- Check for execvp returning ENOTDIR
[dragonfly.git] / sys / kern / vfs_syscalls.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
39 * $FreeBSD: src/sys/kern/vfs_syscalls.c,v 1.151.2.18 2003/04/04 20:35:58 tegge Exp $
8c40f81e 40 * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.73 2005/09/29 20:59:30 dillon Exp $
984263bc
MD
41 */
42
984263bc
MD
43#include <sys/param.h>
44#include <sys/systm.h>
45#include <sys/buf.h>
16df2c74 46#include <sys/conf.h>
984263bc
MD
47#include <sys/sysent.h>
48#include <sys/malloc.h>
49#include <sys/mount.h>
432b8263 50#include <sys/mountctl.h>
984263bc 51#include <sys/sysproto.h>
984263bc
MD
52#include <sys/filedesc.h>
53#include <sys/kernel.h>
54#include <sys/fcntl.h>
55#include <sys/file.h>
56#include <sys/linker.h>
57#include <sys/stat.h>
58#include <sys/unistd.h>
59#include <sys/vnode.h>
60#include <sys/proc.h>
dadab5e9 61#include <sys/namei.h>
21739618 62#include <sys/nlookup.h>
984263bc
MD
63#include <sys/dirent.h>
64#include <sys/extattr.h>
8f6f8622 65#include <sys/kern_syscall.h>
984263bc
MD
66
67#include <machine/limits.h>
1f2de5d4 68#include <vfs/union/union.h>
984263bc
MD
69#include <sys/sysctl.h>
70#include <vm/vm.h>
71#include <vm/vm_object.h>
72#include <vm/vm_zone.h>
73#include <vm/vm_page.h>
74
dadab5e9
MD
75#include <sys/file2.h>
76
e24b948e 77static int checkvp_chdir (struct vnode *vn, struct thread *td);
21739618 78static void checkdirs (struct vnode *olddp, struct namecache *ncp);
402ed7e1 79static int chroot_refuse_vdir_fds (struct filedesc *fdp);
75ffff0d 80static int chroot_visible_mnt(struct mount *mp, struct proc *p);
402ed7e1
RG
81static int getutimes (const struct timeval *, struct timespec *);
82static int setfown (struct vnode *, uid_t, gid_t);
83static int setfmode (struct vnode *, int);
84static int setfflags (struct vnode *, int);
85static int setutimes (struct vnode *, const struct timespec *, int);
984263bc
MD
86static int usermount = 0; /* if 1, non-root can mount fs. */
87
402ed7e1 88int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
984263bc
MD
89
90SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
91
92/*
93 * Virtual File System System Calls
94 */
95
96/*
97 * Mount a file system.
98 */
41c20dac
MD
99/*
100 * mount_args(char *type, char *path, int flags, caddr_t data)
101 */
984263bc
MD
102/* ARGSUSED */
103int
41c20dac 104mount(struct mount_args *uap)
984263bc 105{
dadab5e9
MD
106 struct thread *td = curthread;
107 struct proc *p = td->td_proc;
984263bc 108 struct vnode *vp;
21739618 109 struct namecache *ncp;
984263bc
MD
110 struct mount *mp;
111 struct vfsconf *vfsp;
112 int error, flag = 0, flag2 = 0;
113 struct vattr va;
21739618 114 struct nlookupdata nd;
984263bc 115 char fstypename[MFSNAMELEN];
21739618 116 struct nlcomponent nlc;
984263bc 117
cd0d3e17 118 KKASSERT(p);
84c3a710
HP
119 if (p->p_ucred->cr_prison != NULL)
120 return (EPERM);
dadab5e9 121 if (usermount == 0 && (error = suser(td)))
984263bc
MD
122 return (error);
123 /*
124 * Do not allow NFS export by non-root users.
125 */
ab2eb4eb 126 if (uap->flags & MNT_EXPORTED) {
dadab5e9 127 error = suser(td);
984263bc
MD
128 if (error)
129 return (error);
130 }
131 /*
132 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
133 */
dadab5e9 134 if (suser(td))
ab2eb4eb 135 uap->flags |= MNT_NOSUID | MNT_NODEV;
21739618 136
984263bc 137 /*
21739618 138 * Lookup the requested path and extract the ncp and vnode.
984263bc 139 */
21739618
MD
140 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
141 if (error == 0) {
142 if ((error = nlookup(&nd)) == 0) {
143 if (nd.nl_ncp->nc_vp == NULL)
144 error = ENOENT;
145 }
146 }
147 if (error) {
148 nlookup_done(&nd);
984263bc 149 return (error);
21739618
MD
150 }
151
152 /*
153 * Extract the locked+refd ncp and cleanup the nd structure
154 */
155 ncp = nd.nl_ncp;
156 nd.nl_ncp = NULL;
157 nlookup_done(&nd);
158
159 /*
160 * now we have the locked ref'd ncp and unreferenced vnode.
161 */
162 vp = ncp->nc_vp;
5fd012e0 163 if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0) {
21739618
MD
164 cache_put(ncp);
165 return (error);
166 }
167 cache_unlock(ncp);
168
169 /*
170 * Now we have an unlocked ref'd ncp and a locked ref'd vp
171 */
ab2eb4eb 172 if (uap->flags & MNT_UPDATE) {
984263bc 173 if ((vp->v_flag & VROOT) == 0) {
21739618 174 cache_drop(ncp);
984263bc
MD
175 vput(vp);
176 return (EINVAL);
177 }
178 mp = vp->v_mount;
179 flag = mp->mnt_flag;
180 flag2 = mp->mnt_kern_flag;
181 /*
182 * We only allow the filesystem to be reloaded if it
183 * is currently mounted read-only.
184 */
ab2eb4eb 185 if ((uap->flags & MNT_RELOAD) &&
984263bc 186 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
21739618 187 cache_drop(ncp);
984263bc
MD
188 vput(vp);
189 return (EOPNOTSUPP); /* Needs translation */
190 }
191 /*
192 * Only root, or the user that did the original mount is
193 * permitted to update it.
194 */
195 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
dadab5e9 196 (error = suser(td))) {
21739618 197 cache_drop(ncp);
984263bc
MD
198 vput(vp);
199 return (error);
200 }
861905fb 201 if (vfs_busy(mp, LK_NOWAIT, td)) {
21739618 202 cache_drop(ncp);
984263bc
MD
203 vput(vp);
204 return (EBUSY);
205 }
984263bc
MD
206 if ((vp->v_flag & VMOUNT) != 0 ||
207 vp->v_mountedhere != NULL) {
21739618 208 cache_drop(ncp);
dadab5e9 209 vfs_unbusy(mp, td);
984263bc
MD
210 vput(vp);
211 return (EBUSY);
212 }
213 vp->v_flag |= VMOUNT;
984263bc 214 mp->mnt_flag |=
ab2eb4eb 215 uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
5fd012e0 216 VOP_UNLOCK(vp, 0, td);
984263bc
MD
217 goto update;
218 }
219 /*
220 * If the user is not root, ensure that they own the directory
221 * onto which we are attempting to mount.
222 */
3b568787 223 if ((error = VOP_GETATTR(vp, &va, td)) ||
984263bc 224 (va.va_uid != p->p_ucred->cr_uid &&
dadab5e9 225 (error = suser(td)))) {
21739618 226 cache_drop(ncp);
984263bc
MD
227 vput(vp);
228 return (error);
229 }
3b568787 230 if ((error = vinvalbuf(vp, V_SAVE, td, 0, 0)) != 0) {
21739618 231 cache_drop(ncp);
984263bc
MD
232 vput(vp);
233 return (error);
234 }
235 if (vp->v_type != VDIR) {
21739618 236 cache_drop(ncp);
984263bc
MD
237 vput(vp);
238 return (ENOTDIR);
239 }
ab2eb4eb 240 if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
21739618 241 cache_drop(ncp);
984263bc
MD
242 vput(vp);
243 return (error);
244 }
21739618 245 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
984263bc
MD
246 if (!strcmp(vfsp->vfc_name, fstypename))
247 break;
21739618 248 }
984263bc
MD
249 if (vfsp == NULL) {
250 linker_file_t lf;
251
252 /* Only load modules for root (very important!) */
dadab5e9 253 if ((error = suser(td)) != 0) {
21739618 254 cache_drop(ncp);
984263bc
MD
255 vput(vp);
256 return error;
257 }
258 error = linker_load_file(fstypename, &lf);
259 if (error || lf == NULL) {
21739618 260 cache_drop(ncp);
984263bc
MD
261 vput(vp);
262 if (lf == NULL)
263 error = ENODEV;
264 return error;
265 }
266 lf->userrefs++;
267 /* lookup again, see if the VFS was loaded */
21739618 268 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
984263bc
MD
269 if (!strcmp(vfsp->vfc_name, fstypename))
270 break;
21739618 271 }
984263bc
MD
272 if (vfsp == NULL) {
273 lf->userrefs--;
274 linker_file_unload(lf);
21739618 275 cache_drop(ncp);
984263bc
MD
276 vput(vp);
277 return (ENODEV);
278 }
279 }
984263bc
MD
280 if ((vp->v_flag & VMOUNT) != 0 ||
281 vp->v_mountedhere != NULL) {
21739618 282 cache_drop(ncp);
984263bc
MD
283 vput(vp);
284 return (EBUSY);
285 }
286 vp->v_flag |= VMOUNT;
984263bc
MD
287
288 /*
289 * Allocate and initialize the filesystem.
290 */
0961aa92 291 mp = malloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
984263bc
MD
292 TAILQ_INIT(&mp->mnt_nvnodelist);
293 TAILQ_INIT(&mp->mnt_reservedvnlist);
2281065e 294 TAILQ_INIT(&mp->mnt_jlist);
984263bc 295 mp->mnt_nvnodelistsize = 0;
377d4740 296 lockinit(&mp->mnt_lock, 0, "vfslock", 0, LK_NOPAUSE);
861905fb 297 vfs_busy(mp, LK_NOWAIT, td);
984263bc
MD
298 mp->mnt_op = vfsp->vfc_vfsops;
299 mp->mnt_vfc = vfsp;
300 vfsp->vfc_refcount++;
301 mp->mnt_stat.f_type = vfsp->vfc_typenum;
302 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
303 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
304 mp->mnt_vnodecovered = vp;
305 mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
306 mp->mnt_iosize_max = DFLTPHYS;
5fd012e0 307 VOP_UNLOCK(vp, 0, td);
984263bc
MD
308update:
309 /*
310 * Set the mount level flags.
311 */
ab2eb4eb 312 if (uap->flags & MNT_RDONLY)
984263bc
MD
313 mp->mnt_flag |= MNT_RDONLY;
314 else if (mp->mnt_flag & MNT_RDONLY)
315 mp->mnt_kern_flag |= MNTK_WANTRDWR;
316 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
317 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
318 MNT_NOSYMFOLLOW | MNT_IGNORE |
319 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
ab2eb4eb 320 mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
984263bc
MD
321 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
322 MNT_NOSYMFOLLOW | MNT_IGNORE |
323 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
324 /*
325 * Mount the filesystem.
326 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
fad57d0e 327 * get.
984263bc 328 */
ab2eb4eb 329 error = VFS_MOUNT(mp, uap->path, uap->data, td);
984263bc
MD
330 if (mp->mnt_flag & MNT_UPDATE) {
331 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
332 mp->mnt_flag &= ~MNT_RDONLY;
333 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
334 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
335 if (error) {
336 mp->mnt_flag = flag;
337 mp->mnt_kern_flag = flag2;
338 }
dadab5e9 339 vfs_unbusy(mp, td);
984263bc 340 vp->v_flag &= ~VMOUNT;
984263bc 341 vrele(vp);
21739618 342 cache_drop(ncp);
984263bc
MD
343 return (error);
344 }
5fd012e0 345 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc 346 /*
21739618
MD
347 * Put the new filesystem on the mount list after root. The mount
348 * point gets its own mnt_ncp which is a special ncp linking the
349 * vnode-under to the root of the new mount. The lookup code
350 * detects the mount point going forward and detects the special
351 * mnt_ncp via NCP_MOUNTPT going backwards.
8e005a45
MD
352 *
353 * It is not necessary to invalidate or purge the vnode underneath
354 * because elements under the mount will be given their own glue
355 * namecache record.
984263bc 356 */
984263bc 357 if (!error) {
21739618
MD
358 nlc.nlc_nameptr = "";
359 nlc.nlc_namelen = 0;
360 mp->mnt_ncp = cache_nlookup(ncp, &nlc);
2386d5b6 361 cache_setunresolved(mp->mnt_ncp);
21739618 362 mp->mnt_ncp->nc_flag |= NCF_MOUNTPT;
646a1cda 363 mp->mnt_ncp->nc_mount = mp;
21739618
MD
364 cache_drop(ncp);
365 /* XXX get the root of the fs and cache_setvp(mnt_ncp...) */
984263bc
MD
366 vp->v_flag &= ~VMOUNT;
367 vp->v_mountedhere = mp;
861905fb 368 mountlist_insert(mp, MNTINS_LAST);
21739618
MD
369 checkdirs(vp, mp->mnt_ncp);
370 cache_unlock(mp->mnt_ncp); /* leave ref intact */
5fd012e0 371 VOP_UNLOCK(vp, 0, td);
41a01a4d 372 error = vfs_allocate_syncvnode(mp);
dadab5e9
MD
373 vfs_unbusy(mp, td);
374 if ((error = VFS_START(mp, 0, td)) != 0)
984263bc
MD
375 vrele(vp);
376 } else {
6ddb7618
MD
377 vfs_rm_vnodeops(&mp->mnt_vn_coherency_ops);
378 vfs_rm_vnodeops(&mp->mnt_vn_journal_ops);
379 vfs_rm_vnodeops(&mp->mnt_vn_norm_ops);
0961aa92
MD
380 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
381 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
984263bc 382 vp->v_flag &= ~VMOUNT;
984263bc 383 mp->mnt_vfc->vfc_refcount--;
dadab5e9 384 vfs_unbusy(mp, td);
e38462a2 385 free(mp, M_MOUNT);
21739618 386 cache_drop(ncp);
984263bc
MD
387 vput(vp);
388 }
389 return (error);
390}
391
392/*
393 * Scan all active processes to see if any of them have a current
394 * or root directory onto which the new filesystem has just been
395 * mounted. If so, replace them with the new mount point.
21739618
MD
396 *
397 * The passed ncp is ref'd and locked (from the mount code) and
398 * must be associated with the vnode representing the root of the
399 * mount point.
984263bc
MD
400 */
401static void
21739618 402checkdirs(struct vnode *olddp, struct namecache *ncp)
984263bc
MD
403{
404 struct filedesc *fdp;
405 struct vnode *newdp;
690a3127 406 struct mount *mp;
984263bc
MD
407 struct proc *p;
408
409 if (olddp->v_usecount == 1)
410 return;
690a3127
MD
411 mp = olddp->v_mountedhere;
412 if (VFS_ROOT(mp, &newdp))
984263bc 413 panic("mount: lost mount");
21739618
MD
414 cache_setvp(ncp, newdp);
415
416 if (rootvnode == olddp) {
417 vref(newdp);
418 vfs_cache_setroot(newdp, cache_hold(ncp));
419 }
420
f62004ad 421 FOREACH_PROC_IN_SYSTEM(p) {
984263bc
MD
422 fdp = p->p_fd;
423 if (fdp->fd_cdir == olddp) {
424 vrele(fdp->fd_cdir);
597aea93 425 vref(newdp);
984263bc 426 fdp->fd_cdir = newdp;
690a3127 427 cache_drop(fdp->fd_ncdir);
21739618 428 fdp->fd_ncdir = cache_hold(ncp);
984263bc
MD
429 }
430 if (fdp->fd_rdir == olddp) {
431 vrele(fdp->fd_rdir);
597aea93 432 vref(newdp);
984263bc 433 fdp->fd_rdir = newdp;
690a3127 434 cache_drop(fdp->fd_nrdir);
21739618 435 fdp->fd_nrdir = cache_hold(ncp);
984263bc
MD
436 }
437 }
984263bc
MD
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 */
41c20dac
MD
447/*
448 * umount_args(char *path, int flags)
449 */
984263bc
MD
450/* ARGSUSED */
451int
41c20dac 452unmount(struct unmount_args *uap)
984263bc 453{
dadab5e9
MD
454 struct thread *td = curthread;
455 struct proc *p = td->td_proc;
41c20dac 456 struct vnode *vp;
984263bc
MD
457 struct mount *mp;
458 int error;
fad57d0e 459 struct nlookupdata nd;
984263bc 460
dadab5e9 461 KKASSERT(p);
84c3a710
HP
462 if (p->p_ucred->cr_prison != NULL)
463 return (EPERM);
464 if (usermount == 0 && (error = suser(td)))
465 return (error);
466
fad57d0e 467 vp = NULL;
ab2eb4eb 468 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
469 if (error == 0)
470 error = nlookup(&nd);
471 if (error == 0)
472 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
473 nlookup_done(&nd);
474 if (error)
984263bc 475 return (error);
fad57d0e 476
984263bc
MD
477 mp = vp->v_mount;
478
479 /*
480 * Only root, or the user that did the original mount is
481 * permitted to unmount this filesystem.
482 */
483 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
dadab5e9 484 (error = suser(td))) {
984263bc
MD
485 vput(vp);
486 return (error);
487 }
488
489 /*
490 * Don't allow unmounting the root file system.
491 */
492 if (mp->mnt_flag & MNT_ROOTFS) {
493 vput(vp);
494 return (EINVAL);
495 }
496
497 /*
498 * Must be the root of the filesystem
499 */
500 if ((vp->v_flag & VROOT) == 0) {
501 vput(vp);
502 return (EINVAL);
503 }
504 vput(vp);
ab2eb4eb 505 return (dounmount(mp, uap->flags, td));
984263bc
MD
506}
507
508/*
509 * Do the actual file system unmount.
510 */
861905fb
MD
511static int
512dounmount_interlock(struct mount *mp)
513{
514 if (mp->mnt_kern_flag & MNTK_UNMOUNT)
515 return (EBUSY);
516 mp->mnt_kern_flag |= MNTK_UNMOUNT;
517 return(0);
518}
519
984263bc 520int
dadab5e9 521dounmount(struct mount *mp, int flags, struct thread *td)
984263bc
MD
522{
523 struct vnode *coveredvp;
524 int error;
525 int async_flag;
526
861905fb
MD
527 /*
528 * Exclusive access for unmounting purposes
529 */
530 if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
531 return (error);
532
533 /*
534 * Allow filesystems to detect that a forced unmount is in progress.
535 */
984263bc
MD
536 if (flags & MNT_FORCE)
537 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
861905fb
MD
538 error = lockmgr(&mp->mnt_lock, LK_DRAIN |
539 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), NULL, td);
984263bc
MD
540 if (error) {
541 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
542 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 543 wakeup(mp);
984263bc
MD
544 return (error);
545 }
546
547 if (mp->mnt_flag & MNT_EXPUBLIC)
548 vfs_setpublicfs(NULL, NULL, NULL);
549
550 vfs_msync(mp, MNT_WAIT);
551 async_flag = mp->mnt_flag & MNT_ASYNC;
552 mp->mnt_flag &=~ MNT_ASYNC;
553 cache_purgevfs(mp); /* remove cache entries for this file sys */
554 if (mp->mnt_syncer != NULL)
555 vrele(mp->mnt_syncer);
556 if (((mp->mnt_flag & MNT_RDONLY) ||
3b568787 557 (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) ||
984263bc 558 (flags & MNT_FORCE))
dadab5e9 559 error = VFS_UNMOUNT(mp, flags, td);
984263bc 560 if (error) {
41a01a4d
MD
561 if (mp->mnt_syncer == NULL)
562 vfs_allocate_syncvnode(mp);
984263bc
MD
563 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
564 mp->mnt_flag |= async_flag;
861905fb 565 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_REENABLE, NULL, td);
984263bc 566 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 567 wakeup(mp);
984263bc
MD
568 return (error);
569 }
432b8263
MD
570 /*
571 * Clean up any journals still associated with the mount after
572 * filesystem activity has ceased.
573 */
574 journal_remove_all_journals(mp,
575 ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));
576
861905fb 577 mountlist_remove(mp);
0961aa92
MD
578
579 /*
580 * Remove any installed vnode ops here so the individual VFSs don't
581 * have to.
582 */
6ddb7618
MD
583 vfs_rm_vnodeops(&mp->mnt_vn_coherency_ops);
584 vfs_rm_vnodeops(&mp->mnt_vn_journal_ops);
585 vfs_rm_vnodeops(&mp->mnt_vn_norm_ops);
0961aa92
MD
586 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
587 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
588
984263bc 589 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
7ea21ed1 590 coveredvp->v_mountedhere = NULL;
984263bc 591 vrele(coveredvp);
21739618
MD
592 cache_drop(mp->mnt_ncp);
593 mp->mnt_ncp = NULL;
984263bc
MD
594 }
595 mp->mnt_vfc->vfc_refcount--;
596 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
597 panic("unmount: dangling vnode");
861905fb 598 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, td);
984263bc 599 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2
DR
600 wakeup(mp);
601 free(mp, M_MOUNT);
984263bc
MD
602 return (0);
603}
604
605/*
606 * Sync each mounted filesystem.
607 */
984263bc
MD
608
609#ifdef DEBUG
610static int syncprt = 0;
611SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
6bdbb368 612#endif /* DEBUG */
984263bc 613
861905fb
MD
614static int sync_callback(struct mount *mp, void *data);
615
984263bc
MD
616/* ARGSUSED */
617int
41c20dac 618sync(struct sync_args *uap)
984263bc 619{
861905fb 620 mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
6bdbb368 621#ifdef DEBUG
861905fb
MD
622 /*
623 * print out buffer pool stat information on each sync() call.
624 */
984263bc
MD
625 if (syncprt)
626 vfs_bufstats();
6bdbb368 627#endif /* DEBUG */
984263bc
MD
628 return (0);
629}
630
861905fb
MD
631static
632int
633sync_callback(struct mount *mp, void *data __unused)
634{
635 int asyncflag;
636
637 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
638 asyncflag = mp->mnt_flag & MNT_ASYNC;
639 mp->mnt_flag &= ~MNT_ASYNC;
640 vfs_msync(mp, MNT_NOWAIT);
641 VFS_SYNC(mp, MNT_NOWAIT, curthread);
642 mp->mnt_flag |= asyncflag;
643 }
644 return(0);
645}
646
984263bc
MD
647/* XXX PRISON: could be per prison flag */
648static int prison_quotas;
649#if 0
650SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
651#endif
652
653/*
41c20dac
MD
654 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
655 *
984263bc
MD
656 * Change filesystem quotas.
657 */
984263bc
MD
658/* ARGSUSED */
659int
41c20dac 660quotactl(struct quotactl_args *uap)
984263bc 661{
fad57d0e
MD
662 struct nlookupdata nd;
663 struct thread *td;
664 struct proc *p;
41c20dac 665 struct mount *mp;
984263bc 666 int error;
984263bc 667
fad57d0e
MD
668 td = curthread;
669 p = td->td_proc;
41c20dac 670 if (p->p_ucred->cr_prison && !prison_quotas)
984263bc 671 return (EPERM);
fad57d0e 672
ab2eb4eb 673 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
674 if (error == 0)
675 error = nlookup(&nd);
676 if (error == 0) {
677 mp = nd.nl_ncp->nc_mount;
ab2eb4eb
DR
678 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
679 uap->arg, nd.nl_td);
fad57d0e
MD
680 }
681 nlookup_done(&nd);
682 return (error);
984263bc
MD
683}
684
949ecb9b 685/*
2281065e 686 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
949ecb9b
MD
687 * void *buf, int buflen)
688 *
689 * This function operates on a mount point and executes the specified
690 * operation using the specified control data, and possibly returns data.
691 *
692 * The actual number of bytes stored in the result buffer is returned, 0
693 * if none, otherwise an error is returned.
694 */
695/* ARGSUSED */
696int
697mountctl(struct mountctl_args *uap)
698{
699 struct thread *td = curthread;
700 struct proc *p = td->td_proc;
2281065e
MD
701 struct filedesc *fdp = p->p_fd;
702 struct file *fp;
949ecb9b
MD
703 void *ctl = NULL;
704 void *buf = NULL;
705 char *path = NULL;
706 int error;
707
708 /*
709 * Sanity and permissions checks. We must be root.
710 */
711 KKASSERT(p);
712 if (p->p_ucred->cr_prison != NULL)
713 return (EPERM);
714 if ((error = suser(td)) != 0)
715 return (error);
716
717 /*
718 * Argument length checks
719 */
39b13188 720 if (uap->ctllen < 0 || uap->ctllen > 1024)
949ecb9b 721 return (EINVAL);
39b13188 722 if (uap->buflen < 0 || uap->buflen > 16 * 1024)
949ecb9b
MD
723 return (EINVAL);
724 if (uap->path == NULL)
725 return (EINVAL);
726
727 /*
728 * Allocate the necessary buffers and copyin data
729 */
730 path = zalloc(namei_zone);
731 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
732 if (error)
733 goto done;
734
735 if (uap->ctllen) {
736 ctl = malloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
737 error = copyin(uap->ctl, ctl, uap->ctllen);
738 if (error)
739 goto done;
740 }
741 if (uap->buflen)
742 buf = malloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
743
2281065e
MD
744 /*
745 * Validate the descriptor
746 */
747 if (uap->fd == -1) {
748 fp = NULL;
749 } else if ((u_int)uap->fd >= fdp->fd_nfiles ||
0679adc4 750 (fp = fdp->fd_files[uap->fd].fp) == NULL) {
2281065e
MD
751 error = EBADF;
752 goto done;
753 }
754 if (fp)
755 fhold(fp);
756
949ecb9b
MD
757 /*
758 * Execute the internal kernel function and clean up.
759 */
2281065e
MD
760 error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
761 if (fp)
762 fdrop(fp, td);
949ecb9b
MD
763 if (error == 0 && uap->sysmsg_result > 0)
764 error = copyout(buf, uap->buf, uap->sysmsg_result);
765done:
766 if (path)
767 zfree(namei_zone, path);
768 if (ctl)
769 free(ctl, M_TEMP);
770 if (buf)
771 free(buf, M_TEMP);
772 return (error);
773}
774
775/*
776 * Execute a mount control operation by resolving the path to a mount point
777 * and calling vop_mountctl().
778 */
779int
2281065e
MD
780kern_mountctl(const char *path, int op, struct file *fp,
781 const void *ctl, int ctllen,
949ecb9b
MD
782 void *buf, int buflen, int *res)
783{
949ecb9b
MD
784 struct vnode *vp;
785 struct mount *mp;
786 struct nlookupdata nd;
787 int error;
788
789 *res = 0;
790 vp = NULL;
791 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
792 if (error == 0)
793 error = nlookup(&nd);
794 if (error == 0)
795 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
796 nlookup_done(&nd);
797 if (error)
798 return (error);
799
800 mp = vp->v_mount;
801
802 /*
803 * Must be the root of the filesystem
804 */
805 if ((vp->v_flag & VROOT) == 0) {
806 vput(vp);
807 return (EINVAL);
808 }
2281065e 809 error = vop_mountctl(mp->mnt_vn_use_ops, op, fp, ctl, ctllen,
949ecb9b
MD
810 buf, buflen, res);
811 vput(vp);
812 return (error);
813}
814
984263bc 815int
fad57d0e 816kern_statfs(struct nlookupdata *nd, struct statfs *buf)
984263bc 817{
dadab5e9 818 struct thread *td = curthread;
75ffff0d 819 struct proc *p = td->td_proc;
41c20dac
MD
820 struct mount *mp;
821 struct statfs *sp;
75ffff0d 822 char *fullpath, *freepath;
984263bc 823 int error;
984263bc 824
fad57d0e 825 if ((error = nlookup(nd)) != 0)
984263bc 826 return (error);
fad57d0e 827 mp = nd->nl_ncp->nc_mount;
984263bc 828 sp = &mp->mnt_stat;
fad57d0e 829 if ((error = VFS_STATFS(mp, sp, td)) != 0)
984263bc 830 return (error);
75ffff0d
JS
831
832 error = cache_fullpath(p, mp->mnt_ncp, &fullpath, &freepath);
833 if (error)
834 return(error);
835 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
836 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
837 free(freepath, M_TEMP);
838
984263bc 839 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509
DRJ
840 bcopy(sp, buf, sizeof(*buf));
841 /* Only root should have access to the fsid's. */
842 if (suser(td))
843 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
844 return (0);
984263bc
MD
845}
846
847/*
9697c509 848 * statfs_args(char *path, struct statfs *buf)
41c20dac 849 *
984263bc
MD
850 * Get filesystem statistics.
851 */
984263bc 852int
9697c509
DRJ
853statfs(struct statfs_args *uap)
854{
fad57d0e 855 struct nlookupdata nd;
9697c509
DRJ
856 struct statfs buf;
857 int error;
858
fad57d0e
MD
859 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
860 if (error == 0)
861 error = kern_statfs(&nd, &buf);
862 nlookup_done(&nd);
9697c509
DRJ
863 if (error == 0)
864 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
865 return (error);
866}
867
868int
869kern_fstatfs(int fd, struct statfs *buf)
984263bc 870{
dadab5e9
MD
871 struct thread *td = curthread;
872 struct proc *p = td->td_proc;
984263bc
MD
873 struct file *fp;
874 struct mount *mp;
1fd87d54 875 struct statfs *sp;
75ffff0d 876 char *fullpath, *freepath;
984263bc 877 int error;
984263bc 878
dadab5e9 879 KKASSERT(p);
9697c509
DRJ
880 error = getvnode(p->p_fd, fd, &fp);
881 if (error)
984263bc
MD
882 return (error);
883 mp = ((struct vnode *)fp->f_data)->v_mount;
884 if (mp == NULL)
885 return (EBADF);
886 sp = &mp->mnt_stat;
dadab5e9 887 error = VFS_STATFS(mp, sp, td);
984263bc
MD
888 if (error)
889 return (error);
75ffff0d
JS
890
891 error = cache_fullpath(p, mp->mnt_ncp, &fullpath, &freepath);
892 if (error)
893 return(error);
894 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
895 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
896 free(freepath, M_TEMP);
897
984263bc 898 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509 899 bcopy(sp, buf, sizeof(*buf));
75ffff0d 900
9697c509
DRJ
901 /* Only root should have access to the fsid's. */
902 if (suser(td))
903 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
904 return (0);
905}
906
907/*
908 * fstatfs_args(int fd, struct statfs *buf)
909 *
910 * Get filesystem statistics.
911 */
912int
913fstatfs(struct fstatfs_args *uap)
914{
915 struct statfs buf;
916 int error;
917
918 error = kern_fstatfs(uap->fd, &buf);
919
920 if (error == 0)
921 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
922 return (error);
984263bc
MD
923}
924
925/*
41c20dac
MD
926 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
927 *
984263bc
MD
928 * Get statistics on all filesystems.
929 */
861905fb
MD
930
931struct getfsstat_info {
932 struct statfs *sfsp;
933 long count;
934 long maxcount;
935 int error;
936 int flags;
937 int is_chrooted;
938 struct thread *td;
939 struct proc *p;
940};
941
942static int getfsstat_callback(struct mount *, void *);
943
41c20dac 944/* ARGSUSED */
984263bc 945int
41c20dac 946getfsstat(struct getfsstat_args *uap)
984263bc 947{
dadab5e9 948 struct thread *td = curthread;
75ffff0d 949 struct proc *p = td->td_proc;
861905fb 950 struct getfsstat_info info;
75ffff0d 951
861905fb 952 bzero(&info, sizeof(info));
75ffff0d 953 if (p != NULL && (p->p_fd->fd_nrdir->nc_flag & NCF_ROOT) == 0)
861905fb
MD
954 info.is_chrooted = 1;
955 else
956 info.is_chrooted = 0;
957
958 info.maxcount = uap->bufsize / sizeof(struct statfs);
959 info.sfsp = uap->buf;
960 info.count = 0;
961 info.flags = uap->flags;
962 info.td = td;
963 info.p = p;
964
965 mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
966 if (info.sfsp && info.count > info.maxcount)
967 uap->sysmsg_result = info.maxcount;
75ffff0d 968 else
861905fb
MD
969 uap->sysmsg_result = info.count;
970 return (info.error);
971}
972
973static int
974getfsstat_callback(struct mount *mp, void *data)
975{
976 struct getfsstat_info *info = data;
977 struct statfs *sp;
978 char *freepath;
979 char *fullpath;
980 int error;
981
982 if (info->sfsp && info->count < info->maxcount) {
983 if (info->is_chrooted && !chroot_visible_mnt(mp, info->p))
984 return(0);
985 sp = &mp->mnt_stat;
986
987 /*
988 * If MNT_NOWAIT or MNT_LAZY is specified, do not
989 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
990 * overrides MNT_WAIT.
991 */
992 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
993 (info->flags & MNT_WAIT)) &&
994 (error = VFS_STATFS(mp, sp, info->td))) {
995 return(0);
996 }
997 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
998
999 error = cache_fullpath(info->p, mp->mnt_ncp,
1000 &fullpath, &freepath);
1001 if (error) {
1002 info->error = error;
1003 return(-1);
984263bc 1004 }
861905fb
MD
1005 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1006 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1007 free(freepath, M_TEMP);
1008
1009 error = copyout(sp, info->sfsp, sizeof(*sp));
1010 if (error) {
1011 info->error = error;
1012 return (-1);
984263bc 1013 }
861905fb 1014 ++info->sfsp;
984263bc 1015 }
861905fb
MD
1016 info->count++;
1017 return(0);
984263bc
MD
1018}
1019
1020/*
41c20dac
MD
1021 * fchdir_args(int fd)
1022 *
984263bc
MD
1023 * Change current working directory to a given file descriptor.
1024 */
984263bc
MD
1025/* ARGSUSED */
1026int
41c20dac 1027fchdir(struct fchdir_args *uap)
984263bc 1028{
dadab5e9
MD
1029 struct thread *td = curthread;
1030 struct proc *p = td->td_proc;
41c20dac 1031 struct filedesc *fdp = p->p_fd;
21739618 1032 struct vnode *vp, *ovp;
984263bc
MD
1033 struct mount *mp;
1034 struct file *fp;
21739618
MD
1035 struct namecache *ncp, *oncp;
1036 struct namecache *nct;
984263bc
MD
1037 int error;
1038
ab2eb4eb 1039 if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
984263bc
MD
1040 return (error);
1041 vp = (struct vnode *)fp->f_data;
597aea93 1042 vref(vp);
5fd012e0 1043 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
21739618 1044 if (vp->v_type != VDIR || fp->f_ncp == NULL)
984263bc
MD
1045 error = ENOTDIR;
1046 else
dadab5e9 1047 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td);
984263bc
MD
1048 if (error) {
1049 vput(vp);
1050 return (error);
1051 }
21739618
MD
1052 ncp = cache_hold(fp->f_ncp);
1053 while (!error && (mp = vp->v_mountedhere) != NULL) {
1054 error = nlookup_mp(mp, &nct);
1055 if (error == 0) {
1056 cache_unlock(nct); /* leave ref intact */
1057 vput(vp);
1058 vp = nct->nc_vp;
5fd012e0 1059 error = vget(vp, LK_SHARED, td);
21739618
MD
1060 KKASSERT(error == 0);
1061 cache_drop(ncp);
1062 ncp = nct;
1063 }
1064 }
1065 if (error == 0) {
1066 ovp = fdp->fd_cdir;
1067 oncp = fdp->fd_ncdir;
5fd012e0 1068 VOP_UNLOCK(vp, 0, td); /* leave ref intact */
21739618
MD
1069 fdp->fd_cdir = vp;
1070 fdp->fd_ncdir = ncp;
1071 cache_drop(oncp);
1072 vrele(ovp);
1073 } else {
1074 cache_drop(ncp);
1075 vput(vp);
1076 }
1077 return (error);
984263bc
MD
1078}
1079
9697c509 1080int
21739618 1081kern_chdir(struct nlookupdata *nd)
9697c509
DRJ
1082{
1083 struct thread *td = curthread;
1084 struct proc *p = td->td_proc;
1085 struct filedesc *fdp = p->p_fd;
21739618
MD
1086 struct vnode *vp, *ovp;
1087 struct namecache *oncp;
9697c509
DRJ
1088 int error;
1089
21739618 1090 if ((error = nlookup(nd)) != 0)
9697c509 1091 return (error);
21739618
MD
1092 if ((vp = nd->nl_ncp->nc_vp) == NULL)
1093 return (ENOENT);
5fd012e0 1094 if ((error = vget(vp, LK_SHARED, td)) != 0)
21739618
MD
1095 return (error);
1096
1097 error = checkvp_chdir(vp, td);
5fd012e0 1098 VOP_UNLOCK(vp, 0, td);
21739618
MD
1099 if (error == 0) {
1100 ovp = fdp->fd_cdir;
1101 oncp = fdp->fd_ncdir;
1102 cache_unlock(nd->nl_ncp); /* leave reference intact */
1103 fdp->fd_ncdir = nd->nl_ncp;
1104 fdp->fd_cdir = vp;
1105 cache_drop(oncp);
1106 vrele(ovp);
1107 nd->nl_ncp = NULL;
1108 } else {
1109 vrele(vp);
e24b948e 1110 }
e24b948e 1111 return (error);
9697c509
DRJ
1112}
1113
984263bc 1114/*
41c20dac
MD
1115 * chdir_args(char *path)
1116 *
984263bc
MD
1117 * Change current working directory (``.'').
1118 */
984263bc 1119int
41c20dac 1120chdir(struct chdir_args *uap)
984263bc 1121{
21739618 1122 struct nlookupdata nd;
9697c509 1123 int error;
984263bc 1124
21739618 1125 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e 1126 if (error == 0)
21739618 1127 error = kern_chdir(&nd);
fad57d0e 1128 nlookup_done(&nd);
9697c509 1129 return (error);
984263bc
MD
1130}
1131
1132/*
1133 * Helper function for raised chroot(2) security function: Refuse if
1134 * any filedescriptors are open directories.
1135 */
1136static int
1137chroot_refuse_vdir_fds(fdp)
1138 struct filedesc *fdp;
1139{
1140 struct vnode *vp;
1141 struct file *fp;
1142 int error;
1143 int fd;
1144
1145 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
1146 error = getvnode(fdp, fd, &fp);
1147 if (error)
1148 continue;
1149 vp = (struct vnode *)fp->f_data;
1150 if (vp->v_type != VDIR)
1151 continue;
1152 return(EPERM);
1153 }
1154 return (0);
1155}
1156
1157/*
1158 * This sysctl determines if we will allow a process to chroot(2) if it
1159 * has a directory open:
1160 * 0: disallowed for all processes.
1161 * 1: allowed for processes that were not already chroot(2)'ed.
1162 * 2: allowed for all processes.
1163 */
1164
1165static int chroot_allow_open_directories = 1;
1166
1167SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1168 &chroot_allow_open_directories, 0, "");
1169
e24b948e 1170/*
21739618
MD
1171 * chroot to the specified namecache entry. We obtain the vp from the
1172 * namecache data. The passed ncp must be locked and referenced and will
1173 * remain locked and referenced on return.
e24b948e 1174 */
53dd6631
JS
1175int
1176kern_chroot(struct namecache *ncp)
e24b948e
MD
1177{
1178 struct thread *td = curthread;
1179 struct proc *p = td->td_proc;
1180 struct filedesc *fdp = p->p_fd;
21739618 1181 struct vnode *vp;
e24b948e
MD
1182 int error;
1183
1184 /*
1185 * Only root can chroot
1186 */
1187 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
1188 return (error);
1189
1190 /*
1191 * Disallow open directory descriptors (fchdir() breakouts).
1192 */
1193 if (chroot_allow_open_directories == 0 ||
1194 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1195 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1196 return (error);
1197 }
21739618
MD
1198 if ((vp = ncp->nc_vp) == NULL)
1199 return (ENOENT);
1200
5fd012e0 1201 if ((error = vget(vp, LK_SHARED, td)) != 0)
21739618 1202 return (error);
e24b948e
MD
1203
1204 /*
1205 * Check the validity of vp as a directory to change to and
1206 * associate it with rdir/jdir.
1207 */
21739618 1208 error = checkvp_chdir(vp, td);
5fd012e0 1209 VOP_UNLOCK(vp, 0, td); /* leave reference intact */
21739618 1210 if (error == 0) {
e24b948e 1211 vrele(fdp->fd_rdir);
21739618 1212 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */
690a3127 1213 cache_drop(fdp->fd_nrdir);
21739618 1214 fdp->fd_nrdir = cache_hold(ncp);
e24b948e
MD
1215 if (fdp->fd_jdir == NULL) {
1216 fdp->fd_jdir = vp;
597aea93 1217 vref(fdp->fd_jdir);
21739618 1218 fdp->fd_njdir = cache_hold(ncp);
e24b948e 1219 }
21739618
MD
1220 } else {
1221 vrele(vp);
e24b948e
MD
1222 }
1223 return (error);
1224}
1225
984263bc 1226/*
41c20dac
MD
1227 * chroot_args(char *path)
1228 *
984263bc
MD
1229 * Change notion of root (``/'') directory.
1230 */
984263bc
MD
1231/* ARGSUSED */
1232int
41c20dac 1233chroot(struct chroot_args *uap)
984263bc 1234{
dadab5e9 1235 struct thread *td = curthread;
21739618 1236 struct nlookupdata nd;
e24b948e 1237 int error;
984263bc 1238
e24b948e 1239 KKASSERT(td->td_proc);
21739618 1240 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
53dd6631
JS
1241 if (error) {
1242 nlookup_done(&nd);
1243 return(error);
1244 }
1245 error = nlookup(&nd);
1246 if (error == 0)
1247 error = kern_chroot(nd.nl_ncp);
fad57d0e 1248 nlookup_done(&nd);
53dd6631 1249 return(error);
984263bc
MD
1250}
1251
1252/*
e24b948e
MD
1253 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1254 * determine whether it is legal to chdir to the vnode. The vnode's state
1255 * is not changed by this call.
984263bc 1256 */
e24b948e
MD
1257int
1258checkvp_chdir(struct vnode *vp, struct thread *td)
984263bc 1259{
984263bc
MD
1260 int error;
1261
984263bc
MD
1262 if (vp->v_type != VDIR)
1263 error = ENOTDIR;
1264 else
e24b948e 1265 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred, td);
984263bc
MD
1266 return (error);
1267}
1268
984263bc 1269int
fad57d0e 1270kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
984263bc 1271{
dadab5e9
MD
1272 struct thread *td = curthread;
1273 struct proc *p = td->td_proc;
41c20dac 1274 struct filedesc *fdp = p->p_fd;
9697c509 1275 int cmode, flags;
984263bc 1276 struct file *nfp;
fad57d0e
MD
1277 struct file *fp;
1278 struct vnode *vp;
984263bc
MD
1279 int type, indx, error;
1280 struct flock lf;
984263bc 1281
984263bc
MD
1282 if ((oflags & O_ACCMODE) == O_ACCMODE)
1283 return (EINVAL);
1284 flags = FFLAGS(oflags);
fad57d0e 1285 error = falloc(p, &nfp, NULL);
984263bc
MD
1286 if (error)
1287 return (error);
1288 fp = nfp;
9697c509 1289 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
fad57d0e 1290
984263bc 1291 /*
fad57d0e
MD
1292 * XXX p_dupfd is a real mess. It allows a device to return a
1293 * file descriptor to be duplicated rather then doing the open
1294 * itself.
984263bc 1295 */
fad57d0e 1296 p->p_dupfd = -1;
21739618
MD
1297
1298 /*
fad57d0e
MD
1299 * Call vn_open() to do the lookup and assign the vnode to the
1300 * file pointer. vn_open() does not change the ref count on fp
1301 * and the vnode, on success, will be inherited by the file pointer
1302 * and unlocked.
21739618 1303 */
fad57d0e
MD
1304 nd->nl_flags |= NLC_LOCKVP;
1305 error = vn_open(nd, fp, flags, cmode);
1306 nlookup_done(nd);
984263bc 1307 if (error) {
984263bc
MD
1308 /*
1309 * handle special fdopen() case. bleh. dupfdopen() is
1310 * responsible for dropping the old contents of ofiles[indx]
1311 * if it succeeds.
fad57d0e
MD
1312 *
1313 * Note that if fsetfd() succeeds it will add a ref to fp
0679adc4 1314 * which represents the fd_files[] assignment. We must still
fad57d0e 1315 * drop our reference.
984263bc 1316 */
fad57d0e
MD
1317 if ((error == ENODEV || error == ENXIO) && p->p_dupfd >= 0) {
1318 if (fsetfd(p, fp, &indx) == 0) {
1319 error = dupfdopen(fdp, indx, p->p_dupfd, flags, error);
1320 if (error == 0) {
1321 *res = indx;
1322 fdrop(fp, td); /* our ref */
1323 return (0);
1324 }
0679adc4 1325 if (fdp->fd_files[indx].fp == fp) {
69908319 1326 funsetfd(fdp, indx);
0679adc4 1327 fdrop(fp, td); /* fd_files[] ref */
fad57d0e
MD
1328 }
1329 }
984263bc 1330 }
fad57d0e 1331 fdrop(fp, td); /* our ref */
984263bc
MD
1332 if (error == ERESTART)
1333 error = EINTR;
1334 return (error);
1335 }
fad57d0e
MD
1336
1337 /*
1338 * ref the vnode for ourselves so it can't be ripped out from under
1339 * is. XXX need an ND flag to request that the vnode be returned
1340 * anyway.
1341 */
1342 vp = (struct vnode *)fp->f_data;
1343 vref(vp);
1344 if ((error = fsetfd(p, fp, &indx)) != 0) {
1345 fdrop(fp, td);
1346 vrele(vp);
1347 return (error);
1348 }
1349
1350 /*
1351 * If no error occurs the vp will have been assigned to the file
1352 * pointer.
1353 */
984263bc 1354 p->p_dupfd = 0;
984263bc
MD
1355
1356 /*
1357 * There should be 2 references on the file, one from the descriptor
1358 * table, and one for us.
1359 *
1360 * Handle the case where someone closed the file (via its file
1361 * descriptor) while we were blocked. The end result should look
1362 * like opening the file succeeded but it was immediately closed.
1363 */
1364 if (fp->f_count == 1) {
0679adc4 1365 KASSERT(fdp->fd_files[indx].fp != fp,
984263bc 1366 ("Open file descriptor lost all refs"));
fad57d0e
MD
1367 vrele(vp);
1368 fo_close(fp, td);
dadab5e9 1369 fdrop(fp, td);
9697c509 1370 *res = indx;
984263bc
MD
1371 return 0;
1372 }
1373
984263bc
MD
1374 if (flags & (O_EXLOCK | O_SHLOCK)) {
1375 lf.l_whence = SEEK_SET;
1376 lf.l_start = 0;
1377 lf.l_len = 0;
1378 if (flags & O_EXLOCK)
1379 lf.l_type = F_WRLCK;
1380 else
1381 lf.l_type = F_RDLCK;
1382 type = F_FLOCK;
1383 if ((flags & FNONBLOCK) == 0)
1384 type |= F_WAIT;
fad57d0e 1385
984263bc
MD
1386 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1387 /*
1388 * lock request failed. Normally close the descriptor
1389 * but handle the case where someone might have dup()d
1390 * it when we weren't looking. One reference is
1391 * owned by the descriptor array, the other by us.
1392 */
fad57d0e 1393 vrele(vp);
0679adc4 1394 if (fdp->fd_files[indx].fp == fp) {
69908319 1395 funsetfd(fdp, indx);
dadab5e9 1396 fdrop(fp, td);
984263bc 1397 }
dadab5e9 1398 fdrop(fp, td);
984263bc
MD
1399 return (error);
1400 }
984263bc
MD
1401 fp->f_flag |= FHASLOCK;
1402 }
1403 /* assert that vn_open created a backing object if one is needed */
1404 KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0,
1405 ("open: vmio vnode has no backing object after vn_open"));
984263bc 1406
fad57d0e 1407 vrele(vp);
21739618 1408
984263bc
MD
1409 /*
1410 * release our private reference, leaving the one associated with the
1411 * descriptor table intact.
1412 */
dadab5e9 1413 fdrop(fp, td);
9697c509 1414 *res = indx;
984263bc
MD
1415 return (0);
1416}
1417
984263bc 1418/*
9697c509 1419 * open_args(char *path, int flags, int mode)
41c20dac 1420 *
9697c509
DRJ
1421 * Check permissions, allocate an open file structure,
1422 * and call the device open routine if any.
984263bc 1423 */
984263bc 1424int
9697c509 1425open(struct open_args *uap)
984263bc 1426{
fad57d0e 1427 struct nlookupdata nd;
9697c509
DRJ
1428 int error;
1429
fad57d0e
MD
1430 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1431 if (error == 0) {
1432 error = kern_open(&nd, uap->flags,
1433 uap->mode, &uap->sysmsg_result);
1434 }
1435 nlookup_done(&nd);
9697c509 1436 return (error);
984263bc 1437}
984263bc 1438
984263bc 1439int
fad57d0e 1440kern_mknod(struct nlookupdata *nd, int mode, int dev)
984263bc 1441{
fad57d0e 1442 struct namecache *ncp;
dadab5e9
MD
1443 struct thread *td = curthread;
1444 struct proc *p = td->td_proc;
41c20dac 1445 struct vnode *vp;
984263bc
MD
1446 struct vattr vattr;
1447 int error;
1448 int whiteout = 0;
984263bc 1449
dadab5e9
MD
1450 KKASSERT(p);
1451
9697c509 1452 switch (mode & S_IFMT) {
984263bc
MD
1453 case S_IFCHR:
1454 case S_IFBLK:
dadab5e9 1455 error = suser(td);
984263bc
MD
1456 break;
1457 default:
dadab5e9 1458 error = suser_cred(p->p_ucred, PRISON_ROOT);
984263bc
MD
1459 break;
1460 }
1461 if (error)
1462 return (error);
fad57d0e 1463
984263bc 1464 bwillwrite();
fad57d0e
MD
1465 nd->nl_flags |= NLC_CREATE;
1466 if ((error = nlookup(nd)) != 0)
984263bc 1467 return (error);
fad57d0e
MD
1468 ncp = nd->nl_ncp;
1469 if (ncp->nc_vp)
1470 return (EEXIST);
984263bc 1471
fad57d0e
MD
1472 VATTR_NULL(&vattr);
1473 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1474 vattr.va_rdev = dev;
1475 whiteout = 0;
1476
1477 switch (mode & S_IFMT) {
1478 case S_IFMT: /* used by badsect to flag bad sectors */
1479 vattr.va_type = VBAD;
1480 break;
1481 case S_IFCHR:
1482 vattr.va_type = VCHR;
1483 break;
1484 case S_IFBLK:
1485 vattr.va_type = VBLK;
1486 break;
1487 case S_IFWHT:
1488 whiteout = 1;
1489 break;
1490 default:
1491 error = EINVAL;
1492 break;
984263bc 1493 }
9697c509 1494 if (error == 0) {
fad57d0e
MD
1495 if (whiteout) {
1496 error = VOP_NWHITEOUT(ncp, nd->nl_cred, NAMEI_CREATE);
1497 } else {
1498 vp = NULL;
1499 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
984263bc 1500 if (error == 0)
fad57d0e 1501 vput(vp);
984263bc 1502 }
984263bc 1503 }
9697c509
DRJ
1504 return (error);
1505}
1506
1507/*
1508 * mknod_args(char *path, int mode, int dev)
1509 *
1510 * Create a special file.
1511 */
1512int
1513mknod(struct mknod_args *uap)
1514{
fad57d0e 1515 struct nlookupdata nd;
9697c509
DRJ
1516 int error;
1517
fad57d0e
MD
1518 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1519 if (error == 0)
1520 error = kern_mknod(&nd, uap->mode, uap->dev);
1521 nlookup_done(&nd);
984263bc
MD
1522 return (error);
1523}
1524
984263bc 1525int
fad57d0e 1526kern_mkfifo(struct nlookupdata *nd, int mode)
984263bc 1527{
fad57d0e 1528 struct namecache *ncp;
dadab5e9
MD
1529 struct thread *td = curthread;
1530 struct proc *p = td->td_proc;
984263bc 1531 struct vattr vattr;
fad57d0e 1532 struct vnode *vp;
984263bc 1533 int error;
984263bc
MD
1534
1535 bwillwrite();
fad57d0e
MD
1536
1537 nd->nl_flags |= NLC_CREATE;
1538 if ((error = nlookup(nd)) != 0)
984263bc 1539 return (error);
fad57d0e
MD
1540 ncp = nd->nl_ncp;
1541 if (ncp->nc_vp)
984263bc 1542 return (EEXIST);
fad57d0e 1543
984263bc
MD
1544 VATTR_NULL(&vattr);
1545 vattr.va_type = VFIFO;
136178b3 1546 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
fad57d0e
MD
1547 vp = NULL;
1548 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
984263bc 1549 if (error == 0)
fad57d0e 1550 vput(vp);
136178b3
DRJ
1551 return (error);
1552}
1553
1554/*
1555 * mkfifo_args(char *path, int mode)
1556 *
1557 * Create a named pipe.
1558 */
1559int
1560mkfifo(struct mkfifo_args *uap)
1561{
fad57d0e 1562 struct nlookupdata nd;
136178b3
DRJ
1563 int error;
1564
fad57d0e
MD
1565 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1566 if (error == 0)
1567 error = kern_mkfifo(&nd, uap->mode);
1568 nlookup_done(&nd);
984263bc
MD
1569 return (error);
1570}
1571
8c40f81e
MD
1572static int hardlink_check_uid = 0;
1573SYSCTL_INT(_kern, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1574 &hardlink_check_uid, 0,
1575 "Unprivileged processes cannot create hard links to files owned by other "
1576 "users");
1577static int hardlink_check_gid = 0;
1578SYSCTL_INT(_kern, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1579 &hardlink_check_gid, 0,
1580 "Unprivileged processes cannot create hard links to files owned by other "
1581 "groups");
1582
1583static int
1584can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
1585{
1586 struct vattr va;
1587 int error;
1588
1589 /*
1590 * Shortcut if disabled
1591 */
1592 if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
1593 return (0);
1594
1595 /*
1596 * root cred can always hardlink
1597 */
1598 if (suser_cred(cred, PRISON_ROOT) == 0)
1599 return (0);
1600
1601 /*
1602 * Otherwise only if the originating file is owned by the
1603 * same user or group. Note that any group is allowed if
1604 * the file is owned by the caller.
1605 */
1606 error = VOP_GETATTR(vp, &va, td);
1607 if (error != 0)
1608 return (error);
1609
1610 if (hardlink_check_uid) {
1611 if (cred->cr_uid != va.va_uid)
1612 return (EPERM);
1613 }
1614
1615 if (hardlink_check_gid) {
1616 if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
1617 return (EPERM);
1618 }
1619
1620 return (0);
1621}
1622
984263bc 1623int
fad57d0e 1624kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
984263bc 1625{
dadab5e9 1626 struct thread *td = curthread;
41c20dac 1627 struct vnode *vp;
984263bc
MD
1628 int error;
1629
fad57d0e
MD
1630 /*
1631 * Lookup the source and obtained a locked vnode.
1632 *
1633 * XXX relookup on vget failure / race ?
1634 */
984263bc 1635 bwillwrite();
fad57d0e 1636 if ((error = nlookup(nd)) != 0)
984263bc 1637 return (error);
fad57d0e
MD
1638 vp = nd->nl_ncp->nc_vp;
1639 KKASSERT(vp != NULL);
984263bc 1640 if (vp->v_type == VDIR)
fad57d0e
MD
1641 return (EPERM); /* POSIX */
1642 if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0)
1643 return (error);
1644
1645 /*
1646 * Unlock the source so we can lookup the target without deadlocking
1647 * (XXX vp is locked already, possible other deadlock?). The target
1648 * must not exist.
1649 */
1650 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
1651 nd->nl_flags &= ~NLC_NCPISLOCKED;
1652 cache_unlock(nd->nl_ncp);
1653
1654 linknd->nl_flags |= NLC_CREATE;
1655 if ((error = nlookup(linknd)) != 0) {
1656 vput(vp);
1657 return (error);
984263bc 1658 }
fad57d0e
MD
1659 if (linknd->nl_ncp->nc_vp) {
1660 vput(vp);
1661 return (EEXIST);
1662 }
1663
1664 /*
1665 * Finally run the new API VOP.
1666 */
8c40f81e
MD
1667 error = can_hardlink(vp, td, td->td_proc->p_ucred);
1668 if (error == 0)
1669 error = VOP_NLINK(linknd->nl_ncp, vp, linknd->nl_cred);
fad57d0e 1670 vput(vp);
984263bc
MD
1671 return (error);
1672}
1673
1674/*
9697c509 1675 * link_args(char *path, char *link)
41c20dac 1676 *
9697c509 1677 * Make a hard file link.
984263bc 1678 */
984263bc 1679int
9697c509
DRJ
1680link(struct link_args *uap)
1681{
fad57d0e 1682 struct nlookupdata nd, linknd;
9697c509
DRJ
1683 int error;
1684
fad57d0e
MD
1685 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1686 if (error == 0) {
1687 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
1688 if (error == 0)
1689 error = kern_link(&nd, &linknd);
1690 nlookup_done(&linknd);
1691 }
1692 nlookup_done(&nd);
9697c509
DRJ
1693 return (error);
1694}
1695
1696int
fad57d0e 1697kern_symlink(struct nlookupdata *nd, char *path, int mode)
984263bc 1698{
fad57d0e 1699 struct namecache *ncp;
984263bc 1700 struct vattr vattr;
fad57d0e 1701 struct vnode *vp;
984263bc 1702 int error;
984263bc 1703
984263bc 1704 bwillwrite();
fad57d0e
MD
1705 nd->nl_flags |= NLC_CREATE;
1706 if ((error = nlookup(nd)) != 0)
9697c509 1707 return (error);
fad57d0e
MD
1708 ncp = nd->nl_ncp;
1709 if (ncp->nc_vp)
9697c509 1710 return (EEXIST);
fad57d0e 1711
984263bc 1712 VATTR_NULL(&vattr);
fad57d0e
MD
1713 vattr.va_mode = mode;
1714 error = VOP_NSYMLINK(ncp, &vp, nd->nl_cred, &vattr, path);
984263bc 1715 if (error == 0)
fad57d0e 1716 vput(vp);
9697c509
DRJ
1717 return (error);
1718}
1719
1720/*
1721 * symlink(char *path, char *link)
1722 *
1723 * Make a symbolic link.
1724 */
1725int
1726symlink(struct symlink_args *uap)
1727{
1728 struct thread *td = curthread;
fad57d0e 1729 struct nlookupdata nd;
9697c509
DRJ
1730 char *path;
1731 int error;
fad57d0e 1732 int mode;
9697c509
DRJ
1733
1734 path = zalloc(namei_zone);
1735 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
e24b948e 1736 if (error == 0) {
fad57d0e
MD
1737 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
1738 if (error == 0) {
1739 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
1740 error = kern_symlink(&nd, path, mode);
1741 }
1742 nlookup_done(&nd);
e24b948e 1743 }
984263bc
MD
1744 zfree(namei_zone, path);
1745 return (error);
1746}
1747
1748/*
41c20dac
MD
1749 * undelete_args(char *path)
1750 *
984263bc
MD
1751 * Delete a whiteout from the filesystem.
1752 */
1753/* ARGSUSED */
1754int
41c20dac 1755undelete(struct undelete_args *uap)
984263bc 1756{
fad57d0e 1757 struct nlookupdata nd;
984263bc 1758 int error;
984263bc 1759
ab2eb4eb 1760 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
984263bc 1761 bwillwrite();
fad57d0e
MD
1762 nd.nl_flags |= NLC_DELETE;
1763 if (error == 0)
1764 error = nlookup(&nd);
b8997912 1765 if (error == 0)
fad57d0e
MD
1766 error = VOP_NWHITEOUT(nd.nl_ncp, nd.nl_cred, NAMEI_DELETE);
1767 nlookup_done(&nd);
984263bc
MD
1768 return (error);
1769}
1770
984263bc 1771int
fad57d0e 1772kern_unlink(struct nlookupdata *nd)
984263bc 1773{
fad57d0e 1774 struct namecache *ncp;
984263bc 1775 int error;
984263bc
MD
1776
1777 bwillwrite();
fad57d0e
MD
1778 nd->nl_flags |= NLC_DELETE;
1779 if ((error = nlookup(nd)) != 0)
984263bc 1780 return (error);
fad57d0e
MD
1781 ncp = nd->nl_ncp;
1782 error = VOP_NREMOVE(ncp, nd->nl_cred);
984263bc
MD
1783 return (error);
1784}
1785
1786/*
9697c509 1787 * unlink_args(char *path)
41c20dac 1788 *
9697c509 1789 * Delete a name from the filesystem.
984263bc 1790 */
984263bc 1791int
9697c509
DRJ
1792unlink(struct unlink_args *uap)
1793{
fad57d0e 1794 struct nlookupdata nd;
9697c509
DRJ
1795 int error;
1796
fad57d0e
MD
1797 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1798 if (error == 0)
1799 error = kern_unlink(&nd);
1800 nlookup_done(&nd);
9697c509
DRJ
1801 return (error);
1802}
1803
1804int
5a3fe67d 1805kern_lseek(int fd, off_t offset, int whence, off_t *res)
984263bc 1806{
dadab5e9
MD
1807 struct thread *td = curthread;
1808 struct proc *p = td->td_proc;
41c20dac
MD
1809 struct filedesc *fdp = p->p_fd;
1810 struct file *fp;
984263bc
MD
1811 struct vattr vattr;
1812 int error;
1813
b463648b 1814 if ((u_int)fd >= fdp->fd_nfiles ||
0679adc4 1815 (fp = fdp->fd_files[fd].fp) == NULL)
984263bc
MD
1816 return (EBADF);
1817 if (fp->f_type != DTYPE_VNODE)
1818 return (ESPIPE);
9697c509 1819 switch (whence) {
984263bc 1820 case L_INCR:
9697c509 1821 fp->f_offset += offset;
984263bc
MD
1822 break;
1823 case L_XTND:
3b568787 1824 error=VOP_GETATTR((struct vnode *)fp->f_data, &vattr, td);
984263bc
MD
1825 if (error)
1826 return (error);
9697c509 1827 fp->f_offset = offset + vattr.va_size;
984263bc
MD
1828 break;
1829 case L_SET:
9697c509 1830 fp->f_offset = offset;
984263bc
MD
1831 break;
1832 default:
1833 return (EINVAL);
1834 }
9697c509 1835 *res = fp->f_offset;
984263bc
MD
1836 return (0);
1837}
1838
984263bc 1839/*
9697c509 1840 * lseek_args(int fd, int pad, off_t offset, int whence)
41c20dac 1841 *
9697c509 1842 * Reposition read/write file offset.
984263bc 1843 */
984263bc 1844int
9697c509 1845lseek(struct lseek_args *uap)
984263bc 1846{
984263bc
MD
1847 int error;
1848
9697c509 1849 error = kern_lseek(uap->fd, uap->offset, uap->whence,
5a3fe67d 1850 &uap->sysmsg_offset);
9697c509 1851
984263bc
MD
1852 return (error);
1853}
984263bc 1854
984263bc 1855int
fad57d0e 1856kern_access(struct nlookupdata *nd, int aflags)
984263bc 1857{
dadab5e9 1858 struct thread *td = curthread;
41c20dac 1859 struct vnode *vp;
984263bc 1860 int error, flags;
984263bc 1861
fad57d0e
MD
1862 if ((error = nlookup(nd)) != 0)
1863 return (error);
806dcf9a 1864retry:
fad57d0e 1865 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 1866 if (error)
fad57d0e 1867 return (error);
984263bc
MD
1868
1869 /* Flags == 0 means only check for existence. */
9697c509 1870 if (aflags) {
984263bc 1871 flags = 0;
9697c509 1872 if (aflags & R_OK)
984263bc 1873 flags |= VREAD;
9697c509 1874 if (aflags & W_OK)
984263bc 1875 flags |= VWRITE;
9697c509 1876 if (aflags & X_OK)
984263bc
MD
1877 flags |= VEXEC;
1878 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
fad57d0e 1879 error = VOP_ACCESS(vp, flags, nd->nl_cred, td);
806dcf9a
MD
1880
1881 /*
1882 * If the file handle is stale we have to re-resolve the
1883 * entry. This is a hack at the moment.
1884 */
1885 if (error == ESTALE) {
1886 cache_setunresolved(nd->nl_ncp);
1887 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1888 if (error == 0) {
1889 vput(vp);
1890 vp = NULL;
1891 goto retry;
1892 }
1893 }
984263bc 1894 }
984263bc 1895 vput(vp);
984263bc
MD
1896 return (error);
1897}
1898
984263bc 1899/*
9697c509 1900 * access_args(char *path, int flags)
41c20dac 1901 *
9697c509 1902 * Check access permissions.
984263bc 1903 */
984263bc 1904int
9697c509 1905access(struct access_args *uap)
984263bc 1906{
fad57d0e 1907 struct nlookupdata nd;
9697c509 1908 int error;
984263bc 1909
fad57d0e
MD
1910 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1911 if (error == 0)
1912 error = kern_access(&nd, uap->flags);
1913 nlookup_done(&nd);
984263bc
MD
1914 return (error);
1915}
1916
984263bc 1917int
21739618 1918kern_stat(struct nlookupdata *nd, struct stat *st)
984263bc 1919{
984263bc 1920 int error;
21739618
MD
1921 struct vnode *vp;
1922 thread_t td;
984263bc 1923
21739618 1924 if ((error = nlookup(nd)) != 0)
984263bc 1925 return (error);
806dcf9a 1926again:
21739618
MD
1927 if ((vp = nd->nl_ncp->nc_vp) == NULL)
1928 return (ENOENT);
1929
1930 td = curthread;
5fd012e0 1931 if ((error = vget(vp, LK_SHARED, td)) != 0)
21739618
MD
1932 return (error);
1933 error = vn_stat(vp, st, td);
806dcf9a
MD
1934
1935 /*
1936 * If the file handle is stale we have to re-resolve the entry. This
1937 * is a hack at the moment.
1938 */
1939 if (error == ESTALE) {
1940 cache_setunresolved(nd->nl_ncp);
1941 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1942 if (error == 0) {
1943 vput(vp);
1944 goto again;
1945 }
1946 }
21739618 1947 vput(vp);
984263bc
MD
1948 return (error);
1949}
1950
984263bc 1951/*
41c20dac
MD
1952 * stat_args(char *path, struct stat *ub)
1953 *
984263bc
MD
1954 * Get file status; this version follows links.
1955 */
984263bc 1956int
41c20dac 1957stat(struct stat_args *uap)
984263bc 1958{
21739618 1959 struct nlookupdata nd;
9697c509
DRJ
1960 struct stat st;
1961 int error;
984263bc 1962
21739618
MD
1963 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1964 if (error == 0) {
1965 error = kern_stat(&nd, &st);
1966 if (error == 0)
1967 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 1968 }
fad57d0e 1969 nlookup_done(&nd);
984263bc
MD
1970 return (error);
1971}
1972
1973/*
41c20dac
MD
1974 * lstat_args(char *path, struct stat *ub)
1975 *
984263bc
MD
1976 * Get file status; this version does not follow links.
1977 */
984263bc 1978int
41c20dac 1979lstat(struct lstat_args *uap)
984263bc 1980{
21739618 1981 struct nlookupdata nd;
9697c509
DRJ
1982 struct stat st;
1983 int error;
984263bc 1984
21739618
MD
1985 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1986 if (error == 0) {
1987 error = kern_stat(&nd, &st);
1988 if (error == 0)
1989 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 1990 }
fad57d0e 1991 nlookup_done(&nd);
984263bc
MD
1992 return (error);
1993}
1994
984263bc 1995/*
41c20dac
MD
1996 * pathconf_Args(char *path, int name)
1997 *
984263bc
MD
1998 * Get configurable pathname variables.
1999 */
984263bc
MD
2000/* ARGSUSED */
2001int
41c20dac 2002pathconf(struct pathconf_args *uap)
984263bc 2003{
fad57d0e
MD
2004 struct nlookupdata nd;
2005 struct vnode *vp;
984263bc 2006 int error;
984263bc 2007
fad57d0e 2008 vp = NULL;
ab2eb4eb 2009 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2010 if (error == 0)
2011 error = nlookup(&nd);
2012 if (error == 0)
2013 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
2014 nlookup_done(&nd);
2015 if (error == 0) {
ab2eb4eb 2016 error = VOP_PATHCONF(vp, uap->name, uap->sysmsg_fds);
fad57d0e
MD
2017 vput(vp);
2018 }
984263bc
MD
2019 return (error);
2020}
2021
bbab27f6
DRJ
2022/*
2023 * XXX: daver
2024 * kern_readlink isn't properly split yet. There is a copyin burried
2025 * in VOP_READLINK().
2026 */
984263bc 2027int
fad57d0e 2028kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
984263bc 2029{
dadab5e9
MD
2030 struct thread *td = curthread;
2031 struct proc *p = td->td_proc;
41c20dac 2032 struct vnode *vp;
984263bc
MD
2033 struct iovec aiov;
2034 struct uio auio;
2035 int error;
984263bc 2036
fad57d0e
MD
2037 if ((error = nlookup(nd)) != 0)
2038 return (error);
2039 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 2040 if (error)
984263bc 2041 return (error);
fad57d0e 2042 if (vp->v_type != VLNK) {
984263bc 2043 error = EINVAL;
fad57d0e 2044 } else {
9697c509
DRJ
2045 aiov.iov_base = buf;
2046 aiov.iov_len = count;
984263bc
MD
2047 auio.uio_iov = &aiov;
2048 auio.uio_iovcnt = 1;
2049 auio.uio_offset = 0;
2050 auio.uio_rw = UIO_READ;
2051 auio.uio_segflg = UIO_USERSPACE;
dadab5e9 2052 auio.uio_td = td;
9697c509 2053 auio.uio_resid = count;
984263bc
MD
2054 error = VOP_READLINK(vp, &auio, p->p_ucred);
2055 }
2056 vput(vp);
9697c509
DRJ
2057 *res = count - auio.uio_resid;
2058 return (error);
2059}
2060
2061/*
2062 * readlink_args(char *path, char *buf, int count)
2063 *
2064 * Return target name of a symbolic link.
2065 */
2066int
2067readlink(struct readlink_args *uap)
2068{
fad57d0e 2069 struct nlookupdata nd;
9697c509
DRJ
2070 int error;
2071
fad57d0e
MD
2072 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2073 if (error == 0) {
2074 error = kern_readlink(&nd, uap->buf, uap->count,
2075 &uap->sysmsg_result);
2076 }
2077 nlookup_done(&nd);
984263bc
MD
2078 return (error);
2079}
2080
2081static int
41c20dac 2082setfflags(struct vnode *vp, int flags)
984263bc 2083{
dadab5e9
MD
2084 struct thread *td = curthread;
2085 struct proc *p = td->td_proc;
984263bc
MD
2086 int error;
2087 struct vattr vattr;
2088
2089 /*
2090 * Prevent non-root users from setting flags on devices. When
2091 * a device is reused, users can retain ownership of the device
2092 * if they are allowed to set flags and programs assume that
2093 * chown can't fail when done as root.
2094 */
2095 if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
dadab5e9 2096 ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
984263bc
MD
2097 return (error);
2098
186fccd6
MD
2099 /*
2100 * note: vget is required for any operation that might mod the vnode
2101 * so VINACTIVE is properly cleared.
2102 */
dadab5e9 2103 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
186fccd6
MD
2104 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2105 VATTR_NULL(&vattr);
2106 vattr.va_flags = flags;
2107 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2108 vput(vp);
2109 }
984263bc
MD
2110 return (error);
2111}
2112
2113/*
41c20dac
MD
2114 * chflags(char *path, int flags)
2115 *
984263bc
MD
2116 * Change flags of a file given a path name.
2117 */
984263bc
MD
2118/* ARGSUSED */
2119int
41c20dac 2120chflags(struct chflags_args *uap)
984263bc 2121{
fad57d0e
MD
2122 struct nlookupdata nd;
2123 struct vnode *vp;
984263bc 2124 int error;
984263bc 2125
fad57d0e 2126 vp = NULL;
ab2eb4eb 2127 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2128 /* XXX Add NLC flag indicating modifying operation? */
2129 if (error == 0)
2130 error = nlookup(&nd);
2131 if (error == 0)
2132 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
2133 nlookup_done(&nd);
2134 if (error == 0) {
ab2eb4eb 2135 error = setfflags(vp, uap->flags);
fad57d0e
MD
2136 vrele(vp);
2137 }
2138 return (error);
984263bc
MD
2139}
2140
2141/*
41c20dac
MD
2142 * fchflags_args(int fd, int flags)
2143 *
984263bc
MD
2144 * Change flags of a file given a file descriptor.
2145 */
984263bc
MD
2146/* ARGSUSED */
2147int
41c20dac 2148fchflags(struct fchflags_args *uap)
984263bc 2149{
dadab5e9
MD
2150 struct thread *td = curthread;
2151 struct proc *p = td->td_proc;
984263bc
MD
2152 struct file *fp;
2153 int error;
2154
ab2eb4eb 2155 if ((error = getvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2156 return (error);
ab2eb4eb 2157 return setfflags((struct vnode *) fp->f_data, uap->flags);
984263bc
MD
2158}
2159
2160static int
41c20dac 2161setfmode(struct vnode *vp, int mode)
984263bc 2162{
dadab5e9
MD
2163 struct thread *td = curthread;
2164 struct proc *p = td->td_proc;
984263bc
MD
2165 int error;
2166 struct vattr vattr;
2167
186fccd6
MD
2168 /*
2169 * note: vget is required for any operation that might mod the vnode
2170 * so VINACTIVE is properly cleared.
2171 */
dadab5e9 2172 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
186fccd6
MD
2173 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2174 VATTR_NULL(&vattr);
2175 vattr.va_mode = mode & ALLPERMS;
2176 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2177 vput(vp);
2178 }
984263bc
MD
2179 return error;
2180}
2181
9697c509 2182int
fad57d0e 2183kern_chmod(struct nlookupdata *nd, int mode)
9697c509 2184{
fad57d0e 2185 struct vnode *vp;
9697c509
DRJ
2186 int error;
2187
fad57d0e
MD
2188 /* XXX Add NLC flag indicating modifying operation? */
2189 if ((error = nlookup(nd)) != 0)
9697c509 2190 return (error);
fad57d0e
MD
2191 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2192 return (error);
2193 error = setfmode(vp, mode);
2194 vrele(vp);
2195 return (error);
9697c509
DRJ
2196}
2197
984263bc 2198/*
41c20dac
MD
2199 * chmod_args(char *path, int mode)
2200 *
984263bc
MD
2201 * Change mode of a file given path name.
2202 */
984263bc
MD
2203/* ARGSUSED */
2204int
41c20dac 2205chmod(struct chmod_args *uap)
984263bc 2206{
fad57d0e 2207 struct nlookupdata nd;
9697c509 2208 int error;
984263bc 2209
fad57d0e
MD
2210 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2211 if (error == 0)
2212 error = kern_chmod(&nd, uap->mode);
2213 nlookup_done(&nd);
9697c509 2214 return (error);
984263bc
MD
2215}
2216
2217/*
41c20dac
MD
2218 * lchmod_args(char *path, int mode)
2219 *
984263bc
MD
2220 * Change mode of a file given path name (don't follow links.)
2221 */
984263bc
MD
2222/* ARGSUSED */
2223int
41c20dac 2224lchmod(struct lchmod_args *uap)
984263bc 2225{
fad57d0e 2226 struct nlookupdata nd;
984263bc 2227 int error;
984263bc 2228
fad57d0e
MD
2229 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2230 if (error == 0)
2231 error = kern_chmod(&nd, uap->mode);
2232 nlookup_done(&nd);
2233 return (error);
984263bc
MD
2234}
2235
2236/*
41c20dac
MD
2237 * fchmod_args(int fd, int mode)
2238 *
984263bc
MD
2239 * Change mode of a file given a file descriptor.
2240 */
984263bc
MD
2241/* ARGSUSED */
2242int
41c20dac 2243fchmod(struct fchmod_args *uap)
984263bc 2244{
dadab5e9
MD
2245 struct thread *td = curthread;
2246 struct proc *p = td->td_proc;
984263bc
MD
2247 struct file *fp;
2248 int error;
2249
ab2eb4eb 2250 if ((error = getvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2251 return (error);
ab2eb4eb 2252 return setfmode((struct vnode *)fp->f_data, uap->mode);
984263bc
MD
2253}
2254
2255static int
41c20dac 2256setfown(struct vnode *vp, uid_t uid, gid_t gid)
984263bc 2257{
dadab5e9
MD
2258 struct thread *td = curthread;
2259 struct proc *p = td->td_proc;
984263bc
MD
2260 int error;
2261 struct vattr vattr;
2262
186fccd6
MD
2263 /*
2264 * note: vget is required for any operation that might mod the vnode
2265 * so VINACTIVE is properly cleared.
2266 */
dadab5e9 2267 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
186fccd6
MD
2268 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2269 VATTR_NULL(&vattr);
2270 vattr.va_uid = uid;
2271 vattr.va_gid = gid;
2272 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2273 vput(vp);
2274 }
984263bc
MD
2275 return error;
2276}
2277
9697c509 2278int
fad57d0e 2279kern_chown(struct nlookupdata *nd, int uid, int gid)
9697c509 2280{
fad57d0e 2281 struct vnode *vp;
9697c509
DRJ
2282 int error;
2283
fad57d0e
MD
2284 /* XXX Add NLC flag indicating modifying operation? */
2285 if ((error = nlookup(nd)) != 0)
9697c509 2286 return (error);
fad57d0e
MD
2287 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2288 return (error);
2289 error = setfown(vp, uid, gid);
2290 vrele(vp);
9697c509
DRJ
2291 return (error);
2292}
2293
984263bc 2294/*
41c20dac
MD
2295 * chown(char *path, int uid, int gid)
2296 *
984263bc
MD
2297 * Set ownership given a path name.
2298 */
984263bc 2299int
41c20dac 2300chown(struct chown_args *uap)
984263bc 2301{
fad57d0e 2302 struct nlookupdata nd;
9697c509
DRJ
2303 int error;
2304
fad57d0e
MD
2305 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2306 if (error == 0)
2307 error = kern_chown(&nd, uap->uid, uap->gid);
2308 nlookup_done(&nd);
984263bc
MD
2309 return (error);
2310}
2311
2312/*
41c20dac
MD
2313 * lchown_args(char *path, int uid, int gid)
2314 *
984263bc
MD
2315 * Set ownership given a path name, do not cross symlinks.
2316 */
984263bc 2317int
41c20dac 2318lchown(struct lchown_args *uap)
984263bc 2319{
fad57d0e 2320 struct nlookupdata nd;
984263bc 2321 int error;
9697c509 2322
fad57d0e
MD
2323 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2324 if (error == 0)
2325 error = kern_chown(&nd, uap->uid, uap->gid);
2326 nlookup_done(&nd);
984263bc
MD
2327 return (error);
2328}
2329
2330/*
41c20dac
MD
2331 * fchown_args(int fd, int uid, int gid)
2332 *
984263bc
MD
2333 * Set ownership given a file descriptor.
2334 */
984263bc
MD
2335/* ARGSUSED */
2336int
41c20dac 2337fchown(struct fchown_args *uap)
984263bc 2338{
dadab5e9
MD
2339 struct thread *td = curthread;
2340 struct proc *p = td->td_proc;
984263bc
MD
2341 struct file *fp;
2342 int error;
2343
ab2eb4eb 2344 if ((error = getvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2345 return (error);
41c20dac 2346 return setfown((struct vnode *)fp->f_data,
ab2eb4eb 2347 uap->uid, uap->gid);
984263bc
MD
2348}
2349
2350static int
9697c509 2351getutimes(const struct timeval *tvp, struct timespec *tsp)
984263bc
MD
2352{
2353 struct timeval tv[2];
984263bc 2354
9697c509 2355 if (tvp == NULL) {
984263bc
MD
2356 microtime(&tv[0]);
2357 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2358 tsp[1] = tsp[0];
2359 } else {
9697c509
DRJ
2360 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2361 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
984263bc
MD
2362 }
2363 return 0;
2364}
2365
2366static int
41c20dac 2367setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
984263bc 2368{
dadab5e9
MD
2369 struct thread *td = curthread;
2370 struct proc *p = td->td_proc;
984263bc
MD
2371 int error;
2372 struct vattr vattr;
2373
186fccd6
MD
2374 /*
2375 * note: vget is required for any operation that might mod the vnode
2376 * so VINACTIVE is properly cleared.
2377 */
dadab5e9 2378 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
186fccd6
MD
2379 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2380 VATTR_NULL(&vattr);
2381 vattr.va_atime = ts[0];
2382 vattr.va_mtime = ts[1];
2383 if (nullflag)
2384 vattr.va_vaflags |= VA_UTIMES_NULL;
2385 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2386 vput(vp);
2387 }
984263bc
MD
2388 return error;
2389}
2390
9697c509 2391int
fad57d0e 2392kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
9697c509
DRJ
2393{
2394 struct timespec ts[2];
fad57d0e 2395 struct vnode *vp;
9697c509
DRJ
2396 int error;
2397
fad57d0e 2398 if ((error = getutimes(tptr, ts)) != 0)
9697c509 2399 return (error);
fad57d0e
MD
2400 /* XXX Add NLC flag indicating modifying operation? */
2401 if ((error = nlookup(nd)) != 0)
9697c509 2402 return (error);
fad57d0e
MD
2403 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2404 return (error);
2405 error = setutimes(vp, ts, tptr == NULL);
2406 vrele(vp);
9697c509
DRJ
2407 return (error);
2408}
2409
984263bc 2410/*
41c20dac
MD
2411 * utimes_args(char *path, struct timeval *tptr)
2412 *
984263bc
MD
2413 * Set the access and modification times of a file.
2414 */
984263bc 2415int
41c20dac 2416utimes(struct utimes_args *uap)
984263bc 2417{
9697c509 2418 struct timeval tv[2];
fad57d0e 2419 struct nlookupdata nd;
9697c509
DRJ
2420 int error;
2421
2422 if (uap->tptr) {
2423 error = copyin(uap->tptr, tv, sizeof(tv));
2424 if (error)
2425 return (error);
2426 }
fad57d0e
MD
2427 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2428 if (error == 0)
2429 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2430 nlookup_done(&nd);
984263bc
MD
2431 return (error);
2432}
2433
2434/*
41c20dac
MD
2435 * lutimes_args(char *path, struct timeval *tptr)
2436 *
984263bc
MD
2437 * Set the access and modification times of a file.
2438 */
984263bc 2439int
41c20dac 2440lutimes(struct lutimes_args *uap)
984263bc 2441{
9697c509 2442 struct timeval tv[2];
fad57d0e 2443 struct nlookupdata nd;
9697c509
DRJ
2444 int error;
2445
2446 if (uap->tptr) {
2447 error = copyin(uap->tptr, tv, sizeof(tv));
2448 if (error)
2449 return (error);
2450 }
fad57d0e
MD
2451 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2452 if (error == 0)
2453 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2454 nlookup_done(&nd);
984263bc
MD
2455 return (error);
2456}
2457
984263bc 2458int
9697c509 2459kern_futimes(int fd, struct timeval *tptr)
984263bc 2460{
dadab5e9
MD
2461 struct thread *td = curthread;
2462 struct proc *p = td->td_proc;
984263bc
MD
2463 struct timespec ts[2];
2464 struct file *fp;
984263bc
MD
2465 int error;
2466
9697c509
DRJ
2467 error = getutimes(tptr, ts);
2468 if (error)
984263bc 2469 return (error);
9697c509
DRJ
2470 error = getvnode(p->p_fd, fd, &fp);
2471 if (error)
984263bc 2472 return (error);
9697c509
DRJ
2473 error = setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
2474 return (error);
984263bc
MD
2475}
2476
2477/*
9697c509 2478 * futimes_args(int fd, struct timeval *tptr)
41c20dac 2479 *
9697c509 2480 * Set the access and modification times of a file.
984263bc 2481 */
984263bc 2482int
9697c509
DRJ
2483futimes(struct futimes_args *uap)
2484{
2485 struct timeval tv[2];
2486 int error;
2487
2488 if (uap->tptr) {
2489 error = copyin(uap->tptr, tv, sizeof(tv));
2490 if (error)
2491 return (error);
2492 }
2493
b651c485 2494 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
9697c509
DRJ
2495
2496 return (error);
2497}
2498
2499int
fad57d0e 2500kern_truncate(struct nlookupdata *nd, off_t length)
984263bc 2501{
41c20dac 2502 struct vnode *vp;
984263bc
MD
2503 struct vattr vattr;
2504 int error;
984263bc 2505
9697c509 2506 if (length < 0)
984263bc 2507 return(EINVAL);
fad57d0e
MD
2508 /* XXX Add NLC flag indicating modifying operation? */
2509 if ((error = nlookup(nd)) != 0)
984263bc 2510 return (error);
fad57d0e
MD
2511 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2512 return (error);
2513 VOP_LEASE(vp, nd->nl_td, nd->nl_cred, LEASE_WRITE);
2514 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, nd->nl_td)) != 0) {
2515 vrele(vp);
2516 return (error);
2517 }
2518 if (vp->v_type == VDIR) {
984263bc 2519 error = EISDIR;
fad57d0e
MD
2520 } else if ((error = vn_writechk(vp)) == 0 &&
2521 (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred, nd->nl_td)) == 0) {
984263bc 2522 VATTR_NULL(&vattr);
9697c509 2523 vattr.va_size = length;
fad57d0e 2524 error = VOP_SETATTR(vp, &vattr, nd->nl_cred, nd->nl_td);
984263bc
MD
2525 }
2526 vput(vp);
2527 return (error);
2528}
2529
9697c509
DRJ
2530/*
2531 * truncate(char *path, int pad, off_t length)
2532 *
2533 * Truncate a file given its path name.
2534 */
2535int
2536truncate(struct truncate_args *uap)
2537{
fad57d0e 2538 struct nlookupdata nd;
9697c509
DRJ
2539 int error;
2540
fad57d0e
MD
2541 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2542 if (error == 0)
2543 error = kern_truncate(&nd, uap->length);
2544 nlookup_done(&nd);
9697c509
DRJ
2545 return error;
2546}
2547
984263bc 2548int
8f6f8622 2549kern_ftruncate(int fd, off_t length)
984263bc 2550{
dadab5e9
MD
2551 struct thread *td = curthread;
2552 struct proc *p = td->td_proc;
984263bc
MD
2553 struct vattr vattr;
2554 struct vnode *vp;
2555 struct file *fp;
2556 int error;
2557
8f6f8622 2558 if (length < 0)
984263bc 2559 return(EINVAL);
8f6f8622 2560 if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
984263bc
MD
2561 return (error);
2562 if ((fp->f_flag & FWRITE) == 0)
2563 return (EINVAL);
2564 vp = (struct vnode *)fp->f_data;
dadab5e9 2565 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
5fd012e0 2566 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
2567 if (vp->v_type == VDIR)
2568 error = EISDIR;
2569 else if ((error = vn_writechk(vp)) == 0) {
2570 VATTR_NULL(&vattr);
8f6f8622 2571 vattr.va_size = length;
dadab5e9 2572 error = VOP_SETATTR(vp, &vattr, fp->f_cred, td);
984263bc 2573 }
5fd012e0 2574 VOP_UNLOCK(vp, 0, td);
984263bc
MD
2575 return (error);
2576}
2577
984263bc 2578/*
8f6f8622 2579 * ftruncate_args(int fd, int pad, off_t length)
41c20dac 2580 *
8f6f8622 2581 * Truncate a file given a file descriptor.
984263bc 2582 */
984263bc 2583int
8f6f8622 2584ftruncate(struct ftruncate_args *uap)
984263bc 2585{
8f6f8622 2586 int error;
984263bc 2587
8f6f8622
DRJ
2588 error = kern_ftruncate(uap->fd, uap->length);
2589
2590 return (error);
984263bc
MD
2591}
2592
984263bc 2593/*
41c20dac
MD
2594 * fsync(int fd)
2595 *
984263bc
MD
2596 * Sync an open file.
2597 */
984263bc
MD
2598/* ARGSUSED */
2599int
41c20dac 2600fsync(struct fsync_args *uap)
984263bc 2601{
dadab5e9
MD
2602 struct thread *td = curthread;
2603 struct proc *p = td->td_proc;
41c20dac 2604 struct vnode *vp;
984263bc
MD
2605 struct file *fp;
2606 vm_object_t obj;
2607 int error;
2608
ab2eb4eb 2609 if ((error = getvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc
MD
2610 return (error);
2611 vp = (struct vnode *)fp->f_data;
5fd012e0 2612 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
2613 if (VOP_GETVOBJECT(vp, &obj) == 0)
2614 vm_object_page_clean(obj, 0, 0, 0);
3b568787 2615 if ((error = VOP_FSYNC(vp, MNT_WAIT, td)) == 0 &&
984263bc
MD
2616 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) &&
2617 bioops.io_fsync)
2618 error = (*bioops.io_fsync)(vp);
5fd012e0 2619 VOP_UNLOCK(vp, 0, td);
984263bc
MD
2620 return (error);
2621}
2622
984263bc 2623int
fad57d0e 2624kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
984263bc 2625{
fad57d0e
MD
2626 struct namecache *fncpd;
2627 struct namecache *tncpd;
2628 struct namecache *ncp;
2629 struct mount *mp;
984263bc
MD
2630 int error;
2631
2632 bwillwrite();
fad57d0e 2633 if ((error = nlookup(fromnd)) != 0)
984263bc 2634 return (error);
fad57d0e
MD
2635 if ((fncpd = fromnd->nl_ncp->nc_parent) == NULL)
2636 return (ENOENT);
2637 cache_hold(fncpd);
2638
2639 /*
2640 * unlock the source ncp so we can lookup the target ncp without
2641 * deadlocking. The target may or may not exist so we do not check
2642 * for a target vp like kern_mkdir() and other creation functions do.
2643 *
2644 * The source and target directories are ref'd and rechecked after
2645 * everything is relocked to determine if the source or target file
2646 * has been renamed.
2647 */
2648 KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
2649 fromnd->nl_flags &= ~NLC_NCPISLOCKED;
2650 cache_unlock(fromnd->nl_ncp);
2651
2652 tond->nl_flags |= NLC_CREATE;
2653 if ((error = nlookup(tond)) != 0) {
2654 cache_drop(fncpd);
2655 return (error);
2656 }
2657 if ((tncpd = tond->nl_ncp->nc_parent) == NULL) {
2658 cache_drop(fncpd);
2659 return (ENOENT);
2660 }
2661 cache_hold(tncpd);
2662
2663 /*
2664 * If the source and target are the same there is nothing to do
2665 */
2666 if (fromnd->nl_ncp == tond->nl_ncp) {
2667 cache_drop(fncpd);
2668 cache_drop(tncpd);
2669 return (0);
2670 }
2671
2672 /*
aa40af33
MD
2673 * relock the source ncp. NOTE AFTER RELOCKING: the source ncp
2674 * may have become invalid while it was unlocked, nc_vp and nc_mount
2675 * could be NULL.
fad57d0e
MD
2676 */
2677 if (cache_lock_nonblock(fromnd->nl_ncp) == 0) {
2678 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2679 } else if (fromnd->nl_ncp > tond->nl_ncp) {
2680 cache_lock(fromnd->nl_ncp);
2681 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2682 } else {
2683 cache_unlock(tond->nl_ncp);
2684 cache_lock(fromnd->nl_ncp);
2685 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2686 cache_lock(tond->nl_ncp);
2687 cache_resolve(tond->nl_ncp, tond->nl_cred);
2688 }
2689 fromnd->nl_flags |= NLC_NCPISLOCKED;
2690
2691 /*
2692 * make sure the parent directories linkages are the same
2693 */
2694 if (fncpd != fromnd->nl_ncp->nc_parent ||
2695 tncpd != tond->nl_ncp->nc_parent) {
2696 cache_drop(fncpd);
2697 cache_drop(tncpd);
2698 return (ENOENT);
2699 }
2700
2701 /*
2702 * Both the source and target must be within the same filesystem and
2703 * in the same filesystem as their parent directories within the
2704 * namecache topology.
aa40af33
MD
2705 *
2706 * NOTE: fromnd's nc_mount or nc_vp could be NULL.
fad57d0e
MD
2707 */
2708 mp = fncpd->nc_mount;
2709 if (mp != tncpd->nc_mount || mp != fromnd->nl_ncp->nc_mount ||
2710 mp != tond->nl_ncp->nc_mount) {
2711 cache_drop(fncpd);
2712 cache_drop(tncpd);
2713 return (EXDEV);
2714 }
2715
2716 /*
2717 * If the target exists and either the source or target is a directory,
2718 * then both must be directories.
aa40af33
MD
2719 *
2720 * Due to relocking of the source, fromnd->nl_ncp->nc_vp might have
2721 * become NULL.
fad57d0e
MD
2722 */
2723 if (tond->nl_ncp->nc_vp) {
aa40af33
MD
2724 if (fromnd->nl_ncp->nc_vp == NULL) {
2725 error = ENOENT;
2726 } else if (fromnd->nl_ncp->nc_vp->v_type == VDIR) {
fad57d0e
MD
2727 if (tond->nl_ncp->nc_vp->v_type != VDIR)
2728 error = ENOTDIR;
2729 } else if (tond->nl_ncp->nc_vp->v_type == VDIR) {
984263bc 2730 error = EISDIR;
984263bc
MD
2731 }
2732 }
fad57d0e 2733
984263bc 2734 /*
fad57d0e
MD
2735 * You cannot rename a source into itself or a subdirectory of itself.
2736 * We check this by travsersing the target directory upwards looking
2737 * for a match against the source.
984263bc 2738 */
fad57d0e
MD
2739 if (error == 0) {
2740 for (ncp = tncpd; ncp; ncp = ncp->nc_parent) {
2741 if (fromnd->nl_ncp == ncp) {
2742 error = EINVAL;
2743 break;
2744 }
984263bc 2745 }
fad57d0e
MD
2746 }
2747
2748 cache_drop(fncpd);
2749 cache_drop(tncpd);
bc8fd9e0
MD
2750
2751 /*
2752 * Even though the namespaces are different, they may still represent
2753 * hardlinks to the same file. The filesystem might have a hard time
2754 * with this so we issue a NREMOVE of the source instead of a NRENAME
2755 * when we detect the situation.
2756 */
2757 if (error == 0) {
2758 if (fromnd->nl_ncp->nc_vp == tond->nl_ncp->nc_vp) {
2759 error = VOP_NREMOVE(fromnd->nl_ncp, fromnd->nl_cred);
2760 } else {
2761 error = VOP_NRENAME(fromnd->nl_ncp, tond->nl_ncp,
2762 tond->nl_cred);
2763 }
2764 }
984263bc
MD
2765 return (error);
2766}
2767
2768/*
9697c509 2769 * rename_args(char *from, char *to)
41c20dac 2770 *
9697c509
DRJ
2771 * Rename files. Source and destination must either both be directories,
2772 * or both not be directories. If target is a directory, it must be empty.
984263bc 2773 */
984263bc 2774int
9697c509
DRJ
2775rename(struct rename_args *uap)
2776{
fad57d0e 2777 struct nlookupdata fromnd, tond;
9697c509
DRJ
2778 int error;
2779
fad57d0e
MD
2780 error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
2781 if (error == 0) {
2782 error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
2783 if (error == 0)
2784 error = kern_rename(&fromnd, &tond);
2785 nlookup_done(&tond);
2786 }
2787 nlookup_done(&fromnd);
9697c509
DRJ
2788 return (error);
2789}
2790
2791int
fad57d0e 2792kern_mkdir(struct nlookupdata *nd, int mode)
984263bc 2793{
dadab5e9
MD
2794 struct thread *td = curthread;
2795 struct proc *p = td->td_proc;
fad57d0e 2796 struct namecache *ncp;
41c20dac 2797 struct vnode *vp;
984263bc
MD
2798 struct vattr vattr;
2799 int error;
984263bc
MD
2800
2801 bwillwrite();
fad57d0e
MD
2802 nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE;
2803 if ((error = nlookup(nd)) != 0)
984263bc 2804 return (error);
fad57d0e
MD
2805
2806 ncp = nd->nl_ncp;
2807 if (ncp->nc_vp)
984263bc 2808 return (EEXIST);
fad57d0e 2809
984263bc
MD
2810 VATTR_NULL(&vattr);
2811 vattr.va_type = VDIR;
9697c509 2812 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
fad57d0e
MD
2813
2814 vp = NULL;
2815 error = VOP_NMKDIR(ncp, &vp, p->p_ucred, &vattr);
9697c509 2816 if (error == 0)
fad57d0e 2817 vput(vp);
984263bc
MD
2818 return (error);
2819}
2820
2821/*
9697c509 2822 * mkdir_args(char *path, int mode)
41c20dac 2823 *
9697c509 2824 * Make a directory file.
984263bc 2825 */
984263bc
MD
2826/* ARGSUSED */
2827int
9697c509
DRJ
2828mkdir(struct mkdir_args *uap)
2829{
fad57d0e 2830 struct nlookupdata nd;
9697c509
DRJ
2831 int error;
2832
fad57d0e
MD
2833 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2834 if (error == 0)
2835 error = kern_mkdir(&nd, uap->mode);
2836 nlookup_done(&nd);
9697c509
DRJ
2837 return (error);
2838}
2839
2840int
fad57d0e 2841kern_rmdir(struct nlookupdata *nd)
984263bc 2842{
fad57d0e 2843 struct namecache *ncp;
984263bc 2844 int error;
984263bc
MD
2845
2846 bwillwrite();
fad57d0e
MD
2847 nd->nl_flags |= NLC_DELETE;
2848 if ((error = nlookup(nd)) != 0)
984263bc 2849 return (error);
fad57d0e
MD
2850
2851 ncp = nd->nl_ncp;
2852 error = VOP_NRMDIR(ncp, nd->nl_cred);
984263bc
MD
2853 return (error);
2854}
2855
984263bc 2856/*
9697c509 2857 * rmdir_args(char *path)
41c20dac 2858 *
9697c509 2859 * Remove a directory file.
984263bc 2860 */
9697c509 2861/* ARGSUSED */
984263bc 2862int
9697c509 2863rmdir(struct rmdir_args *uap)
984263bc 2864{
fad57d0e 2865 struct nlookupdata nd;
9697c509
DRJ
2866 int error;
2867
fad57d0e
MD
2868 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2869 if (error == 0)
2870 error = kern_rmdir(&nd);
2871 nlookup_done(&nd);
984263bc
MD
2872 return (error);
2873}
984263bc 2874
984263bc 2875int
53b02c48
JS
2876kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res,
2877 enum uio_seg direction)
984263bc 2878{
dadab5e9
MD
2879 struct thread *td = curthread;
2880 struct proc *p = td->td_proc;
984263bc
MD
2881 struct vnode *vp;
2882 struct file *fp;
2883 struct uio auio;
2884 struct iovec aiov;
2885 long loff;
2886 int error, eofflag;
2887
9697c509 2888 if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
984263bc
MD
2889 return (error);
2890 if ((fp->f_flag & FREAD) == 0)
2891 return (EBADF);
2892 vp = (struct vnode *)fp->f_data;
2893unionread:
2894 if (vp->v_type != VDIR)
2895 return (EINVAL);
9697c509
DRJ
2896 aiov.iov_base = buf;
2897 aiov.iov_len = count;
984263bc
MD
2898 auio.uio_iov = &aiov;
2899 auio.uio_iovcnt = 1;
2900 auio.uio_rw = UIO_READ;
53b02c48 2901 auio.uio_segflg = direction;
dadab5e9 2902 auio.uio_td = td;
9697c509 2903 auio.uio_resid = count;
5fd012e0
MD
2904 /* vn_lock(vp, LK_SHARED | LK_RETRY, td); */
2905 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
2906 loff = auio.uio_offset = fp->f_offset;
2907 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
2908 fp->f_offset = auio.uio_offset;
5fd012e0 2909 VOP_UNLOCK(vp, 0, td);
984263bc
MD
2910 if (error)
2911 return (error);
9697c509 2912 if (count == auio.uio_resid) {
984263bc 2913 if (union_dircheckp) {
dadab5e9 2914 error = union_dircheckp(td, &vp, fp);
984263bc
MD
2915 if (error == -1)
2916 goto unionread;
2917 if (error)
2918 return (error);
2919 }
2920 if ((vp->v_flag & VROOT) &&
2921 (vp->v_mount->mnt_flag & MNT_UNION)) {
2922 struct vnode *tvp = vp;
2923 vp = vp->v_mount->mnt_vnodecovered;
597aea93 2924 vref(vp);
fbb4eeab 2925 fp->f_data = vp;
984263bc
MD
2926 fp->f_offset = 0;
2927 vrele(tvp);
2928 goto unionread;
2929 }
2930 }
9697c509
DRJ
2931 if (basep) {
2932 *basep = loff;
984263bc 2933 }
9697c509
DRJ
2934 *res = count - auio.uio_resid;
2935 return (error);
2936}
2937
2938/*
2939 * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
2940 *
2941 * Read a block of directory entries in a file system independent format.
2942 */
2943int
2944getdirentries(struct getdirentries_args *uap)
2945{
2946 long base;
2947 int error;
2948
2949 error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base,
53b02c48 2950 &uap->sysmsg_result, UIO_USERSPACE);
9697c509
DRJ
2951
2952 if (error == 0)
2953 error = copyout(&base, uap->basep, sizeof(*uap->basep));
984263bc
MD
2954 return (error);
2955}
41c20dac
MD
2956
2957/*
2958 * getdents_args(int fd, char *buf, size_t count)
2959 */
984263bc 2960int
41c20dac 2961getdents(struct getdents_args *uap)
984263bc 2962{
9697c509
DRJ
2963 int error;
2964
2965 error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL,
53b02c48 2966 &uap->sysmsg_result, UIO_USERSPACE);
41c20dac 2967
9697c509 2968 return (error);
984263bc
MD
2969}
2970
2971/*
41c20dac
MD
2972 * umask(int newmask)
2973 *
984263bc
MD
2974 * Set the mode mask for creation of filesystem nodes.
2975 *
2976 * MP SAFE
2977 */
984263bc 2978int
41c20dac 2979umask(struct umask_args *uap)
984263bc 2980{
dadab5e9
MD
2981 struct thread *td = curthread;
2982 struct proc *p = td->td_proc;
41c20dac 2983 struct filedesc *fdp;
984263bc
MD
2984
2985 fdp = p->p_fd;
c7114eea 2986 uap->sysmsg_result = fdp->fd_cmask;
ab2eb4eb 2987 fdp->fd_cmask = uap->newmask & ALLPERMS;
984263bc
MD
2988 return (0);
2989}
2990
2991/*
41c20dac
MD
2992 * revoke(char *path)
2993 *
984263bc
MD
2994 * Void all references to file by ripping underlying filesystem
2995 * away from vnode.
2996 */
984263bc
MD
2997/* ARGSUSED */
2998int
41c20dac 2999revoke(struct revoke_args *uap)
984263bc 3000{
dadab5e9 3001 struct thread *td = curthread;
fad57d0e 3002 struct nlookupdata nd;
984263bc 3003 struct vattr vattr;
fad57d0e
MD
3004 struct vnode *vp;
3005 struct ucred *cred;
984263bc 3006 int error;
984263bc 3007
fad57d0e 3008 vp = NULL;
ab2eb4eb 3009 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
3010 if (error == 0)
3011 error = nlookup(&nd);
3012 if (error == 0)
3013 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
3014 cred = crhold(nd.nl_cred);
3015 nlookup_done(&nd);
3016 if (error == 0) {
3017 if (vp->v_type != VCHR && vp->v_type != VBLK)
3018 error = EINVAL;
3019 if (error == 0)
3020 error = VOP_GETATTR(vp, &vattr, td);
3021 if (error == 0 && cred->cr_uid != vattr.va_uid)
3022 error = suser_cred(cred, PRISON_ROOT);
3023 if (error == 0 && count_udev(vp->v_udev) > 0) {
3024 if ((error = vx_lock(vp)) == 0) {
3025 VOP_REVOKE(vp, REVOKEALL);
3026 vx_unlock(vp);
3027 }
5fd012e0 3028 }
fad57d0e 3029 vrele(vp);
5fd012e0 3030 }
fad57d0e 3031 crfree(cred);
984263bc
MD
3032 return (error);
3033}
3034
3035/*
3036 * Convert a user file descriptor to a kernel file entry.
3037 */
3038int
41c20dac 3039getvnode(struct filedesc *fdp, int fd, struct file **fpp)
984263bc
MD
3040{
3041 struct file *fp;
3042
3043 if ((u_int)fd >= fdp->fd_nfiles ||
0679adc4 3044 (fp = fdp->fd_files[fd].fp) == NULL)
984263bc
MD
3045 return (EBADF);
3046 if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO)
3047 return (EINVAL);
3048 *fpp = fp;
3049 return (0);
3050}
3051/*
41c20dac
MD
3052 * getfh_args(char *fname, fhandle_t *fhp)
3053 *
984263bc
MD
3054 * Get (NFS) file handle
3055 */
984263bc 3056int
41c20dac 3057getfh(struct getfh_args *uap)
984263bc 3058{
dadab5e9 3059 struct thread *td = curthread;
fad57d0e 3060 struct nlookupdata nd;
984263bc 3061 fhandle_t fh;
41c20dac 3062 struct vnode *vp;
984263bc
MD
3063 int error;
3064
3065 /*
3066 * Must be super user
3067 */
fad57d0e 3068 if ((error = suser(td)) != 0)
984263bc 3069 return (error);
fad57d0e
MD
3070
3071 vp = NULL;
3072 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
3073 if (error == 0)
3074 error = nlookup(&nd);
3075 if (error == 0)
3076 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3077 nlookup_done(&nd);
3078 if (error == 0) {
3079 bzero(&fh, sizeof(fh));
3080 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3081 error = VFS_VPTOFH(vp, &fh.fh_fid);
3082 vput(vp);
3083 if (error == 0)
3084 error = copyout(&fh, uap->fhp, sizeof(fh));
3085 }
984263bc
MD
3086 return (error);
3087}
3088
3089/*
41c20dac
MD
3090 * fhopen_args(const struct fhandle *u_fhp, int flags)
3091 *
984263bc
MD
3092 * syscall for the rpc.lockd to use to translate a NFS file handle into
3093 * an open descriptor.
3094 *
3095 * warning: do not remove the suser() call or this becomes one giant
3096 * security hole.
3097 */
984263bc 3098int
41c20dac 3099fhopen(struct fhopen_args *uap)
984263bc 3100{
dadab5e9
MD
3101 struct thread *td = curthread;
3102 struct proc *p = td->td_proc;
984263bc
MD
3103 struct mount *mp;
3104 struct vnode *vp;
3105 struct fhandle fhp;
3106 struct vattr vat;
3107 struct vattr *vap = &vat;
3108 struct flock lf;
41c20dac 3109 struct filedesc *fdp = p->p_fd;
984263bc
MD
3110 int fmode, mode, error, type;
3111 struct file *nfp;
fad57d0e 3112 struct file *fp;
984263bc
MD
3113 int indx;
3114
3115 /*
3116 * Must be super user
3117 */
dadab5e9 3118 error = suser(td);
984263bc
MD
3119 if (error)
3120 return (error);
3121
ab2eb4eb 3122 fmode = FFLAGS(uap->flags);
984263bc
MD
3123 /* why not allow a non-read/write open for our lockd? */
3124 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3125 return (EINVAL);
ab2eb4eb 3126 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
984263bc
MD
3127 if (error)
3128 return(error);
3129 /* find the mount point */
3130 mp = vfs_getvfs(&fhp.fh_fsid);
3131 if (mp == NULL)
3132 return (ESTALE);
3133 /* now give me my vnode, it gets returned to me locked */
3134 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
3135 if (error)
3136 return (error);
3137 /*
3138 * from now on we have to make sure not
3139 * to forget about the vnode
3140 * any error that causes an abort must vput(vp)
3141 * just set error = err and 'goto bad;'.
3142 */
3143
3144 /*
3145 * from vn_open
3146 */
3147 if (vp->v_type == VLNK) {
3148 error = EMLINK;
3149 goto bad;
3150 }
3151 if (vp->v_type == VSOCK) {
3152 error = EOPNOTSUPP;
3153 goto bad;
3154 }
3155 mode = 0;
3156 if (fmode & (FWRITE | O_TRUNC)) {
3157 if (vp->v_type == VDIR) {
3158 error = EISDIR;
3159 goto bad;
3160 }
3161 error = vn_writechk(vp);
3162 if (error)
3163 goto bad;
3164 mode |= VWRITE;
3165 }
3166 if (fmode & FREAD)
3167 mode |= VREAD;
3168 if (mode) {
dadab5e9 3169 error = VOP_ACCESS(vp, mode, p->p_ucred, td);
984263bc
MD
3170 if (error)
3171 goto bad;
3172 }
3173 if (fmode & O_TRUNC) {
5fd012e0 3174 VOP_UNLOCK(vp, 0, td); /* XXX */
dadab5e9 3175 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
5fd012e0 3176 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); /* XXX */
984263bc
MD
3177 VATTR_NULL(vap);
3178 vap->va_size = 0;
dadab5e9 3179 error = VOP_SETATTR(vp, vap, p->p_ucred, td);
984263bc
MD
3180 if (error)
3181 goto bad;
3182 }
fad57d0e 3183
984263bc 3184 /*
fad57d0e
MD
3185 * VOP_OPEN needs the file pointer so it can potentially override
3186 * it.
3187 *
3188 * WARNING! no f_ncp will be associated when fhopen()ing a directory.
3189 * XXX
984263bc 3190 */
fad57d0e
MD
3191 if ((error = falloc(p, &nfp, NULL)) != 0)
3192 goto bad;
3193 fp = nfp;
3194
fbb4eeab 3195 fp->f_type = DTYPE_VNODE;
fad57d0e
MD
3196 fp->f_flag = fmode & FMASK;
3197 fp->f_ops = &vnode_fileops;
fbb4eeab 3198 fp->f_data = vp;
fad57d0e
MD
3199
3200 error = VOP_OPEN(vp, fmode, p->p_ucred, fp, td);
3201 if (error) {
3202 /*
3203 * setting f_ops this way prevents VOP_CLOSE from being
3204 * called or fdrop() releasing the vp from v_data. Since
3205 * the VOP_OPEN failed we don't want to VOP_CLOSE.
3206 */
3207 fp->f_ops = &badfileops;
3208 fp->f_data = NULL;
3209 fdrop(fp, td);
3210 goto bad;
984263bc
MD
3211 }
3212 if (fmode & FWRITE)
3213 vp->v_writecount++;
3214
3215 /*
fad57d0e
MD
3216 * The fp now owns a reference on the vnode. We still have our own
3217 * ref+lock.
3218 */
3219 vref(vp);
3220
3221 /*
3222 * Make sure that a VM object is created for VMIO support. If this
3223 * fails just fdrop() normally to clean up.
984263bc 3224 */
fad57d0e
MD
3225 if (vn_canvmio(vp) == TRUE) {
3226 if ((error = vfs_object_create(vp, td)) != 0) {
3227 fdrop(fp, td);
3228 goto bad;
3229 }
3230 }
984263bc 3231
fad57d0e
MD
3232 /*
3233 * The open was successful, associate it with a file descriptor.
3234 */
3235 if ((error = fsetfd(p, fp, &indx)) != 0) {
984263bc
MD
3236 if (fmode & FWRITE)
3237 vp->v_writecount--;
fad57d0e 3238 fdrop(fp, td);
984263bc
MD
3239 goto bad;
3240 }
984263bc 3241
984263bc
MD
3242 if (fmode & (O_EXLOCK | O_SHLOCK)) {
3243 lf.l_whence = SEEK_SET;
3244 lf.l_start = 0;
3245 lf.l_len = 0;
3246 if (fmode & O_EXLOCK)
3247 lf.l_type = F_WRLCK;
3248 else
3249 lf.l_type = F_RDLCK;
3250 type = F_FLOCK;
3251 if ((fmode & FNONBLOCK) == 0)
3252 type |= F_WAIT;
5fd012e0 3253 VOP_UNLOCK(vp, 0, td);
984263bc
MD
3254 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
3255 /*
3256 * lock request failed. Normally close the descriptor
3257 * but handle the case where someone might have dup()d
3258 * or close()d it when we weren't looking.
3259 */
0679adc4 3260 if (fdp->fd_files[indx].fp == fp) {
69908319 3261 funsetfd(fdp, indx);
dadab5e9 3262 fdrop(fp, td);
984263bc
MD
3263 }
3264
3265 /*
3266 * release our private reference.
3267 */
dadab5e9 3268 fdrop(fp, td);
fad57d0e 3269 vrele(vp);
984263bc
MD
3270 return (error);
3271 }
5fd012e0 3272 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
3273 fp->f_flag |= FHASLOCK;
3274 }
3275 if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0))
3b568787 3276 vfs_object_create(vp, td);
984263bc 3277
fad57d0e 3278 vput(vp);
dadab5e9 3279 fdrop(fp, td);
c7114eea 3280 uap->sysmsg_result = indx;
984263bc
MD
3281 return (0);
3282
3283bad:
3284 vput(vp);
3285 return (error);
3286}
3287
41c20dac
MD
3288/*
3289 * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
3290 */
984263bc 3291int
41c20dac 3292fhstat(struct fhstat_args *uap)
984263bc 3293{
dadab5e9 3294 struct thread *td = curthread;
984263bc
MD
3295 struct stat sb;
3296 fhandle_t fh;
3297 struct mount *mp;
3298 struct vnode *vp;
3299 int error;
3300
3301 /*
3302 * Must be super user
3303 */
dadab5e9 3304 error = suser(td);
984263bc
MD
3305 if (error)
3306 return (error);