vfs_syscalls.c revision 1.217.2.3 1 /* $NetBSD: vfs_syscalls.c,v 1.217.2.3 2005/07/02 15:51:26 tron 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. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95
37 */
38
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.217.2.3 2005/07/02 15:51:26 tron Exp $");
41
42 #include "opt_compat_netbsd.h"
43 #include "opt_compat_43.h"
44 #include "opt_ktrace.h"
45 #include "fss.h"
46
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/namei.h>
50 #include <sys/filedesc.h>
51 #include <sys/kernel.h>
52 #include <sys/file.h>
53 #include <sys/stat.h>
54 #include <sys/vnode.h>
55 #include <sys/mount.h>
56 #include <sys/proc.h>
57 #include <sys/uio.h>
58 #include <sys/malloc.h>
59 #include <sys/dirent.h>
60 #include <sys/extattr.h>
61 #include <sys/sysctl.h>
62 #include <sys/sa.h>
63 #include <sys/syscallargs.h>
64 #ifdef KTRACE
65 #include <sys/ktrace.h>
66 #endif
67 #ifdef VERIFIED_EXEC
68 #include <sys/verified_exec.h>
69 #endif
70
71 #include <miscfs/genfs/genfs.h>
72 #include <miscfs/syncfs/syncfs.h>
73
74 #if NFSS > 0
75 #include <dev/fssvar.h>
76 #endif
77
78 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
79
80 static int change_dir(struct nameidata *, struct proc *);
81 static int change_flags(struct vnode *, u_long, struct proc *);
82 static int change_mode(struct vnode *, int, struct proc *p);
83 static int change_owner(struct vnode *, uid_t, gid_t, struct proc *, int);
84 static int change_utimes(struct vnode *vp, const struct timeval *,
85 struct proc *p);
86 static int rename_files(const char *, const char *, struct proc *, int);
87
88 void checkdirs(struct vnode *);
89
90 int dovfsusermount = 0;
91
92 /*
93 * Virtual File System System Calls
94 */
95
96 /*
97 * Mount a file system.
98 */
99
100 #if defined(COMPAT_09) || defined(COMPAT_43)
101 /*
102 * This table is used to maintain compatibility with 4.3BSD
103 * and NetBSD 0.9 mount syscalls. Note, the order is important!
104 *
105 * Do not modify this table. It should only contain filesystems
106 * supported by NetBSD 0.9 and 4.3BSD.
107 */
108 const char * const mountcompatnames[] = {
109 NULL, /* 0 = MOUNT_NONE */
110 MOUNT_FFS, /* 1 = MOUNT_UFS */
111 MOUNT_NFS, /* 2 */
112 MOUNT_MFS, /* 3 */
113 MOUNT_MSDOS, /* 4 */
114 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */
115 MOUNT_FDESC, /* 6 */
116 MOUNT_KERNFS, /* 7 */
117 NULL, /* 8 = MOUNT_DEVFS */
118 MOUNT_AFS, /* 9 */
119 };
120 const int nmountcompatnames = sizeof(mountcompatnames) /
121 sizeof(mountcompatnames[0]);
122 #endif /* COMPAT_09 || COMPAT_43 */
123
124 /* ARGSUSED */
125 int
126 sys_mount(l, v, retval)
127 struct lwp *l;
128 void *v;
129 register_t *retval;
130 {
131 struct sys_mount_args /* {
132 syscallarg(const char *) type;
133 syscallarg(const char *) path;
134 syscallarg(int) flags;
135 syscallarg(void *) data;
136 } */ *uap = v;
137 struct proc *p = l->l_proc;
138 struct vnode *vp;
139 struct mount *mp;
140 int error, flag = 0;
141 char fstypename[MFSNAMELEN];
142 struct vattr va;
143 struct nameidata nd;
144 struct vfsops *vfs;
145
146 /*
147 * if MNT_GETARGS is specified, it should be only flag.
148 */
149
150 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 &&
151 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) {
152 return EINVAL;
153 }
154
155 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 &&
156 (error = suser(p->p_ucred, &p->p_acflag)))
157 return (error);
158 /*
159 * Get vnode to be covered
160 */
161 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE,
162 SCARG(uap, path), p);
163 if ((error = namei(&nd)) != 0)
164 return (error);
165 vp = nd.ni_vp;
166 /*
167 * A lookup in VFS_MOUNT might result in an attempt to
168 * lock this vnode again, so make the lock recursive.
169 */
170 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
171 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) {
172 if ((vp->v_flag & VROOT) == 0) {
173 vput(vp);
174 return (EINVAL);
175 }
176 mp = vp->v_mount;
177 flag = mp->mnt_flag;
178 vfs = mp->mnt_op;
179 /*
180 * We only allow the filesystem to be reloaded if it
181 * is currently mounted read-only.
182 */
183 if ((SCARG(uap, flags) & MNT_RELOAD) &&
184 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
185 vput(vp);
186 return (EOPNOTSUPP); /* Needs translation */
187 }
188 /*
189 * In "highly secure" mode, don't let the caller do anything
190 * but downgrade a filesystem from read-write to read-only.
191 * (see also below; MNT_UPDATE or MNT_GETARGS is required.)
192 */
193 if (securelevel >= 2 &&
194 SCARG(uap, flags) != MNT_GETARGS &&
195 SCARG(uap, flags) !=
196 (mp->mnt_flag | MNT_RDONLY |
197 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) {
198 vput(vp);
199 return (EPERM);
200 }
201 mp->mnt_flag |= SCARG(uap, flags) &
202 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
203 /*
204 * Only root, or the user that did the original mount is
205 * permitted to update it.
206 */
207 if ((mp->mnt_flag & MNT_GETARGS) == 0 &&
208 mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
209 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
210 vput(vp);
211 return (error);
212 }
213 /*
214 * Do not allow NFS export by non-root users. For non-root
215 * users, silently enforce MNT_NOSUID and MNT_NODEV, and
216 * MNT_NOEXEC if mount point is already MNT_NOEXEC.
217 */
218 if (p->p_ucred->cr_uid != 0) {
219 if (SCARG(uap, flags) & MNT_EXPORTED) {
220 vput(vp);
221 return (EPERM);
222 }
223 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
224 if (flag & MNT_NOEXEC)
225 SCARG(uap, flags) |= MNT_NOEXEC;
226 }
227 if (vfs_busy(mp, LK_NOWAIT, 0)) {
228 vput(vp);
229 return (EPERM);
230 }
231 goto update;
232 } else {
233 if (securelevel >= 2) {
234 vput(vp);
235 return (EPERM);
236 }
237 }
238 /*
239 * If the user is not root, ensure that they own the directory
240 * onto which we are attempting to mount.
241 */
242 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 ||
243 (va.va_uid != p->p_ucred->cr_uid &&
244 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) {
245 vput(vp);
246 return (error);
247 }
248 /*
249 * Do not allow NFS export by non-root users. For non-root users,
250 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the
251 * mount point is already MNT_NOEXEC.
252 */
253 if (p->p_ucred->cr_uid != 0) {
254 if (SCARG(uap, flags) & MNT_EXPORTED) {
255 vput(vp);
256 return (EPERM);
257 }
258 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
259 if (vp->v_mount->mnt_flag & MNT_NOEXEC)
260 SCARG(uap, flags) |= MNT_NOEXEC;
261 }
262 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) {
263 vput(vp);
264 return (error);
265 }
266 if (vp->v_type != VDIR) {
267 vput(vp);
268 return (ENOTDIR);
269 }
270 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
271 if (error) {
272 #if defined(COMPAT_09) || defined(COMPAT_43)
273 /*
274 * Historically filesystem types were identified by number.
275 * If we get an integer for the filesystem type instead of a
276 * string, we check to see if it matches one of the historic
277 * filesystem types.
278 */
279 u_long fsindex = (u_long)SCARG(uap, type);
280 if (fsindex >= nmountcompatnames ||
281 mountcompatnames[fsindex] == NULL) {
282 vput(vp);
283 return (ENODEV);
284 }
285 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN);
286 #else
287 vput(vp);
288 return (error);
289 #endif
290 }
291 #ifdef COMPAT_10
292 /* Accept `ufs' as an alias for `ffs'. */
293 if (!strncmp(fstypename, "ufs", MFSNAMELEN))
294 strncpy(fstypename, "ffs", MFSNAMELEN);
295 #endif
296 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) {
297 vput(vp);
298 return (ENODEV);
299 }
300 if (vp->v_mountedhere != NULL) {
301 vput(vp);
302 return (EBUSY);
303 }
304
305 /*
306 * Allocate and initialize the file system.
307 */
308 mp = (struct mount *)malloc((u_long)sizeof(struct mount),
309 M_MOUNT, M_WAITOK);
310 memset((char *)mp, 0, (u_long)sizeof(struct mount));
311 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
312 simple_lock_init(&mp->mnt_slock);
313 (void)vfs_busy(mp, LK_NOWAIT, 0);
314 mp->mnt_op = vfs;
315 vfs->vfs_refcount++;
316 mp->mnt_vnodecovered = vp;
317 mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
318 mp->mnt_unmounter = NULL;
319 mp->mnt_leaf = mp;
320
321 /*
322 * The underlying file system may refuse the mount for
323 * various reasons. Allow the user to force it to happen.
324 */
325 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE;
326 update:
327 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) {
328 /*
329 * Set the mount level flags.
330 */
331 if (SCARG(uap, flags) & MNT_RDONLY)
332 mp->mnt_flag |= MNT_RDONLY;
333 else if (mp->mnt_flag & MNT_RDONLY)
334 mp->mnt_iflag |= IMNT_WANTRDWR;
335 mp->mnt_flag &=
336 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
337 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
338 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
339 mp->mnt_flag |= SCARG(uap, flags) &
340 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
341 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
342 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
343 MNT_IGNORE);
344 }
345 /*
346 * Mount the filesystem.
347 */
348 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
349 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) {
350 if (mp->mnt_iflag & IMNT_WANTRDWR)
351 mp->mnt_flag &= ~MNT_RDONLY;
352 if (error)
353 mp->mnt_flag = flag;
354 mp->mnt_flag &=~
355 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
356 mp->mnt_iflag &=~ IMNT_WANTRDWR;
357 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
358 if (mp->mnt_syncer == NULL)
359 error = vfs_allocate_syncvnode(mp);
360 } else {
361 if (mp->mnt_syncer != NULL)
362 vfs_deallocate_syncvnode(mp);
363 }
364 vfs_unbusy(mp);
365 VOP_UNLOCK(vp, 0);
366 vrele(vp);
367 return (error);
368 }
369 /*
370 * Put the new filesystem on the mount list after root.
371 */
372 cache_purge(vp);
373 if (!error) {
374 mp->mnt_flag &=~
375 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS);
376 mp->mnt_iflag &=~ IMNT_WANTRDWR;
377 vp->v_mountedhere = mp;
378 simple_lock(&mountlist_slock);
379 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
380 simple_unlock(&mountlist_slock);
381 checkdirs(vp);
382 VOP_UNLOCK(vp, 0);
383 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
384 error = vfs_allocate_syncvnode(mp);
385 vfs_unbusy(mp);
386 (void) VFS_STATVFS(mp, &mp->mnt_stat, p);
387 if ((error = VFS_START(mp, 0, p)))
388 vrele(vp);
389 } else {
390 vp->v_mountedhere = (struct mount *)0;
391 vfs->vfs_refcount--;
392 vfs_unbusy(mp);
393 free(mp, M_MOUNT);
394 vput(vp);
395 }
396 return (error);
397 }
398
399 /*
400 * Scan all active processes to see if any of them have a current
401 * or root directory onto which the new filesystem has just been
402 * mounted. If so, replace them with the new mount point.
403 */
404 void
405 checkdirs(olddp)
406 struct vnode *olddp;
407 {
408 struct cwdinfo *cwdi;
409 struct vnode *newdp;
410 struct proc *p;
411
412 if (olddp->v_usecount == 1)
413 return;
414 if (VFS_ROOT(olddp->v_mountedhere, &newdp))
415 panic("mount: lost mount");
416 proclist_lock_read();
417 PROCLIST_FOREACH(p, &allproc) {
418 cwdi = p->p_cwdi;
419 if (!cwdi)
420 continue;
421 if (cwdi->cwdi_cdir == olddp) {
422 vrele(cwdi->cwdi_cdir);
423 VREF(newdp);
424 cwdi->cwdi_cdir = newdp;
425 }
426 if (cwdi->cwdi_rdir == olddp) {
427 vrele(cwdi->cwdi_rdir);
428 VREF(newdp);
429 cwdi->cwdi_rdir = newdp;
430 }
431 }
432 proclist_unlock_read();
433 if (rootvnode == olddp) {
434 vrele(rootvnode);
435 VREF(newdp);
436 rootvnode = newdp;
437 }
438 vput(newdp);
439 }
440
441 /*
442 * Unmount a file system.
443 *
444 * Note: unmount takes a path to the vnode mounted on as argument,
445 * not special file (as before).
446 */
447 /* ARGSUSED */
448 int
449 sys_unmount(l, v, retval)
450 struct lwp *l;
451 void *v;
452 register_t *retval;
453 {
454 struct sys_unmount_args /* {
455 syscallarg(const char *) path;
456 syscallarg(int) flags;
457 } */ *uap = v;
458 struct proc *p = l->l_proc;
459 struct vnode *vp;
460 struct mount *mp;
461 int error;
462 struct nameidata nd;
463
464 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
465 SCARG(uap, path), p);
466 if ((error = namei(&nd)) != 0)
467 return (error);
468 vp = nd.ni_vp;
469 mp = vp->v_mount;
470
471 /*
472 * Only root, or the user that did the original mount is
473 * permitted to unmount this filesystem.
474 */
475 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
476 (error = suser(p->p_ucred, &p->p_acflag)) != 0) {
477 vput(vp);
478 return (error);
479 }
480
481 /*
482 * Don't allow unmounting the root file system.
483 */
484 if (mp->mnt_flag & MNT_ROOTFS) {
485 vput(vp);
486 return (EINVAL);
487 }
488
489 /*
490 * Must be the root of the filesystem
491 */
492 if ((vp->v_flag & VROOT) == 0) {
493 vput(vp);
494 return (EINVAL);
495 }
496 vput(vp);
497
498 /*
499 * XXX Freeze syncer. Must do this before locking the
500 * mount point. See dounmount() for details.
501 */
502 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL);
503
504 if (vfs_busy(mp, 0, 0)) {
505 lockmgr(&syncer_lock, LK_RELEASE, NULL);
506 return (EBUSY);
507 }
508
509 return (dounmount(mp, SCARG(uap, flags), p));
510 }
511
512 /*
513 * Do the actual file system unmount. File system is assumed to have been
514 * marked busy by the caller.
515 */
516 int
517 dounmount(mp, flags, p)
518 struct mount *mp;
519 int flags;
520 struct proc *p;
521 {
522 struct vnode *coveredvp;
523 int error;
524 int async;
525 int used_syncer;
526
527 simple_lock(&mountlist_slock);
528 vfs_unbusy(mp);
529 used_syncer = (mp->mnt_syncer != NULL);
530
531 /*
532 * XXX Syncer must be frozen when we get here. This should really
533 * be done on a per-mountpoint basis, but especially the softdep
534 * code possibly called from the syncer doens't exactly work on a
535 * per-mountpoint basis, so the softdep code would become a maze
536 * of vfs_busy() calls.
537 *
538 * The caller of dounmount() must acquire syncer_lock because
539 * the syncer itself acquires locks in syncer_lock -> vfs_busy
540 * order, and we must preserve that order to avoid deadlock.
541 *
542 * So, if the file system did not use the syncer, now is
543 * the time to release the syncer_lock.
544 */
545 if (used_syncer == 0)
546 lockmgr(&syncer_lock, LK_RELEASE, NULL);
547
548 mp->mnt_iflag |= IMNT_UNMOUNT;
549 mp->mnt_unmounter = p;
550 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock);
551 vn_start_write(NULL, &mp, V_WAIT);
552
553 if (mp->mnt_flag & MNT_EXPUBLIC)
554 vfs_setpublicfs(NULL, NULL, NULL);
555 async = mp->mnt_flag & MNT_ASYNC;
556 mp->mnt_flag &= ~MNT_ASYNC;
557 cache_purgevfs(mp); /* remove cache entries for this file sys */
558 if (mp->mnt_syncer != NULL)
559 vfs_deallocate_syncvnode(mp);
560 error = 0;
561 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
562 #if NFSS > 0
563 error = fss_umount_hook(mp, (flags & MNT_FORCE));
564 #endif
565 if (error == 0)
566 error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p);
567 }
568 if (error == 0 || (flags & MNT_FORCE))
569 error = VFS_UNMOUNT(mp, flags, p);
570 vn_finished_write(mp, 0);
571 simple_lock(&mountlist_slock);
572 if (error) {
573 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
574 (void) vfs_allocate_syncvnode(mp);
575 mp->mnt_iflag &= ~IMNT_UNMOUNT;
576 mp->mnt_unmounter = NULL;
577 mp->mnt_flag |= async;
578 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
579 &mountlist_slock);
580 if (used_syncer)
581 lockmgr(&syncer_lock, LK_RELEASE, NULL);
582 simple_lock(&mp->mnt_slock);
583 while (mp->mnt_wcnt > 0) {
584 wakeup(mp);
585 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1",
586 0, &mp->mnt_slock);
587 }
588 simple_unlock(&mp->mnt_slock);
589 return (error);
590 }
591 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
592 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
593 coveredvp->v_mountedhere = NULL;
594 vrele(coveredvp);
595 }
596 mp->mnt_op->vfs_refcount--;
597 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL)
598 panic("unmount: dangling vnode");
599 mp->mnt_iflag |= IMNT_GONE;
600 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock);
601 if (used_syncer)
602 lockmgr(&syncer_lock, LK_RELEASE, NULL);
603 simple_lock(&mp->mnt_slock);
604 while (mp->mnt_wcnt > 0) {
605 wakeup(mp);
606 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock);
607 }
608 simple_unlock(&mp->mnt_slock);
609 free(mp, M_MOUNT);
610 return (0);
611 }
612
613 /*
614 * Sync each mounted filesystem.
615 */
616 #ifdef DEBUG
617 int syncprt = 0;
618 struct ctldebug debug0 = { "syncprt", &syncprt };
619 #endif
620
621 /* ARGSUSED */
622 int
623 sys_sync(l, v, retval)
624 struct lwp *l;
625 void *v;
626 register_t *retval;
627 {
628 struct mount *mp, *nmp;
629 int asyncflag;
630 struct proc *p = l == NULL ? &proc0 : l->l_proc;
631
632 simple_lock(&mountlist_slock);
633 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
634 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
635 nmp = mp->mnt_list.cqe_prev;
636 continue;
637 }
638 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
639 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
640 asyncflag = mp->mnt_flag & MNT_ASYNC;
641 mp->mnt_flag &= ~MNT_ASYNC;
642 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
643 if (asyncflag)
644 mp->mnt_flag |= MNT_ASYNC;
645 vn_finished_write(mp, 0);
646 }
647 simple_lock(&mountlist_slock);
648 nmp = mp->mnt_list.cqe_prev;
649 vfs_unbusy(mp);
650
651 }
652 simple_unlock(&mountlist_slock);
653 #ifdef DEBUG
654 if (syncprt)
655 vfs_bufstats();
656 #endif /* DEBUG */
657 return (0);
658 }
659
660 /*
661 * Change filesystem quotas.
662 */
663 /* ARGSUSED */
664 int
665 sys_quotactl(l, v, retval)
666 struct lwp *l;
667 void *v;
668 register_t *retval;
669 {
670 struct sys_quotactl_args /* {
671 syscallarg(const char *) path;
672 syscallarg(int) cmd;
673 syscallarg(int) uid;
674 syscallarg(caddr_t) arg;
675 } */ *uap = v;
676 struct proc *p = l->l_proc;
677 struct mount *mp;
678 int error;
679 struct nameidata nd;
680
681 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
682 if ((error = namei(&nd)) != 0)
683 return (error);
684 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH);
685 vrele(nd.ni_vp);
686 if (error)
687 return (error);
688 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
689 SCARG(uap, arg), p);
690 vn_finished_write(mp, 0);
691 return (error);
692 }
693
694 int
695 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags,
696 int root)
697 {
698 struct cwdinfo *cwdi = p->p_cwdi;
699 int error = 0;
700
701 /*
702 * If MNT_NOWAIT or MNT_LAZY is specified, do not
703 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
704 * overrides MNT_NOWAIT.
705 */
706 if (flags == MNT_NOWAIT || flags == MNT_LAZY ||
707 (flags != MNT_WAIT && flags != 0)) {
708 memcpy(sp, &mp->mnt_stat, sizeof(*sp));
709 goto done;
710 }
711
712 /* Get the filesystem stats now */
713 memset(sp, 0, sizeof(*sp));
714 if ((error = VFS_STATVFS(mp, sp, p)) != 0) {
715 return error;
716 }
717
718 if (cwdi->cwdi_rdir == NULL)
719 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
720 done:
721 if (cwdi->cwdi_rdir != NULL) {
722 size_t len;
723 char *bp;
724 char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
725 if (!path)
726 return ENOMEM;
727
728 bp = path + MAXPATHLEN;
729 *--bp = '\0';
730 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
731 MAXPATHLEN / 2, 0, p);
732 if (error) {
733 free(path, M_TEMP);
734 return error;
735 }
736 len = strlen(bp);
737 /*
738 * for mount points that are below our root, we can see
739 * them, so we fix up the pathname and return them. The
740 * rest we cannot see, so we don't allow viewing the
741 * data.
742 */
743 if (strncmp(bp, sp->f_mntonname, len) == 0) {
744 strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
745 sizeof(sp->f_mntonname));
746 if (sp->f_mntonname[0] == '\0')
747 (void)strlcpy(sp->f_mntonname, "/",
748 sizeof(sp->f_mntonname));
749 } else {
750 if (root)
751 (void)strlcpy(sp->f_mntonname, "/",
752 sizeof(sp->f_mntonname));
753 else
754 error = EPERM;
755 }
756 free(path, M_TEMP);
757 }
758 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
759 return error;
760 }
761
762 /*
763 * Get filesystem statistics.
764 */
765 /* ARGSUSED */
766 int
767 sys_statvfs1(l, v, retval)
768 struct lwp *l;
769 void *v;
770 register_t *retval;
771 {
772 struct sys_statvfs1_args /* {
773 syscallarg(const char *) path;
774 syscallarg(struct statvfs *) buf;
775 syscallarg(int) flags;
776 } */ *uap = v;
777 struct proc *p = l->l_proc;
778 struct mount *mp;
779 struct statvfs sbuf;
780 int error;
781 struct nameidata nd;
782
783 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
784 if ((error = namei(&nd)) != 0)
785 return error;
786 mp = nd.ni_vp->v_mount;
787 vrele(nd.ni_vp);
788 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
789 return error;
790 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
791 }
792
793 /*
794 * Get filesystem statistics.
795 */
796 /* ARGSUSED */
797 int
798 sys_fstatvfs1(l, v, retval)
799 struct lwp *l;
800 void *v;
801 register_t *retval;
802 {
803 struct sys_fstatvfs1_args /* {
804 syscallarg(int) fd;
805 syscallarg(struct statvfs *) buf;
806 syscallarg(int) flags;
807 } */ *uap = v;
808 struct proc *p = l->l_proc;
809 struct file *fp;
810 struct mount *mp;
811 struct statvfs sbuf;
812 int error;
813
814 /* getvnode() will use the descriptor for us */
815 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
816 return (error);
817 mp = ((struct vnode *)fp->f_data)->v_mount;
818 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0)
819 goto out;
820 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
821 out:
822 FILE_UNUSE(fp, p);
823 return error;
824 }
825
826
827 /*
828 * Get statistics on all filesystems.
829 */
830 int
831 sys_getvfsstat(l, v, retval)
832 struct lwp *l;
833 void *v;
834 register_t *retval;
835 {
836 struct sys_getvfsstat_args /* {
837 syscallarg(struct statvfs *) buf;
838 syscallarg(size_t) bufsize;
839 syscallarg(int) flags;
840 } */ *uap = v;
841 int root = 0;
842 struct proc *p = l->l_proc;
843 struct mount *mp, *nmp;
844 struct statvfs sbuf;
845 struct statvfs *sfsp;
846 size_t count, maxcount;
847 int error = 0;
848
849 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs);
850 sfsp = SCARG(uap, buf);
851 simple_lock(&mountlist_slock);
852 count = 0;
853 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
854 mp = nmp) {
855 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
856 nmp = CIRCLEQ_NEXT(mp, mnt_list);
857 continue;
858 }
859 if (sfsp && count < maxcount) {
860 error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0);
861 if (error) {
862 simple_lock(&mountlist_slock);
863 nmp = CIRCLEQ_NEXT(mp, mnt_list);
864 vfs_unbusy(mp);
865 continue;
866 }
867 error = copyout(&sbuf, sfsp, sizeof(*sfsp));
868 if (error) {
869 vfs_unbusy(mp);
870 return (error);
871 }
872 sfsp++;
873 root |= strcmp(sbuf.f_mntonname, "/") == 0;
874 }
875 count++;
876 simple_lock(&mountlist_slock);
877 nmp = CIRCLEQ_NEXT(mp, mnt_list);
878 vfs_unbusy(mp);
879 }
880 simple_unlock(&mountlist_slock);
881 if (root == 0 && p->p_cwdi->cwdi_rdir) {
882 /*
883 * fake a root entry
884 */
885 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p,
886 SCARG(uap, flags), 1)) != 0)
887 return error;
888 if (sfsp)
889 error = copyout(&sbuf, sfsp, sizeof(*sfsp));
890 count++;
891 }
892 if (sfsp && count > maxcount)
893 *retval = maxcount;
894 else
895 *retval = count;
896 return error;
897 }
898
899 /*
900 * Change current working directory to a given file descriptor.
901 */
902 /* ARGSUSED */
903 int
904 sys_fchdir(l, v, retval)
905 struct lwp *l;
906 void *v;
907 register_t *retval;
908 {
909 struct sys_fchdir_args /* {
910 syscallarg(int) fd;
911 } */ *uap = v;
912 struct proc *p = l->l_proc;
913 struct filedesc *fdp = p->p_fd;
914 struct cwdinfo *cwdi = p->p_cwdi;
915 struct vnode *vp, *tdp;
916 struct mount *mp;
917 struct file *fp;
918 int error;
919
920 /* getvnode() will use the descriptor for us */
921 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
922 return (error);
923 vp = (struct vnode *)fp->f_data;
924
925 VREF(vp);
926 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
927 if (vp->v_type != VDIR)
928 error = ENOTDIR;
929 else
930 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
931 while (!error && (mp = vp->v_mountedhere) != NULL) {
932 if (vfs_busy(mp, 0, 0))
933 continue;
934 error = VFS_ROOT(mp, &tdp);
935 vfs_unbusy(mp);
936 if (error)
937 break;
938 vput(vp);
939 vp = tdp;
940 }
941 if (error) {
942 vput(vp);
943 goto out;
944 }
945 VOP_UNLOCK(vp, 0);
946
947 /*
948 * Disallow changing to a directory not under the process's
949 * current root directory (if there is one).
950 */
951 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) {
952 vrele(vp);
953 error = EPERM; /* operation not permitted */
954 goto out;
955 }
956
957 vrele(cwdi->cwdi_cdir);
958 cwdi->cwdi_cdir = vp;
959 out:
960 FILE_UNUSE(fp, p);
961 return (error);
962 }
963
964 /*
965 * Change this process's notion of the root directory to a given file descriptor.
966 */
967
968 int
969 sys_fchroot(l, v, retval)
970 struct lwp *l;
971 void *v;
972 register_t *retval;
973 {
974 struct sys_fchroot_args *uap = v;
975 struct proc *p = l->l_proc;
976 struct filedesc *fdp = p->p_fd;
977 struct cwdinfo *cwdi = p->p_cwdi;
978 struct vnode *vp;
979 struct file *fp;
980 int error;
981
982 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
983 return error;
984 /* getvnode() will use the descriptor for us */
985 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
986 return error;
987 vp = (struct vnode *) fp->f_data;
988 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
989 if (vp->v_type != VDIR)
990 error = ENOTDIR;
991 else
992 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
993 VOP_UNLOCK(vp, 0);
994 if (error)
995 goto out;
996 VREF(vp);
997
998 /*
999 * Prevent escaping from chroot by putting the root under
1000 * the working directory. Silently chdir to / if we aren't
1001 * already there.
1002 */
1003 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
1004 /*
1005 * XXX would be more failsafe to change directory to a
1006 * deadfs node here instead
1007 */
1008 vrele(cwdi->cwdi_cdir);
1009 VREF(vp);
1010 cwdi->cwdi_cdir = vp;
1011 }
1012
1013 if (cwdi->cwdi_rdir != NULL)
1014 vrele(cwdi->cwdi_rdir);
1015 cwdi->cwdi_rdir = vp;
1016 out:
1017 FILE_UNUSE(fp, p);
1018 return (error);
1019 }
1020
1021
1022
1023 /*
1024 * Change current working directory (``.'').
1025 */
1026 /* ARGSUSED */
1027 int
1028 sys_chdir(l, v, retval)
1029 struct lwp *l;
1030 void *v;
1031 register_t *retval;
1032 {
1033 struct sys_chdir_args /* {
1034 syscallarg(const char *) path;
1035 } */ *uap = v;
1036 struct proc *p = l->l_proc;
1037 struct cwdinfo *cwdi = p->p_cwdi;
1038 int error;
1039 struct nameidata nd;
1040
1041 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1042 SCARG(uap, path), p);
1043 if ((error = change_dir(&nd, p)) != 0)
1044 return (error);
1045 vrele(cwdi->cwdi_cdir);
1046 cwdi->cwdi_cdir = nd.ni_vp;
1047 return (0);
1048 }
1049
1050 /*
1051 * Change notion of root (``/'') directory.
1052 */
1053 /* ARGSUSED */
1054 int
1055 sys_chroot(l, v, retval)
1056 struct lwp *l;
1057 void *v;
1058 register_t *retval;
1059 {
1060 struct sys_chroot_args /* {
1061 syscallarg(const char *) path;
1062 } */ *uap = v;
1063 struct proc *p = l->l_proc;
1064 struct cwdinfo *cwdi = p->p_cwdi;
1065 struct vnode *vp;
1066 int error;
1067 struct nameidata nd;
1068
1069 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1070 return (error);
1071 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1072 SCARG(uap, path), p);
1073 if ((error = change_dir(&nd, p)) != 0)
1074 return (error);
1075 if (cwdi->cwdi_rdir != NULL)
1076 vrele(cwdi->cwdi_rdir);
1077 vp = nd.ni_vp;
1078 cwdi->cwdi_rdir = vp;
1079
1080 /*
1081 * Prevent escaping from chroot by putting the root under
1082 * the working directory. Silently chdir to / if we aren't
1083 * already there.
1084 */
1085 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) {
1086 /*
1087 * XXX would be more failsafe to change directory to a
1088 * deadfs node here instead
1089 */
1090 vrele(cwdi->cwdi_cdir);
1091 VREF(vp);
1092 cwdi->cwdi_cdir = vp;
1093 }
1094
1095 return (0);
1096 }
1097
1098 /*
1099 * Common routine for chroot and chdir.
1100 */
1101 static int
1102 change_dir(ndp, p)
1103 struct nameidata *ndp;
1104 struct proc *p;
1105 {
1106 struct vnode *vp;
1107 int error;
1108
1109 if ((error = namei(ndp)) != 0)
1110 return (error);
1111 vp = ndp->ni_vp;
1112 if (vp->v_type != VDIR)
1113 error = ENOTDIR;
1114 else
1115 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
1116
1117 if (error)
1118 vput(vp);
1119 else
1120 VOP_UNLOCK(vp, 0);
1121 return (error);
1122 }
1123
1124 /*
1125 * Check permissions, allocate an open file structure,
1126 * and call the device open routine if any.
1127 */
1128 int
1129 sys_open(l, v, retval)
1130 struct lwp *l;
1131 void *v;
1132 register_t *retval;
1133 {
1134 struct sys_open_args /* {
1135 syscallarg(const char *) path;
1136 syscallarg(int) flags;
1137 syscallarg(int) mode;
1138 } */ *uap = v;
1139 struct proc *p = l->l_proc;
1140 struct cwdinfo *cwdi = p->p_cwdi;
1141 struct filedesc *fdp = p->p_fd;
1142 struct file *fp;
1143 struct vnode *vp;
1144 int flags, cmode;
1145 int type, indx, error;
1146 struct flock lf;
1147 struct nameidata nd;
1148
1149 flags = FFLAGS(SCARG(uap, flags));
1150 if ((flags & (FREAD | FWRITE)) == 0)
1151 return (EINVAL);
1152 /* falloc() will use the file descriptor for us */
1153 if ((error = falloc(p, &fp, &indx)) != 0)
1154 return (error);
1155 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1156 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1157 l->l_dupfd = -indx - 1; /* XXX check for fdopen */
1158 if ((error = vn_open(&nd, flags, cmode)) != 0) {
1159 FILE_UNUSE(fp, p);
1160 fdp->fd_ofiles[indx] = NULL;
1161 ffree(fp);
1162 if ((error == EDUPFD || error == EMOVEFD) &&
1163 l->l_dupfd >= 0 && /* XXX from fdopen */
1164 (error =
1165 dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) {
1166 *retval = indx;
1167 return (0);
1168 }
1169 if (error == ERESTART)
1170 error = EINTR;
1171 fdremove(fdp, indx);
1172 return (error);
1173 }
1174 l->l_dupfd = 0;
1175 vp = nd.ni_vp;
1176 fp->f_flag = flags & FMASK;
1177 fp->f_type = DTYPE_VNODE;
1178 fp->f_ops = &vnops;
1179 fp->f_data = vp;
1180 if (flags & (O_EXLOCK | O_SHLOCK)) {
1181 lf.l_whence = SEEK_SET;
1182 lf.l_start = 0;
1183 lf.l_len = 0;
1184 if (flags & O_EXLOCK)
1185 lf.l_type = F_WRLCK;
1186 else
1187 lf.l_type = F_RDLCK;
1188 type = F_FLOCK;
1189 if ((flags & FNONBLOCK) == 0)
1190 type |= F_WAIT;
1191 VOP_UNLOCK(vp, 0);
1192 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1193 if (error) {
1194 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1195 FILE_UNUSE(fp, p);
1196 ffree(fp);
1197 fdremove(fdp, indx);
1198 return (error);
1199 }
1200 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1201 fp->f_flag |= FHASLOCK;
1202 }
1203 VOP_UNLOCK(vp, 0);
1204 *retval = indx;
1205 FILE_SET_MATURE(fp);
1206 FILE_UNUSE(fp, p);
1207 return (0);
1208 }
1209
1210 /*
1211 * Get file handle system call
1212 */
1213 int
1214 sys_getfh(l, v, retval)
1215 struct lwp *l;
1216 void *v;
1217 register_t *retval;
1218 {
1219 struct sys_getfh_args /* {
1220 syscallarg(char *) fname;
1221 syscallarg(fhandle_t *) fhp;
1222 } */ *uap = v;
1223 struct proc *p = l->l_proc;
1224 struct vnode *vp;
1225 fhandle_t fh;
1226 int error;
1227 struct nameidata nd;
1228
1229 /*
1230 * Must be super user
1231 */
1232 error = suser(p->p_ucred, &p->p_acflag);
1233 if (error)
1234 return (error);
1235 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1236 SCARG(uap, fname), p);
1237 error = namei(&nd);
1238 if (error)
1239 return (error);
1240 vp = nd.ni_vp;
1241 memset(&fh, 0, sizeof(fh));
1242 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
1243 error = VFS_VPTOFH(vp, &fh.fh_fid);
1244 vput(vp);
1245 if (error)
1246 return (error);
1247 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1248 return (error);
1249 }
1250
1251 /*
1252 * Open a file given a file handle.
1253 *
1254 * Check permissions, allocate an open file structure,
1255 * and call the device open routine if any.
1256 */
1257 int
1258 sys_fhopen(l, v, retval)
1259 struct lwp *l;
1260 void *v;
1261 register_t *retval;
1262 {
1263 struct sys_fhopen_args /* {
1264 syscallarg(const fhandle_t *) fhp;
1265 syscallarg(int) flags;
1266 } */ *uap = v;
1267 struct proc *p = l->l_proc;
1268 struct filedesc *fdp = p->p_fd;
1269 struct file *fp;
1270 struct vnode *vp = NULL;
1271 struct mount *mp;
1272 struct ucred *cred = p->p_ucred;
1273 int flags;
1274 struct file *nfp;
1275 int type, indx, error=0;
1276 struct flock lf;
1277 struct vattr va;
1278 fhandle_t fh;
1279
1280 /*
1281 * Must be super user
1282 */
1283 if ((error = suser(p->p_ucred, &p->p_acflag)))
1284 return (error);
1285
1286 flags = FFLAGS(SCARG(uap, flags));
1287 if ((flags & (FREAD | FWRITE)) == 0)
1288 return (EINVAL);
1289 if ((flags & O_CREAT))
1290 return (EINVAL);
1291 /* falloc() will use the file descriptor for us */
1292 if ((error = falloc(p, &nfp, &indx)) != 0)
1293 return (error);
1294 fp = nfp;
1295 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1296 goto bad;
1297
1298 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1299 error = ESTALE;
1300 goto bad;
1301 }
1302
1303 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1304 vp = NULL; /* most likely unnecessary sanity for bad: */
1305 goto bad;
1306 }
1307
1308 /* Now do an effective vn_open */
1309
1310 if (vp->v_type == VSOCK) {
1311 error = EOPNOTSUPP;
1312 goto bad;
1313 }
1314 if (flags & FREAD) {
1315 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1316 goto bad;
1317 }
1318 if (flags & (FWRITE | O_TRUNC)) {
1319 if (vp->v_type == VDIR) {
1320 error = EISDIR;
1321 goto bad;
1322 }
1323 if ((error = vn_writechk(vp)) != 0 ||
1324 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1325 goto bad;
1326 }
1327 if (flags & O_TRUNC) {
1328 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
1329 goto bad;
1330 VOP_UNLOCK(vp, 0); /* XXX */
1331 VOP_LEASE(vp, p, cred, LEASE_WRITE);
1332 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
1333 VATTR_NULL(&va);
1334 va.va_size = 0;
1335 error = VOP_SETATTR(vp, &va, cred, p);
1336 vn_finished_write(mp, 0);
1337 if (error)
1338 goto bad;
1339 }
1340 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1341 goto bad;
1342 if (vp->v_type == VREG &&
1343 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
1344 error = EIO;
1345 goto bad;
1346 }
1347 if (flags & FWRITE)
1348 vp->v_writecount++;
1349
1350 /* done with modified vn_open, now finish what sys_open does. */
1351
1352 fp->f_flag = flags & FMASK;
1353 fp->f_type = DTYPE_VNODE;
1354 fp->f_ops = &vnops;
1355 fp->f_data = vp;
1356 if (flags & (O_EXLOCK | O_SHLOCK)) {
1357 lf.l_whence = SEEK_SET;
1358 lf.l_start = 0;
1359 lf.l_len = 0;
1360 if (flags & O_EXLOCK)
1361 lf.l_type = F_WRLCK;
1362 else
1363 lf.l_type = F_RDLCK;
1364 type = F_FLOCK;
1365 if ((flags & FNONBLOCK) == 0)
1366 type |= F_WAIT;
1367 VOP_UNLOCK(vp, 0);
1368 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1369 if (error) {
1370 (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
1371 FILE_UNUSE(fp, p);
1372 ffree(fp);
1373 fdremove(fdp, indx);
1374 return (error);
1375 }
1376 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1377 fp->f_flag |= FHASLOCK;
1378 }
1379 VOP_UNLOCK(vp, 0);
1380 *retval = indx;
1381 FILE_SET_MATURE(fp);
1382 FILE_UNUSE(fp, p);
1383 return (0);
1384
1385 bad:
1386 FILE_UNUSE(fp, p);
1387 ffree(fp);
1388 fdremove(fdp, indx);
1389 if (vp != NULL)
1390 vput(vp);
1391 return (error);
1392 }
1393
1394 /* ARGSUSED */
1395 int
1396 sys_fhstat(l, v, retval)
1397 struct lwp *l;
1398 void *v;
1399 register_t *retval;
1400 {
1401 struct sys_fhstat_args /* {
1402 syscallarg(const fhandle_t *) fhp;
1403 syscallarg(struct stat *) sb;
1404 } */ *uap = v;
1405 struct proc *p = l->l_proc;
1406 struct stat sb;
1407 int error;
1408 fhandle_t fh;
1409 struct mount *mp;
1410 struct vnode *vp;
1411
1412 /*
1413 * Must be super user
1414 */
1415 if ((error = suser(p->p_ucred, &p->p_acflag)))
1416 return (error);
1417
1418 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1419 return (error);
1420
1421 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1422 return (ESTALE);
1423 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1424 return (error);
1425 error = vn_stat(vp, &sb, p);
1426 vput(vp);
1427 if (error)
1428 return (error);
1429 error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1430 return (error);
1431 }
1432
1433 /* ARGSUSED */
1434 int
1435 sys_fhstatvfs1(l, v, retval)
1436 struct lwp *l;
1437 void *v;
1438 register_t *retval;
1439 {
1440 struct sys_fhstatvfs1_args /*
1441 syscallarg(const fhandle_t *) fhp;
1442 syscallarg(struct statvfs *) buf;
1443 syscallarg(int) flags;
1444 } */ *uap = v;
1445 struct proc *p = l->l_proc;
1446 struct statvfs sbuf;
1447 fhandle_t fh;
1448 struct mount *mp;
1449 struct vnode *vp;
1450 int error;
1451
1452 /*
1453 * Must be super user
1454 */
1455 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1456 return error;
1457
1458 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1459 return error;
1460
1461 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1462 return ESTALE;
1463 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1464 return error;
1465
1466 mp = vp->v_mount;
1467 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) {
1468 vput(vp);
1469 return error;
1470 }
1471 vput(vp);
1472 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf));
1473 }
1474
1475 /*
1476 * Create a special file.
1477 */
1478 /* ARGSUSED */
1479 int
1480 sys_mknod(l, v, retval)
1481 struct lwp *l;
1482 void *v;
1483 register_t *retval;
1484 {
1485 struct sys_mknod_args /* {
1486 syscallarg(const char *) path;
1487 syscallarg(int) mode;
1488 syscallarg(int) dev;
1489 } */ *uap = v;
1490 struct proc *p = l->l_proc;
1491 struct vnode *vp;
1492 struct mount *mp;
1493 struct vattr vattr;
1494 int error;
1495 int whiteout = 0;
1496 struct nameidata nd;
1497
1498 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1499 return (error);
1500 restart:
1501 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1502 if ((error = namei(&nd)) != 0)
1503 return (error);
1504 vp = nd.ni_vp;
1505 if (vp != NULL)
1506 error = EEXIST;
1507 else {
1508 VATTR_NULL(&vattr);
1509 vattr.va_mode =
1510 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1511 vattr.va_rdev = SCARG(uap, dev);
1512 whiteout = 0;
1513
1514 switch (SCARG(uap, mode) & S_IFMT) {
1515 case S_IFMT: /* used by badsect to flag bad sectors */
1516 vattr.va_type = VBAD;
1517 break;
1518 case S_IFCHR:
1519 vattr.va_type = VCHR;
1520 break;
1521 case S_IFBLK:
1522 vattr.va_type = VBLK;
1523 break;
1524 case S_IFWHT:
1525 whiteout = 1;
1526 break;
1527 default:
1528 error = EINVAL;
1529 break;
1530 }
1531 }
1532 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1533 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1534 if (nd.ni_dvp == vp)
1535 vrele(nd.ni_dvp);
1536 else
1537 vput(nd.ni_dvp);
1538 if (vp)
1539 vrele(vp);
1540 if ((error = vn_start_write(NULL, &mp,
1541 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1542 return (error);
1543 goto restart;
1544 }
1545 if (!error) {
1546 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1547 if (whiteout) {
1548 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1549 if (error)
1550 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1551 vput(nd.ni_dvp);
1552 } else {
1553 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1554 &nd.ni_cnd, &vattr);
1555 if (error == 0)
1556 vput(nd.ni_vp);
1557 }
1558 } else {
1559 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1560 if (nd.ni_dvp == vp)
1561 vrele(nd.ni_dvp);
1562 else
1563 vput(nd.ni_dvp);
1564 if (vp)
1565 vrele(vp);
1566 }
1567 vn_finished_write(mp, 0);
1568 return (error);
1569 }
1570
1571 /*
1572 * Create a named pipe.
1573 */
1574 /* ARGSUSED */
1575 int
1576 sys_mkfifo(l, v, retval)
1577 struct lwp *l;
1578 void *v;
1579 register_t *retval;
1580 {
1581 struct sys_mkfifo_args /* {
1582 syscallarg(const char *) path;
1583 syscallarg(int) mode;
1584 } */ *uap = v;
1585 struct proc *p = l->l_proc;
1586 struct mount *mp;
1587 struct vattr vattr;
1588 int error;
1589 struct nameidata nd;
1590
1591 restart:
1592 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1593 if ((error = namei(&nd)) != 0)
1594 return (error);
1595 if (nd.ni_vp != NULL) {
1596 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1597 if (nd.ni_dvp == nd.ni_vp)
1598 vrele(nd.ni_dvp);
1599 else
1600 vput(nd.ni_dvp);
1601 vrele(nd.ni_vp);
1602 return (EEXIST);
1603 }
1604 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1605 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1606 if (nd.ni_dvp == nd.ni_vp)
1607 vrele(nd.ni_dvp);
1608 else
1609 vput(nd.ni_dvp);
1610 if (nd.ni_vp)
1611 vrele(nd.ni_vp);
1612 if ((error = vn_start_write(NULL, &mp,
1613 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1614 return (error);
1615 goto restart;
1616 }
1617 VATTR_NULL(&vattr);
1618 vattr.va_type = VFIFO;
1619 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1620 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1621 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1622 if (error == 0)
1623 vput(nd.ni_vp);
1624 vn_finished_write(mp, 0);
1625 return (error);
1626 }
1627
1628 /*
1629 * Make a hard file link.
1630 */
1631 /* ARGSUSED */
1632 int
1633 sys_link(l, v, retval)
1634 struct lwp *l;
1635 void *v;
1636 register_t *retval;
1637 {
1638 struct sys_link_args /* {
1639 syscallarg(const char *) path;
1640 syscallarg(const char *) link;
1641 } */ *uap = v;
1642 struct proc *p = l->l_proc;
1643 struct vnode *vp;
1644 struct mount *mp;
1645 struct nameidata nd;
1646 int error;
1647
1648 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1649 if ((error = namei(&nd)) != 0)
1650 return (error);
1651 vp = nd.ni_vp;
1652 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
1653 vrele(vp);
1654 return (error);
1655 }
1656 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1657 if ((error = namei(&nd)) != 0)
1658 goto out;
1659 if (nd.ni_vp) {
1660 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1661 if (nd.ni_dvp == nd.ni_vp)
1662 vrele(nd.ni_dvp);
1663 else
1664 vput(nd.ni_dvp);
1665 vrele(nd.ni_vp);
1666 error = EEXIST;
1667 goto out;
1668 }
1669 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1670 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1671 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1672 out:
1673 vrele(vp);
1674 vn_finished_write(mp, 0);
1675 return (error);
1676 }
1677
1678 /*
1679 * Make a symbolic link.
1680 */
1681 /* ARGSUSED */
1682 int
1683 sys_symlink(l, v, retval)
1684 struct lwp *l;
1685 void *v;
1686 register_t *retval;
1687 {
1688 struct sys_symlink_args /* {
1689 syscallarg(const char *) path;
1690 syscallarg(const char *) link;
1691 } */ *uap = v;
1692 struct proc *p = l->l_proc;
1693 struct mount *mp;
1694 struct vattr vattr;
1695 char *path;
1696 int error;
1697 struct nameidata nd;
1698
1699 path = PNBUF_GET();
1700 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1701 if (error)
1702 goto out;
1703 restart:
1704 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1705 if ((error = namei(&nd)) != 0)
1706 goto out;
1707 if (nd.ni_vp) {
1708 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1709 if (nd.ni_dvp == nd.ni_vp)
1710 vrele(nd.ni_dvp);
1711 else
1712 vput(nd.ni_dvp);
1713 vrele(nd.ni_vp);
1714 error = EEXIST;
1715 goto out;
1716 }
1717 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1718 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1719 if (nd.ni_dvp == nd.ni_vp)
1720 vrele(nd.ni_dvp);
1721 else
1722 vput(nd.ni_dvp);
1723 if ((error = vn_start_write(NULL, &mp,
1724 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1725 return (error);
1726 goto restart;
1727 }
1728 VATTR_NULL(&vattr);
1729 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
1730 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1731 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1732 if (error == 0)
1733 vput(nd.ni_vp);
1734 vn_finished_write(mp, 0);
1735 out:
1736 PNBUF_PUT(path);
1737 return (error);
1738 }
1739
1740 /*
1741 * Delete a whiteout from the filesystem.
1742 */
1743 /* ARGSUSED */
1744 int
1745 sys_undelete(l, v, retval)
1746 struct lwp *l;
1747 void *v;
1748 register_t *retval;
1749 {
1750 struct sys_undelete_args /* {
1751 syscallarg(const char *) path;
1752 } */ *uap = v;
1753 struct proc *p = l->l_proc;
1754 int error;
1755 struct mount *mp;
1756 struct nameidata nd;
1757
1758 restart:
1759 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1760 SCARG(uap, path), p);
1761 error = namei(&nd);
1762 if (error)
1763 return (error);
1764
1765 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1766 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1767 if (nd.ni_dvp == nd.ni_vp)
1768 vrele(nd.ni_dvp);
1769 else
1770 vput(nd.ni_dvp);
1771 if (nd.ni_vp)
1772 vrele(nd.ni_vp);
1773 return (EEXIST);
1774 }
1775 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1776 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1777 if (nd.ni_dvp == nd.ni_vp)
1778 vrele(nd.ni_dvp);
1779 else
1780 vput(nd.ni_dvp);
1781 if ((error = vn_start_write(NULL, &mp,
1782 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1783 return (error);
1784 goto restart;
1785 }
1786 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1787 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1788 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1789 vput(nd.ni_dvp);
1790 vn_finished_write(mp, 0);
1791 return (error);
1792 }
1793
1794 /*
1795 * Delete a name from the filesystem.
1796 */
1797 /* ARGSUSED */
1798 int
1799 sys_unlink(l, v, retval)
1800 struct lwp *l;
1801 void *v;
1802 register_t *retval;
1803 {
1804 struct sys_unlink_args /* {
1805 syscallarg(const char *) path;
1806 } */ *uap = v;
1807 struct proc *p = l->l_proc;
1808 struct mount *mp;
1809 struct vnode *vp;
1810 int error;
1811 struct nameidata nd;
1812
1813 restart:
1814 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1815 SCARG(uap, path), p);
1816 if ((error = namei(&nd)) != 0)
1817 return (error);
1818 vp = nd.ni_vp;
1819
1820 /*
1821 * The root of a mounted filesystem cannot be deleted.
1822 */
1823 if (vp->v_flag & VROOT) {
1824 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1825 if (nd.ni_dvp == vp)
1826 vrele(nd.ni_dvp);
1827 else
1828 vput(nd.ni_dvp);
1829 vput(vp);
1830 error = EBUSY;
1831 goto out;
1832 }
1833
1834 #ifdef VERIFIED_EXEC
1835 /* Handle remove requests for veriexec entries. */
1836 if ((error = veriexec_removechk(p, vp, nd.ni_dirp)) != 0) {
1837 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1838 if (nd.ni_dvp == vp)
1839 vrele(nd.ni_dvp);
1840 else
1841 vput(nd.ni_dvp);
1842 vput(vp);
1843 goto out;
1844 }
1845 #endif
1846
1847 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1848 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1849 if (nd.ni_dvp == vp)
1850 vrele(nd.ni_dvp);
1851 else
1852 vput(nd.ni_dvp);
1853 vput(vp);
1854 if ((error = vn_start_write(NULL, &mp,
1855 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
1856 return (error);
1857 goto restart;
1858 }
1859 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1860 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1861 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1862 vn_finished_write(mp, 0);
1863 out:
1864 return (error);
1865 }
1866
1867 /*
1868 * Reposition read/write file offset.
1869 */
1870 int
1871 sys_lseek(l, v, retval)
1872 struct lwp *l;
1873 void *v;
1874 register_t *retval;
1875 {
1876 struct sys_lseek_args /* {
1877 syscallarg(int) fd;
1878 syscallarg(int) pad;
1879 syscallarg(off_t) offset;
1880 syscallarg(int) whence;
1881 } */ *uap = v;
1882 struct proc *p = l->l_proc;
1883 struct ucred *cred = p->p_ucred;
1884 struct filedesc *fdp = p->p_fd;
1885 struct file *fp;
1886 struct vnode *vp;
1887 struct vattr vattr;
1888 off_t newoff;
1889 int error;
1890
1891 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1892 return (EBADF);
1893
1894 FILE_USE(fp);
1895
1896 vp = (struct vnode *)fp->f_data;
1897 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1898 error = ESPIPE;
1899 goto out;
1900 }
1901
1902 switch (SCARG(uap, whence)) {
1903 case SEEK_CUR:
1904 newoff = fp->f_offset + SCARG(uap, offset);
1905 break;
1906 case SEEK_END:
1907 error = VOP_GETATTR(vp, &vattr, cred, p);
1908 if (error)
1909 goto out;
1910 newoff = SCARG(uap, offset) + vattr.va_size;
1911 break;
1912 case SEEK_SET:
1913 newoff = SCARG(uap, offset);
1914 break;
1915 default:
1916 error = EINVAL;
1917 goto out;
1918 }
1919 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0)
1920 goto out;
1921
1922 *(off_t *)retval = fp->f_offset = newoff;
1923 out:
1924 FILE_UNUSE(fp, p);
1925 return (error);
1926 }
1927
1928 /*
1929 * Positional read system call.
1930 */
1931 int
1932 sys_pread(l, v, retval)
1933 struct lwp *l;
1934 void *v;
1935 register_t *retval;
1936 {
1937 struct sys_pread_args /* {
1938 syscallarg(int) fd;
1939 syscallarg(void *) buf;
1940 syscallarg(size_t) nbyte;
1941 syscallarg(off_t) offset;
1942 } */ *uap = v;
1943 struct proc *p = l->l_proc;
1944 struct filedesc *fdp = p->p_fd;
1945 struct file *fp;
1946 struct vnode *vp;
1947 off_t offset;
1948 int error, fd = SCARG(uap, fd);
1949
1950 if ((fp = fd_getfile(fdp, fd)) == NULL)
1951 return (EBADF);
1952
1953 if ((fp->f_flag & FREAD) == 0) {
1954 simple_unlock(&fp->f_slock);
1955 return (EBADF);
1956 }
1957
1958 FILE_USE(fp);
1959
1960 vp = (struct vnode *)fp->f_data;
1961 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
1962 error = ESPIPE;
1963 goto out;
1964 }
1965
1966 offset = SCARG(uap, offset);
1967
1968 /*
1969 * XXX This works because no file systems actually
1970 * XXX take any action on the seek operation.
1971 */
1972 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
1973 goto out;
1974
1975 /* dofileread() will unuse the descriptor for us */
1976 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
1977 &offset, 0, retval));
1978
1979 out:
1980 FILE_UNUSE(fp, p);
1981 return (error);
1982 }
1983
1984 /*
1985 * Positional scatter read system call.
1986 */
1987 int
1988 sys_preadv(l, v, retval)
1989 struct lwp *l;
1990 void *v;
1991 register_t *retval;
1992 {
1993 struct sys_preadv_args /* {
1994 syscallarg(int) fd;
1995 syscallarg(const struct iovec *) iovp;
1996 syscallarg(int) iovcnt;
1997 syscallarg(off_t) offset;
1998 } */ *uap = v;
1999 struct proc *p = l->l_proc;
2000 struct filedesc *fdp = p->p_fd;
2001 struct file *fp;
2002 struct vnode *vp;
2003 off_t offset;
2004 int error, fd = SCARG(uap, fd);
2005
2006 if ((fp = fd_getfile(fdp, fd)) == NULL)
2007 return (EBADF);
2008
2009 if ((fp->f_flag & FREAD) == 0) {
2010 simple_unlock(&fp->f_slock);
2011 return (EBADF);
2012 }
2013
2014 FILE_USE(fp);
2015
2016 vp = (struct vnode *)fp->f_data;
2017 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2018 error = ESPIPE;
2019 goto out;
2020 }
2021
2022 offset = SCARG(uap, offset);
2023
2024 /*
2025 * XXX This works because no file systems actually
2026 * XXX take any action on the seek operation.
2027 */
2028 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2029 goto out;
2030
2031 /* dofilereadv() will unuse the descriptor for us */
2032 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2033 &offset, 0, retval));
2034
2035 out:
2036 FILE_UNUSE(fp, p);
2037 return (error);
2038 }
2039
2040 /*
2041 * Positional write system call.
2042 */
2043 int
2044 sys_pwrite(l, v, retval)
2045 struct lwp *l;
2046 void *v;
2047 register_t *retval;
2048 {
2049 struct sys_pwrite_args /* {
2050 syscallarg(int) fd;
2051 syscallarg(const void *) buf;
2052 syscallarg(size_t) nbyte;
2053 syscallarg(off_t) offset;
2054 } */ *uap = v;
2055 struct proc *p = l->l_proc;
2056 struct filedesc *fdp = p->p_fd;
2057 struct file *fp;
2058 struct vnode *vp;
2059 off_t offset;
2060 int error, fd = SCARG(uap, fd);
2061
2062 if ((fp = fd_getfile(fdp, fd)) == NULL)
2063 return (EBADF);
2064
2065 if ((fp->f_flag & FWRITE) == 0) {
2066 simple_unlock(&fp->f_slock);
2067 return (EBADF);
2068 }
2069
2070 FILE_USE(fp);
2071
2072 vp = (struct vnode *)fp->f_data;
2073 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2074 error = ESPIPE;
2075 goto out;
2076 }
2077
2078 offset = SCARG(uap, offset);
2079
2080 /*
2081 * XXX This works because no file systems actually
2082 * XXX take any action on the seek operation.
2083 */
2084 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2085 goto out;
2086
2087 /* dofilewrite() will unuse the descriptor for us */
2088 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2089 &offset, 0, retval));
2090
2091 out:
2092 FILE_UNUSE(fp, p);
2093 return (error);
2094 }
2095
2096 /*
2097 * Positional gather write system call.
2098 */
2099 int
2100 sys_pwritev(l, v, retval)
2101 struct lwp *l;
2102 void *v;
2103 register_t *retval;
2104 {
2105 struct sys_pwritev_args /* {
2106 syscallarg(int) fd;
2107 syscallarg(const struct iovec *) iovp;
2108 syscallarg(int) iovcnt;
2109 syscallarg(off_t) offset;
2110 } */ *uap = v;
2111 struct proc *p = l->l_proc;
2112 struct filedesc *fdp = p->p_fd;
2113 struct file *fp;
2114 struct vnode *vp;
2115 off_t offset;
2116 int error, fd = SCARG(uap, fd);
2117
2118 if ((fp = fd_getfile(fdp, fd)) == NULL)
2119 return (EBADF);
2120
2121 if ((fp->f_flag & FWRITE) == 0) {
2122 simple_unlock(&fp->f_slock);
2123 return (EBADF);
2124 }
2125
2126 FILE_USE(fp);
2127
2128 vp = (struct vnode *)fp->f_data;
2129 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2130 error = ESPIPE;
2131 goto out;
2132 }
2133
2134 offset = SCARG(uap, offset);
2135
2136 /*
2137 * XXX This works because no file systems actually
2138 * XXX take any action on the seek operation.
2139 */
2140 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2141 goto out;
2142
2143 /* dofilewritev() will unuse the descriptor for us */
2144 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2145 &offset, 0, retval));
2146
2147 out:
2148 FILE_UNUSE(fp, p);
2149 return (error);
2150 }
2151
2152 /*
2153 * Check access permissions.
2154 */
2155 int
2156 sys_access(l, v, retval)
2157 struct lwp *l;
2158 void *v;
2159 register_t *retval;
2160 {
2161 struct sys_access_args /* {
2162 syscallarg(const char *) path;
2163 syscallarg(int) flags;
2164 } */ *uap = v;
2165 struct proc *p = l->l_proc;
2166 struct ucred *cred;
2167 struct vnode *vp;
2168 int error, flags;
2169 struct nameidata nd;
2170
2171 cred = crdup(p->p_ucred);
2172 cred->cr_uid = p->p_cred->p_ruid;
2173 cred->cr_gid = p->p_cred->p_rgid;
2174 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2175 SCARG(uap, path), p);
2176 /* Override default credentials */
2177 nd.ni_cnd.cn_cred = cred;
2178 if ((error = namei(&nd)) != 0)
2179 goto out;
2180 vp = nd.ni_vp;
2181
2182 /* Flags == 0 means only check for existence. */
2183 if (SCARG(uap, flags)) {
2184 flags = 0;
2185 if (SCARG(uap, flags) & R_OK)
2186 flags |= VREAD;
2187 if (SCARG(uap, flags) & W_OK)
2188 flags |= VWRITE;
2189 if (SCARG(uap, flags) & X_OK)
2190 flags |= VEXEC;
2191
2192 error = VOP_ACCESS(vp, flags, cred, p);
2193 if (!error && (flags & VWRITE))
2194 error = vn_writechk(vp);
2195 }
2196 vput(vp);
2197 out:
2198 crfree(cred);
2199 return (error);
2200 }
2201
2202 /*
2203 * Get file status; this version follows links.
2204 */
2205 /* ARGSUSED */
2206 int
2207 sys___stat13(l, v, retval)
2208 struct lwp *l;
2209 void *v;
2210 register_t *retval;
2211 {
2212 struct sys___stat13_args /* {
2213 syscallarg(const char *) path;
2214 syscallarg(struct stat *) ub;
2215 } */ *uap = v;
2216 struct proc *p = l->l_proc;
2217 struct stat sb;
2218 int error;
2219 struct nameidata nd;
2220
2221 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2222 SCARG(uap, path), p);
2223 if ((error = namei(&nd)) != 0)
2224 return (error);
2225 error = vn_stat(nd.ni_vp, &sb, p);
2226 vput(nd.ni_vp);
2227 if (error)
2228 return (error);
2229 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2230 return (error);
2231 }
2232
2233 /*
2234 * Get file status; this version does not follow links.
2235 */
2236 /* ARGSUSED */
2237 int
2238 sys___lstat13(l, v, retval)
2239 struct lwp *l;
2240 void *v;
2241 register_t *retval;
2242 {
2243 struct sys___lstat13_args /* {
2244 syscallarg(const char *) path;
2245 syscallarg(struct stat *) ub;
2246 } */ *uap = v;
2247 struct proc *p = l->l_proc;
2248 struct stat sb;
2249 int error;
2250 struct nameidata nd;
2251
2252 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2253 SCARG(uap, path), p);
2254 if ((error = namei(&nd)) != 0)
2255 return (error);
2256 error = vn_stat(nd.ni_vp, &sb, p);
2257 vput(nd.ni_vp);
2258 if (error)
2259 return (error);
2260 error = copyout(&sb, SCARG(uap, ub), sizeof(sb));
2261 return (error);
2262 }
2263
2264 /*
2265 * Get configurable pathname variables.
2266 */
2267 /* ARGSUSED */
2268 int
2269 sys_pathconf(l, v, retval)
2270 struct lwp *l;
2271 void *v;
2272 register_t *retval;
2273 {
2274 struct sys_pathconf_args /* {
2275 syscallarg(const char *) path;
2276 syscallarg(int) name;
2277 } */ *uap = v;
2278 struct proc *p = l->l_proc;
2279 int error;
2280 struct nameidata nd;
2281
2282 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2283 SCARG(uap, path), p);
2284 if ((error = namei(&nd)) != 0)
2285 return (error);
2286 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2287 vput(nd.ni_vp);
2288 return (error);
2289 }
2290
2291 /*
2292 * Return target name of a symbolic link.
2293 */
2294 /* ARGSUSED */
2295 int
2296 sys_readlink(l, v, retval)
2297 struct lwp *l;
2298 void *v;
2299 register_t *retval;
2300 {
2301 struct sys_readlink_args /* {
2302 syscallarg(const char *) path;
2303 syscallarg(char *) buf;
2304 syscallarg(size_t) count;
2305 } */ *uap = v;
2306 struct proc *p = l->l_proc;
2307 struct vnode *vp;
2308 struct iovec aiov;
2309 struct uio auio;
2310 int error;
2311 struct nameidata nd;
2312
2313 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
2314 SCARG(uap, path), p);
2315 if ((error = namei(&nd)) != 0)
2316 return (error);
2317 vp = nd.ni_vp;
2318 if (vp->v_type != VLNK)
2319 error = EINVAL;
2320 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2321 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) {
2322 aiov.iov_base = SCARG(uap, buf);
2323 aiov.iov_len = SCARG(uap, count);
2324 auio.uio_iov = &aiov;
2325 auio.uio_iovcnt = 1;
2326 auio.uio_offset = 0;
2327 auio.uio_rw = UIO_READ;
2328 auio.uio_segflg = UIO_USERSPACE;
2329 auio.uio_procp = p;
2330 auio.uio_resid = SCARG(uap, count);
2331 error = VOP_READLINK(vp, &auio, p->p_ucred);
2332 }
2333 vput(vp);
2334 *retval = SCARG(uap, count) - auio.uio_resid;
2335 return (error);
2336 }
2337
2338 /*
2339 * Change flags of a file given a path name.
2340 */
2341 /* ARGSUSED */
2342 int
2343 sys_chflags(l, v, retval)
2344 struct lwp *l;
2345 void *v;
2346 register_t *retval;
2347 {
2348 struct sys_chflags_args /* {
2349 syscallarg(const char *) path;
2350 syscallarg(u_long) flags;
2351 } */ *uap = v;
2352 struct proc *p = l->l_proc;
2353 struct vnode *vp;
2354 int error;
2355 struct nameidata nd;
2356
2357 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2358 if ((error = namei(&nd)) != 0)
2359 return (error);
2360 vp = nd.ni_vp;
2361 error = change_flags(vp, SCARG(uap, flags), p);
2362 vput(vp);
2363 return (error);
2364 }
2365
2366 /*
2367 * Change flags of a file given a file descriptor.
2368 */
2369 /* ARGSUSED */
2370 int
2371 sys_fchflags(l, v, retval)
2372 struct lwp *l;
2373 void *v;
2374 register_t *retval;
2375 {
2376 struct sys_fchflags_args /* {
2377 syscallarg(int) fd;
2378 syscallarg(u_long) flags;
2379 } */ *uap = v;
2380 struct proc *p = l->l_proc;
2381 struct vnode *vp;
2382 struct file *fp;
2383 int error;
2384
2385 /* getvnode() will use the descriptor for us */
2386 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2387 return (error);
2388 vp = (struct vnode *)fp->f_data;
2389 error = change_flags(vp, SCARG(uap, flags), p);
2390 VOP_UNLOCK(vp, 0);
2391 FILE_UNUSE(fp, p);
2392 return (error);
2393 }
2394
2395 /*
2396 * Change flags of a file given a path name; this version does
2397 * not follow links.
2398 */
2399 int
2400 sys_lchflags(l, v, retval)
2401 struct lwp *l;
2402 void *v;
2403 register_t *retval;
2404 {
2405 struct sys_lchflags_args /* {
2406 syscallarg(const char *) path;
2407 syscallarg(u_long) flags;
2408 } */ *uap = v;
2409 struct proc *p = l->l_proc;
2410 struct vnode *vp;
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 vp = nd.ni_vp;
2418 error = change_flags(vp, SCARG(uap, flags), p);
2419 vput(vp);
2420 return (error);
2421 }
2422
2423 /*
2424 * Common routine to change flags of a file.
2425 */
2426 int
2427 change_flags(vp, flags, p)
2428 struct vnode *vp;
2429 u_long flags;
2430 struct proc *p;
2431 {
2432 struct mount *mp;
2433 struct vattr vattr;
2434 int error;
2435
2436 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2437 return (error);
2438 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2439 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2440 /*
2441 * Non-superusers cannot change the flags on devices, even if they
2442 * own them.
2443 */
2444 if (suser(p->p_ucred, &p->p_acflag) != 0) {
2445 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2446 goto out;
2447 if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2448 error = EINVAL;
2449 goto out;
2450 }
2451 }
2452 VATTR_NULL(&vattr);
2453 vattr.va_flags = flags;
2454 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2455 out:
2456 vn_finished_write(mp, 0);
2457 return (error);
2458 }
2459
2460 /*
2461 * Change mode of a file given path name; this version follows links.
2462 */
2463 /* ARGSUSED */
2464 int
2465 sys_chmod(l, v, retval)
2466 struct lwp *l;
2467 void *v;
2468 register_t *retval;
2469 {
2470 struct sys_chmod_args /* {
2471 syscallarg(const char *) path;
2472 syscallarg(int) mode;
2473 } */ *uap = v;
2474 struct proc *p = l->l_proc;
2475 int error;
2476 struct nameidata nd;
2477
2478 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2479 if ((error = namei(&nd)) != 0)
2480 return (error);
2481
2482 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2483
2484 vrele(nd.ni_vp);
2485 return (error);
2486 }
2487
2488 /*
2489 * Change mode of a file given a file descriptor.
2490 */
2491 /* ARGSUSED */
2492 int
2493 sys_fchmod(l, v, retval)
2494 struct lwp *l;
2495 void *v;
2496 register_t *retval;
2497 {
2498 struct sys_fchmod_args /* {
2499 syscallarg(int) fd;
2500 syscallarg(int) mode;
2501 } */ *uap = v;
2502 struct proc *p = l->l_proc;
2503 struct file *fp;
2504 int error;
2505
2506 /* getvnode() will use the descriptor for us */
2507 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2508 return (error);
2509
2510 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p);
2511 FILE_UNUSE(fp, p);
2512 return (error);
2513 }
2514
2515 /*
2516 * Change mode of a file given path name; this version does not follow links.
2517 */
2518 /* ARGSUSED */
2519 int
2520 sys_lchmod(l, v, retval)
2521 struct lwp *l;
2522 void *v;
2523 register_t *retval;
2524 {
2525 struct sys_lchmod_args /* {
2526 syscallarg(const char *) path;
2527 syscallarg(int) mode;
2528 } */ *uap = v;
2529 struct proc *p = l->l_proc;
2530 int error;
2531 struct nameidata nd;
2532
2533 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2534 if ((error = namei(&nd)) != 0)
2535 return (error);
2536
2537 error = change_mode(nd.ni_vp, SCARG(uap, mode), p);
2538
2539 vrele(nd.ni_vp);
2540 return (error);
2541 }
2542
2543 /*
2544 * Common routine to set mode given a vnode.
2545 */
2546 static int
2547 change_mode(vp, mode, p)
2548 struct vnode *vp;
2549 int mode;
2550 struct proc *p;
2551 {
2552 struct mount *mp;
2553 struct vattr vattr;
2554 int error;
2555
2556 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2557 return (error);
2558 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2559 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2560 VATTR_NULL(&vattr);
2561 vattr.va_mode = mode & ALLPERMS;
2562 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2563 VOP_UNLOCK(vp, 0);
2564 vn_finished_write(mp, 0);
2565 return (error);
2566 }
2567
2568 /*
2569 * Set ownership given a path name; this version follows links.
2570 */
2571 /* ARGSUSED */
2572 int
2573 sys_chown(l, v, retval)
2574 struct lwp *l;
2575 void *v;
2576 register_t *retval;
2577 {
2578 struct sys_chown_args /* {
2579 syscallarg(const char *) path;
2580 syscallarg(uid_t) uid;
2581 syscallarg(gid_t) gid;
2582 } */ *uap = v;
2583 struct proc *p = l->l_proc;
2584 int error;
2585 struct nameidata nd;
2586
2587 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2588 if ((error = namei(&nd)) != 0)
2589 return (error);
2590
2591 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2592
2593 vrele(nd.ni_vp);
2594 return (error);
2595 }
2596
2597 /*
2598 * Set ownership given a path name; this version follows links.
2599 * Provides POSIX semantics.
2600 */
2601 /* ARGSUSED */
2602 int
2603 sys___posix_chown(l, v, retval)
2604 struct lwp *l;
2605 void *v;
2606 register_t *retval;
2607 {
2608 struct sys_chown_args /* {
2609 syscallarg(const char *) path;
2610 syscallarg(uid_t) uid;
2611 syscallarg(gid_t) gid;
2612 } */ *uap = v;
2613 struct proc *p = l->l_proc;
2614 int error;
2615 struct nameidata nd;
2616
2617 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2618 if ((error = namei(&nd)) != 0)
2619 return (error);
2620
2621 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2622
2623 vrele(nd.ni_vp);
2624 return (error);
2625 }
2626
2627 /*
2628 * Set ownership given a file descriptor.
2629 */
2630 /* ARGSUSED */
2631 int
2632 sys_fchown(l, v, retval)
2633 struct lwp *l;
2634 void *v;
2635 register_t *retval;
2636 {
2637 struct sys_fchown_args /* {
2638 syscallarg(int) fd;
2639 syscallarg(uid_t) uid;
2640 syscallarg(gid_t) gid;
2641 } */ *uap = v;
2642 struct proc *p = l->l_proc;
2643 int error;
2644 struct file *fp;
2645
2646 /* getvnode() will use the descriptor for us */
2647 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2648 return (error);
2649
2650 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2651 SCARG(uap, gid), p, 0);
2652 FILE_UNUSE(fp, p);
2653 return (error);
2654 }
2655
2656 /*
2657 * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2658 */
2659 /* ARGSUSED */
2660 int
2661 sys___posix_fchown(l, v, retval)
2662 struct lwp *l;
2663 void *v;
2664 register_t *retval;
2665 {
2666 struct sys_fchown_args /* {
2667 syscallarg(int) fd;
2668 syscallarg(uid_t) uid;
2669 syscallarg(gid_t) gid;
2670 } */ *uap = v;
2671 struct proc *p = l->l_proc;
2672 int error;
2673 struct file *fp;
2674
2675 /* getvnode() will use the descriptor for us */
2676 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2677 return (error);
2678
2679 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2680 SCARG(uap, gid), p, 1);
2681 FILE_UNUSE(fp, p);
2682 return (error);
2683 }
2684
2685 /*
2686 * Set ownership given a path name; this version does not follow links.
2687 */
2688 /* ARGSUSED */
2689 int
2690 sys_lchown(l, v, retval)
2691 struct lwp *l;
2692 void *v;
2693 register_t *retval;
2694 {
2695 struct sys_lchown_args /* {
2696 syscallarg(const char *) path;
2697 syscallarg(uid_t) uid;
2698 syscallarg(gid_t) gid;
2699 } */ *uap = v;
2700 struct proc *p = l->l_proc;
2701 int error;
2702 struct nameidata nd;
2703
2704 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2705 if ((error = namei(&nd)) != 0)
2706 return (error);
2707
2708 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0);
2709
2710 vrele(nd.ni_vp);
2711 return (error);
2712 }
2713
2714 /*
2715 * Set ownership given a path name; this version does not follow links.
2716 * Provides POSIX/XPG semantics.
2717 */
2718 /* ARGSUSED */
2719 int
2720 sys___posix_lchown(l, v, retval)
2721 struct lwp *l;
2722 void *v;
2723 register_t *retval;
2724 {
2725 struct sys_lchown_args /* {
2726 syscallarg(const char *) path;
2727 syscallarg(uid_t) uid;
2728 syscallarg(gid_t) gid;
2729 } */ *uap = v;
2730 struct proc *p = l->l_proc;
2731 int error;
2732 struct nameidata nd;
2733
2734 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2735 if ((error = namei(&nd)) != 0)
2736 return (error);
2737
2738 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1);
2739
2740 vrele(nd.ni_vp);
2741 return (error);
2742 }
2743
2744 /*
2745 * Common routine to set ownership given a vnode.
2746 */
2747 static int
2748 change_owner(vp, uid, gid, p, posix_semantics)
2749 struct vnode *vp;
2750 uid_t uid;
2751 gid_t gid;
2752 struct proc *p;
2753 int posix_semantics;
2754 {
2755 struct mount *mp;
2756 struct vattr vattr;
2757 mode_t newmode;
2758 int error;
2759
2760 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2761 return (error);
2762 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2763 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2764 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2765 goto out;
2766
2767 #define CHANGED(x) ((int)(x) != -1)
2768 newmode = vattr.va_mode;
2769 if (posix_semantics) {
2770 /*
2771 * POSIX/XPG semantics: if the caller is not the super-user,
2772 * clear set-user-id and set-group-id bits. Both POSIX and
2773 * the XPG consider the behaviour for calls by the super-user
2774 * implementation-defined; we leave the set-user-id and set-
2775 * group-id settings intact in that case.
2776 */
2777 if (suser(p->p_ucred, NULL) != 0)
2778 newmode &= ~(S_ISUID | S_ISGID);
2779 } else {
2780 /*
2781 * NetBSD semantics: when changing owner and/or group,
2782 * clear the respective bit(s).
2783 */
2784 if (CHANGED(uid))
2785 newmode &= ~S_ISUID;
2786 if (CHANGED(gid))
2787 newmode &= ~S_ISGID;
2788 }
2789 /* Update va_mode iff altered. */
2790 if (vattr.va_mode == newmode)
2791 newmode = VNOVAL;
2792
2793 VATTR_NULL(&vattr);
2794 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2795 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2796 vattr.va_mode = newmode;
2797 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2798 #undef CHANGED
2799
2800 out:
2801 VOP_UNLOCK(vp, 0);
2802 vn_finished_write(mp, 0);
2803 return (error);
2804 }
2805
2806 /*
2807 * Set the access and modification times given a path name; this
2808 * version follows links.
2809 */
2810 /* ARGSUSED */
2811 int
2812 sys_utimes(l, v, retval)
2813 struct lwp *l;
2814 void *v;
2815 register_t *retval;
2816 {
2817 struct sys_utimes_args /* {
2818 syscallarg(const char *) path;
2819 syscallarg(const struct timeval *) tptr;
2820 } */ *uap = v;
2821 struct proc *p = l->l_proc;
2822 int error;
2823 struct nameidata nd;
2824
2825 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2826 if ((error = namei(&nd)) != 0)
2827 return (error);
2828
2829 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2830
2831 vrele(nd.ni_vp);
2832 return (error);
2833 }
2834
2835 /*
2836 * Set the access and modification times given a file descriptor.
2837 */
2838 /* ARGSUSED */
2839 int
2840 sys_futimes(l, v, retval)
2841 struct lwp *l;
2842 void *v;
2843 register_t *retval;
2844 {
2845 struct sys_futimes_args /* {
2846 syscallarg(int) fd;
2847 syscallarg(const struct timeval *) tptr;
2848 } */ *uap = v;
2849 struct proc *p = l->l_proc;
2850 int error;
2851 struct file *fp;
2852
2853 /* getvnode() will use the descriptor for us */
2854 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2855 return (error);
2856
2857 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p);
2858 FILE_UNUSE(fp, p);
2859 return (error);
2860 }
2861
2862 /*
2863 * Set the access and modification times given a path name; this
2864 * version does not follow links.
2865 */
2866 /* ARGSUSED */
2867 int
2868 sys_lutimes(l, v, retval)
2869 struct lwp *l;
2870 void *v;
2871 register_t *retval;
2872 {
2873 struct sys_lutimes_args /* {
2874 syscallarg(const char *) path;
2875 syscallarg(const struct timeval *) tptr;
2876 } */ *uap = v;
2877 struct proc *p = l->l_proc;
2878 int error;
2879 struct nameidata nd;
2880
2881 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2882 if ((error = namei(&nd)) != 0)
2883 return (error);
2884
2885 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p);
2886
2887 vrele(nd.ni_vp);
2888 return (error);
2889 }
2890
2891 /*
2892 * Common routine to set access and modification times given a vnode.
2893 */
2894 static int
2895 change_utimes(vp, tptr, p)
2896 struct vnode *vp;
2897 const struct timeval *tptr;
2898 struct proc *p;
2899 {
2900 struct timeval tv[2];
2901 struct mount *mp;
2902 struct vattr vattr;
2903 int error;
2904
2905 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2906 return (error);
2907 VATTR_NULL(&vattr);
2908 if (tptr == NULL) {
2909 microtime(&tv[0]);
2910 tv[1] = tv[0];
2911 vattr.va_vaflags |= VA_UTIMES_NULL;
2912 } else {
2913 error = copyin(tptr, tv, sizeof(tv));
2914 if (error)
2915 goto out;
2916 }
2917 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2918 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2919 vattr.va_atime.tv_sec = tv[0].tv_sec;
2920 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2921 vattr.va_mtime.tv_sec = tv[1].tv_sec;
2922 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2923 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2924 VOP_UNLOCK(vp, 0);
2925 out:
2926 vn_finished_write(mp, 0);
2927 return (error);
2928 }
2929
2930 /*
2931 * Truncate a file given its path name.
2932 */
2933 /* ARGSUSED */
2934 int
2935 sys_truncate(l, v, retval)
2936 struct lwp *l;
2937 void *v;
2938 register_t *retval;
2939 {
2940 struct sys_truncate_args /* {
2941 syscallarg(const char *) path;
2942 syscallarg(int) pad;
2943 syscallarg(off_t) length;
2944 } */ *uap = v;
2945 struct proc *p = l->l_proc;
2946 struct vnode *vp;
2947 struct mount *mp;
2948 struct vattr vattr;
2949 int error;
2950 struct nameidata nd;
2951
2952 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2953 if ((error = namei(&nd)) != 0)
2954 return (error);
2955 vp = nd.ni_vp;
2956 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
2957 vrele(vp);
2958 return (error);
2959 }
2960 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2961 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2962 if (vp->v_type == VDIR)
2963 error = EISDIR;
2964 else if ((error = vn_writechk(vp)) == 0 &&
2965 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2966 VATTR_NULL(&vattr);
2967 vattr.va_size = SCARG(uap, length);
2968 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2969 }
2970 vput(vp);
2971 vn_finished_write(mp, 0);
2972 return (error);
2973 }
2974
2975 /*
2976 * Truncate a file given a file descriptor.
2977 */
2978 /* ARGSUSED */
2979 int
2980 sys_ftruncate(l, v, retval)
2981 struct lwp *l;
2982 void *v;
2983 register_t *retval;
2984 {
2985 struct sys_ftruncate_args /* {
2986 syscallarg(int) fd;
2987 syscallarg(int) pad;
2988 syscallarg(off_t) length;
2989 } */ *uap = v;
2990 struct proc *p = l->l_proc;
2991 struct mount *mp;
2992 struct vattr vattr;
2993 struct vnode *vp;
2994 struct file *fp;
2995 int error;
2996
2997 /* getvnode() will use the descriptor for us */
2998 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2999 return (error);
3000 if ((fp->f_flag & FWRITE) == 0) {
3001 error = EINVAL;
3002 goto out;
3003 }
3004 vp = (struct vnode *)fp->f_data;
3005 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3006 FILE_UNUSE(fp, p);
3007 return (error);
3008 }
3009 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3010 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3011 if (vp->v_type == VDIR)
3012 error = EISDIR;
3013 else if ((error = vn_writechk(vp)) == 0) {
3014 VATTR_NULL(&vattr);
3015 vattr.va_size = SCARG(uap, length);
3016 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
3017 }
3018 VOP_UNLOCK(vp, 0);
3019 vn_finished_write(mp, 0);
3020 out:
3021 FILE_UNUSE(fp, p);
3022 return (error);
3023 }
3024
3025 /*
3026 * Sync an open file.
3027 */
3028 /* ARGSUSED */
3029 int
3030 sys_fsync(l, v, retval)
3031 struct lwp *l;
3032 void *v;
3033 register_t *retval;
3034 {
3035 struct sys_fsync_args /* {
3036 syscallarg(int) fd;
3037 } */ *uap = v;
3038 struct proc *p = l->l_proc;
3039 struct vnode *vp;
3040 struct mount *mp;
3041 struct file *fp;
3042 int error;
3043
3044 /* getvnode() will use the descriptor for us */
3045 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3046 return (error);
3047 vp = (struct vnode *)fp->f_data;
3048 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3049 FILE_UNUSE(fp, p);
3050 return (error);
3051 }
3052 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3053 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p);
3054 if (error == 0 && bioops.io_fsync != NULL &&
3055 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3056 (*bioops.io_fsync)(vp, 0);
3057 VOP_UNLOCK(vp, 0);
3058 vn_finished_write(mp, 0);
3059 FILE_UNUSE(fp, p);
3060 return (error);
3061 }
3062
3063 /*
3064 * Sync a range of file data. API modeled after that found in AIX.
3065 *
3066 * FDATASYNC indicates that we need only save enough metadata to be able
3067 * to re-read the written data. Note we duplicate AIX's requirement that
3068 * the file be open for writing.
3069 */
3070 /* ARGSUSED */
3071 int
3072 sys_fsync_range(l, v, retval)
3073 struct lwp *l;
3074 void *v;
3075 register_t *retval;
3076 {
3077 struct sys_fsync_range_args /* {
3078 syscallarg(int) fd;
3079 syscallarg(int) flags;
3080 syscallarg(off_t) start;
3081 syscallarg(int) length;
3082 } */ *uap = v;
3083 struct proc *p = l->l_proc;
3084 struct vnode *vp;
3085 struct file *fp;
3086 int flags, nflags;
3087 off_t s, e, len;
3088 int error;
3089
3090 /* getvnode() will use the descriptor for us */
3091 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3092 return (error);
3093
3094 if ((fp->f_flag & FWRITE) == 0) {
3095 FILE_UNUSE(fp, p);
3096 return (EBADF);
3097 }
3098
3099 flags = SCARG(uap, flags);
3100 if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3101 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3102 return (EINVAL);
3103 }
3104 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3105 if (flags & FDATASYNC)
3106 nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3107 else
3108 nflags = FSYNC_WAIT;
3109 if (flags & FDISKSYNC)
3110 nflags |= FSYNC_CACHE;
3111
3112 len = SCARG(uap, length);
3113 /* If length == 0, we do the whole file, and s = l = 0 will do that */
3114 if (len) {
3115 s = SCARG(uap, start);
3116 e = s + len;
3117 if (e < s) {
3118 FILE_UNUSE(fp, p);
3119 return (EINVAL);
3120 }
3121 } else {
3122 e = 0;
3123 s = 0;
3124 }
3125
3126 vp = (struct vnode *)fp->f_data;
3127 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3128 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p);
3129
3130 if (error == 0 && bioops.io_fsync != NULL &&
3131 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3132 (*bioops.io_fsync)(vp, nflags);
3133
3134 VOP_UNLOCK(vp, 0);
3135 FILE_UNUSE(fp, p);
3136 return (error);
3137 }
3138
3139 /*
3140 * Sync the data of an open file.
3141 */
3142 /* ARGSUSED */
3143 int
3144 sys_fdatasync(l, v, retval)
3145 struct lwp *l;
3146 void *v;
3147 register_t *retval;
3148 {
3149 struct sys_fdatasync_args /* {
3150 syscallarg(int) fd;
3151 } */ *uap = v;
3152 struct proc *p = l->l_proc;
3153 struct vnode *vp;
3154 struct file *fp;
3155 int error;
3156
3157 /* getvnode() will use the descriptor for us */
3158 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3159 return (error);
3160 if ((fp->f_flag & FWRITE) == 0) {
3161 FILE_UNUSE(fp, p);
3162 return (EBADF);
3163 }
3164 vp = (struct vnode *)fp->f_data;
3165 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3166 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p);
3167 VOP_UNLOCK(vp, 0);
3168 FILE_UNUSE(fp, p);
3169 return (error);
3170 }
3171
3172 /*
3173 * Rename files, (standard) BSD semantics frontend.
3174 */
3175 /* ARGSUSED */
3176 int
3177 sys_rename(l, v, retval)
3178 struct lwp *l;
3179 void *v;
3180 register_t *retval;
3181 {
3182 struct sys_rename_args /* {
3183 syscallarg(const char *) from;
3184 syscallarg(const char *) to;
3185 } */ *uap = v;
3186 struct proc *p = l->l_proc;
3187
3188 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0));
3189 }
3190
3191 /*
3192 * Rename files, POSIX semantics frontend.
3193 */
3194 /* ARGSUSED */
3195 int
3196 sys___posix_rename(l, v, retval)
3197 struct lwp *l;
3198 void *v;
3199 register_t *retval;
3200 {
3201 struct sys___posix_rename_args /* {
3202 syscallarg(const char *) from;
3203 syscallarg(const char *) to;
3204 } */ *uap = v;
3205 struct proc *p = l->l_proc;
3206
3207 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1));
3208 }
3209
3210 /*
3211 * Rename files. Source and destination must either both be directories,
3212 * or both not be directories. If target is a directory, it must be empty.
3213 * If `from' and `to' refer to the same object, the value of the `retain'
3214 * argument is used to determine whether `from' will be
3215 *
3216 * (retain == 0) deleted unless `from' and `to' refer to the same
3217 * object in the file system's name space (BSD).
3218 * (retain == 1) always retained (POSIX).
3219 */
3220 static int
3221 rename_files(from, to, p, retain)
3222 const char *from, *to;
3223 struct proc *p;
3224 int retain;
3225 {
3226 struct mount *mp = NULL;
3227 struct vnode *tvp, *fvp, *tdvp;
3228 struct nameidata fromnd, tond;
3229 int error;
3230
3231 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
3232 from, p);
3233 if ((error = namei(&fromnd)) != 0)
3234 return (error);
3235 fvp = fromnd.ni_vp;
3236 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH);
3237 if (error != 0) {
3238 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3239 vrele(fromnd.ni_dvp);
3240 vrele(fvp);
3241 if (fromnd.ni_startdir)
3242 vrele(fromnd.ni_startdir);
3243 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3244 return (error);
3245 }
3246 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3247 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p);
3248 if ((error = namei(&tond)) != 0) {
3249 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3250 vrele(fromnd.ni_dvp);
3251 vrele(fvp);
3252 goto out1;
3253 }
3254 tdvp = tond.ni_dvp;
3255 tvp = tond.ni_vp;
3256
3257 if (tvp != NULL) {
3258 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3259 error = ENOTDIR;
3260 goto out;
3261 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3262 error = EISDIR;
3263 goto out;
3264 }
3265 }
3266
3267 if (fvp == tdvp)
3268 error = EINVAL;
3269
3270 /*
3271 * Source and destination refer to the same object.
3272 */
3273 if (fvp == tvp) {
3274 if (retain)
3275 error = -1;
3276 else if (fromnd.ni_dvp == tdvp &&
3277 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3278 !memcmp(fromnd.ni_cnd.cn_nameptr,
3279 tond.ni_cnd.cn_nameptr,
3280 fromnd.ni_cnd.cn_namelen))
3281 error = -1;
3282 }
3283
3284 out:
3285 if (!error) {
3286 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
3287 if (fromnd.ni_dvp != tdvp)
3288 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3289 if (tvp) {
3290 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
3291 }
3292 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3293 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3294 } else {
3295 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3296 if (tdvp == tvp)
3297 vrele(tdvp);
3298 else
3299 vput(tdvp);
3300 if (tvp)
3301 vput(tvp);
3302 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3303 vrele(fromnd.ni_dvp);
3304 vrele(fvp);
3305 }
3306 vrele(tond.ni_startdir);
3307 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3308 out1:
3309 vn_finished_write(mp, 0);
3310 if (fromnd.ni_startdir)
3311 vrele(fromnd.ni_startdir);
3312 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3313 return (error == -1 ? 0 : error);
3314 }
3315
3316 /*
3317 * Make a directory file.
3318 */
3319 /* ARGSUSED */
3320 int
3321 sys_mkdir(l, v, retval)
3322 struct lwp *l;
3323 void *v;
3324 register_t *retval;
3325 {
3326 struct sys_mkdir_args /* {
3327 syscallarg(const char *) path;
3328 syscallarg(int) mode;
3329 } */ *uap = v;
3330 struct proc *p = l->l_proc;
3331 struct mount *mp;
3332 struct vnode *vp;
3333 struct vattr vattr;
3334 int error;
3335 struct nameidata nd;
3336
3337 restart:
3338 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE,
3339 SCARG(uap, path), p);
3340 if ((error = namei(&nd)) != 0)
3341 return (error);
3342 vp = nd.ni_vp;
3343 if (vp != NULL) {
3344 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3345 if (nd.ni_dvp == vp)
3346 vrele(nd.ni_dvp);
3347 else
3348 vput(nd.ni_dvp);
3349 vrele(vp);
3350 return (EEXIST);
3351 }
3352 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3353 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3354 if (nd.ni_dvp == vp)
3355 vrele(nd.ni_dvp);
3356 else
3357 vput(nd.ni_dvp);
3358 if ((error = vn_start_write(NULL, &mp,
3359 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3360 return (error);
3361 goto restart;
3362 }
3363 VATTR_NULL(&vattr);
3364 vattr.va_type = VDIR;
3365 vattr.va_mode =
3366 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3367 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3368 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3369 if (!error)
3370 vput(nd.ni_vp);
3371 vn_finished_write(mp, 0);
3372 return (error);
3373 }
3374
3375 /*
3376 * Remove a directory file.
3377 */
3378 /* ARGSUSED */
3379 int
3380 sys_rmdir(l, v, retval)
3381 struct lwp *l;
3382 void *v;
3383 register_t *retval;
3384 {
3385 struct sys_rmdir_args /* {
3386 syscallarg(const char *) path;
3387 } */ *uap = v;
3388 struct proc *p = l->l_proc;
3389 struct mount *mp;
3390 struct vnode *vp;
3391 int error;
3392 struct nameidata nd;
3393
3394 restart:
3395 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
3396 SCARG(uap, path), p);
3397 if ((error = namei(&nd)) != 0)
3398 return (error);
3399 vp = nd.ni_vp;
3400 if (vp->v_type != VDIR) {
3401 error = ENOTDIR;
3402 goto out;
3403 }
3404 /*
3405 * No rmdir "." please.
3406 */
3407 if (nd.ni_dvp == vp) {
3408 error = EINVAL;
3409 goto out;
3410 }
3411 /*
3412 * The root of a mounted filesystem cannot be deleted.
3413 */
3414 if (vp->v_flag & VROOT) {
3415 error = EBUSY;
3416 goto out;
3417 }
3418 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3419 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3420 if (nd.ni_dvp == vp)
3421 vrele(nd.ni_dvp);
3422 else
3423 vput(nd.ni_dvp);
3424 vput(vp);
3425 if ((error = vn_start_write(NULL, &mp,
3426 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
3427 return (error);
3428 goto restart;
3429 }
3430 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
3431 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3432 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3433 vn_finished_write(mp, 0);
3434 return (error);
3435
3436 out:
3437 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3438 if (nd.ni_dvp == vp)
3439 vrele(nd.ni_dvp);
3440 else
3441 vput(nd.ni_dvp);
3442 vput(vp);
3443 return (error);
3444 }
3445
3446 /*
3447 * Read a block of directory entries in a file system independent format.
3448 */
3449 int
3450 sys_getdents(l, v, retval)
3451 struct lwp *l;
3452 void *v;
3453 register_t *retval;
3454 {
3455 struct sys_getdents_args /* {
3456 syscallarg(int) fd;
3457 syscallarg(char *) buf;
3458 syscallarg(size_t) count;
3459 } */ *uap = v;
3460 struct proc *p = l->l_proc;
3461 struct file *fp;
3462 int error, done;
3463
3464 /* getvnode() will use the descriptor for us */
3465 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3466 return (error);
3467 if ((fp->f_flag & FREAD) == 0) {
3468 error = EBADF;
3469 goto out;
3470 }
3471 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3472 SCARG(uap, count), &done, p, 0, 0);
3473 #ifdef KTRACE
3474 if (!error && KTRPOINT(p, KTR_GENIO)) {
3475 struct iovec iov;
3476 iov.iov_base = SCARG(uap, buf);
3477 iov.iov_len = done;
3478 ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0);
3479 }
3480 #endif
3481 *retval = done;
3482 out:
3483 FILE_UNUSE(fp, p);
3484 return (error);
3485 }
3486
3487 /*
3488 * Set the mode mask for creation of filesystem nodes.
3489 */
3490 int
3491 sys_umask(l, v, retval)
3492 struct lwp *l;
3493 void *v;
3494 register_t *retval;
3495 {
3496 struct sys_umask_args /* {
3497 syscallarg(mode_t) newmask;
3498 } */ *uap = v;
3499 struct proc *p = l->l_proc;
3500 struct cwdinfo *cwdi;
3501
3502 cwdi = p->p_cwdi;
3503 *retval = cwdi->cwdi_cmask;
3504 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3505 return (0);
3506 }
3507
3508 /*
3509 * Void all references to file by ripping underlying filesystem
3510 * away from vnode.
3511 */
3512 /* ARGSUSED */
3513 int
3514 sys_revoke(l, v, retval)
3515 struct lwp *l;
3516 void *v;
3517 register_t *retval;
3518 {
3519 struct sys_revoke_args /* {
3520 syscallarg(const char *) path;
3521 } */ *uap = v;
3522 struct proc *p = l->l_proc;
3523 struct mount *mp;
3524 struct vnode *vp;
3525 struct vattr vattr;
3526 int error;
3527 struct nameidata nd;
3528
3529 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3530 if ((error = namei(&nd)) != 0)
3531 return (error);
3532 vp = nd.ni_vp;
3533 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3534 goto out;
3535 if (p->p_ucred->cr_uid != vattr.va_uid &&
3536 (error = suser(p->p_ucred, &p->p_acflag)) != 0)
3537 goto out;
3538 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3539 goto out;
3540 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
3541 VOP_REVOKE(vp, REVOKEALL);
3542 vn_finished_write(mp, 0);
3543 out:
3544 vrele(vp);
3545 return (error);
3546 }
3547
3548 /*
3549 * Convert a user file descriptor to a kernel file entry.
3550 */
3551 int
3552 getvnode(fdp, fd, fpp)
3553 struct filedesc *fdp;
3554 int fd;
3555 struct file **fpp;
3556 {
3557 struct vnode *vp;
3558 struct file *fp;
3559
3560 if ((fp = fd_getfile(fdp, fd)) == NULL)
3561 return (EBADF);
3562
3563 FILE_USE(fp);
3564
3565 if (fp->f_type != DTYPE_VNODE) {
3566 FILE_UNUSE(fp, NULL);
3567 return (EINVAL);
3568 }
3569
3570 vp = (struct vnode *)fp->f_data;
3571 if (vp->v_type == VBAD) {
3572 FILE_UNUSE(fp, NULL);
3573 return (EBADF);
3574 }
3575
3576 *fpp = fp;
3577 return (0);
3578 }
3579
3580 /*
3581 * Push extended attribute configuration information into the VFS.
3582 *
3583 * NOTE: Not all file systems that support extended attributes will
3584 * require the use of this system call.
3585 */
3586 int
3587 sys_extattrctl(struct lwp *l, void *v, register_t *retval)
3588 {
3589 struct sys_extattrctl_args /* {
3590 syscallarg(const char *) path;
3591 syscallarg(int) cmd;
3592 syscallarg(const char *) filename;
3593 syscallarg(int) attrnamespace;
3594 syscallarg(const char *) attrname;
3595 } */ *uap = v;
3596 struct proc *p = l->l_proc;
3597 struct vnode *vp;
3598 struct nameidata nd;
3599 struct mount *mp;
3600 char attrname[EXTATTR_MAXNAMELEN];
3601 int error;
3602
3603 if (SCARG(uap, attrname) != NULL) {
3604 error = copyinstr(SCARG(uap, attrname), attrname,
3605 sizeof(attrname), NULL);
3606 if (error)
3607 return (error);
3608 }
3609
3610 vp = NULL;
3611 if (SCARG(uap, filename) != NULL) {
3612 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
3613 SCARG(uap, filename), p);
3614 error = namei(&nd);
3615 if (error)
3616 return (error);
3617 vp = nd.ni_vp;
3618 }
3619
3620 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3621 error = namei(&nd);
3622 if (error) {
3623 if (vp != NULL)
3624 vput(vp);
3625 return (error);
3626 }
3627
3628 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH);
3629 if (error) {
3630 if (vp != NULL)
3631 vput(vp);
3632 return (error);
3633 }
3634
3635 error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), vp,
3636 SCARG(uap, attrnamespace),
3637 SCARG(uap, attrname) != NULL ? attrname : NULL, p);
3638
3639 vn_finished_write(mp, 0);
3640
3641 if (vp != NULL)
3642 vrele(vp);
3643
3644 return (error);
3645 }
3646
3647 /*
3648 * Set a named extended attribute on a file or directory.
3649 */
3650 static int
3651 extattr_set_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3652 const void *data, size_t nbytes, struct proc *p, register_t *retval)
3653 {
3654 struct mount *mp;
3655 struct uio auio;
3656 struct iovec aiov;
3657 ssize_t cnt;
3658 int error;
3659
3660 error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3661 if (error)
3662 return (error);
3663 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3664 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3665
3666 aiov.iov_base = (caddr_t) data; /* XXX kills const */
3667 aiov.iov_len = nbytes;
3668 auio.uio_iov = &aiov;
3669 auio.uio_iovcnt = 1;
3670 auio.uio_offset = 0;
3671 if (nbytes > INT_MAX) {
3672 error = EINVAL;
3673 goto done;
3674 }
3675 auio.uio_resid = nbytes;
3676 auio.uio_rw = UIO_WRITE;
3677 auio.uio_segflg = UIO_USERSPACE;
3678 auio.uio_procp = p;
3679 cnt = nbytes;
3680
3681 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio,
3682 p->p_ucred, p);
3683 cnt -= auio.uio_resid;
3684 retval[0] = cnt;
3685
3686 done:
3687 VOP_UNLOCK(vp, 0);
3688 vn_finished_write(mp, 0);
3689 return (error);
3690 }
3691
3692 int
3693 sys_extattr_set_fd(struct lwp *l, void *v, register_t *retval)
3694 {
3695 struct sys_extattr_set_fd_args /* {
3696 syscallarg(int) fd;
3697 syscallarg(int) attrnamespace;
3698 syscallarg(const char *) attrname;
3699 syscallarg(const void *) data;
3700 syscallarg(size_t) nbytes;
3701 } */ *uap = v;
3702 struct proc *p = l->l_proc;
3703 struct file *fp;
3704 struct vnode *vp;
3705 char attrname[EXTATTR_MAXNAMELEN];
3706 int error;
3707
3708 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3709 NULL);
3710 if (error)
3711 return (error);
3712
3713 error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
3714 if (error)
3715 return (error);
3716 vp = (struct vnode *) fp->f_data;
3717
3718 error = extattr_set_vp(vp, SCARG(uap, attrnamespace), attrname,
3719 SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3720
3721 FILE_UNUSE(fp, p);
3722 return (error);
3723 }
3724
3725 int
3726 sys_extattr_set_file(struct lwp *l, void *v, register_t *retval)
3727 {
3728 struct sys_extattr_set_file_args /* {
3729 syscallarg(const char *) path;
3730 syscallarg(int) attrnamespace;
3731 syscallarg(const char *) attrname;
3732 syscallarg(const void *) data;
3733 syscallarg(size_t) nbytes;
3734 } */ *uap = v;
3735 struct proc *p = l->l_proc;
3736 struct nameidata nd;
3737 char attrname[EXTATTR_MAXNAMELEN];
3738 int error;
3739
3740 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3741 NULL);
3742 if (error)
3743 return (error);
3744
3745 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3746 error = namei(&nd);
3747 if (error)
3748 return (error);
3749
3750 error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3751 SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3752
3753 vrele(nd.ni_vp);
3754 return (error);
3755 }
3756
3757 int
3758 sys_extattr_set_link(struct lwp *l, void *v, register_t *retval)
3759 {
3760 struct sys_extattr_set_link_args /* {
3761 syscallarg(const char *) path;
3762 syscallarg(int) attrnamespace;
3763 syscallarg(const char *) attrname;
3764 syscallarg(const void *) data;
3765 syscallarg(size_t) nbytes;
3766 } */ *uap = v;
3767 struct proc *p = l->l_proc;
3768 struct nameidata nd;
3769 char attrname[EXTATTR_MAXNAMELEN];
3770 int error;
3771
3772 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3773 NULL);
3774 if (error)
3775 return (error);
3776
3777 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3778 error = namei(&nd);
3779 if (error)
3780 return (error);
3781
3782 error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3783 SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3784
3785 vrele(nd.ni_vp);
3786 return (error);
3787 }
3788
3789 /*
3790 * Get a named extended attribute on a file or directory.
3791 */
3792 static int
3793 extattr_get_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3794 void *data, size_t nbytes, struct proc *p, register_t *retval)
3795 {
3796 struct uio auio, *auiop;
3797 struct iovec aiov;
3798 ssize_t cnt;
3799 size_t size, *sizep;
3800 int error;
3801
3802 VOP_LEASE(vp, p, p->p_ucred, LEASE_READ);
3803 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3804
3805 /*
3806 * Slightly unusual semantics: if the user provides a NULL data
3807 * pointer, they don't want to receive the data, just the maximum
3808 * read length.
3809 */
3810 auiop = NULL;
3811 sizep = NULL;
3812 cnt = 0;
3813 if (data != NULL) {
3814 aiov.iov_base = data;
3815 aiov.iov_len = nbytes;
3816 auio.uio_iov = &aiov;
3817 auio.uio_offset = 0;
3818 if (nbytes > INT_MAX) {
3819 error = EINVAL;
3820 goto done;
3821 }
3822 auio.uio_resid = nbytes;
3823 auio.uio_rw = UIO_READ;
3824 auio.uio_segflg = UIO_USERSPACE;
3825 auio.uio_procp = p;
3826 auiop = &auio;
3827 cnt = nbytes;
3828 } else
3829 sizep = &size;
3830
3831 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, auiop, sizep,
3832 p->p_ucred, p);
3833
3834 if (auiop != NULL) {
3835 cnt -= auio.uio_resid;
3836 retval[0] = cnt;
3837 } else
3838 retval[0] = size;
3839
3840 done:
3841 VOP_UNLOCK(vp, 0);
3842 return (error);
3843 }
3844
3845 int
3846 sys_extattr_get_fd(struct lwp *l, void *v, register_t *retval)
3847 {
3848 struct sys_extattr_get_fd_args /* {
3849 syscallarg(int) fd;
3850 syscallarg(int) attrnamespace;
3851 syscallarg(const char *) attrname;
3852 syscallarg(void *) data;
3853 syscallarg(size_t) nbytes;
3854 } */ *uap = v;
3855 struct proc *p = l->l_proc;
3856 struct file *fp;
3857 struct vnode *vp;
3858 char attrname[EXTATTR_MAXNAMELEN];
3859 int error;
3860
3861 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3862 NULL);
3863 if (error)
3864 return (error);
3865
3866 error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
3867 if (error)
3868 return (error);
3869 vp = (struct vnode *) fp->f_data;
3870
3871 error = extattr_get_vp(vp, SCARG(uap, attrnamespace), attrname,
3872 SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3873
3874 FILE_UNUSE(fp, p);
3875 return (error);
3876 }
3877
3878 int
3879 sys_extattr_get_file(struct lwp *l, void *v, register_t *retval)
3880 {
3881 struct sys_extattr_get_file_args /* {
3882 syscallarg(const char *) path;
3883 syscallarg(int) attrnamespace;
3884 syscallarg(const char *) attrname;
3885 syscallarg(void *) data;
3886 syscallarg(size_t) nbytes;
3887 } */ *uap = v;
3888 struct proc *p = l->l_proc;
3889 struct nameidata nd;
3890 char attrname[EXTATTR_MAXNAMELEN];
3891 int error;
3892
3893 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3894 NULL);
3895 if (error)
3896 return (error);
3897
3898 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3899 error = namei(&nd);
3900 if (error)
3901 return (error);
3902
3903 error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3904 SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3905
3906 vrele(nd.ni_vp);
3907 return (error);
3908 }
3909
3910 int
3911 sys_extattr_get_link(struct lwp *l, void *v, register_t *retval)
3912 {
3913 struct sys_extattr_get_link_args /* {
3914 syscallarg(const char *) path;
3915 syscallarg(int) attrnamespace;
3916 syscallarg(const char *) attrname;
3917 syscallarg(void *) data;
3918 syscallarg(size_t) nbytes;
3919 } */ *uap = v;
3920 struct proc *p = l->l_proc;
3921 struct nameidata nd;
3922 char attrname[EXTATTR_MAXNAMELEN];
3923 int error;
3924
3925 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3926 NULL);
3927 if (error)
3928 return (error);
3929
3930 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3931 error = namei(&nd);
3932 if (error)
3933 return (error);
3934
3935 error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3936 SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3937
3938 vrele(nd.ni_vp);
3939 return (error);
3940 }
3941
3942 /*
3943 * Delete a named extended attribute on a file or directory.
3944 */
3945 static int
3946 extattr_delete_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3947 struct proc *p)
3948 {
3949 struct mount *mp;
3950 int error;
3951
3952 error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3953 if (error)
3954 return (error);
3955 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3956 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3957
3958 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, p->p_ucred, p);
3959 if (error == EOPNOTSUPP)
3960 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
3961 p->p_ucred, p);
3962
3963 VOP_UNLOCK(vp, 0);
3964 vn_finished_write(mp, 0);
3965 return (error);
3966 }
3967
3968 int
3969 sys_extattr_delete_fd(struct lwp *l, void *v, register_t *retval)
3970 {
3971 struct sys_extattr_delete_fd_args /* {
3972 syscallarg(int) fd;
3973 syscallarg(int) attrnamespace;
3974 syscallarg(const char *) attrname;
3975 } */ *uap = v;
3976 struct proc *p = l->l_proc;
3977 struct file *fp;
3978 struct vnode *vp;
3979 char attrname[EXTATTR_MAXNAMELEN];
3980 int error;
3981
3982 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
3983 NULL);
3984 if (error)
3985 return (error);
3986
3987 error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
3988 if (error)
3989 return (error);
3990 vp = (struct vnode *) fp->f_data;
3991
3992 error = extattr_delete_vp(vp, SCARG(uap, attrnamespace), attrname, p);
3993
3994 FILE_UNUSE(fp, p);
3995 return (error);
3996 }
3997
3998 int
3999 sys_extattr_delete_file(struct lwp *l, void *v, register_t *retval)
4000 {
4001 struct sys_extattr_delete_file_args /* {
4002 syscallarg(const char *) path;
4003 syscallarg(int) attrnamespace;
4004 syscallarg(const char *) attrname;
4005 } */ *uap = v;
4006 struct proc *p = l->l_proc;
4007 struct nameidata nd;
4008 char attrname[EXTATTR_MAXNAMELEN];
4009 int error;
4010
4011 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
4012 NULL);
4013 if (error)
4014 return (error);
4015
4016 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4017 error = namei(&nd);
4018 if (error)
4019 return (error);
4020
4021 error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
4022 p);
4023
4024 vrele(nd.ni_vp);
4025 return (error);
4026 }
4027
4028 int
4029 sys_extattr_delete_link(struct lwp *l, void *v, register_t *retval)
4030 {
4031 struct sys_extattr_delete_link_args /* {
4032 syscallarg(const char *) path;
4033 syscallarg(int) attrnamespace;
4034 syscallarg(const char *) attrname;
4035 } */ *uap = v;
4036 struct proc *p = l->l_proc;
4037 struct nameidata nd;
4038 char attrname[EXTATTR_MAXNAMELEN];
4039 int error;
4040
4041 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname),
4042 NULL);
4043 if (error)
4044 return (error);
4045
4046 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4047 error = namei(&nd);
4048 if (error)
4049 return (error);
4050
4051 error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
4052 p);
4053
4054 vrele(nd.ni_vp);
4055 return (error);
4056 }
4057
4058 /*
4059 * Retrieve a list of extended attributes on a file or directory.
4060 */
4061 static int
4062 extattr_list_vp(struct vnode *vp, int attrnamespace, void *data, size_t nbytes,
4063 struct proc *p, register_t *retval)
4064 {
4065 struct uio auio, *auiop;
4066 size_t size, *sizep;
4067 struct iovec aiov;
4068 ssize_t cnt;
4069 int error;
4070
4071 VOP_LEASE(vp, p, p->p_ucred, LEASE_READ);
4072 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4073
4074 auiop = NULL;
4075 sizep = NULL;
4076 cnt = 0;
4077 if (data != NULL) {
4078 aiov.iov_base = data;
4079 aiov.iov_len = nbytes;
4080 auio.uio_iov = &aiov;
4081 auio.uio_offset = 0;
4082 if (nbytes > INT_MAX) {
4083 error = EINVAL;
4084 goto done;
4085 }
4086 auio.uio_resid = nbytes;
4087 auio.uio_rw = UIO_READ;
4088 auio.uio_segflg = UIO_USERSPACE;
4089 auio.uio_procp = p;
4090 auiop = &auio;
4091 cnt = nbytes;
4092 } else
4093 sizep = &size;
4094
4095 error = VOP_LISTEXTATTR(vp, attrnamespace, auiop, sizep,
4096 p->p_ucred, p);
4097
4098 if (auiop != NULL) {
4099 cnt -= auio.uio_resid;
4100 retval[0] = cnt;
4101 } else
4102 retval[0] = size;
4103
4104 done:
4105 VOP_UNLOCK(vp, 0);
4106 return (error);
4107 }
4108
4109 int
4110 sys_extattr_list_fd(struct lwp *l, void *v, register_t *retval)
4111 {
4112 struct sys_extattr_list_fd_args /* {
4113 syscallarg(int) fd;
4114 syscallarg(int) attrnamespace;
4115 syscallarg(void *) data;
4116 syscallarg(size_t) nbytes;
4117 } */ *uap = v;
4118 struct proc *p = l->l_proc;
4119 struct file *fp;
4120 struct vnode *vp;
4121 int error;
4122
4123 error = getvnode(p->p_fd, SCARG(uap, fd), &fp);
4124 if (error)
4125 return (error);
4126 vp = (struct vnode *) fp->f_data;
4127
4128 error = extattr_list_vp(vp, SCARG(uap, attrnamespace),
4129 SCARG(uap, data), SCARG(uap, nbytes), p, retval);
4130
4131 FILE_UNUSE(fp, p);
4132 return (error);
4133 }
4134
4135 int
4136 sys_extattr_list_file(struct lwp *l, void *v, register_t *retval)
4137 {
4138 struct sys_extattr_list_file_args /* {
4139 syscallarg(const char *) path;
4140 syscallarg(int) attrnamespace;
4141 syscallarg(void *) data;
4142 syscallarg(size_t) nbytes;
4143 } */ *uap = v;
4144 struct proc *p = l->l_proc;
4145 struct nameidata nd;
4146 int error;
4147
4148 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4149 error = namei(&nd);
4150 if (error)
4151 return (error);
4152
4153 error = extattr_list_vp(nd.ni_vp, SCARG(uap, attrnamespace),
4154 SCARG(uap, data), SCARG(uap, nbytes), p, retval);
4155
4156 vrele(nd.ni_vp);
4157 return (error);
4158 }
4159
4160 int
4161 sys_extattr_list_link(struct lwp *l, void *v, register_t *retval)
4162 {
4163 struct sys_extattr_list_link_args /* {
4164 syscallarg(const char *) path;
4165 syscallarg(int) attrnamespace;
4166 syscallarg(void *) data;
4167 syscallarg(size_t) nbytes;
4168 } */ *uap = v;
4169 struct proc *p = l->l_proc;
4170 struct nameidata nd;
4171 int error;
4172
4173 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
4174 error = namei(&nd);
4175 if (error)
4176 return (error);
4177
4178 error = extattr_list_vp(nd.ni_vp, SCARG(uap, attrnamespace),
4179 SCARG(uap, data), SCARG(uap, nbytes), p, retval);
4180
4181 vrele(nd.ni_vp);
4182 return (error);
4183 }
4184