, p);
1064 ffree(fp);
1065 fdremove(fdp, indx);
1066 return (error);
1067 }
1068 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1069 fp->f_flag |= FHASLOCK;
1070 }
1071 VOP_UNLOCK(vp, 0);
1072 *retval = indx;
1073 FILE_SET_MATURE(fp);
1074 FILE_UNUSE(fp, p);
1075 return (0);
1076 }
1077
1078 /*
1079 * Get file handle system call
1080 */
1081 int
1082 sys_getfh(l, v, retval)
1083 struct lwp *l;
1084 void *v;
1085 register_t *retval;
1086 {
1087 struct sys_getfh_args /* {
1088 syscallarg(char *) fname;
1089 syscallarg(fhandle_t *) fhp;
1090 } */ *uap = v;
1091 struct proc *p = l->l_proc;
1092 struct vnode *vp;
1093 fhandle_t fh;
1094 int error;
1095 struct nameidata nd;
1096
1097 /*
1098 * Must be super user
1099 */
1100 error = suser(p->p_ucred, &p->p_acflag);
1101 if (error)
1102 return (error);
1103 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1104 SCARG(uap, fname), p);
1105 error = namei(&nd);
1106 if (error)
1107 return (error);
1108 vp = nd.ni_vp;
1109 memset((caddr_t)&fh, 0, sizeof(fh));
1110 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1111 error = VFS_VPTOFH(vp, &fh.fh_fid);
1112 vput(vp);
1113 if (error)
1114 return (error);
1115 error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1116 return (error);
1117 }
1118
1119 /*
1120 * Open a file given a file handle.
1121 *
1122 * Check permissions, allocate an open file structure,
1123 * and call the device open routine if any.
1124 */
1125 int
1126 sys_fhopen(l, v, retval)
1127 struct lwp *l;
1128 void *v;
1129 register_t *retval;
1130 {
1131 struct sys_fhopen_args /* {
1132 syscallarg(const fhandle_t *) fhp;
1133 syscallarg(int) flags;
1134 } */ *uap = v;
1135 struct proc *p = l->l_proc;
1136 struct filedesc *fdp = p->p_fd;
1137 struct file *fp;
1138 struct vnode *vp = NULL;
1139 struct mount *mp;
1140 struct ucred *cred = p->p_ucred;
1141 int flags;
1142 struct file *nfp;
1143 int type, indx, error=0;
1144 struct flock lf;
1145 struct vattr va;
1146 fhandle_t fh;
1147
1148 /*
1149 * Must be super user
1150 */
1151 if ((error = suser(p->p_ucred, &p->p_acflag)))
1152 return (error);
1153
1154 flags = FFLAGS(SCARG(uap, flags));
1155 if ((flags & (FREAD | FWRITE)) == 0)
1156 return (EINVAL);
1157 if ((flags & O_CREAT))
1158 return (EINVAL);
1159 /* falloc() will use the file descriptor for us */
1160 if ((error = falloc(p, &nfp, &indx)) != 0)
1161 return (error);
1162 fp = nfp;
1163 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1164 goto bad;
1165
1166 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1167 error = ESTALE;
1168 goto bad;
1169 }
1170
1171 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1172 vp = NULL; /* most likely unnecessary sanity for bad: */
1173 goto bad;
1174 }
1175
1176 /* Now do an effective vn_open */
1177
1178 if (vp->v_type == VSOCK) {
1179 error = EOPNOTSUPP;
1180 goto bad;
1181 }
1182 if (flags & FREAD) {
1183 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1184 goto bad;
1185 }
1186 if (flags & (FWRITE | O_TRUNC)) {
1187 if (vp->v_type == VDIR) {
1188 error = EISDIR;
1189 goto bad;
1190 }
1191 if ((error = vn_writechk(vp)) != 0 ||
1192 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1193 goto bad;
1194 }
1195 if (flags & O_TRUNC) {
1196 VOP_UNLOCK(vp, 0); /* XXX */
1197 VOP_LEASE(vp, p, cred, LEASE_WRITE);
1198 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
1199 VATTR_NULL(&va);
1200 va.va_size = 0;
1201 if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1202 goto bad;
1203 }
1204 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1205 goto bad;
1206 if (vp->v_type == VREG &&
1207 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1208 error = EIO;
1209 goto bad;
1210 }
1211 if (flags & FWRITE)
1212 vp->v_writecount++;
1213
1214 /* done with modified vn_open, now finish what sys_open does. */
1215
1216 fp->f_flag = flags & FMASK;
1217 fp->f_type = DTYPE_VNODE;
1218 fp->f_ops = &vnops;
1219 fp->f_data = (caddr_t)vp;
1220 if (flags & (O_EXLOCK | O_SHLOCK)) {
1221 lf.l_whence = SEEK_SET;
1222 lf.l_start = 0;
1223 lf.l_len = 0;
1224 if (flags & O_EXLOCK)
1225 lf.l_type = F_WRLCK;
1226 else
1227 lf.l_type = F_RDLCK;
1228 type = F_FLOCK;
1229 if ((flags & FNONBLOCK) == 0)
1230 type |= F_WAIT;
1231 VOP_UNLOCK(vp, 0);
1232 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1233 if (error) {
1234 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1235 FILE_UNUSE(fp, p);
1236 ffree(fp);
1237 fdremove(fdp, indx);
1238 return (error);
1239 }
1240 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1241 fp->f_flag |= FHASLOCK;
1242 }
1243 VOP_UNLOCK(vp, 0);
1244 *retval = indx;
1245 FILE_SET_MATURE(fp);
1246 FILE_UNUSE(fp, p);
1247 return (0);
1248
1249 bad:
1250 FILE_UNUSE(fp, p);
1251 ffree(fp);
1252 fdremove(fdp, indx);
1253 if (vp != NULL)
1254 vput(vp);
1255 return (error);
1256 }
1257
1258 /* ARGSUSED */
1259 int
1260 sys_fhstat(l, v, retval)
1261 struct lwp *l;
1262 void *v;
1263 register_t *retval;
1264 {
1265 struct sys_fhstat_args /* {
1266 syscallarg(const fhandle_t *) fhp;
1267 syscallarg(struct stat *) sb;
1268 } */ *uap = v;
1269 struct proc *p = l->l_proc;
1270 struct stat sb;
1271 int error;
1272 fhandle_t fh;
1273 struct mount *mp;
1274 struct vnode *vp;
1275
1276 /*
1277 * Must be super user
1278 */
1279 if ((error = suser(p->p_ucred, &p->p_acflag)))
1280 return (error);
1281
1282 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1283 return (error);
1284
1285 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1286 return (ESTALE);
1287 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1288 return (error);
1289 error = vn_stat(vp, &sb, p);
1290 vput(vp);
1291 if (error)
1292 return (error);
1293 error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1294 return (error);
1295 }
1296
1297 /* ARGSUSED */
1298 int
1299 sys_fhstatfs(l, v, retval)
1300 struct lwp *l;
1301 void *v;
1302 register_t *retval;
1303 {
1304 struct sys_fhstatfs_args /*
1305 syscallarg(const fhandle_t *) fhp;
1306 syscallarg(struct statfs *) buf;
1307 } */ *uap = v;
1308 struct proc *p = l->l_proc;
1309 struct statfs sp;
1310 fhandle_t fh;
1311 struct mount *mp;
1312 struct vnode *vp;
1313 int error;
1314
1315 /*
1316 * Must be super user
1317 */
1318 if ((error = suser(p->p_ucred, &p->p_acflag)))
1319 return (error);
1320
1321 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1322 return (error);
1323
1324 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1325 return (ESTALE);
1326 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1327 return (error);
1328 mp = vp->v_mount;
1329 vput(vp);
1330 if ((error = VFS_STATFS(mp, &sp, p)) != 0)
1331 return (error);
1332 sp.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1333 sp.f_oflags = sp.f_flags & 0xffff;
1334 return (copyout(&sp, SCARG(uap, buf), sizeof(sp)));
1335 }
1336
1337 /*
1338 * Create a special file.
1339 */
1340 /* ARGSUSED */
1341 int
1342 sys_mknod(l, v, retval)
1343 struct lwp *l;
1344 void *v;
1345 register_t *retval;
1346 {
1347 struct sys_mknod_args /* {
1348 syscallarg(const char *) path;
1349 syscallarg(int) mode;
1350 syscallarg(int) dev;
1351 } */ *uap = v;
1352 struct proc *p = l->l_proc;
1353 struct vnode *vp;
1354 struct vattr vattr;
1355 int error;
1356 int whiteout = 0;
1357 struct nameidata nd;
1358
1359 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1360 return (error);
1361 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1362 if ((error = namei(&nd)) != 0)
1363 return (error);
1364 vp = nd.ni_vp;
1365 if (vp != NULL)
1366 error = EEXIST;
1367 else {
1368 VATTR_NULL(&vattr);
1369 vattr.va_mode =
1370 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1371 vattr.va_rdev = SCARG(uap, dev);
1372 whiteout = 0;
1373
1374 switch (SCARG(uap, mode) & S_IFMT) {
1375 case S_IFMT: /* used by badsect to flag bad sectors */
1376 vattr.va_type = VBAD;
1377 break;
1378 case S_IFCHR:
1379 vattr.va_type = VCHR;
1380 break;
1381 case S_IFBLK:
1382 vattr.va_type = VBLK;
1383 break;
1384 case S_IFWHT:
1385 whiteout = 1;
1386 break;
1387 default:
1388 error = EINVAL;
1389 break;
1390 }
1391 }
1392 if (!error) {
1393 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1394 if (whiteout) {
1395 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1396 if (error)
1397 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1398 vput(nd.ni_dvp);
1399 } else {
1400 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1401 &nd.ni_cnd, &vattr);
1402 if (error == 0)
1403 vput(nd.ni_vp);
1404 }
1405 } else {
1406 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1407 if (nd.ni_dvp == vp)
1408 vrele(nd.ni_dvp);
1409 else
1410 vput(nd.ni_dvp);
1411 if (vp)
1412 vrele(vp);
1413 }
1414 return (error);
1415 }
1416
1417 /*
1418 * Create a named pipe.
1419 */
1420 /* ARGSUSED */
1421 int
1422 sys_mkfifo(l, v, retval)
1423 struct lwp *l;
1424 void *v;
1425 register_t *retval;
1426 {
1427 struct sys_mkfifo_args /* {
1428 syscallarg(const char *) path;
1429 syscallarg(int) mode;
1430 } */ *uap = v;
1431 struct proc *p = l->l_proc;
1432 struct vattr vattr;
1433 int error;
1434 struct nameidata nd;
1435
1436 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1437 if ((error = namei(&nd)) != 0)
1438 return (error);
1439 if (nd.ni_vp != NULL) {
1440 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1441 if (nd.ni_dvp == nd.ni_vp)
1442 vrele(nd.ni_dvp);
1443 else
1444 vput(nd.ni_dvp);
1445 vrele(nd.ni_vp);
1446 return (EEXIST);
1447 }
1448 VATTR_NULL(&vattr);
1449 vattr.va_type = VFIFO;
1450 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1451 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1452 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1453 if (error == 0)
1454 vput(nd.ni_vp);
1455 return (error);
1456 }
1457
1458 /*
1459 * Make a hard file link.
1460 */
1461 /* ARGSUSED */
1462 int
1463 sys_link(l, v, retval)
1464 struct lwp *l;
1465 void *v;
1466 register_t *retval;
1467 {
1468 struct sys_link_args /* {
1469 syscallarg(const char *) path;
1470 syscallarg(const char *) link;
1471 } */ *uap = v;
1472 struct proc *p = l->l_proc;
1473 struct vnode *vp;
1474 struct nameidata nd;
1475 int error;
1476
1477 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1478 if ((error = namei(&nd)) != 0)
1479 return (error);
1480 vp = nd.ni_vp;
1481 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1482 if ((error = namei(&nd)) != 0)
1483 goto out;
1484 if (nd.ni_vp) {
1485 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1486 if (nd.ni_dvp == nd.ni_vp)
1487 vrele(nd.ni_dvp);
1488 else
1489 vput(nd.ni_dvp);
1490 vrele(nd.ni_vp);
1491 error = EEXIST;
1492 goto out;
1493 }
1494 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1495 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1496 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1497 out:
1498 vrele(vp);
1499 return (error);
1500 }
1501
1502 /*
1503 * Make a symbolic link.
1504 */
1505 /* ARGSUSED */
1506 int
1507 sys_symlink(l, v, retval)
1508 struct lwp *l;
1509 void *v;
1510 register_t *retval;
1511 {
1512 struct sys_symlink_args /* {
1513 syscallarg(const char *) path;
1514 syscallarg(const char *) link;
1515 } */ *uap = v;
1516 struct proc *p = l->l_proc;
1517 struct vattr vattr;
1518 char *path;
1519 int error;
1520 struct nameidata nd;
1521
1522 path = PNBUF_GET();
1523 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1524 if (error)
1525 goto out;
1526 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1527 if ((error = namei(&nd)) != 0)
1528 goto out;
1529 if (nd.ni_vp) {
1530 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1531 if (nd.ni_dvp == nd.ni_vp)
1532 vrele(nd.ni_dvp);
1533 else
1534 vput(nd.ni_dvp);
1535 vrele(nd.ni_vp);
1536 error = EEXIST;
1537 goto out;
1538 }
1539 VATTR_NULL(&vattr);
1540 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1541 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1542 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1543 if (error == 0)
1544 vput(nd.ni_vp);
1545 out:
1546 PNBUF_PUT(path);
1547 return (error);
1548 }
1549
1550 /*
1551 * Delete a whiteout from the filesystem.
1552 */
1553 /* ARGSUSED */
1554 int
1555 sys_undelete(l, v, retval)
1556 struct lwp *l;
1557 void *v;
1558 register_t *retval;
1559 {
1560 struct sys_undelete_args /* {
1561 syscallarg(const char *) path;
1562 } */ *uap = v;
1563 struct proc *p = l->l_proc;
1564 int error;
1565 struct nameidata nd;
1566
1567 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1568 SCARG(uap, path), p);
1569 error = namei(&nd);
1570 if (error)
1571 return (error);
1572
1573 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1574 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1575 if (nd.ni_dvp == nd.ni_vp)
1576 vrele(nd.ni_dvp);
1577 else
1578 vput(nd.ni_dvp);
1579 if (nd.ni_vp)
1580 vrele(nd.ni_vp);
1581 return (EEXIST);
1582 }
1583
1584 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1585 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1586 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1587 vput(nd.ni_dvp);
1588 return (error);
1589 }
1590
1591 /*
1592 * Delete a name from the filesystem.
1593 */
1594 /* ARGSUSED */
1595 int
1596 sys_unlink(l, v, retval)
1597 struct lwp *l;
1598 void *v;
1599 register_t *retval;
1600 {
1601 struct sys_unlink_args /* {
1602 syscallarg(const char *) path;
1603 } */ *uap = v;
1604 struct proc *p = l->l_proc;
1605 struct vnode *vp;
1606 int error;
1607 struct nameidata nd;
1608
1609 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1610 SCARG(uap, path), p);
1611 if ((error = namei(&nd)) != 0)
1612 return (error);
1613 vp = nd.ni_vp;
1614
1615 /*
1616 * The root of a mounted filesystem cannot be deleted.
1617 */
1618 if (vp->v_flag & VROOT) {
1619 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1620 if (nd.ni_dvp == vp)
1621 vrele(nd.ni_dvp);
1622 else
1623 vput(nd.ni_dvp);
1624 vput(vp);
1625 error = EBUSY;
1626 goto out;
1627 }
1628
1629 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1630 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1631 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1632 out:
1633 return (error);
1634 }
1635
1636 /*
1637 * Reposition read/write file offset.
1638 */
1639 int
1640 sys_lseek(l, v, retval)
1641 struct lwp *l;
1642 void *v;
1643 register_t *retval;
1644 {
1645 struct sys_lseek_args /* {
1646 syscallarg(int) fd;
1647 syscallarg(int) pad;
1648 syscallarg(off_t) offset;
1649 syscallarg(int) whence;
1650 } */ *uap = v;
1651 struct proc *p = l->l_proc;
1652 struct ucred *cred = p->p_ucred;
1653 struct filedesc *fdp = p->p_fd;
1654 struct file *fp;
1655 struct vnode *vp;
1656 struct vattr vattr;
1657 off_t newoff;
1658 int error;
1659
1660 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1661 return (EBADF);
1662
1663 FILE_USE(fp);
1664
1665 vp = (struct vnode *)fp->f_data;
1666 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1667 error = ESPIPE;
1668 goto out;
1669 }
1670
1671 switch (SCARG(uap, whence)) {
1672 case SEEK_CUR:
1673 newoff = fp->f_offset + SCARG(uap, offset);
1674 break;
1675 case SEEK_END:
1676 error = VOP_GETATTR(vp, &vattr, cred, p);
1677 if (error)
1678 goto out;
1679 newoff = SCARG(uap, offset) + vattr.va_size;
1680 break;
1681 case SEEK_SET:
1682 newoff = SCARG(uap, offset);
1683 break;
1684 default:
1685 error = EINVAL;
1686 goto out;
1687 }
1688 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1689 goto out;
1690
1691 *(off_t *)retval = fp->f_offset = newoff;
1692 out:
1693 FILE_UNUSE(fp, p);
1694 return (error);
1695 }
1696
1697 /*
1698 * Positional read system call.
1699 */
1700 int
1701 sys_pread(l, v, retval)
1702 struct lwp *l;
1703 void *v;
1704 register_t *retval;
1705 {
1706 struct sys_pread_args /* {
1707 syscallarg(int) fd;
1708 syscallarg(void *) buf;
1709 syscallarg(size_t) nbyte;
1710 syscallarg(off_t) offset;
1711 } */ *uap = v;
1712 struct proc *p = l->l_proc;
1713 struct filedesc *fdp = p->p_fd;
1714 struct file *fp;
1715 struct vnode *vp;
1716 off_t offset;
1717 int error, fd = SCARG(uap, fd);
1718
1719 if ((fp = fd_getfile(fdp, fd)) == NULL)
1720 return (EBADF);
1721
1722 if ((fp->f_flag & FREAD) == 0)
1723 return (EBADF);
1724
1725 FILE_USE(fp);
1726
1727 vp = (struct vnode *)fp->f_data;
1728 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1729 error = ESPIPE;
1730 goto out;
1731 }
1732
1733 offset = SCARG(uap, offset);
1734
1735 /*
1736 * XXX This works because no file systems actually
1737 * XXX take any action on the seek operation.
1738 */
1739 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1740 goto out;
1741
1742 /* dofileread() will unuse the descriptor for us */
1743 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1744 &offset, 0, retval));
1745
1746 out:
1747 FILE_UNUSE(fp, p);
1748 return (error);
1749 }
1750
1751 /*
1752 * Positional scatter read system call.
1753 */
1754 int
1755 sys_preadv(l, v, retval)
1756 struct lwp *l;
1757 void *v;
1758 register_t *retval;
1759 {
1760 struct sys_preadv_args /* {
1761 syscallarg(int) fd;
1762 syscallarg(const struct iovec *) iovp;
1763 syscallarg(int) iovcnt;
1764 syscallarg(off_t) offset;
1765 } */ *uap = v;
1766 struct proc *p = l->l_proc;
1767 struct filedesc *fdp = p->p_fd;
1768 struct file *fp;
1769 struct vnode *vp;
1770 off_t offset;
1771 int error, fd = SCARG(uap, fd);
1772
1773 if ((fp = fd_getfile(fdp, fd)) == NULL)
1774 return (EBADF);
1775
1776 if ((fp->f_flag & FREAD) == 0)
1777 return (EBADF);
1778
1779 FILE_USE(fp);
1780
1781 vp = (struct vnode *)fp->f_data;
1782 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1783 error = ESPIPE;
1784 goto out;
1785 }
1786
1787 offset = SCARG(uap, offset);
1788
1789 /*
1790 * XXX This works because no file systems actually
1791 * XXX take any action on the seek operation.
1792 */
1793 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1794 goto out;
1795
1796 /* dofilereadv() will unuse the descriptor for us */
1797 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1798 &offset, 0, retval));
1799
1800 out:
1801 FILE_UNUSE(fp, p);
1802 return (error);
1803 }
1804
1805 /*
1806 * Positional write system call.
1807 */
1808 int
1809 sys_pwrite(l, v, retval)
1810 struct lwp *l;
1811 void *v;
1812 register_t *retval;
1813 {
1814 struct sys_pwrite_args /* {
1815 syscallarg(int) fd;
1816 syscallarg(const void *) buf;
1817 syscallarg(size_t) nbyte;
1818 syscallarg(off_t) offset;
1819 } */ *uap = v;
1820 struct proc *p = l->l_proc;
1821 struct filedesc *fdp = p->p_fd;
1822 struct file *fp;
1823 struct vnode *vp;
1824 off_t offset;
1825 int error, fd = SCARG(uap, fd);
1826
1827 if ((fp = fd_getfile(fdp, fd)) == NULL)
1828 return (EBADF);
1829
1830 if ((fp->f_flag & FWRITE) == 0)
1831 return (EBADF);
1832
1833 FILE_USE(fp);
1834
1835 vp = (struct vnode *)fp->f_data;
1836 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1837 error = ESPIPE;
1838 goto out;
1839 }
1840
1841 offset = SCARG(uap, offset);
1842
1843 /*
1844 * XXX This works because no file systems actually
1845 * XXX take any action on the seek operation.
1846 */
1847 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1848 goto out;
1849
1850 /* dofilewrite() will unuse the descriptor for us */
1851 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1852 &offset, 0, retval));
1853
1854 out:
1855 FILE_UNUSE(fp, p);
1856 return (error);
1857 }
1858
1859 /*
1860 * Positional gather write system call.
1861 */
1862 int
1863 sys_pwritev(l, v, retval)
1864 struct lwp *l;
1865 void *v;
1866 register_t *retval;
1867 {
1868 struct sys_pwritev_args /* {
1869 syscallarg(int) fd;
1870 syscallarg(const struct iovec *) iovp;
1871 syscallarg(int) iovcnt;
1872 syscallarg(off_t) offset;
1873 } */ *uap = v;
1874 struct proc *p = l->l_proc;
1875 struct filedesc *fdp = p->p_fd;
1876 struct file *fp;
1877 struct vnode *vp;
1878 off_t offset;
1879 int error, fd = SCARG(uap, fd);
1880
1881 if ((fp = fd_getfile(fdp, fd)) == NULL)
1882 return (EBADF);
1883
1884 if ((fp->f_flag & FWRITE) == 0)
1885 return (EBADF);
1886
1887 FILE_USE(fp);
1888
1889 vp = (struct vnode *)fp->f_data;
1890 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1891 error = ESPIPE;
1892 goto out;
1893 }
1894
1895 offset = SCARG(uap, offset);
1896
1897 /*
1898 * XXX This works because no file systems actually
1899 * XXX take any action on the seek operation.
1900 */
1901 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1902 goto out;
1903
1904 /* dofilewritev() will unuse the descriptor for us */
1905 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
1906 &offset, 0, retval));
1907
1908 out:
1909 FILE_UNUSE(fp, p);
1910 return (error);
1911 }
1912
1913 /*
1914 * Check access permissions.
1915 */
1916 int
1917 sys_access(l, v, retval)
1918 struct lwp *l;
1919 void *v;
1920 register_t *retval;
1921 {
1922 struct sys_access_args /* {
1923 syscallarg(const char *) path;
1924 syscallarg(int) flags;
1925 } */ *uap = v;
1926 struct ucred *cred = crget();
1927 struct vnode *vp;
1928 int error, flags;
1929 struct nameidata nd;
1930
1931 (void)memcpy(cred, p->p_ucred, sizeof(*cred));
1932 cred->cr_ref = 1;
1933 cred->cr_uid = p->p_cred->p_ruid;
1934 cred->cr_gid = p->p_cred->p_rgid;
1935 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1936 SCARG(uap, path), p);
1937 /* Override default credentials */
1938 nd.ni_cnd.cn_cred = cred;
1939 if ((error = namei(&nd)) != 0)
1940 goto out;
1941 vp = nd.ni_vp;
1942
1943 /* Flags == 0 means only check for existence. */
1944 if (SCARG(uap, flags)) {
1945 flags = 0;
1946 if (SCARG(uap, flags) & R_OK)
1947 flags |= VREAD;
1948 if (SCARG(uap, flags) & W_OK)
1949 flags |= VWRITE;
1950 if (SCARG(uap, flags) & X_OK)
1951 flags |= VEXEC;
1952
1953 error = VOP_ACCESS(vp, flags, cred, p);
1954 if (!error && (flags & VWRITE))
1955 error = vn_writechk(vp);
1956 }
1957 vput(vp);
1958 out:
1959 crfree(cred);
1960 return (error);
1961 }
1962
1963 /*
1964 * Get file status; this version follows links.
1965 */
1966 /* ARGSUSED */
1967 int
1968 sys___stat13(l, v, retval)
1969 struct lwp *l;
1970 void *v;
1971 register_t *retval;
1972 {
1973 struct sys___stat13_args /* {
1974 syscallarg(const char *) path;
1975 syscallarg(struct stat *) ub;
1976 } */ *uap = v;
1977 struct proc *p = l->l_proc;
1978 struct stat sb;
1979 int error;
1980 struct nameidata nd;
1981
1982 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1983 SCARG(uap, path), p);
1984 if ((error = namei(&nd)) != 0)
1985 return (error);
1986 error = vn_stat(nd.ni_vp, &sb, p);
1987 vput(nd.ni_vp);
1988 if (error)
1989 return (error);
1990 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
1991 return (error);
1992 }
1993
1994 /*
1995 * Get file status; this version does not follow links.
1996 */
1997 /* ARGSUSED */
1998 int
1999 sys___lstat13(l, v, retval)
2000 struct lwp *l;
2001 void *v;
2002 register_t *retval;
2003 {
2004 struct sys___lstat13_args /* {
2005 syscallarg(const char *) path;
2006 syscallarg(struct stat *) ub;
2007 } */ *uap = v;
2008 struct proc *p = l->l_proc;
2009 struct stat sb;
2010 int error;
2011 struct nameidata nd;
2012
2013 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2014 SCARG(uap, path), p);
2015 if ((error = namei(&nd)) != 0)
2016 return (error);
2017 error = vn_stat(nd.ni_vp, &sb, p);
2018 vput(nd.ni_vp);
2019 if (error)
2020 return (error);
2021 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2022 return (error);
2023 }
2024
2025 /*
2026 * Get configurable pathname variables.
2027 */
2028 /* ARGSUSED */
2029 int
2030 sys_pathconf(l, v, retval)
2031 struct lwp *l;
2032 void *v;
2033 register_t *retval;
2034 {
2035 struct sys_pathconf_args /* {
2036 syscallarg(const char *) path;
2037 syscallarg(int) name;
2038 } */ *uap = v;
2039 struct proc *p = l->l_proc;
2040 int error;
2041 struct nameidata nd;
2042
2043 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2044 SCARG(uap, path), p);
2045 if ((error = namei(&nd)) != 0)
2046 return (error);
2047 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2048 vput(nd.ni_vp);
2049 return (error);
2050 }
2051
2052 /*
2053 * Return target name of a symbolic link.
2054 */
2055 /* ARGSUSED */
2056 int
2057 sys_readlink(l, v, retval)
2058 struct lwp *l;
2059 void *v;
2060 register_t *retval;
2061 {
2062 struct sys_readlink_args /* {
2063 syscallarg(const char *) path;
2064 syscallarg(char *) buf;
2065 syscallarg(size_t) count;
2066 } */ *uap = v;
2067 struct proc *p = l->l_proc;
2068 struct vnode *vp;
2069 struct iovec aiov;
2070 struct uio auio;
2071 int error;
2072 struct nameidata nd;
2073
2074 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2075 SCARG(uap, path), p);
2076 if ((error = namei(&nd)) != 0)
2077 return (error);
2078 vp = nd.ni_vp;
2079 if (vp->v_type != VLNK)
2080 error = EINVAL;
2081 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2082 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2083 aiov.iov_base = SCARG(uap, buf);
2084 aiov.iov_len = SCARG(uap, count);
2085 auio.uio_iov = &aiov;
2086 auio.uio_iovcnt = 1;
2087 auio.uio_offset = 0;
2088 auio.uio_rw = UIO_READ;
2089 auio.uio_segflg = UIO_USERSPACE;
2090 auio.uio_procp = p;
2091 auio.uio_resid = SCARG(uap, count);
2092 error = VOP_READLINK(vp, &auio, p->p_ucred);
2093 }
2094 vput(vp);
2095 *retval = SCARG(uap, count) - auio.uio_resid;
2096 return (error);
2097 }
2098
2099 /*
2100 * Change flags of a file given a path name.
2101 */
2102 /* ARGSUSED */
2103 int
2104 sys_chflags(l, v, retval)
2105 struct lwp *l;
2106 void *v;
2107 register_t *retval;
2108 {
2109 struct sys_chflags_args /* {
2110 syscallarg(const char *) path;
2111 syscallarg(u_long) flags;
2112 } */ *uap = v;
2113 struct proc *p = l->l_proc;
2114 struct vnode *vp;
2115 int error;
2116 struct nameidata nd;
2117
2118 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2119 if ((error = namei(&nd)) != 0)
2120 return (error);
2121 vp = nd.ni_vp;
2122 error = change_flags(vp, SCARG(uap, flags), p);
2123 vput(vp);
2124 return (error);
2125 }
2126
2127 /*
2128 * Change flags of a file given a file descriptor.
2129 */
2130 /* ARGSUSED */
2131 int
2132 sys_fchflags(l, v, retval)
2133 struct lwp *l;
2134 void *v;
2135 register_t *retval;
2136 {
2137 struct sys_fchflags_args /* {
2138 syscallarg(int) fd;
2139 syscallarg(u_long) flags;
2140 } */ *uap = v;
2141 struct proc *p = l->l_proc;
2142 struct vnode *vp;
2143 struct file *fp;
2144 int error;
2145
2146 /* getvnode() will use the descriptor for us */
2147 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2148 return (error);
2149 vp = (struct vnode *)fp->f_data;
2150 error = change_flags(vp, SCARG(uap, flags), p);
2151 VOP_UNLOCK(vp, 0);
2152 FILE_UNUSE(fp, p);
2153 return (error);
2154 }
2155
2156 /*
2157 * Change flags of a file given a path name; this version does
2158 * not follow links.
2159 */
2160 int
2161 sys_lchflags(l, v, retval)
2162 struct lwp *l;
2163 void *v;
2164 register_t *retval;
2165 {
2166 struct sys_lchflags_args /* {
2167 syscallarg(const char *) path;
2168 syscallarg(u_long) flags;
2169 } */ *uap = v;
2170 struct proc *p = l->l_proc;
2171 struct vnode *vp;
2172 int error;
2173 struct nameidata nd;
2174
2175 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2176 if ((error = namei(&nd)) != 0)
2177 return (error);
2178 vp = nd.ni_vp;
2179 error = change_flags(vp, SCARG(uap, flags), p);
2180 vput(vp);
2181 return (error);
2182 }
2183
2184 /*
2185 * Common routine to change flags of a file.
2186 */
2187 int
2188 change_flags(vp, flags, p)
2189 struct vnode *vp;
2190 u_long flags;
2191 struct proc *p;
2192 {
2193 struct vattr vattr;
2194 int error;
2195
2196 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2197 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2198 /*
2199 * Non-superusers cannot change the flags on devices, even if they
2200 * own them.
2201 */
2202 if (suser(p->p_ucred, &p->p_acflag) != 0) {
2203 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2204 goto out;
2205 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2206 error = EINVAL;
2207 goto out;
2208 }
2209 }
2210 VATTR_NULL(&vattr);
2211 vattr.va_flags = flags;
2212 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2213 out:
2214 return (error);
2215 }
2216
2217 /*
2218 * Change mode of a file given path name; this version follows links.
2219 */
2220 /* ARGSUSED */
2221 int
2222 sys_chmod(l, v, retval)
2223 struct lwp *l;
2224 void *v;
2225 register_t *retval;
2226 {
2227 struct sys_chmod_args /* {
2228 syscallarg(const char *) path;
2229 syscallarg(int) mode;
2230 } */ *uap = v;
2231 struct proc *p = l->l_proc;
2232 int error;
2233 struct nameidata nd;
2234
2235 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2236 if ((error = namei(&nd)) != 0)
2237 return (error);
2238
2239 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2240
2241 vrele(nd.ni_vp);
2242 return (error);
2243 }
2244
2245 /*
2246 * Change mode of a file given a file descriptor.
2247 */
2248 /* ARGSUSED */
2249 int
2250 sys_fchmod(l, v, retval)
2251 struct lwp *l;
2252 void *v;
2253 register_t *retval;
2254 {
2255 struct sys_fchmod_args /* {
2256 syscallarg(int) fd;
2257 syscallarg(int) mode;
2258 } */ *uap = v;
2259 struct proc *p = l->l_proc;
2260 struct file *fp;
2261 int error;
2262
2263 /* getvnode() will use the descriptor for us */
2264 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2265 return (error);
2266
2267 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2268 FILE_UNUSE(fp, p);
2269 return (error);
2270 }
2271
2272 /*
2273 * Change mode of a file given path name; this version does not follow links.
2274 */
2275 /* ARGSUSED */
2276 int
2277 sys_lchmod(l, v, retval)
2278 struct lwp *l;
2279 void *v;
2280 register_t *retval;
2281 {
2282 struct sys_lchmod_args /* {
2283 syscallarg(const char *) path;
2284 syscallarg(int) mode;
2285 } */ *uap = v;
2286 struct proc *p = l->l_proc;
2287 int error;
2288 struct nameidata nd;
2289
2290 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2291 if ((error = namei(&nd)) != 0)
2292 return (error);
2293
2294 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2295
2296 vrele(nd.ni_vp);
2297 return (error);
2298 }
2299
2300 /*
2301 * Common routine to set mode given a vnode.
2302 */
2303 static int
2304 change_mode(vp, mode, p)
2305 struct vnode *vp;
2306 int mode;
2307 struct proc *p;
2308 {
2309 struct vattr vattr;
2310 int error;
2311
2312 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2313 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2314 VATTR_NULL(&vattr);
2315 vattr.va_mode = mode & ALLPERMS;
2316 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2317 VOP_UNLOCK(vp, 0);
2318 return (error);
2319 }
2320
2321 /*
2322 * Set ownership given a path name; this version follows links.
2323 */
2324 /* ARGSUSED */
2325 int
2326 sys_chown(l, v, retval)
2327 struct lwp *l;
2328 void *v;
2329 register_t *retval;
2330 {
2331 struct sys_chown_args /* {
2332 syscallarg(const char *) path;
2333 syscallarg(uid_t) uid;
2334 syscallarg(gid_t) gid;
2335 } */ *uap = v;
2336 struct proc *p = l->l_proc;
2337 int error;
2338 struct nameidata nd;
2339
2340 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2341 if ((error = namei(&nd)) != 0)
2342 return (error);
2343
2344 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2345
2346 vrele(nd.ni_vp);
2347 return (error);
2348 }
2349
2350 /*
2351 * Set ownership given a path name; this version follows links.
2352 * Provides POSIX semantics.
2353 */
2354 /* ARGSUSED */
2355 int
2356 sys___posix_chown(l, v, retval)
2357 struct lwp *l;
2358 void *v;
2359 register_t *retval;
2360 {
2361 struct sys_chown_args /* {
2362 syscallarg(const char *) path;
2363 syscallarg(uid_t) uid;
2364 syscallarg(gid_t) gid;
2365 } */ *uap = v;
2366 struct proc *p = l->l_proc;
2367 int error;
2368 struct nameidata nd;
2369
2370 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2371 if ((error = namei(&nd)) != 0)
2372 return (error);
2373
2374 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2375
2376 vrele(nd.ni_vp);
2377 return (error);
2378 }
2379
2380 /*
2381 * Set ownership given a file descriptor.
2382 */
2383 /* ARGSUSED */
2384 int
2385 sys_fchown(l, v, retval)
2386 struct lwp *l;
2387 void *v;
2388 register_t *retval;
2389 {
2390 struct sys_fchown_args /* {
2391 syscallarg(int) fd;
2392 syscallarg(uid_t) uid;
2393 syscallarg(gid_t) gid;
2394 } */ *uap = v;
2395 struct proc *p = l->l_proc;
2396 int error;
2397 struct file *fp;
2398
2399 /* getvnode() will use the descriptor for us */
2400 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2401 return (error);
2402
2403 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2404 SCARG(uap, gid), p, 0);
2405 FILE_UNUSE(fp, p);
2406 return (error);
2407 }
2408
2409 /*
2410 * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2411 */
2412 /* ARGSUSED */
2413 int
2414 sys___posix_fchown(l, v, retval)
2415 struct lwp *l;
2416 void *v;
2417 register_t *retval;
2418 {
2419 struct sys_fchown_args /* {
2420 syscallarg(int) fd;
2421 syscallarg(uid_t) uid;
2422 syscallarg(gid_t) gid;
2423 } */ *uap = v;
2424 struct proc *p = l->l_proc;
2425 int error;
2426 struct file *fp;
2427
2428 /* getvnode() will use the descriptor for us */
2429 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2430 return (error);
2431
2432 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2433 SCARG(uap, gid), p, 1);
2434 FILE_UNUSE(fp, p);
2435 return (error);
2436 }
2437
2438 /*
2439 * Set ownership given a path name; this version does not follow links.
2440 */
2441 /* ARGSUSED */
2442 int
2443 sys_lchown(l, v, retval)
2444 struct lwp *l;
2445 void *v;
2446 register_t *retval;
2447 {
2448 struct sys_lchown_args /* {
2449 syscallarg(const char *) path;
2450 syscallarg(uid_t) uid;
2451 syscallarg(gid_t) gid;
2452 } */ *uap = v;
2453 struct proc *p = l->l_proc;
2454 int error;
2455 struct nameidata nd;
2456
2457 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2458 if ((error = namei(&nd)) != 0)
2459 return (error);
2460
2461 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2462
2463 vrele(nd.ni_vp);
2464 return (error);
2465 }
2466
2467 /*
2468 * Set ownership given a path name; this version does not follow links.
2469 * Provides POSIX/XPG semantics.
2470 */
2471 /* ARGSUSED */
2472 int
2473 sys___posix_lchown(l, v, retval)
2474 struct lwp *l;
2475 void *v;
2476 register_t *retval;
2477 {
2478 struct sys_lchown_args /* {
2479 syscallarg(const char *) path;
2480 syscallarg(uid_t) uid;
2481 syscallarg(gid_t) gid;
2482 } */ *uap = v;
2483 struct proc *p = l->l_proc;
2484 int error;
2485 struct nameidata nd;
2486
2487 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2488 if ((error = namei(&nd)) != 0)
2489 return (error);
2490
2491 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2492
2493 vrele(nd.ni_vp);
2494 return (error);
2495 }
2496
2497 /*
2498 * Common routine to set ownership given a vnode.
2499 */
2500 static int
2501 change_owner(vp, uid, gid, p, posix_semantics)
2502 struct vnode *vp;
2503 uid_t uid;
2504 gid_t gid;
2505 struct proc *p;
2506 int posix_semantics;
2507 {
2508 struct vattr vattr;
2509 mode_t newmode;
2510 int error;
2511
2512 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2513 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2514 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2515 goto out;
2516
2517 #define CHANGED(x) ((x) != -1)
2518 newmode = vattr.va_mode;
2519 if (posix_semantics) {
2520 /*
2521 * POSIX/XPG semantics: if the caller is not the super-user,
2522 * clear set-user-id and set-group-id bits. Both POSIX and
2523 * the XPG consider the behaviour for calls by the super-user
2524 * implementation-defined; we leave the set-user-id and set-
2525 * group-id settings intact in that case.
2526 */
2527 if (suser(p->p_ucred, NULL) != 0)
2528 newmode &= ~(S_ISUID | S_ISGID);
2529 } else {
2530 /*
2531 * NetBSD semantics: when changing owner and/or group,
2532 * clear the respective bit(s).
2533 */
2534 if (CHANGED(uid))
2535 newmode &= ~S_ISUID;
2536 if (CHANGED(gid))
2537 newmode &= ~S_ISGID;
2538 }
2539 /* Update va_mode iff altered. */
2540 if (vattr.va_mode == newmode)
2541 newmode = VNOVAL;
2542
2543 VATTR_NULL(&vattr);
2544 vattr.va_uid = CHANGED(uid) ? uid : VNOVAL;
2545 vattr.va_gid = CHANGED(gid) ? gid : VNOVAL;
2546 vattr.va_mode = newmode;
2547 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2548 #undef CHANGED
2549
2550 out:
2551 VOP_UNLOCK(vp, 0);
2552 return (error);
2553 }
2554
2555 /*
2556 * Set the access and modification times given a path name; this
2557 * version follows links.
2558 */
2559 /* ARGSUSED */
2560 int
2561 sys_utimes(l, v, retval)
2562 struct lwp *l;
2563 void *v;
2564 register_t *retval;
2565 {
2566 struct sys_utimes_args /* {
2567 syscallarg(const char *) path;
2568 syscallarg(const struct timeval *) tptr;
2569 } */ *uap = v;
2570 struct proc *p = l->l_proc;
2571 int error;
2572 struct nameidata nd;
2573
2574 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2575 if ((error = namei(&nd)) != 0)
2576 return (error);
2577
2578 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2579
2580 vrele(nd.ni_vp);
2581 return (error);
2582 }
2583
2584 /*
2585 * Set the access and modification times given a file descriptor.
2586 */
2587 /* ARGSUSED */
2588 int
2589 sys_futimes(l, v, retval)
2590 struct lwp *l;
2591 void *v;
2592 register_t *retval;
2593 {
2594 struct sys_futimes_args /* {
2595 syscallarg(int) fd;
2596 syscallarg(const struct timeval *) tptr;
2597 } */ *uap = v;
2598 struct proc *p = l->l_proc;
2599 int error;
2600 struct file *fp;
2601
2602 /* getvnode() will use the descriptor for us */
2603 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2604 return (error);
2605
2606 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2607 FILE_UNUSE(fp, p);
2608 return (error);
2609 }
2610
2611 /*
2612 * Set the access and modification times given a path name; this
2613 * version does not follow links.
2614 */
2615 /* ARGSUSED */
2616 int
2617 sys_lutimes(l, v, retval)
2618 struct lwp *l;
2619 void *v;
2620 register_t *retval;
2621 {
2622 struct sys_lutimes_args /* {
2623 syscallarg(const char *) path;
2624 syscallarg(const struct timeval *) tptr;
2625 } */ *uap = v;
2626 struct proc *p = l->l_proc;
2627 int error;
2628 struct nameidata nd;
2629
2630 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2631 if ((error = namei(&nd)) != 0)
2632 return (error);
2633
2634 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2635
2636 vrele(nd.ni_vp);
2637 return (error);
2638 }
2639
2640 /*
2641 * Common routine to set access and modification times given a vnode.
2642 */
2643 static int
2644 change_utimes(vp, tptr, p)
2645 struct vnode *vp;
2646 const struct timeval *tptr;
2647 struct proc *p;
2648 {
2649 struct timeval tv[2];
2650 struct vattr vattr;
2651 int error;
2652
2653 VATTR_NULL(&vattr);
2654 if (tptr == NULL) {
2655 microtime(&tv[0]);
2656 tv[1] = tv[0];
2657 vattr.va_vaflags |= VA_UTIMES_NULL;
2658 } else {
2659 error = copyin(tptr, tv, sizeof(tv));
2660 if (error)
2661 return (error);
2662 }
2663 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2664 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2665 vattr.va_atime.tv_sec = tv[0].tv_sec;
2666 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2667 vattr.va_mtime.tv_sec = tv[1].tv_sec;
2668 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2669 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2670 VOP_UNLOCK(vp, 0);
2671 return (error);
2672 }
2673
2674 /*
2675 * Truncate a file given its path name.
2676 */
2677 /* ARGSUSED */
2678 int
2679 sys_truncate(l, v, retval)
2680 struct lwp *l;
2681 void *v;
2682 register_t *retval;
2683 {
2684 struct sys_truncate_args /* {
2685 syscallarg(const char *) path;
2686 syscallarg(int) pad;
2687 syscallarg(off_t) length;
2688 } */ *uap = v;
2689 struct proc *p = l->l_proc;
2690 struct vnode *vp;
2691 struct vattr vattr;
2692 int error;
2693 struct nameidata nd;
2694
2695 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2696 if ((error = namei(&nd)) != 0)
2697 return (error);
2698 vp = nd.ni_vp;
2699 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2700 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2701 if (vp->v_type == VDIR)
2702 error = EISDIR;
2703 else if ((error = vn_writechk(vp)) == 0 &&
2704 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2705 VATTR_NULL(&vattr);
2706 vattr.va_size = SCARG(uap, length);
2707 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2708 }
2709 vput(vp);
2710 return (error);
2711 }
2712
2713 /*
2714 * Truncate a file given a file descriptor.
2715 */
2716 /* ARGSUSED */
2717 int
2718 sys_ftruncate(l, v, retval)
2719 struct lwp *l;
2720 void *v;
2721 register_t *retval;
2722 {
2723 struct sys_ftruncate_args /* {
2724 syscallarg(int) fd;
2725 syscallarg(int) pad;
2726 syscallarg(off_t) length;
2727 } */ *uap = v;
2728 struct proc *p = l->l_proc;
2729 struct vattr vattr;
2730 struct vnode *vp;
2731 struct file *fp;
2732 int error;
2733
2734 /* getvnode() will use the descriptor for us */
2735 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2736 return (error);
2737 if ((fp->f_flag & FWRITE) == 0) {
2738 error = EINVAL;
2739 goto out;
2740 }
2741 vp = (struct vnode *)fp->f_data;
2742 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2743 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2744 if (vp->v_type == VDIR)
2745 error = EISDIR;
2746 else if ((error = vn_writechk(vp)) == 0) {
2747 VATTR_NULL(&vattr);
2748 vattr.va_size = SCARG(uap, length);
2749 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2750 }
2751 VOP_UNLOCK(vp, 0);
2752 out:
2753 FILE_UNUSE(fp, p);
2754 return (error);
2755 }
2756
2757 /*
2758 * Sync an open file.
2759 */
2760 /* ARGSUSED */
2761 int
2762 sys_fsync(l, v, retval)
2763 struct lwp *l;
2764 void *v;
2765 register_t *retval;
2766 {
2767 struct sys_fsync_args /* {
2768 syscallarg(int) fd;
2769 } */ *uap = v;
2770 struct proc *p = l->l_proc;
2771 struct vnode *vp;
2772 struct file *fp;
2773 int error;
2774
2775 /* getvnode() will use the descriptor for us */
2776 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2777 return (error);
2778 vp = (struct vnode *)fp->f_data;
2779 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2780 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
2781 if (error == 0 && bioops.io_fsync != NULL &&
2782 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2783 (*bioops.io_fsync)(vp);
2784 VOP_UNLOCK(vp, 0);
2785 FILE_UNUSE(fp, p);
2786 return (error);
2787 }
2788
2789 /*
2790 * Sync the data of an open file.
2791 */
2792 /* ARGSUSED */
2793 int
2794 sys_fdatasync(l, v, retval)
2795 struct lwp *l;
2796 void *v;
2797 register_t *retval;
2798 {
2799 struct sys_fdatasync_args /* {
2800 syscallarg(int) fd;
2801 } */ *uap = v;
2802 struct proc *p = l->l_proc;
2803 struct vnode *vp;
2804 struct file *fp;
2805 int error;
2806
2807 /* getvnode() will use the descriptor for us */
2808 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2809 return (error);
2810 vp = (struct vnode *)fp->f_data;
2811 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2812 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
2813 VOP_UNLOCK(vp, 0);
2814 FILE_UNUSE(fp, p);
2815 return (error);
2816 }
2817
2818 /*
2819 * Rename files, (standard) BSD semantics frontend.
2820 */
2821 /* ARGSUSED */
2822 int
2823 sys_rename(l, v, retval)
2824 struct lwp *l;
2825 void *v;
2826 register_t *retval;
2827 {
2828 struct sys_rename_args /* {
2829 syscallarg(const char *) from;
2830 syscallarg(const char *) to;
2831 } */ *uap = v;
2832 struct proc *p = l->l_proc;
2833
2834 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
2835 }
2836
2837 /*
2838 * Rename files, POSIX semantics frontend.
2839 */
2840 /* ARGSUSED */
2841 int
2842 sys___posix_rename(l, v, retval)
2843 struct lwp *l;
2844 void *v;
2845 register_t *retval;
2846 {
2847 struct sys___posix_rename_args /* {
2848 syscallarg(const char *) from;
2849 syscallarg(const char *) to;
2850 } */ *uap = v;
2851 struct proc *p = l->l_proc;
2852
2853 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
2854 }
2855
2856 /*
2857 * Rename files. Source and destination must either both be directories,
2858 * or both not be directories. If target is a directory, it must be empty.
2859 * If `from' and `to' refer to the same object, the value of the `retain'
2860 * argument is used to determine whether `from' will be
2861 *
2862 * (retain == 0) deleted unless `from' and `to' refer to the same
2863 * object in the file system's name space (BSD).
2864 * (retain == 1) always retained (POSIX).
2865 */
2866 static int
2867 rename_files(from, to, p, retain)
2868 const char *from, *to;
2869 struct proc *p;
2870 int retain;
2871 {
2872 struct vnode *tvp, *fvp, *tdvp;
2873 struct nameidata fromnd, tond;
2874 int error;
2875
2876 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2877 from, p);
2878 if ((error = namei(&fromnd)) != 0)
2879 return (error);
2880 fvp = fromnd.ni_vp;
2881 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART,
2882 UIO_USERSPACE, to, p);
2883 if ((error = namei(&tond)) != 0) {
2884 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2885 vrele(fromnd.ni_dvp);
2886 vrele(fvp);
2887 goto out1;
2888 }
2889 tdvp = tond.ni_dvp;
2890 tvp = tond.ni_vp;
2891
2892 if (tvp != NULL) {
2893 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2894 error = ENOTDIR;
2895 goto out;
2896 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2897 error = EISDIR;
2898 goto out;
2899 }
2900 }
2901
2902 if (fvp == tdvp)
2903 error = EINVAL;
2904
2905 /*
2906 * Source and destination refer to the same object.
2907 */
2908 if (fvp == tvp) {
2909 if (retain)
2910 error = -1;
2911 else if (fromnd.ni_dvp == tdvp &&
2912 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2913 !memcmp(fromnd.ni_cnd.cn_nameptr,
2914 tond.ni_cnd.cn_nameptr,
2915 fromnd.ni_cnd.cn_namelen))
2916 error = -1;
2917 }
2918
2919 out:
2920 if (!error) {
2921 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2922 if (fromnd.ni_dvp != tdvp)
2923 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2924 if (tvp) {
2925 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2926 }
2927 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2928 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2929 } else {
2930 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2931 if (tdvp == tvp)
2932 vrele(tdvp);
2933 else
2934 vput(tdvp);
2935 if (tvp)
2936 vput(tvp);
2937 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2938 vrele(fromnd.ni_dvp);
2939 vrele(fvp);
2940 }
2941 vrele(tond.ni_startdir);
2942 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2943 out1:
2944 if (fromnd.ni_startdir)
2945 vrele(fromnd.ni_startdir);
2946 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2947 return (error == -1 ? 0 : error);
2948 }
2949
2950 /*
2951 * Make a directory file.
2952 */
2953 /* ARGSUSED */
2954 int
2955 sys_mkdir(l, v, retval)
2956 struct lwp *l;
2957 void *v;
2958 register_t *retval;
2959 {
2960 struct sys_mkdir_args /* {
2961 syscallarg(const char *) path;
2962 syscallarg(int) mode;
2963 } */ *uap = v;
2964 struct proc *p = l->l_proc;
2965 struct vnode *vp;
2966 struct vattr vattr;
2967 int error;
2968 struct nameidata nd;
2969
2970 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
2971 if ((error = namei(&nd)) != 0)
2972 return (error);
2973 vp = nd.ni_vp;
2974 if (vp != NULL) {
2975 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2976 if (nd.ni_dvp == vp)
2977 vrele(nd.ni_dvp);
2978 else
2979 vput(nd.ni_dvp);
2980 vrele(vp);
2981 return (EEXIST);
2982 }
2983 VATTR_NULL(&vattr);
2984 vattr.va_type = VDIR;
2985 vattr.va_mode =
2986 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
2987 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2988 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2989 if (!error)
2990 vput(nd.ni_vp);
2991 return (error);
2992 }
2993
2994 /*
2995 * Remove a directory file.
2996 */
2997 /* ARGSUSED */
2998 int
2999 sys_rmdir(l, v, retval)
3000 struct lwp *l;
3001 void *v;
3002 register_t *retval;
3003 {
3004 struct sys_rmdir_args /* {
3005 syscallarg(const char *) path;
3006 } */ *uap = v;
3007 struct proc *p = l->l_proc;
3008 struct vnode *vp;
3009 int error;
3010 struct nameidata nd;
3011
3012 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3013 SCARG(uap, path), p);
3014 if ((error = namei(&nd)) != 0)
3015 return (error);
3016 vp = nd.ni_vp;
3017 if (vp->v_type != VDIR) {
3018 error = ENOTDIR;
3019 goto out;
3020 }
3021 /*
3022 * No rmdir "." please.
3023 */
3024 if (nd.ni_dvp == vp) {
3025 error = EINVAL;
3026 goto out;
3027 }
3028 /*
3029 * The root of a mounted filesystem cannot be deleted.
3030 */
3031 if (vp->v_flag & VROOT)
3032 error = EBUSY;
3033 out:
3034 if (!error) {
3035 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3036 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3037 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3038 } else {
3039 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3040 if (nd.ni_dvp == vp)
3041 vrele(nd.ni_dvp);
3042 else
3043 vput(nd.ni_dvp);
3044 vput(vp);
3045 }
3046 return (error);
3047 }
3048
3049 /*
3050 * Read a block of directory entries in a file system independent format.
3051 */
3052 int
3053 sys_getdents(l, v, retval)
3054 struct lwp *l;
3055 void *v;
3056 register_t *retval;
3057 {
3058 struct sys_getdents_args /* {
3059 syscallarg(int) fd;
3060 syscallarg(char *) buf;
3061 syscallarg(size_t) count;
3062 } */ *uap = v;
3063 struct proc *p = l->l_proc;
3064 struct file *fp;
3065 int error, done;
3066
3067 /* getvnode() will use the descriptor for us */
3068 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3069 return (error);
3070 if ((fp->f_flag & FREAD) == 0) {
3071 error = EBADF;
3072 goto out;
3073 }
3074 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3075 SCARG(uap, count), &done, p, 0, 0);
3076 *retval = done;
3077 out:
3078 FILE_UNUSE(fp, p);
3079 return (error);
3080 }
3081
3082 /*
3083 * Set the mode mask for creation of filesystem nodes.
3084 */
3085 int
3086 sys_umask(l, v, retval)
3087 struct lwp *l;
3088 void *v;
3089 register_t *retval;
3090 {
3091 struct sys_umask_args /* {
3092 syscallarg(mode_t) newmask;
3093 } */ *uap = v;
3094 struct proc *p = l->l_proc;
3095 struct cwdinfo *cwdi;
3096
3097 cwdi = p->p_cwdi;
3098 *retval = cwdi->cwdi_cmask;
3099 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3100 return (0);
3101 }
3102
3103 /*
3104 * Void all references to file by ripping underlying filesystem
3105 * away from vnode.
3106 */
3107 /* ARGSUSED */
3108 int
3109 sys_revoke(l, v, retval)
3110 struct lwp *l;
3111 void *v;
3112 register_t *retval;
3113 {
3114 struct sys_revoke_args /* {
3115 syscallarg(const char *) path;
3116 } */ *uap = v;
3117 struct proc *p = l->l_proc;
3118 struct vnode *vp;
3119 struct vattr vattr;
3120 int error;
3121 struct nameidata nd;
3122
3123 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3124 if ((error = namei(&nd)) != 0)
3125 return (error);
3126 vp = nd.ni_vp;
3127 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3128 goto out;
3129 if (p->p_ucred->cr_uid != vattr.va_uid &&
3130 (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3131 goto out;
3132 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3133 VOP_REVOKE(vp, REVOKEALL);
3134 out:
3135 vrele(vp);
3136 return (error);
3137 }
3138
3139 /*
3140 * Convert a user file descriptor to a kernel file entry.
3141 */
3142 int
3143 getvnode(fdp, fd, fpp)
3144 struct filedesc *fdp;
3145 int fd;
3146 struct file **fpp;
3147 {
3148 struct vnode *vp;
3149 struct file *fp;
3150
3151 if ((fp = fd_getfile(fdp, fd)) == NULL)
3152 return (EBADF);
3153
3154 FILE_USE(fp);
3155
3156 if (fp->f_type != DTYPE_VNODE) {
3157 FILE_UNUSE(fp, NULL);
3158 return (EINVAL);
3159 }
3160
3161 vp = (struct vnode *)fp->f_data;
3162 if (vp->v_type == VBAD) {
3163 FILE_UNUSE(fp, NULL);
3164 return (EBADF);
3165 }
3166
3167 *fpp = fp;
3168 return (0);
3169 }
3170