Add a fairly bad hack to detect ripouts that might occur during a list
[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 $
753fd850 40 * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.96 2006/06/05 07:26:10 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>
8fa76237 65#include <sys/spinlock.h>
8f6f8622 66#include <sys/kern_syscall.h>
70aac194 67#include <sys/objcache.h>
984263bc
MD
68
69#include <machine/limits.h>
1f2de5d4 70#include <vfs/union/union.h>
984263bc
MD
71#include <sys/sysctl.h>
72#include <vm/vm.h>
73#include <vm/vm_object.h>
984263bc
MD
74#include <vm/vm_page.h>
75
dadab5e9 76#include <sys/file2.h>
8fa76237 77#include <sys/spinlock2.h>
dadab5e9 78
e24b948e 79static int checkvp_chdir (struct vnode *vn, struct thread *td);
21739618 80static void checkdirs (struct vnode *olddp, struct namecache *ncp);
402ed7e1 81static int chroot_refuse_vdir_fds (struct filedesc *fdp);
75ffff0d 82static int chroot_visible_mnt(struct mount *mp, struct proc *p);
402ed7e1
RG
83static int getutimes (const struct timeval *, struct timespec *);
84static int setfown (struct vnode *, uid_t, gid_t);
85static int setfmode (struct vnode *, int);
86static int setfflags (struct vnode *, int);
87static int setutimes (struct vnode *, const struct timespec *, int);
984263bc
MD
88static int usermount = 0; /* if 1, non-root can mount fs. */
89
402ed7e1 90int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
984263bc
MD
91
92SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
93
94/*
95 * Virtual File System System Calls
96 */
97
98/*
99 * Mount a file system.
100 */
41c20dac
MD
101/*
102 * mount_args(char *type, char *path, int flags, caddr_t data)
103 */
984263bc
MD
104/* ARGSUSED */
105int
753fd850 106sys_mount(struct mount_args *uap)
984263bc 107{
dadab5e9
MD
108 struct thread *td = curthread;
109 struct proc *p = td->td_proc;
984263bc 110 struct vnode *vp;
21739618 111 struct namecache *ncp;
984263bc
MD
112 struct mount *mp;
113 struct vfsconf *vfsp;
114 int error, flag = 0, flag2 = 0;
115 struct vattr va;
21739618 116 struct nlookupdata nd;
984263bc 117 char fstypename[MFSNAMELEN];
21739618 118 struct nlcomponent nlc;
acde96db 119 struct ucred *cred = p->p_ucred;
984263bc 120
cd0d3e17 121 KKASSERT(p);
acde96db 122 if (cred->cr_prison != NULL)
84c3a710 123 return (EPERM);
dadab5e9 124 if (usermount == 0 && (error = suser(td)))
984263bc
MD
125 return (error);
126 /*
127 * Do not allow NFS export by non-root users.
128 */
ab2eb4eb 129 if (uap->flags & MNT_EXPORTED) {
dadab5e9 130 error = suser(td);
984263bc
MD
131 if (error)
132 return (error);
133 }
134 /*
135 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
136 */
dadab5e9 137 if (suser(td))
ab2eb4eb 138 uap->flags |= MNT_NOSUID | MNT_NODEV;
21739618 139
984263bc 140 /*
21739618 141 * Lookup the requested path and extract the ncp and vnode.
984263bc 142 */
21739618
MD
143 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
144 if (error == 0) {
145 if ((error = nlookup(&nd)) == 0) {
146 if (nd.nl_ncp->nc_vp == NULL)
147 error = ENOENT;
148 }
149 }
150 if (error) {
151 nlookup_done(&nd);
984263bc 152 return (error);
21739618
MD
153 }
154
155 /*
156 * Extract the locked+refd ncp and cleanup the nd structure
157 */
158 ncp = nd.nl_ncp;
159 nd.nl_ncp = NULL;
160 nlookup_done(&nd);
161
162 /*
163 * now we have the locked ref'd ncp and unreferenced vnode.
164 */
165 vp = ncp->nc_vp;
87de5057 166 if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
21739618
MD
167 cache_put(ncp);
168 return (error);
169 }
170 cache_unlock(ncp);
171
172 /*
173 * Now we have an unlocked ref'd ncp and a locked ref'd vp
174 */
ab2eb4eb 175 if (uap->flags & MNT_UPDATE) {
984263bc 176 if ((vp->v_flag & VROOT) == 0) {
21739618 177 cache_drop(ncp);
984263bc
MD
178 vput(vp);
179 return (EINVAL);
180 }
181 mp = vp->v_mount;
182 flag = mp->mnt_flag;
183 flag2 = mp->mnt_kern_flag;
184 /*
185 * We only allow the filesystem to be reloaded if it
186 * is currently mounted read-only.
187 */
ab2eb4eb 188 if ((uap->flags & MNT_RELOAD) &&
984263bc 189 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
21739618 190 cache_drop(ncp);
984263bc
MD
191 vput(vp);
192 return (EOPNOTSUPP); /* Needs translation */
193 }
194 /*
195 * Only root, or the user that did the original mount is
196 * permitted to update it.
197 */
acde96db 198 if (mp->mnt_stat.f_owner != cred->cr_uid &&
dadab5e9 199 (error = suser(td))) {
21739618 200 cache_drop(ncp);
984263bc
MD
201 vput(vp);
202 return (error);
203 }
f9642f56 204 if (vfs_busy(mp, LK_NOWAIT)) {
21739618 205 cache_drop(ncp);
984263bc
MD
206 vput(vp);
207 return (EBUSY);
208 }
984263bc
MD
209 if ((vp->v_flag & VMOUNT) != 0 ||
210 vp->v_mountedhere != NULL) {
21739618 211 cache_drop(ncp);
f9642f56 212 vfs_unbusy(mp);
984263bc
MD
213 vput(vp);
214 return (EBUSY);
215 }
216 vp->v_flag |= VMOUNT;
984263bc 217 mp->mnt_flag |=
ab2eb4eb 218 uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
ca466bae 219 VOP_UNLOCK(vp, 0);
984263bc
MD
220 goto update;
221 }
222 /*
223 * If the user is not root, ensure that they own the directory
224 * onto which we are attempting to mount.
225 */
87de5057 226 if ((error = VOP_GETATTR(vp, &va)) ||
acde96db 227 (va.va_uid != cred->cr_uid && (error = suser(td)))) {
21739618 228 cache_drop(ncp);
984263bc
MD
229 vput(vp);
230 return (error);
231 }
87de5057 232 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
21739618 233 cache_drop(ncp);
984263bc
MD
234 vput(vp);
235 return (error);
236 }
237 if (vp->v_type != VDIR) {
21739618 238 cache_drop(ncp);
984263bc
MD
239 vput(vp);
240 return (ENOTDIR);
241 }
ab2eb4eb 242 if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
21739618 243 cache_drop(ncp);
984263bc
MD
244 vput(vp);
245 return (error);
246 }
21739618 247 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
984263bc
MD
248 if (!strcmp(vfsp->vfc_name, fstypename))
249 break;
21739618 250 }
984263bc
MD
251 if (vfsp == NULL) {
252 linker_file_t lf;
253
254 /* Only load modules for root (very important!) */
dadab5e9 255 if ((error = suser(td)) != 0) {
21739618 256 cache_drop(ncp);
984263bc
MD
257 vput(vp);
258 return error;
259 }
260 error = linker_load_file(fstypename, &lf);
261 if (error || lf == NULL) {
21739618 262 cache_drop(ncp);
984263bc
MD
263 vput(vp);
264 if (lf == NULL)
265 error = ENODEV;
266 return error;
267 }
268 lf->userrefs++;
269 /* lookup again, see if the VFS was loaded */
21739618 270 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
984263bc
MD
271 if (!strcmp(vfsp->vfc_name, fstypename))
272 break;
21739618 273 }
984263bc
MD
274 if (vfsp == NULL) {
275 lf->userrefs--;
276 linker_file_unload(lf);
21739618 277 cache_drop(ncp);
984263bc
MD
278 vput(vp);
279 return (ENODEV);
280 }
281 }
984263bc
MD
282 if ((vp->v_flag & VMOUNT) != 0 ||
283 vp->v_mountedhere != NULL) {
21739618 284 cache_drop(ncp);
984263bc
MD
285 vput(vp);
286 return (EBUSY);
287 }
288 vp->v_flag |= VMOUNT;
984263bc
MD
289
290 /*
291 * Allocate and initialize the filesystem.
292 */
0961aa92 293 mp = malloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
984263bc
MD
294 TAILQ_INIT(&mp->mnt_nvnodelist);
295 TAILQ_INIT(&mp->mnt_reservedvnlist);
2281065e 296 TAILQ_INIT(&mp->mnt_jlist);
984263bc 297 mp->mnt_nvnodelistsize = 0;
f2770c70 298 lockinit(&mp->mnt_lock, "vfslock", 0, LK_NOPAUSE);
f9642f56 299 vfs_busy(mp, LK_NOWAIT);
984263bc
MD
300 mp->mnt_op = vfsp->vfc_vfsops;
301 mp->mnt_vfc = vfsp;
302 vfsp->vfc_refcount++;
303 mp->mnt_stat.f_type = vfsp->vfc_typenum;
304 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
305 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
306 mp->mnt_vnodecovered = vp;
acde96db 307 mp->mnt_stat.f_owner = cred->cr_uid;
984263bc 308 mp->mnt_iosize_max = DFLTPHYS;
ca466bae 309 VOP_UNLOCK(vp, 0);
984263bc
MD
310update:
311 /*
312 * Set the mount level flags.
313 */
ab2eb4eb 314 if (uap->flags & MNT_RDONLY)
984263bc
MD
315 mp->mnt_flag |= MNT_RDONLY;
316 else if (mp->mnt_flag & MNT_RDONLY)
317 mp->mnt_kern_flag |= MNTK_WANTRDWR;
318 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
319 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
320 MNT_NOSYMFOLLOW | MNT_IGNORE |
321 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
ab2eb4eb 322 mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
984263bc
MD
323 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
324 MNT_NOSYMFOLLOW | MNT_IGNORE |
325 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
326 /*
327 * Mount the filesystem.
328 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
fad57d0e 329 * get.
984263bc 330 */
acde96db 331 error = VFS_MOUNT(mp, uap->path, uap->data, cred);
984263bc
MD
332 if (mp->mnt_flag & MNT_UPDATE) {
333 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
334 mp->mnt_flag &= ~MNT_RDONLY;
335 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
336 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
337 if (error) {
338 mp->mnt_flag = flag;
339 mp->mnt_kern_flag = flag2;
340 }
f9642f56 341 vfs_unbusy(mp);
984263bc 342 vp->v_flag &= ~VMOUNT;
984263bc 343 vrele(vp);
21739618 344 cache_drop(ncp);
984263bc
MD
345 return (error);
346 }
ca466bae 347 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc 348 /*
21739618
MD
349 * Put the new filesystem on the mount list after root. The mount
350 * point gets its own mnt_ncp which is a special ncp linking the
351 * vnode-under to the root of the new mount. The lookup code
352 * detects the mount point going forward and detects the special
353 * mnt_ncp via NCP_MOUNTPT going backwards.
8e005a45
MD
354 *
355 * It is not necessary to invalidate or purge the vnode underneath
356 * because elements under the mount will be given their own glue
357 * namecache record.
984263bc 358 */
984263bc 359 if (!error) {
21739618
MD
360 nlc.nlc_nameptr = "";
361 nlc.nlc_namelen = 0;
362 mp->mnt_ncp = cache_nlookup(ncp, &nlc);
2386d5b6 363 cache_setunresolved(mp->mnt_ncp);
21739618 364 mp->mnt_ncp->nc_flag |= NCF_MOUNTPT;
646a1cda 365 mp->mnt_ncp->nc_mount = mp;
21739618
MD
366 cache_drop(ncp);
367 /* XXX get the root of the fs and cache_setvp(mnt_ncp...) */
984263bc
MD
368 vp->v_flag &= ~VMOUNT;
369 vp->v_mountedhere = mp;
861905fb 370 mountlist_insert(mp, MNTINS_LAST);
21739618
MD
371 checkdirs(vp, mp->mnt_ncp);
372 cache_unlock(mp->mnt_ncp); /* leave ref intact */
ca466bae 373 VOP_UNLOCK(vp, 0);
41a01a4d 374 error = vfs_allocate_syncvnode(mp);
f9642f56 375 vfs_unbusy(mp);
acde96db 376 if ((error = VFS_START(mp, 0)) != 0)
984263bc
MD
377 vrele(vp);
378 } else {
6ddb7618
MD
379 vfs_rm_vnodeops(&mp->mnt_vn_coherency_ops);
380 vfs_rm_vnodeops(&mp->mnt_vn_journal_ops);
381 vfs_rm_vnodeops(&mp->mnt_vn_norm_ops);
0961aa92
MD
382 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
383 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
984263bc 384 vp->v_flag &= ~VMOUNT;
984263bc 385 mp->mnt_vfc->vfc_refcount--;
f9642f56 386 vfs_unbusy(mp);
e38462a2 387 free(mp, M_MOUNT);
21739618 388 cache_drop(ncp);
984263bc
MD
389 vput(vp);
390 }
391 return (error);
392}
393
394/*
395 * Scan all active processes to see if any of them have a current
396 * or root directory onto which the new filesystem has just been
397 * mounted. If so, replace them with the new mount point.
21739618
MD
398 *
399 * The passed ncp is ref'd and locked (from the mount code) and
400 * must be associated with the vnode representing the root of the
401 * mount point.
984263bc 402 */
8fa76237
MD
403struct checkdirs_info {
404 struct vnode *olddp;
405 struct vnode *newdp;
406 struct namecache *ncp;
407};
408
409static int checkdirs_callback(struct proc *p, void *data);
410
984263bc 411static void
21739618 412checkdirs(struct vnode *olddp, struct namecache *ncp)
984263bc 413{
8fa76237 414 struct checkdirs_info info;
984263bc 415 struct vnode *newdp;
690a3127 416 struct mount *mp;
984263bc
MD
417
418 if (olddp->v_usecount == 1)
419 return;
690a3127
MD
420 mp = olddp->v_mountedhere;
421 if (VFS_ROOT(mp, &newdp))
984263bc 422 panic("mount: lost mount");
21739618
MD
423 cache_setvp(ncp, newdp);
424
425 if (rootvnode == olddp) {
426 vref(newdp);
427 vfs_cache_setroot(newdp, cache_hold(ncp));
428 }
429
8fa76237
MD
430 info.olddp = olddp;
431 info.newdp = newdp;
432 info.ncp = ncp;
433 allproc_scan(checkdirs_callback, &info);
434 vput(newdp);
435}
436
437/*
438 * NOTE: callback is not MP safe because the scanned process's filedesc
439 * structure can be ripped out from under us, amoung other things.
440 */
441static int
442checkdirs_callback(struct proc *p, void *data)
443{
444 struct checkdirs_info *info = data;
445 struct filedesc *fdp;
446 struct namecache *ncdrop1;
447 struct namecache *ncdrop2;
448 struct vnode *vprele1;
449 struct vnode *vprele2;
450
451 if ((fdp = p->p_fd) != NULL) {
452 ncdrop1 = NULL;
453 ncdrop2 = NULL;
454 vprele1 = NULL;
455 vprele2 = NULL;
456
457 /*
458 * MPUNSAFE - XXX fdp can be pulled out from under a
459 * foreign process.
460 *
461 * A shared filedesc is ok, we don't have to copy it
462 * because we are making this change globally.
463 */
464 spin_lock_wr(&fdp->fd_spin);
465 if (fdp->fd_cdir == info->olddp) {
466 vprele1 = fdp->fd_cdir;
467 vref(info->newdp);
468 fdp->fd_cdir = info->newdp;
469 ncdrop1 = fdp->fd_ncdir;
470 fdp->fd_ncdir = cache_hold(info->ncp);
984263bc 471 }
8fa76237
MD
472 if (fdp->fd_rdir == info->olddp) {
473 vprele2 = fdp->fd_rdir;
474 vref(info->newdp);
475 fdp->fd_rdir = info->newdp;
476 ncdrop2 = fdp->fd_nrdir;
477 fdp->fd_nrdir = cache_hold(info->ncp);
984263bc 478 }
8fa76237
MD
479 spin_unlock_wr(&fdp->fd_spin);
480 if (ncdrop1)
481 cache_drop(ncdrop1);
482 if (ncdrop2)
483 cache_drop(ncdrop2);
484 if (vprele1)
485 vrele(vprele1);
486 if (vprele2)
487 vrele(vprele2);
984263bc 488 }
8fa76237 489 return(0);
984263bc
MD
490}
491
492/*
493 * Unmount a file system.
494 *
495 * Note: unmount takes a path to the vnode mounted on as argument,
496 * not special file (as before).
497 */
41c20dac
MD
498/*
499 * umount_args(char *path, int flags)
500 */
984263bc
MD
501/* ARGSUSED */
502int
753fd850 503sys_unmount(struct unmount_args *uap)
984263bc 504{
dadab5e9
MD
505 struct thread *td = curthread;
506 struct proc *p = td->td_proc;
d2293868 507 struct mount *mp = NULL;
984263bc 508 int error;
fad57d0e 509 struct nlookupdata nd;
984263bc 510
dadab5e9 511 KKASSERT(p);
84c3a710
HP
512 if (p->p_ucred->cr_prison != NULL)
513 return (EPERM);
514 if (usermount == 0 && (error = suser(td)))
515 return (error);
516
ab2eb4eb 517 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
518 if (error == 0)
519 error = nlookup(&nd);
fad57d0e 520 if (error)
d2293868 521 goto out;
fad57d0e 522
d2293868 523 mp = nd.nl_ncp->nc_mount;
984263bc
MD
524
525 /*
526 * Only root, or the user that did the original mount is
527 * permitted to unmount this filesystem.
528 */
529 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
d2293868
MD
530 (error = suser(td)))
531 goto out;
984263bc
MD
532
533 /*
534 * Don't allow unmounting the root file system.
535 */
536 if (mp->mnt_flag & MNT_ROOTFS) {
d2293868
MD
537 error = EINVAL;
538 goto out;
984263bc
MD
539 }
540
541 /*
542 * Must be the root of the filesystem
543 */
d2293868
MD
544 if (! (nd.nl_ncp->nc_flag & NCF_MOUNTPT)) {
545 error = EINVAL;
546 goto out;
984263bc 547 }
d2293868
MD
548
549out:
550 nlookup_done(&nd);
551 if (error)
552 return (error);
acde96db 553 return (dounmount(mp, uap->flags));
984263bc
MD
554}
555
556/*
557 * Do the actual file system unmount.
558 */
861905fb
MD
559static int
560dounmount_interlock(struct mount *mp)
561{
562 if (mp->mnt_kern_flag & MNTK_UNMOUNT)
563 return (EBUSY);
564 mp->mnt_kern_flag |= MNTK_UNMOUNT;
565 return(0);
566}
567
984263bc 568int
acde96db 569dounmount(struct mount *mp, int flags)
984263bc
MD
570{
571 struct vnode *coveredvp;
572 int error;
573 int async_flag;
2df9419c 574 int lflags;
984263bc 575
861905fb
MD
576 /*
577 * Exclusive access for unmounting purposes
578 */
579 if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
580 return (error);
581
582 /*
583 * Allow filesystems to detect that a forced unmount is in progress.
584 */
984263bc
MD
585 if (flags & MNT_FORCE)
586 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
2df9419c 587 lflags = LK_EXCLUSIVE | ((flags & MNT_FORCE) ? 0 : LK_NOWAIT);
df4f70a6 588 error = lockmgr(&mp->mnt_lock, lflags);
984263bc
MD
589 if (error) {
590 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
591 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 592 wakeup(mp);
984263bc
MD
593 return (error);
594 }
595
596 if (mp->mnt_flag & MNT_EXPUBLIC)
597 vfs_setpublicfs(NULL, NULL, NULL);
598
599 vfs_msync(mp, MNT_WAIT);
600 async_flag = mp->mnt_flag & MNT_ASYNC;
601 mp->mnt_flag &=~ MNT_ASYNC;
602 cache_purgevfs(mp); /* remove cache entries for this file sys */
603 if (mp->mnt_syncer != NULL)
604 vrele(mp->mnt_syncer);
605 if (((mp->mnt_flag & MNT_RDONLY) ||
87de5057 606 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) ||
984263bc 607 (flags & MNT_FORCE))
acde96db 608 error = VFS_UNMOUNT(mp, flags);
984263bc 609 if (error) {
41a01a4d
MD
610 if (mp->mnt_syncer == NULL)
611 vfs_allocate_syncvnode(mp);
984263bc
MD
612 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
613 mp->mnt_flag |= async_flag;
df4f70a6 614 lockmgr(&mp->mnt_lock, LK_RELEASE);
984263bc 615 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 616 wakeup(mp);
984263bc
MD
617 return (error);
618 }
432b8263
MD
619 /*
620 * Clean up any journals still associated with the mount after
621 * filesystem activity has ceased.
622 */
623 journal_remove_all_journals(mp,
624 ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));
625
861905fb 626 mountlist_remove(mp);
0961aa92
MD
627
628 /*
629 * Remove any installed vnode ops here so the individual VFSs don't
630 * have to.
631 */
6ddb7618
MD
632 vfs_rm_vnodeops(&mp->mnt_vn_coherency_ops);
633 vfs_rm_vnodeops(&mp->mnt_vn_journal_ops);
634 vfs_rm_vnodeops(&mp->mnt_vn_norm_ops);
0961aa92
MD
635 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
636 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
637
984263bc 638 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
7ea21ed1 639 coveredvp->v_mountedhere = NULL;
984263bc 640 vrele(coveredvp);
21739618
MD
641 cache_drop(mp->mnt_ncp);
642 mp->mnt_ncp = NULL;
984263bc
MD
643 }
644 mp->mnt_vfc->vfc_refcount--;
645 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
646 panic("unmount: dangling vnode");
df4f70a6 647 lockmgr(&mp->mnt_lock, LK_RELEASE);
984263bc 648 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2
DR
649 wakeup(mp);
650 free(mp, M_MOUNT);
984263bc
MD
651 return (0);
652}
653
654/*
655 * Sync each mounted filesystem.
656 */
984263bc
MD
657
658#ifdef DEBUG
659static int syncprt = 0;
660SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
6bdbb368 661#endif /* DEBUG */
984263bc 662
861905fb
MD
663static int sync_callback(struct mount *mp, void *data);
664
984263bc
MD
665/* ARGSUSED */
666int
753fd850 667sys_sync(struct sync_args *uap)
984263bc 668{
861905fb 669 mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
6bdbb368 670#ifdef DEBUG
861905fb
MD
671 /*
672 * print out buffer pool stat information on each sync() call.
673 */
984263bc
MD
674 if (syncprt)
675 vfs_bufstats();
6bdbb368 676#endif /* DEBUG */
984263bc
MD
677 return (0);
678}
679
861905fb
MD
680static
681int
682sync_callback(struct mount *mp, void *data __unused)
683{
684 int asyncflag;
685
686 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
687 asyncflag = mp->mnt_flag & MNT_ASYNC;
688 mp->mnt_flag &= ~MNT_ASYNC;
689 vfs_msync(mp, MNT_NOWAIT);
87de5057 690 VFS_SYNC(mp, MNT_NOWAIT);
861905fb
MD
691 mp->mnt_flag |= asyncflag;
692 }
693 return(0);
694}
695
984263bc
MD
696/* XXX PRISON: could be per prison flag */
697static int prison_quotas;
698#if 0
699SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
700#endif
701
702/*
41c20dac
MD
703 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
704 *
984263bc
MD
705 * Change filesystem quotas.
706 */
984263bc
MD
707/* ARGSUSED */
708int
753fd850 709sys_quotactl(struct quotactl_args *uap)
984263bc 710{
fad57d0e
MD
711 struct nlookupdata nd;
712 struct thread *td;
713 struct proc *p;
41c20dac 714 struct mount *mp;
984263bc 715 int error;
984263bc 716
fad57d0e
MD
717 td = curthread;
718 p = td->td_proc;
41c20dac 719 if (p->p_ucred->cr_prison && !prison_quotas)
984263bc 720 return (EPERM);
fad57d0e 721
ab2eb4eb 722 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
723 if (error == 0)
724 error = nlookup(&nd);
725 if (error == 0) {
726 mp = nd.nl_ncp->nc_mount;
ab2eb4eb 727 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
acde96db 728 uap->arg, nd.nl_cred);
fad57d0e
MD
729 }
730 nlookup_done(&nd);
731 return (error);
984263bc
MD
732}
733
949ecb9b 734/*
2281065e 735 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
949ecb9b
MD
736 * void *buf, int buflen)
737 *
738 * This function operates on a mount point and executes the specified
739 * operation using the specified control data, and possibly returns data.
740 *
741 * The actual number of bytes stored in the result buffer is returned, 0
742 * if none, otherwise an error is returned.
743 */
744/* ARGSUSED */
745int
753fd850 746sys_mountctl(struct mountctl_args *uap)
949ecb9b
MD
747{
748 struct thread *td = curthread;
749 struct proc *p = td->td_proc;
2281065e 750 struct file *fp;
949ecb9b
MD
751 void *ctl = NULL;
752 void *buf = NULL;
753 char *path = NULL;
754 int error;
755
756 /*
757 * Sanity and permissions checks. We must be root.
758 */
759 KKASSERT(p);
760 if (p->p_ucred->cr_prison != NULL)
761 return (EPERM);
762 if ((error = suser(td)) != 0)
763 return (error);
764
765 /*
766 * Argument length checks
767 */
39b13188 768 if (uap->ctllen < 0 || uap->ctllen > 1024)
949ecb9b 769 return (EINVAL);
39b13188 770 if (uap->buflen < 0 || uap->buflen > 16 * 1024)
949ecb9b
MD
771 return (EINVAL);
772 if (uap->path == NULL)
773 return (EINVAL);
774
775 /*
776 * Allocate the necessary buffers and copyin data
777 */
70aac194 778 path = objcache_get(namei_oc, M_WAITOK);
949ecb9b
MD
779 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
780 if (error)
781 goto done;
782
783 if (uap->ctllen) {
784 ctl = malloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
785 error = copyin(uap->ctl, ctl, uap->ctllen);
786 if (error)
787 goto done;
788 }
789 if (uap->buflen)
790 buf = malloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
791
2281065e
MD
792 /*
793 * Validate the descriptor
794 */
fa541be6
MD
795 fp = holdfp(p->p_fd, uap->fd, -1);
796 if (fp == NULL) {
2281065e
MD
797 error = EBADF;
798 goto done;
799 }
2281065e 800
949ecb9b
MD
801 /*
802 * Execute the internal kernel function and clean up.
803 */
2281065e
MD
804 error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
805 if (fp)
9f87144f 806 fdrop(fp);
949ecb9b
MD
807 if (error == 0 && uap->sysmsg_result > 0)
808 error = copyout(buf, uap->buf, uap->sysmsg_result);
809done:
810 if (path)
70aac194 811 objcache_put(namei_oc, path);
949ecb9b
MD
812 if (ctl)
813 free(ctl, M_TEMP);
814 if (buf)
815 free(buf, M_TEMP);
816 return (error);
817}
818
819/*
820 * Execute a mount control operation by resolving the path to a mount point
821 * and calling vop_mountctl().
822 */
823int
2281065e
MD
824kern_mountctl(const char *path, int op, struct file *fp,
825 const void *ctl, int ctllen,
949ecb9b
MD
826 void *buf, int buflen, int *res)
827{
949ecb9b
MD
828 struct vnode *vp;
829 struct mount *mp;
830 struct nlookupdata nd;
831 int error;
832
833 *res = 0;
834 vp = NULL;
835 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
836 if (error == 0)
837 error = nlookup(&nd);
838 if (error == 0)
839 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
840 nlookup_done(&nd);
841 if (error)
842 return (error);
843
844 mp = vp->v_mount;
845
846 /*
847 * Must be the root of the filesystem
848 */
849 if ((vp->v_flag & VROOT) == 0) {
850 vput(vp);
851 return (EINVAL);
852 }
2281065e 853 error = vop_mountctl(mp->mnt_vn_use_ops, op, fp, ctl, ctllen,
949ecb9b
MD
854 buf, buflen, res);
855 vput(vp);
856 return (error);
857}
858
984263bc 859int
fad57d0e 860kern_statfs(struct nlookupdata *nd, struct statfs *buf)
984263bc 861{
dadab5e9 862 struct thread *td = curthread;
75ffff0d 863 struct proc *p = td->td_proc;
41c20dac
MD
864 struct mount *mp;
865 struct statfs *sp;
75ffff0d 866 char *fullpath, *freepath;
984263bc 867 int error;
984263bc 868
fad57d0e 869 if ((error = nlookup(nd)) != 0)
984263bc 870 return (error);
fad57d0e 871 mp = nd->nl_ncp->nc_mount;
984263bc 872 sp = &mp->mnt_stat;
acde96db 873 if ((error = VFS_STATFS(mp, sp, nd->nl_cred)) != 0)
984263bc 874 return (error);
75ffff0d
JS
875
876 error = cache_fullpath(p, mp->mnt_ncp, &fullpath, &freepath);
877 if (error)
878 return(error);
879 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
880 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
881 free(freepath, M_TEMP);
882
984263bc 883 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509
DRJ
884 bcopy(sp, buf, sizeof(*buf));
885 /* Only root should have access to the fsid's. */
886 if (suser(td))
887 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
888 return (0);
984263bc
MD
889}
890
891/*
9697c509 892 * statfs_args(char *path, struct statfs *buf)
41c20dac 893 *
984263bc
MD
894 * Get filesystem statistics.
895 */
984263bc 896int
753fd850 897sys_statfs(struct statfs_args *uap)
9697c509 898{
fad57d0e 899 struct nlookupdata nd;
9697c509
DRJ
900 struct statfs buf;
901 int error;
902
fad57d0e
MD
903 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
904 if (error == 0)
905 error = kern_statfs(&nd, &buf);
906 nlookup_done(&nd);
9697c509
DRJ
907 if (error == 0)
908 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
909 return (error);
910}
911
912int
913kern_fstatfs(int fd, struct statfs *buf)
984263bc 914{
dadab5e9
MD
915 struct thread *td = curthread;
916 struct proc *p = td->td_proc;
984263bc
MD
917 struct file *fp;
918 struct mount *mp;
1fd87d54 919 struct statfs *sp;
75ffff0d 920 char *fullpath, *freepath;
984263bc 921 int error;
984263bc 922
dadab5e9 923 KKASSERT(p);
5b287bba 924 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc
MD
925 return (error);
926 mp = ((struct vnode *)fp->f_data)->v_mount;
5b287bba
MD
927 if (mp == NULL) {
928 error = EBADF;
929 goto done;
930 }
931 if (fp->f_cred == NULL) {
932 error = EINVAL;
933 goto done;
934 }
984263bc 935 sp = &mp->mnt_stat;
5b287bba
MD
936 if ((error = VFS_STATFS(mp, sp, fp->f_cred)) != 0)
937 goto done;
75ffff0d 938
5b287bba
MD
939 if ((error = cache_fullpath(p, mp->mnt_ncp, &fullpath, &freepath)) != 0)
940 goto done;
75ffff0d
JS
941 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
942 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
943 free(freepath, M_TEMP);
944
984263bc 945 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509 946 bcopy(sp, buf, sizeof(*buf));
75ffff0d 947
9697c509
DRJ
948 /* Only root should have access to the fsid's. */
949 if (suser(td))
950 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
5b287bba
MD
951 error = 0;
952done:
953 fdrop(fp);
954 return (error);
9697c509
DRJ
955}
956
957/*
958 * fstatfs_args(int fd, struct statfs *buf)
959 *
960 * Get filesystem statistics.
961 */
962int
753fd850 963sys_fstatfs(struct fstatfs_args *uap)
9697c509
DRJ
964{
965 struct statfs buf;
966 int error;
967
968 error = kern_fstatfs(uap->fd, &buf);
969
970 if (error == 0)
971 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
972 return (error);
984263bc
MD
973}
974
975/*
41c20dac
MD
976 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
977 *
984263bc
MD
978 * Get statistics on all filesystems.
979 */
861905fb
MD
980
981struct getfsstat_info {
982 struct statfs *sfsp;
983 long count;
984 long maxcount;
985 int error;
986 int flags;
987 int is_chrooted;
861905fb
MD
988 struct proc *p;
989};
990
991static int getfsstat_callback(struct mount *, void *);
992
41c20dac 993/* ARGSUSED */
984263bc 994int
753fd850 995sys_getfsstat(struct getfsstat_args *uap)
984263bc 996{
dadab5e9 997 struct thread *td = curthread;
75ffff0d 998 struct proc *p = td->td_proc;
861905fb 999 struct getfsstat_info info;
75ffff0d 1000
861905fb 1001 bzero(&info, sizeof(info));
75ffff0d 1002 if (p != NULL && (p->p_fd->fd_nrdir->nc_flag & NCF_ROOT) == 0)
861905fb
MD
1003 info.is_chrooted = 1;
1004 else
1005 info.is_chrooted = 0;
1006
1007 info.maxcount = uap->bufsize / sizeof(struct statfs);
1008 info.sfsp = uap->buf;
1009 info.count = 0;
1010 info.flags = uap->flags;
861905fb
MD
1011 info.p = p;
1012
1013 mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
1014 if (info.sfsp && info.count > info.maxcount)
1015 uap->sysmsg_result = info.maxcount;
75ffff0d 1016 else
861905fb
MD
1017 uap->sysmsg_result = info.count;
1018 return (info.error);
1019}
1020
1021static int
1022getfsstat_callback(struct mount *mp, void *data)
1023{
1024 struct getfsstat_info *info = data;
1025 struct statfs *sp;
1026 char *freepath;
1027 char *fullpath;
1028 int error;
1029
1030 if (info->sfsp && info->count < info->maxcount) {
1031 if (info->is_chrooted && !chroot_visible_mnt(mp, info->p))
1032 return(0);
1033 sp = &mp->mnt_stat;
1034
1035 /*
1036 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1037 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1038 * overrides MNT_WAIT.
1039 */
1040 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1041 (info->flags & MNT_WAIT)) &&
acde96db 1042 (error = VFS_STATFS(mp, sp, info->p->p_ucred))) {
861905fb
MD
1043 return(0);
1044 }
1045 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1046
1047 error = cache_fullpath(info->p, mp->mnt_ncp,
1048 &fullpath, &freepath);
1049 if (error) {
1050 info->error = error;
1051 return(-1);
984263bc 1052 }
861905fb
MD
1053 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1054 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1055 free(freepath, M_TEMP);
1056
1057 error = copyout(sp, info->sfsp, sizeof(*sp));
1058 if (error) {
1059 info->error = error;
1060 return (-1);
984263bc 1061 }
861905fb 1062 ++info->sfsp;
984263bc 1063 }
861905fb
MD
1064 info->count++;
1065 return(0);
984263bc
MD
1066}
1067
1068/*
41c20dac
MD
1069 * fchdir_args(int fd)
1070 *
984263bc
MD
1071 * Change current working directory to a given file descriptor.
1072 */
984263bc
MD
1073/* ARGSUSED */
1074int
753fd850 1075sys_fchdir(struct fchdir_args *uap)
984263bc 1076{
dadab5e9
MD
1077 struct thread *td = curthread;
1078 struct proc *p = td->td_proc;
41c20dac 1079 struct filedesc *fdp = p->p_fd;
21739618 1080 struct vnode *vp, *ovp;
984263bc
MD
1081 struct mount *mp;
1082 struct file *fp;
21739618
MD
1083 struct namecache *ncp, *oncp;
1084 struct namecache *nct;
984263bc
MD
1085 int error;
1086
5b287bba 1087 if ((error = holdvnode(fdp, uap->fd, &fp)) != 0)
984263bc
MD
1088 return (error);
1089 vp = (struct vnode *)fp->f_data;
597aea93 1090 vref(vp);
ca466bae 1091 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
21739618 1092 if (vp->v_type != VDIR || fp->f_ncp == NULL)
984263bc
MD
1093 error = ENOTDIR;
1094 else
87de5057 1095 error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
984263bc
MD
1096 if (error) {
1097 vput(vp);
5b287bba 1098 fdrop(fp);
984263bc
MD
1099 return (error);
1100 }
21739618
MD
1101 ncp = cache_hold(fp->f_ncp);
1102 while (!error && (mp = vp->v_mountedhere) != NULL) {
1103 error = nlookup_mp(mp, &nct);
1104 if (error == 0) {
1105 cache_unlock(nct); /* leave ref intact */
1106 vput(vp);
1107 vp = nct->nc_vp;
87de5057 1108 error = vget(vp, LK_SHARED);
21739618
MD
1109 KKASSERT(error == 0);
1110 cache_drop(ncp);
1111 ncp = nct;
1112 }
1113 }
1114 if (error == 0) {
1115 ovp = fdp->fd_cdir;
1116 oncp = fdp->fd_ncdir;
ca466bae 1117 VOP_UNLOCK(vp, 0); /* leave ref intact */
21739618
MD
1118 fdp->fd_cdir = vp;
1119 fdp->fd_ncdir = ncp;
1120 cache_drop(oncp);
1121 vrele(ovp);
1122 } else {
1123 cache_drop(ncp);
1124 vput(vp);
1125 }
5b287bba 1126 fdrop(fp);
21739618 1127 return (error);
984263bc
MD
1128}
1129
9697c509 1130int
21739618 1131kern_chdir(struct nlookupdata *nd)
9697c509
DRJ
1132{
1133 struct thread *td = curthread;
1134 struct proc *p = td->td_proc;
1135 struct filedesc *fdp = p->p_fd;
21739618
MD
1136 struct vnode *vp, *ovp;
1137 struct namecache *oncp;
9697c509
DRJ
1138 int error;
1139
21739618 1140 if ((error = nlookup(nd)) != 0)
9697c509 1141 return (error);
21739618
MD
1142 if ((vp = nd->nl_ncp->nc_vp) == NULL)
1143 return (ENOENT);
87de5057 1144 if ((error = vget(vp, LK_SHARED)) != 0)
21739618
MD
1145 return (error);
1146
1147 error = checkvp_chdir(vp, td);
ca466bae 1148 VOP_UNLOCK(vp, 0);
21739618
MD
1149 if (error == 0) {
1150 ovp = fdp->fd_cdir;
1151 oncp = fdp->fd_ncdir;
1152 cache_unlock(nd->nl_ncp); /* leave reference intact */
1153 fdp->fd_ncdir = nd->nl_ncp;
1154 fdp->fd_cdir = vp;
1155 cache_drop(oncp);
1156 vrele(ovp);
1157 nd->nl_ncp = NULL;
1158 } else {
1159 vrele(vp);
e24b948e 1160 }
e24b948e 1161 return (error);
9697c509
DRJ
1162}
1163
984263bc 1164/*
41c20dac
MD
1165 * chdir_args(char *path)
1166 *
984263bc
MD
1167 * Change current working directory (``.'').
1168 */
984263bc 1169int
753fd850 1170sys_chdir(struct chdir_args *uap)
984263bc 1171{
21739618 1172 struct nlookupdata nd;
9697c509 1173 int error;
984263bc 1174
21739618 1175 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e 1176 if (error == 0)
21739618 1177 error = kern_chdir(&nd);
fad57d0e 1178 nlookup_done(&nd);
9697c509 1179 return (error);
984263bc
MD
1180}
1181
1182/*
1183 * Helper function for raised chroot(2) security function: Refuse if
1184 * any filedescriptors are open directories.
1185 */
1186static int
1187chroot_refuse_vdir_fds(fdp)
1188 struct filedesc *fdp;
1189{
1190 struct vnode *vp;
1191 struct file *fp;
1192 int error;
1193 int fd;
1194
1195 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
5b287bba 1196 if ((error = holdvnode(fdp, fd, &fp)) != 0)
984263bc
MD
1197 continue;
1198 vp = (struct vnode *)fp->f_data;
5b287bba
MD
1199 if (vp->v_type != VDIR) {
1200 fdrop(fp);
984263bc 1201 continue;
5b287bba
MD
1202 }
1203 fdrop(fp);
984263bc
MD
1204 return(EPERM);
1205 }
1206 return (0);
1207}
1208
1209/*
1210 * This sysctl determines if we will allow a process to chroot(2) if it
1211 * has a directory open:
1212 * 0: disallowed for all processes.
1213 * 1: allowed for processes that were not already chroot(2)'ed.
1214 * 2: allowed for all processes.
1215 */
1216
1217static int chroot_allow_open_directories = 1;
1218
1219SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1220 &chroot_allow_open_directories, 0, "");
1221
e24b948e 1222/*
21739618
MD
1223 * chroot to the specified namecache entry. We obtain the vp from the
1224 * namecache data. The passed ncp must be locked and referenced and will
1225 * remain locked and referenced on return.
e24b948e 1226 */
53dd6631
JS
1227int
1228kern_chroot(struct namecache *ncp)
e24b948e
MD
1229{
1230 struct thread *td = curthread;
1231 struct proc *p = td->td_proc;
1232 struct filedesc *fdp = p->p_fd;
21739618 1233 struct vnode *vp;
e24b948e
MD
1234 int error;
1235
1236 /*
1237 * Only root can chroot
1238 */
1239 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
1240 return (error);
1241
1242 /*
1243 * Disallow open directory descriptors (fchdir() breakouts).
1244 */
1245 if (chroot_allow_open_directories == 0 ||
1246 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1247 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1248 return (error);
1249 }
21739618
MD
1250 if ((vp = ncp->nc_vp) == NULL)
1251 return (ENOENT);
1252
87de5057 1253 if ((error = vget(vp, LK_SHARED)) != 0)
21739618 1254 return (error);
e24b948e
MD
1255
1256 /*
1257 * Check the validity of vp as a directory to change to and
1258 * associate it with rdir/jdir.
1259 */
21739618 1260 error = checkvp_chdir(vp, td);
ca466bae 1261 VOP_UNLOCK(vp, 0); /* leave reference intact */
21739618 1262 if (error == 0) {
e24b948e 1263 vrele(fdp->fd_rdir);
21739618 1264 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */
690a3127 1265 cache_drop(fdp->fd_nrdir);
21739618 1266 fdp->fd_nrdir = cache_hold(ncp);
e24b948e
MD
1267 if (fdp->fd_jdir == NULL) {
1268 fdp->fd_jdir = vp;
597aea93 1269 vref(fdp->fd_jdir);
21739618 1270 fdp->fd_njdir = cache_hold(ncp);
e24b948e 1271 }
21739618
MD
1272 } else {
1273 vrele(vp);
e24b948e
MD
1274 }
1275 return (error);
1276}
1277
984263bc 1278/*
41c20dac
MD
1279 * chroot_args(char *path)
1280 *
984263bc
MD
1281 * Change notion of root (``/'') directory.
1282 */
984263bc
MD
1283/* ARGSUSED */
1284int
753fd850 1285sys_chroot(struct chroot_args *uap)
984263bc 1286{
dadab5e9 1287 struct thread *td = curthread;
21739618 1288 struct nlookupdata nd;
e24b948e 1289 int error;
984263bc 1290
e24b948e 1291 KKASSERT(td->td_proc);
21739618 1292 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
53dd6631
JS
1293 if (error) {
1294 nlookup_done(&nd);
1295 return(error);
1296 }
1297 error = nlookup(&nd);
1298 if (error == 0)
1299 error = kern_chroot(nd.nl_ncp);
fad57d0e 1300 nlookup_done(&nd);
53dd6631 1301 return(error);
984263bc
MD
1302}
1303
1304/*
e24b948e
MD
1305 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1306 * determine whether it is legal to chdir to the vnode. The vnode's state
1307 * is not changed by this call.
984263bc 1308 */
e24b948e
MD
1309int
1310checkvp_chdir(struct vnode *vp, struct thread *td)
984263bc 1311{
984263bc
MD
1312 int error;
1313
984263bc
MD
1314 if (vp->v_type != VDIR)
1315 error = ENOTDIR;
1316 else
87de5057 1317 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred);
984263bc
MD
1318 return (error);
1319}
1320
984263bc 1321int
fad57d0e 1322kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
984263bc 1323{
dadab5e9
MD
1324 struct thread *td = curthread;
1325 struct proc *p = td->td_proc;
fbfe4e7d 1326 struct lwp *lp = td->td_lwp;
41c20dac 1327 struct filedesc *fdp = p->p_fd;
9697c509 1328 int cmode, flags;
984263bc 1329 struct file *nfp;
fad57d0e
MD
1330 struct file *fp;
1331 struct vnode *vp;
984263bc
MD
1332 int type, indx, error;
1333 struct flock lf;
984263bc 1334
984263bc
MD
1335 if ((oflags & O_ACCMODE) == O_ACCMODE)
1336 return (EINVAL);
1337 flags = FFLAGS(oflags);
fad57d0e 1338 error = falloc(p, &nfp, NULL);
984263bc
MD
1339 if (error)
1340 return (error);
1341 fp = nfp;
9697c509 1342 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
fad57d0e 1343
984263bc 1344 /*
fad57d0e
MD
1345 * XXX p_dupfd is a real mess. It allows a device to return a
1346 * file descriptor to be duplicated rather then doing the open
1347 * itself.
984263bc 1348 */
fbfe4e7d 1349 lp->lwp_dupfd = -1;
21739618
MD
1350
1351 /*
fad57d0e
MD
1352 * Call vn_open() to do the lookup and assign the vnode to the
1353 * file pointer. vn_open() does not change the ref count on fp
1354 * and the vnode, on success, will be inherited by the file pointer
1355 * and unlocked.
21739618 1356 */
fad57d0e
MD
1357 nd->nl_flags |= NLC_LOCKVP;
1358 error = vn_open(nd, fp, flags, cmode);
1359 nlookup_done(nd);
984263bc 1360 if (error) {
984263bc
MD
1361 /*
1362 * handle special fdopen() case. bleh. dupfdopen() is
1363 * responsible for dropping the old contents of ofiles[indx]
1364 * if it succeeds.
fad57d0e 1365 *
259b8ea0
MD
1366 * Note that fsetfd() will add a ref to fp which represents
1367 * the fd_files[] assignment. We must still drop our
1368 * reference.
984263bc 1369 */
fbfe4e7d 1370 if ((error == ENODEV || error == ENXIO) && lp->lwp_dupfd >= 0) {
259b8ea0
MD
1371 if (fdalloc(p, 0, &indx) == 0) {
1372 error = dupfdopen(p, indx, lp->lwp_dupfd, flags, error);
fad57d0e
MD
1373 if (error == 0) {
1374 *res = indx;
9f87144f 1375 fdrop(fp); /* our ref */
fad57d0e
MD
1376 return (0);
1377 }
259b8ea0 1378 fsetfd(p, NULL, indx);
fad57d0e 1379 }
984263bc 1380 }
9f87144f 1381 fdrop(fp); /* our ref */
984263bc
MD
1382 if (error == ERESTART)
1383 error = EINTR;
1384 return (error);
1385 }
fad57d0e
MD
1386
1387 /*
1388 * ref the vnode for ourselves so it can't be ripped out from under
1389 * is. XXX need an ND flag to request that the vnode be returned
1390 * anyway.
259b8ea0
MD
1391 *
1392 * Reserve a file descriptor but do not assign it until the open
1393 * succeeds.
fad57d0e
MD
1394 */
1395 vp = (struct vnode *)fp->f_data;
1396 vref(vp);
259b8ea0 1397 if ((error = fdalloc(p, 0, &indx)) != 0) {
9f87144f 1398 fdrop(fp);
fad57d0e
MD
1399 vrele(vp);
1400 return (error);
1401 }
1402
1403 /*
1404 * If no error occurs the vp will have been assigned to the file
1405 * pointer.
1406 */
fbfe4e7d 1407 lp->lwp_dupfd = 0;
984263bc 1408
984263bc
MD
1409 if (flags & (O_EXLOCK | O_SHLOCK)) {
1410 lf.l_whence = SEEK_SET;
1411 lf.l_start = 0;
1412 lf.l_len = 0;
1413 if (flags & O_EXLOCK)
1414 lf.l_type = F_WRLCK;
1415 else
1416 lf.l_type = F_RDLCK;
71c18fe3
MD
1417 if (flags & FNONBLOCK)
1418 type = 0;
1419 else
1420 type = F_WAIT;
fad57d0e 1421
984263bc
MD
1422 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1423 /*
259b8ea0
MD
1424 * lock request failed. Clean up the reserved
1425 * descriptor.
984263bc 1426 */
fad57d0e 1427 vrele(vp);
259b8ea0 1428 fsetfd(p, NULL, indx);
9f87144f 1429 fdrop(fp);
984263bc
MD
1430 return (error);
1431 }
984263bc
MD
1432 fp->f_flag |= FHASLOCK;
1433 }
7540ab49
MD
1434#if 0
1435 /*
1436 * Assert that all regular file vnodes were created with a object.
1437 */
1438 KASSERT(vp->v_type != VREG || vp->v_object != NULL,
1439 ("open: regular file has no backing object after vn_open"));
1440#endif
984263bc 1441
fad57d0e 1442 vrele(vp);
21739618 1443
984263bc
MD
1444 /*
1445 * release our private reference, leaving the one associated with the
1446 * descriptor table intact.
1447 */
259b8ea0 1448 fsetfd(p, fp, indx);
9f87144f 1449 fdrop(fp);
9697c509 1450 *res = indx;
984263bc
MD
1451 return (0);
1452}
1453
984263bc 1454/*
9697c509 1455 * open_args(char *path, int flags, int mode)
41c20dac 1456 *
9697c509
DRJ
1457 * Check permissions, allocate an open file structure,
1458 * and call the device open routine if any.
984263bc 1459 */
984263bc 1460int
753fd850 1461sys_open(struct open_args *uap)
984263bc 1462{
fad57d0e 1463 struct nlookupdata nd;
9697c509
DRJ
1464 int error;
1465
fad57d0e
MD
1466 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1467 if (error == 0) {
1468 error = kern_open(&nd, uap->flags,
1469 uap->mode, &uap->sysmsg_result);
1470 }
1471 nlookup_done(&nd);
9697c509 1472 return (error);
984263bc 1473}
984263bc 1474
984263bc 1475int
fad57d0e 1476kern_mknod(struct nlookupdata *nd, int mode, int dev)
984263bc 1477{
fad57d0e 1478 struct namecache *ncp;
dadab5e9
MD
1479 struct thread *td = curthread;
1480 struct proc *p = td->td_proc;
41c20dac 1481 struct vnode *vp;
984263bc
MD
1482 struct vattr vattr;
1483 int error;
1484 int whiteout = 0;
984263bc 1485
dadab5e9
MD
1486 KKASSERT(p);
1487
9697c509 1488 switch (mode & S_IFMT) {
984263bc
MD
1489 case S_IFCHR:
1490 case S_IFBLK:
dadab5e9 1491 error = suser(td);
984263bc
MD
1492 break;
1493 default:
dadab5e9 1494 error = suser_cred(p->p_ucred, PRISON_ROOT);
984263bc
MD
1495 break;
1496 }
1497 if (error)
1498 return (error);
fad57d0e 1499
984263bc 1500 bwillwrite();
fad57d0e
MD
1501 nd->nl_flags |= NLC_CREATE;
1502 if ((error = nlookup(nd)) != 0)
984263bc 1503 return (error);
fad57d0e
MD
1504 ncp = nd->nl_ncp;
1505 if (ncp->nc_vp)
1506 return (EEXIST);
984263bc 1507
fad57d0e
MD
1508 VATTR_NULL(&vattr);
1509 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1510 vattr.va_rdev = dev;
1511 whiteout = 0;
1512
1513 switch (mode & S_IFMT) {
1514 case S_IFMT: /* used by badsect to flag bad sectors */
1515 vattr.va_type = VBAD;
1516 break;
1517 case S_IFCHR:
1518 vattr.va_type = VCHR;
1519 break;
1520 case S_IFBLK:
1521 vattr.va_type = VBLK;
1522 break;
1523 case S_IFWHT:
1524 whiteout = 1;
1525 break;
1526 default:
1527 error = EINVAL;
1528 break;
984263bc 1529 }
9697c509 1530 if (error == 0) {
fad57d0e
MD
1531 if (whiteout) {
1532 error = VOP_NWHITEOUT(ncp, nd->nl_cred, NAMEI_CREATE);
1533 } else {
1534 vp = NULL;
1535 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
984263bc 1536 if (error == 0)
fad57d0e 1537 vput(vp);
984263bc 1538 }
984263bc 1539 }
9697c509
DRJ
1540 return (error);
1541}
1542
1543/*
1544 * mknod_args(char *path, int mode, int dev)
1545 *
1546 * Create a special file.
1547 */
1548int
753fd850 1549sys_mknod(struct mknod_args *uap)
9697c509 1550{
fad57d0e 1551 struct nlookupdata nd;
9697c509
DRJ
1552 int error;
1553
fad57d0e
MD
1554 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1555 if (error == 0)
1556 error = kern_mknod(&nd, uap->mode, uap->dev);
1557 nlookup_done(&nd);
984263bc
MD
1558 return (error);
1559}
1560
984263bc 1561int
fad57d0e 1562kern_mkfifo(struct nlookupdata *nd, int mode)
984263bc 1563{
fad57d0e 1564 struct namecache *ncp;
dadab5e9
MD
1565 struct thread *td = curthread;
1566 struct proc *p = td->td_proc;
984263bc 1567 struct vattr vattr;
fad57d0e 1568 struct vnode *vp;
984263bc 1569 int error;
984263bc
MD
1570
1571 bwillwrite();
fad57d0e
MD
1572
1573 nd->nl_flags |= NLC_CREATE;
1574 if ((error = nlookup(nd)) != 0)
984263bc 1575 return (error);
fad57d0e
MD
1576 ncp = nd->nl_ncp;
1577 if (ncp->nc_vp)
984263bc 1578 return (EEXIST);
fad57d0e 1579
984263bc
MD
1580 VATTR_NULL(&vattr);
1581 vattr.va_type = VFIFO;
136178b3 1582 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
fad57d0e
MD
1583 vp = NULL;
1584 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
984263bc 1585 if (error == 0)
fad57d0e 1586 vput(vp);
136178b3
DRJ
1587 return (error);
1588}
1589
1590/*
1591 * mkfifo_args(char *path, int mode)
1592 *
1593 * Create a named pipe.
1594 */
1595int
753fd850 1596sys_mkfifo(struct mkfifo_args *uap)
136178b3 1597{
fad57d0e 1598 struct nlookupdata nd;
136178b3
DRJ
1599 int error;
1600
fad57d0e
MD
1601 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1602 if (error == 0)
1603 error = kern_mkfifo(&nd, uap->mode);
1604 nlookup_done(&nd);
984263bc
MD
1605 return (error);
1606}
1607
8c40f81e
MD
1608static int hardlink_check_uid = 0;
1609SYSCTL_INT(_kern, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1610 &hardlink_check_uid, 0,
1611 "Unprivileged processes cannot create hard links to files owned by other "
1612 "users");
1613static int hardlink_check_gid = 0;
1614SYSCTL_INT(_kern, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1615 &hardlink_check_gid, 0,
1616 "Unprivileged processes cannot create hard links to files owned by other "
1617 "groups");
1618
1619static int
1620can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
1621{
1622 struct vattr va;
1623 int error;
1624
1625 /*
1626 * Shortcut if disabled
1627 */
1628 if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
1629 return (0);
1630
1631 /*
1632 * root cred can always hardlink
1633 */
1634 if (suser_cred(cred, PRISON_ROOT) == 0)
1635 return (0);
1636
1637 /*
1638 * Otherwise only if the originating file is owned by the
1639 * same user or group. Note that any group is allowed if
1640 * the file is owned by the caller.
1641 */
87de5057 1642 error = VOP_GETATTR(vp, &va);
8c40f81e
MD
1643 if (error != 0)
1644 return (error);
1645
1646 if (hardlink_check_uid) {
1647 if (cred->cr_uid != va.va_uid)
1648 return (EPERM);
1649 }
1650
1651 if (hardlink_check_gid) {
1652 if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
1653 return (EPERM);
1654 }
1655
1656 return (0);
1657}
1658
984263bc 1659int
fad57d0e 1660kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
984263bc 1661{
dadab5e9 1662 struct thread *td = curthread;
41c20dac 1663 struct vnode *vp;
984263bc
MD
1664 int error;
1665
fad57d0e
MD
1666 /*
1667 * Lookup the source and obtained a locked vnode.
1668 *
1669 * XXX relookup on vget failure / race ?
1670 */
984263bc 1671 bwillwrite();
fad57d0e 1672 if ((error = nlookup(nd)) != 0)
984263bc 1673 return (error);
fad57d0e
MD
1674 vp = nd->nl_ncp->nc_vp;
1675 KKASSERT(vp != NULL);
984263bc 1676 if (vp->v_type == VDIR)
fad57d0e 1677 return (EPERM); /* POSIX */
87de5057 1678 if ((error = vget(vp, LK_EXCLUSIVE)) != 0)
fad57d0e
MD
1679 return (error);
1680
1681 /*
1682 * Unlock the source so we can lookup the target without deadlocking
1683 * (XXX vp is locked already, possible other deadlock?). The target
1684 * must not exist.
1685 */
1686 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
1687 nd->nl_flags &= ~NLC_NCPISLOCKED;
1688 cache_unlock(nd->nl_ncp);
1689
1690 linknd->nl_flags |= NLC_CREATE;
1691 if ((error = nlookup(linknd)) != 0) {
1692 vput(vp);
1693 return (error);
984263bc 1694 }
fad57d0e
MD
1695 if (linknd->nl_ncp->nc_vp) {
1696 vput(vp);
1697 return (EEXIST);
1698 }
1699
1700 /*
1701 * Finally run the new API VOP.
1702 */
8c40f81e
MD
1703 error = can_hardlink(vp, td, td->td_proc->p_ucred);
1704 if (error == 0)
1705 error = VOP_NLINK(linknd->nl_ncp, vp, linknd->nl_cred);
fad57d0e 1706 vput(vp);
984263bc
MD
1707 return (error);
1708}
1709
1710/*
9697c509 1711 * link_args(char *path, char *link)
41c20dac 1712 *
9697c509 1713 * Make a hard file link.
984263bc 1714 */
984263bc 1715int
753fd850 1716sys_link(struct link_args *uap)
9697c509 1717{
fad57d0e 1718 struct nlookupdata nd, linknd;
9697c509
DRJ
1719 int error;
1720
fad57d0e
MD
1721 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1722 if (error == 0) {
1723 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
1724 if (error == 0)
1725 error = kern_link(&nd, &linknd);
1726 nlookup_done(&linknd);
1727 }
1728 nlookup_done(&nd);
9697c509
DRJ
1729 return (error);
1730}
1731
1732int
fad57d0e 1733kern_symlink(struct nlookupdata *nd, char *path, int mode)
984263bc 1734{
fad57d0e 1735 struct namecache *ncp;
984263bc 1736 struct vattr vattr;
fad57d0e 1737 struct vnode *vp;
984263bc 1738 int error;
984263bc 1739
984263bc 1740 bwillwrite();
fad57d0e
MD
1741 nd->nl_flags |= NLC_CREATE;
1742 if ((error = nlookup(nd)) != 0)
9697c509 1743 return (error);
fad57d0e
MD
1744 ncp = nd->nl_ncp;
1745 if (ncp->nc_vp)
9697c509 1746 return (EEXIST);
fad57d0e 1747
984263bc 1748 VATTR_NULL(&vattr);
fad57d0e
MD
1749 vattr.va_mode = mode;
1750 error = VOP_NSYMLINK(ncp, &vp, nd->nl_cred, &vattr, path);
984263bc 1751 if (error == 0)
fad57d0e 1752 vput(vp);
9697c509
DRJ
1753 return (error);
1754}
1755
1756/*
1757 * symlink(char *path, char *link)
1758 *
1759 * Make a symbolic link.
1760 */
1761int
753fd850 1762sys_symlink(struct symlink_args *uap)
9697c509
DRJ
1763{
1764 struct thread *td = curthread;
fad57d0e 1765 struct nlookupdata nd;
9697c509
DRJ
1766 char *path;
1767 int error;
fad57d0e 1768 int mode;
9697c509 1769
70aac194 1770 path = objcache_get(namei_oc, M_WAITOK);
9697c509 1771 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
e24b948e 1772 if (error == 0) {
fad57d0e
MD
1773 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
1774 if (error == 0) {
1775 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
1776 error = kern_symlink(&nd, path, mode);
1777 }
1778 nlookup_done(&nd);
e24b948e 1779 }
70aac194 1780 objcache_put(namei_oc, path);
984263bc
MD
1781 return (error);
1782}
1783
1784/*
41c20dac
MD
1785 * undelete_args(char *path)
1786 *
984263bc
MD
1787 * Delete a whiteout from the filesystem.
1788 */
1789/* ARGSUSED */
1790int
753fd850 1791sys_undelete(struct undelete_args *uap)
984263bc 1792{
fad57d0e 1793 struct nlookupdata nd;
984263bc 1794 int error;
984263bc 1795
ab2eb4eb 1796 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
984263bc 1797 bwillwrite();
fad57d0e
MD
1798 nd.nl_flags |= NLC_DELETE;
1799 if (error == 0)
1800 error = nlookup(&nd);
b8997912 1801 if (error == 0)
fad57d0e
MD
1802 error = VOP_NWHITEOUT(nd.nl_ncp, nd.nl_cred, NAMEI_DELETE);
1803 nlookup_done(&nd);
984263bc
MD
1804 return (error);
1805}
1806
984263bc 1807int
fad57d0e 1808kern_unlink(struct nlookupdata *nd)
984263bc 1809{
fad57d0e 1810 struct namecache *ncp;
984263bc 1811 int error;
984263bc
MD
1812
1813 bwillwrite();
fad57d0e
MD
1814 nd->nl_flags |= NLC_DELETE;
1815 if ((error = nlookup(nd)) != 0)
984263bc 1816 return (error);
fad57d0e
MD
1817 ncp = nd->nl_ncp;
1818 error = VOP_NREMOVE(ncp, nd->nl_cred);
984263bc
MD
1819 return (error);
1820}
1821
1822/*
9697c509 1823 * unlink_args(char *path)
41c20dac 1824 *
9697c509 1825 * Delete a name from the filesystem.
984263bc 1826 */
984263bc 1827int
753fd850 1828sys_unlink(struct unlink_args *uap)
9697c509 1829{
fad57d0e 1830 struct nlookupdata nd;
9697c509
DRJ
1831 int error;
1832
fad57d0e
MD
1833 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1834 if (error == 0)
1835 error = kern_unlink(&nd);
1836 nlookup_done(&nd);
9697c509
DRJ
1837 return (error);
1838}
1839
1840int
5a3fe67d 1841kern_lseek(int fd, off_t offset, int whence, off_t *res)
984263bc 1842{
dadab5e9
MD
1843 struct thread *td = curthread;
1844 struct proc *p = td->td_proc;
41c20dac 1845 struct file *fp;
984263bc
MD
1846 struct vattr vattr;
1847 int error;
1848
228b401d
MD
1849 fp = holdfp(p->p_fd, fd, -1);
1850 if (fp == NULL)
984263bc 1851 return (EBADF);
228b401d
MD
1852 if (fp->f_type != DTYPE_VNODE) {
1853 error = ESPIPE;
1854 goto done;
1855 }
1856
9697c509 1857 switch (whence) {
984263bc 1858 case L_INCR:
9697c509 1859 fp->f_offset += offset;
228b401d 1860 error = 0;
984263bc
MD
1861 break;
1862 case L_XTND:
228b401d
MD
1863 error = VOP_GETATTR((struct vnode *)fp->f_data, &vattr);
1864 if (error == 0)
1865 fp->f_offset = offset + vattr.va_size;
984263bc
MD
1866 break;
1867 case L_SET:
9697c509 1868 fp->f_offset = offset;
228b401d 1869 error = 0;
984263bc
MD
1870 break;
1871 default:
228b401d
MD
1872 error = EINVAL;
1873 break;
984263bc 1874 }
9697c509 1875 *res = fp->f_offset;
228b401d
MD
1876done:
1877 fdrop(fp);
1878 return (error);
984263bc
MD
1879}
1880
984263bc 1881/*
9697c509 1882 * lseek_args(int fd, int pad, off_t offset, int whence)
41c20dac 1883 *
9697c509 1884 * Reposition read/write file offset.
984263bc 1885 */
984263bc 1886int
753fd850 1887sys_lseek(struct lseek_args *uap)
984263bc 1888{
984263bc
MD
1889 int error;
1890
9697c509 1891 error = kern_lseek(uap->fd, uap->offset, uap->whence,
5a3fe67d 1892 &uap->sysmsg_offset);
9697c509 1893
984263bc
MD
1894 return (error);
1895}
984263bc 1896
984263bc 1897int
fad57d0e 1898kern_access(struct nlookupdata *nd, int aflags)
984263bc 1899{
41c20dac 1900 struct vnode *vp;
984263bc 1901 int error, flags;
984263bc 1902
fad57d0e
MD
1903 if ((error = nlookup(nd)) != 0)
1904 return (error);
806dcf9a 1905retry:
fad57d0e 1906 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 1907 if (error)
fad57d0e 1908 return (error);
984263bc
MD
1909
1910 /* Flags == 0 means only check for existence. */
9697c509 1911 if (aflags) {
984263bc 1912 flags = 0;
9697c509 1913 if (aflags & R_OK)
984263bc 1914 flags |= VREAD;
9697c509 1915 if (aflags & W_OK)
984263bc 1916 flags |= VWRITE;
9697c509 1917 if (aflags & X_OK)
984263bc
MD
1918 flags |= VEXEC;
1919 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
87de5057 1920 error = VOP_ACCESS(vp, flags, nd->nl_cred);
806dcf9a
MD
1921
1922 /*
1923 * If the file handle is stale we have to re-resolve the
1924 * entry. This is a hack at the moment.
1925 */
1926 if (error == ESTALE) {
1927 cache_setunresolved(nd->nl_ncp);
1928 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1929 if (error == 0) {
1930 vput(vp);
1931 vp = NULL;
1932 goto retry;
1933 }
1934 }
984263bc 1935 }
984263bc 1936 vput(vp);
984263bc
MD
1937 return (error);
1938}
1939
984263bc 1940/*
9697c509 1941 * access_args(char *path, int flags)
41c20dac 1942 *
9697c509 1943 * Check access permissions.
984263bc 1944 */
984263bc 1945int
753fd850 1946sys_access(struct access_args *uap)
984263bc 1947{
fad57d0e 1948 struct nlookupdata nd;
9697c509 1949 int error;
984263bc 1950
fad57d0e
MD
1951 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1952 if (error == 0)
1953 error = kern_access(&nd, uap->flags);
1954 nlookup_done(&nd);
984263bc
MD
1955 return (error);
1956}
1957
984263bc 1958int
21739618 1959kern_stat(struct nlookupdata *nd, struct stat *st)
984263bc 1960{
984263bc 1961 int error;
21739618
MD
1962 struct vnode *vp;
1963 thread_t td;
984263bc 1964
21739618 1965 if ((error = nlookup(nd)) != 0)
984263bc 1966 return (error);
806dcf9a 1967again:
21739618
MD
1968 if ((vp = nd->nl_ncp->nc_vp) == NULL)
1969 return (ENOENT);
1970
1971 td = curthread;
87de5057 1972 if ((error = vget(vp, LK_SHARED)) != 0)
21739618 1973 return (error);
87de5057 1974 error = vn_stat(vp, st, nd->nl_cred);
806dcf9a
MD
1975
1976 /*
1977 * If the file handle is stale we have to re-resolve the entry. This
1978 * is a hack at the moment.
1979 */
1980 if (error == ESTALE) {
1981 cache_setunresolved(nd->nl_ncp);
1982 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1983 if (error == 0) {
1984 vput(vp);
1985 goto again;
1986 }
1987 }
21739618 1988 vput(vp);
984263bc
MD
1989 return (error);
1990}
1991
984263bc 1992/*
41c20dac
MD
1993 * stat_args(char *path, struct stat *ub)
1994 *
984263bc
MD
1995 * Get file status; this version follows links.
1996 */
984263bc 1997int
753fd850 1998sys_stat(struct stat_args *uap)
984263bc 1999{
21739618 2000 struct nlookupdata nd;
9697c509
DRJ
2001 struct stat st;
2002 int error;
984263bc 2003
21739618
MD
2004 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2005 if (error == 0) {
2006 error = kern_stat(&nd, &st);
2007 if (error == 0)
2008 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 2009 }
fad57d0e 2010 nlookup_done(&nd);
984263bc
MD
2011 return (error);
2012}
2013
2014/*
41c20dac
MD
2015 * lstat_args(char *path, struct stat *ub)
2016 *
984263bc
MD
2017 * Get file status; this version does not follow links.
2018 */
984263bc 2019int
753fd850 2020sys_lstat(struct lstat_args *uap)
984263bc 2021{
21739618 2022 struct nlookupdata nd;
9697c509
DRJ
2023 struct stat st;
2024 int error;
984263bc 2025
21739618
MD
2026 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2027 if (error == 0) {
2028 error = kern_stat(&nd, &st);
2029 if (error == 0)
2030 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 2031 }
fad57d0e 2032 nlookup_done(&nd);
984263bc
MD
2033 return (error);
2034}
2035
984263bc 2036/*
41c20dac
MD
2037 * pathconf_Args(char *path, int name)
2038 *
984263bc
MD
2039 * Get configurable pathname variables.
2040 */
984263bc
MD
2041/* ARGSUSED */
2042int
753fd850 2043sys_pathconf(struct pathconf_args *uap)
984263bc 2044{
fad57d0e
MD
2045 struct nlookupdata nd;
2046 struct vnode *vp;
984263bc 2047 int error;
984263bc 2048
fad57d0e 2049 vp = NULL;
ab2eb4eb 2050 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2051 if (error == 0)
2052 error = nlookup(&nd);
2053 if (error == 0)
2054 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
2055 nlookup_done(&nd);
2056 if (error == 0) {
ab2eb4eb 2057 error = VOP_PATHCONF(vp, uap->name, uap->sysmsg_fds);
fad57d0e
MD
2058 vput(vp);
2059 }
984263bc
MD
2060 return (error);
2061}
2062
bbab27f6
DRJ
2063/*
2064 * XXX: daver
2065 * kern_readlink isn't properly split yet. There is a copyin burried
2066 * in VOP_READLINK().
2067 */
984263bc 2068int
fad57d0e 2069kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
984263bc 2070{
dadab5e9
MD
2071 struct thread *td = curthread;
2072 struct proc *p = td->td_proc;
41c20dac 2073 struct vnode *vp;
984263bc
MD
2074 struct iovec aiov;
2075 struct uio auio;
2076 int error;
984263bc 2077
fad57d0e
MD
2078 if ((error = nlookup(nd)) != 0)
2079 return (error);
2080 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 2081 if (error)
984263bc 2082 return (error);
fad57d0e 2083 if (vp->v_type != VLNK) {
984263bc 2084 error = EINVAL;
fad57d0e 2085 } else {
9697c509
DRJ
2086 aiov.iov_base = buf;
2087 aiov.iov_len = count;
984263bc
MD
2088 auio.uio_iov = &aiov;
2089 auio.uio_iovcnt = 1;
2090 auio.uio_offset = 0;
2091 auio.uio_rw = UIO_READ;
2092 auio.uio_segflg = UIO_USERSPACE;
dadab5e9 2093 auio.uio_td = td;
9697c509 2094 auio.uio_resid = count;
984263bc
MD
2095 error = VOP_READLINK(vp, &auio, p->p_ucred);
2096 }
2097 vput(vp);
9697c509
DRJ
2098 *res = count - auio.uio_resid;
2099 return (error);
2100}
2101
2102/*
2103 * readlink_args(char *path, char *buf, int count)
2104 *
2105 * Return target name of a symbolic link.
2106 */
2107int
753fd850 2108sys_readlink(struct readlink_args *uap)
9697c509 2109{
fad57d0e 2110 struct nlookupdata nd;
9697c509
DRJ
2111 int error;
2112
fad57d0e
MD
2113 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2114 if (error == 0) {
2115 error = kern_readlink(&nd, uap->buf, uap->count,
2116 &uap->sysmsg_result);
2117 }
2118 nlookup_done(&nd);
984263bc
MD
2119 return (error);
2120}
2121
2122static int
41c20dac 2123setfflags(struct vnode *vp, int flags)
984263bc 2124{
dadab5e9
MD
2125 struct thread *td = curthread;
2126 struct proc *p = td->td_proc;
984263bc
MD
2127 int error;
2128 struct vattr vattr;
2129
2130 /*
2131 * Prevent non-root users from setting flags on devices. When
2132 * a device is reused, users can retain ownership of the device
2133 * if they are allowed to set flags and programs assume that
2134 * chown can't fail when done as root.
2135 */
2136 if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
dadab5e9 2137 ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
984263bc
MD
2138 return (error);
2139
186fccd6
MD
2140 /*
2141 * note: vget is required for any operation that might mod the vnode
2142 * so VINACTIVE is properly cleared.
2143 */
87de5057 2144 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2145 VATTR_NULL(&vattr);
2146 vattr.va_flags = flags;
87de5057 2147 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2148 vput(vp);
2149 }
984263bc
MD
2150 return (error);
2151}
2152
2153/*
41c20dac
MD
2154 * chflags(char *path, int flags)
2155 *
984263bc
MD
2156 * Change flags of a file given a path name.
2157 */
984263bc
MD
2158/* ARGSUSED */
2159int
753fd850 2160sys_chflags(struct chflags_args *uap)
984263bc 2161{
fad57d0e
MD
2162 struct nlookupdata nd;
2163 struct vnode *vp;
984263bc 2164 int error;
984263bc 2165
fad57d0e 2166 vp = NULL;
ab2eb4eb 2167 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2168 /* XXX Add NLC flag indicating modifying operation? */
2169 if (error == 0)
2170 error = nlookup(&nd);
2171 if (error == 0)
2172 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
2173 nlookup_done(&nd);
2174 if (error == 0) {
ab2eb4eb 2175 error = setfflags(vp, uap->flags);
fad57d0e
MD
2176 vrele(vp);
2177 }
2178 return (error);
984263bc
MD
2179}
2180
2181/*
41c20dac
MD
2182 * fchflags_args(int fd, int flags)
2183 *
984263bc
MD
2184 * Change flags of a file given a file descriptor.
2185 */
984263bc
MD
2186/* ARGSUSED */
2187int
753fd850 2188sys_fchflags(struct fchflags_args *uap)
984263bc 2189{
dadab5e9
MD
2190 struct thread *td = curthread;
2191 struct proc *p = td->td_proc;
984263bc
MD
2192 struct file *fp;
2193 int error;
2194
5b287bba 2195 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2196 return (error);
5b287bba
MD
2197 error = setfflags((struct vnode *) fp->f_data, uap->flags);
2198 fdrop(fp);
2199 return (error);
984263bc
MD
2200}
2201
2202static int
41c20dac 2203setfmode(struct vnode *vp, int mode)
984263bc 2204{
dadab5e9
MD
2205 struct thread *td = curthread;
2206 struct proc *p = td->td_proc;
984263bc
MD
2207 int error;
2208 struct vattr vattr;
2209
186fccd6
MD
2210 /*
2211 * note: vget is required for any operation that might mod the vnode
2212 * so VINACTIVE is properly cleared.
2213 */
87de5057 2214 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2215 VATTR_NULL(&vattr);
2216 vattr.va_mode = mode & ALLPERMS;
87de5057 2217 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2218 vput(vp);
2219 }
984263bc
MD
2220 return error;
2221}
2222
9697c509 2223int
fad57d0e 2224kern_chmod(struct nlookupdata *nd, int mode)
9697c509 2225{
fad57d0e 2226 struct vnode *vp;
9697c509
DRJ
2227 int error;
2228
fad57d0e
MD
2229 /* XXX Add NLC flag indicating modifying operation? */
2230 if ((error = nlookup(nd)) != 0)
9697c509 2231 return (error);
fad57d0e
MD
2232 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2233 return (error);
2234 error = setfmode(vp, mode);
2235 vrele(vp);
2236 return (error);
9697c509
DRJ
2237}
2238
984263bc 2239/*
41c20dac
MD
2240 * chmod_args(char *path, int mode)
2241 *
984263bc
MD
2242 * Change mode of a file given path name.
2243 */
984263bc
MD
2244/* ARGSUSED */
2245int
753fd850 2246sys_chmod(struct chmod_args *uap)
984263bc 2247{
fad57d0e 2248 struct nlookupdata nd;
9697c509 2249 int error;
984263bc 2250
fad57d0e
MD
2251 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2252 if (error == 0)
2253 error = kern_chmod(&nd, uap->mode);
2254 nlookup_done(&nd);
9697c509 2255 return (error);
984263bc
MD
2256}
2257
2258/*
41c20dac
MD
2259 * lchmod_args(char *path, int mode)
2260 *
984263bc
MD
2261 * Change mode of a file given path name (don't follow links.)
2262 */
984263bc
MD
2263/* ARGSUSED */
2264int
753fd850 2265sys_lchmod(struct lchmod_args *uap)
984263bc 2266{
fad57d0e 2267 struct nlookupdata nd;
984263bc 2268 int error;
984263bc 2269
fad57d0e
MD
2270 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2271 if (error == 0)
2272 error = kern_chmod(&nd, uap->mode);
2273 nlookup_done(&nd);
2274 return (error);
984263bc
MD
2275}
2276
2277/*
41c20dac
MD
2278 * fchmod_args(int fd, int mode)
2279 *
984263bc
MD
2280 * Change mode of a file given a file descriptor.
2281 */
984263bc
MD
2282/* ARGSUSED */
2283int
753fd850 2284sys_fchmod(struct fchmod_args *uap)
984263bc 2285{
dadab5e9
MD
2286 struct thread *td = curthread;
2287 struct proc *p = td->td_proc;
984263bc
MD
2288 struct file *fp;
2289 int error;
2290
5b287bba 2291 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2292 return (error);
5b287bba
MD
2293 error = setfmode((struct vnode *)fp->f_data, uap->mode);
2294 fdrop(fp);
2295 return (error);
984263bc
MD
2296}
2297
2298static int
41c20dac 2299setfown(struct vnode *vp, uid_t uid, gid_t gid)
984263bc 2300{
dadab5e9
MD
2301 struct thread *td = curthread;
2302 struct proc *p = td->td_proc;
984263bc
MD
2303 int error;
2304 struct vattr vattr;
2305
186fccd6
MD
2306 /*
2307 * note: vget is required for any operation that might mod the vnode
2308 * so VINACTIVE is properly cleared.
2309 */
87de5057 2310 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2311 VATTR_NULL(&vattr);
2312 vattr.va_uid = uid;
2313 vattr.va_gid = gid;
87de5057 2314 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2315 vput(vp);
2316 }
984263bc
MD
2317 return error;
2318}
2319
9697c509 2320int
fad57d0e 2321kern_chown(struct nlookupdata *nd, int uid, int gid)
9697c509 2322{
fad57d0e 2323 struct vnode *vp;
9697c509
DRJ
2324 int error;
2325
fad57d0e
MD
2326 /* XXX Add NLC flag indicating modifying operation? */
2327 if ((error = nlookup(nd)) != 0)
9697c509 2328 return (error);
fad57d0e
MD
2329 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2330 return (error);
2331 error = setfown(vp, uid, gid);
2332 vrele(vp);
9697c509
DRJ
2333 return (error);
2334}
2335
984263bc 2336/*
41c20dac
MD
2337 * chown(char *path, int uid, int gid)
2338 *
984263bc
MD
2339 * Set ownership given a path name.
2340 */
984263bc 2341int
753fd850 2342sys_chown(struct chown_args *uap)
984263bc 2343{
fad57d0e 2344 struct nlookupdata nd;
9697c509
DRJ
2345 int error;
2346
fad57d0e
MD
2347 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2348 if (error == 0)
2349 error = kern_chown(&nd, uap->uid, uap->gid);
2350 nlookup_done(&nd);
984263bc
MD
2351 return (error);
2352}
2353
2354/*
41c20dac
MD
2355 * lchown_args(char *path, int uid, int gid)
2356 *
984263bc
MD
2357 * Set ownership given a path name, do not cross symlinks.
2358 */
984263bc 2359int
753fd850 2360sys_lchown(struct lchown_args *uap)
984263bc 2361{
fad57d0e 2362 struct nlookupdata nd;
984263bc 2363 int error;
9697c509 2364
fad57d0e
MD
2365 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2366 if (error == 0)
2367 error = kern_chown(&nd, uap->uid, uap->gid);
2368 nlookup_done(&nd);
984263bc
MD
2369 return (error);
2370}
2371
2372/*
41c20dac
MD
2373 * fchown_args(int fd, int uid, int gid)
2374 *
984263bc
MD
2375 * Set ownership given a file descriptor.
2376 */
984263bc
MD
2377/* ARGSUSED */
2378int
753fd850 2379sys_fchown(struct fchown_args *uap)
984263bc 2380{
dadab5e9
MD
2381 struct thread *td = curthread;
2382 struct proc *p = td->td_proc;
984263bc
MD
2383 struct file *fp;
2384 int error;
2385
5b287bba 2386 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2387 return (error);
5b287bba
MD
2388 error = setfown((struct vnode *)fp->f_data, uap->uid, uap->gid);
2389 fdrop(fp);
2390 return (error);
984263bc
MD
2391}
2392
2393static int
9697c509 2394getutimes(const struct timeval *tvp, struct timespec *tsp)
984263bc
MD
2395{
2396 struct timeval tv[2];
984263bc 2397
9697c509 2398 if (tvp == NULL) {
984263bc
MD
2399 microtime(&tv[0]);
2400 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2401 tsp[1] = tsp[0];
2402 } else {
9697c509
DRJ
2403 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2404 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
984263bc
MD
2405 }
2406 return 0;
2407}
2408
2409static int
41c20dac 2410setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
984263bc 2411{
dadab5e9
MD
2412 struct thread *td = curthread;
2413 struct proc *p = td->td_proc;
984263bc
MD
2414 int error;
2415 struct vattr vattr;
2416
186fccd6
MD
2417 /*
2418 * note: vget is required for any operation that might mod the vnode
2419 * so VINACTIVE is properly cleared.
2420 */
87de5057 2421 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2422 VATTR_NULL(&vattr);
2423 vattr.va_atime = ts[0];
2424 vattr.va_mtime = ts[1];
2425 if (nullflag)
2426 vattr.va_vaflags |= VA_UTIMES_NULL;
87de5057 2427 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2428 vput(vp);
2429 }
984263bc
MD
2430 return error;
2431}
2432
9697c509 2433int
fad57d0e 2434kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
9697c509
DRJ
2435{
2436 struct timespec ts[2];
fad57d0e 2437 struct vnode *vp;
9697c509
DRJ
2438 int error;
2439
fad57d0e 2440 if ((error = getutimes(tptr, ts)) != 0)
9697c509 2441 return (error);
fad57d0e
MD
2442 /* XXX Add NLC flag indicating modifying operation? */
2443 if ((error = nlookup(nd)) != 0)
9697c509 2444 return (error);
fad57d0e
MD
2445 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2446 return (error);
2447 error = setutimes(vp, ts, tptr == NULL);
2448 vrele(vp);
9697c509
DRJ
2449 return (error);
2450}
2451
984263bc 2452/*
41c20dac
MD
2453 * utimes_args(char *path, struct timeval *tptr)
2454 *
984263bc
MD
2455 * Set the access and modification times of a file.
2456 */
984263bc 2457int
753fd850 2458sys_utimes(struct utimes_args *uap)
984263bc 2459{
9697c509 2460 struct timeval tv[2];
fad57d0e 2461 struct nlookupdata nd;
9697c509
DRJ
2462 int error;
2463
2464 if (uap->tptr) {
2465 error = copyin(uap->tptr, tv, sizeof(tv));
2466 if (error)
2467 return (error);
2468 }
fad57d0e
MD
2469 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2470 if (error == 0)
2471 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2472 nlookup_done(&nd);
984263bc
MD
2473 return (error);
2474}
2475
2476/*
41c20dac
MD
2477 * lutimes_args(char *path, struct timeval *tptr)
2478 *
984263bc
MD
2479 * Set the access and modification times of a file.
2480 */
984263bc 2481int
753fd850 2482sys_lutimes(struct lutimes_args *uap)
984263bc 2483{
9697c509 2484 struct timeval tv[2];
fad57d0e 2485 struct nlookupdata nd;
9697c509
DRJ
2486 int error;
2487
2488 if (uap->tptr) {
2489 error = copyin(uap->tptr, tv, sizeof(tv));
2490 if (error)
2491 return (error);
2492 }
fad57d0e
MD
2493 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2494 if (error == 0)
2495 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2496 nlookup_done(&nd);
984263bc
MD
2497 return (error);
2498}
2499
984263bc 2500int
9697c509 2501kern_futimes(int fd, struct timeval *tptr)
984263bc 2502{
dadab5e9
MD
2503 struct thread *td = curthread;
2504 struct proc *p = td->td_proc;
984263bc
MD
2505 struct timespec ts[2];
2506 struct file *fp;
984263bc
MD
2507 int error;
2508
9697c509
DRJ
2509 error = getutimes(tptr, ts);
2510 if (error)
984263bc 2511 return (error);
5b287bba 2512 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc 2513 return (error);
9697c509 2514 error = setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
5b287bba 2515 fdrop(fp);
9697c509 2516 return (error);
984263bc
MD
2517}
2518
2519/*
9697c509 2520 * futimes_args(int fd, struct timeval *tptr)
41c20dac 2521 *
9697c509 2522 * Set the access and modification times of a file.
984263bc 2523 */
984263bc 2524int
753fd850 2525sys_futimes(struct futimes_args *uap)
9697c509
DRJ
2526{
2527 struct timeval tv[2];
2528 int error;
2529
2530 if (uap->tptr) {
2531 error = copyin(uap->tptr, tv, sizeof(tv));
2532 if (error)
2533 return (error);
2534 }
2535
b651c485 2536 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
9697c509
DRJ
2537
2538 return (error);
2539}
2540
2541int
fad57d0e 2542kern_truncate(struct nlookupdata *nd, off_t length)
984263bc 2543{
41c20dac 2544 struct vnode *vp;
984263bc
MD
2545 struct vattr vattr;
2546 int error;
984263bc 2547
9697c509 2548 if (length < 0)
984263bc 2549 return(EINVAL);
fad57d0e
MD
2550 /* XXX Add NLC flag indicating modifying operation? */
2551 if ((error = nlookup(nd)) != 0)
984263bc 2552 return (error);
fad57d0e
MD
2553 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2554 return (error);
ca466bae 2555 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) != 0) {
fad57d0e
MD
2556 vrele(vp);
2557 return (error);
2558 }
2559 if (vp->v_type == VDIR) {
984263bc 2560 error = EISDIR;
fad57d0e 2561 } else if ((error = vn_writechk(vp)) == 0 &&
87de5057 2562 (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred)) == 0) {
984263bc 2563 VATTR_NULL(&vattr);
9697c509 2564 vattr.va_size = length;
87de5057 2565 error = VOP_SETATTR(vp, &vattr, nd->nl_cred);
984263bc
MD
2566 }
2567 vput(vp);
2568 return (error);
2569}
2570
9697c509
DRJ
2571/*
2572 * truncate(char *path, int pad, off_t length)
2573 *
2574 * Truncate a file given its path name.
2575 */
2576int
753fd850 2577sys_truncate(struct truncate_args *uap)
9697c509 2578{
fad57d0e 2579 struct nlookupdata nd;
9697c509
DRJ
2580 int error;
2581
fad57d0e
MD
2582 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2583 if (error == 0)
2584 error = kern_truncate(&nd, uap->length);
2585 nlookup_done(&nd);
9697c509
DRJ
2586 return error;
2587}
2588
984263bc 2589int
8f6f8622 2590kern_ftruncate(int fd, off_t length)
984263bc 2591{
dadab5e9
MD
2592 struct thread *td = curthread;
2593 struct proc *p = td->td_proc;
984263bc
MD
2594 struct vattr vattr;
2595 struct vnode *vp;
2596 struct file *fp;
2597 int error;
2598
8f6f8622 2599 if (length < 0)
984263bc 2600 return(EINVAL);
5b287bba 2601 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc 2602 return (error);
5b287bba
MD
2603 if ((fp->f_flag & FWRITE) == 0) {
2604 error = EINVAL;
2605 goto done;
2606 }
984263bc 2607 vp = (struct vnode *)fp->f_data;
ca466bae 2608 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
5b287bba 2609 if (vp->v_type == VDIR) {
984263bc 2610 error = EISDIR;
5b287bba 2611 } else if ((error = vn_writechk(vp)) == 0) {
984263bc 2612 VATTR_NULL(&vattr);
8f6f8622 2613 vattr.va_size = length;
87de5057 2614 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
984263bc 2615 }
ca466bae 2616 VOP_UNLOCK(vp, 0);
5b287bba
MD
2617done:
2618 fdrop(fp);
984263bc
MD
2619 return (error);
2620}
2621
984263bc 2622/*
8f6f8622 2623 * ftruncate_args(int fd, int pad, off_t length)
41c20dac 2624 *
8f6f8622 2625 * Truncate a file given a file descriptor.
984263bc 2626 */
984263bc 2627int
753fd850 2628sys_ftruncate(struct ftruncate_args *uap)
984263bc 2629{
8f6f8622 2630 int error;
984263bc 2631
8f6f8622
DRJ
2632 error = kern_ftruncate(uap->fd, uap->length);
2633
2634 return (error);
984263bc
MD
2635}
2636
984263bc 2637/*
41c20dac
MD
2638 * fsync(int fd)
2639 *
984263bc
MD
2640 * Sync an open file.
2641 */
984263bc
MD
2642/* ARGSUSED */
2643int
753fd850 2644sys_fsync(struct fsync_args *uap)
984263bc 2645{
dadab5e9
MD
2646 struct thread *td = curthread;
2647 struct proc *p = td->td_proc;
41c20dac 2648 struct vnode *vp;
984263bc
MD
2649 struct file *fp;
2650 vm_object_t obj;
2651 int error;
2652
5b287bba 2653 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc
MD
2654 return (error);
2655 vp = (struct vnode *)fp->f_data;
ca466bae 2656 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
7540ab49 2657 if ((obj = vp->v_object) != NULL)
984263bc 2658 vm_object_page_clean(obj, 0, 0, 0);
87de5057 2659 if ((error = VOP_FSYNC(vp, MNT_WAIT)) == 0 &&
984263bc 2660 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) &&
5b287bba 2661 bioops.io_fsync) {
984263bc 2662 error = (*bioops.io_fsync)(vp);
5b287bba 2663 }
ca466bae 2664 VOP_UNLOCK(vp, 0);
5b287bba 2665 fdrop(fp);
984263bc
MD
2666 return (error);
2667}
2668
984263bc 2669int
fad57d0e 2670kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
984263bc 2671{
fad57d0e
MD
2672 struct namecache *fncpd;
2673 struct namecache *tncpd;
2674 struct namecache *ncp;
2675 struct mount *mp;
984263bc
MD
2676 int error;
2677
2678 bwillwrite();
fad57d0e 2679 if ((error = nlookup(fromnd)) != 0)
984263bc 2680 return (error);
fad57d0e
MD
2681 if ((fncpd = fromnd->nl_ncp->nc_parent) == NULL)
2682 return (ENOENT);
2683 cache_hold(fncpd);
2684
2685 /*
2686 * unlock the source ncp so we can lookup the target ncp without
2687 * deadlocking. The target may or may not exist so we do not check
2688 * for a target vp like kern_mkdir() and other creation functions do.
2689 *
2690 * The source and target directories are ref'd and rechecked after
2691 * everything is relocked to determine if the source or target file
2692 * has been renamed.
2693 */
2694 KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
2695 fromnd->nl_flags &= ~NLC_NCPISLOCKED;
2696 cache_unlock(fromnd->nl_ncp);
2697
2698 tond->nl_flags |= NLC_CREATE;
2699 if ((error = nlookup(tond)) != 0) {
2700 cache_drop(fncpd);
2701 return (error);
2702 }
2703 if ((tncpd = tond->nl_ncp->nc_parent) == NULL) {
2704 cache_drop(fncpd);
2705 return (ENOENT);
2706 }
2707 cache_hold(tncpd);
2708
2709 /*
2710 * If the source and target are the same there is nothing to do
2711 */
2712 if (fromnd->nl_ncp == tond->nl_ncp) {
2713 cache_drop(fncpd);
2714 cache_drop(tncpd);
2715 return (0);
2716 }
2717
2718 /*
aa40af33
MD
2719 * relock the source ncp. NOTE AFTER RELOCKING: the source ncp
2720 * may have become invalid while it was unlocked, nc_vp and nc_mount
2721 * could be NULL.
fad57d0e
MD
2722 */
2723 if (cache_lock_nonblock(fromnd->nl_ncp) == 0) {
2724 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2725 } else if (fromnd->nl_ncp > tond->nl_ncp) {
2726 cache_lock(fromnd->nl_ncp);
2727 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2728 } else {
2729 cache_unlock(tond->nl_ncp);
2730 cache_lock(fromnd->nl_ncp);
2731 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2732 cache_lock(tond->nl_ncp);
2733 cache_resolve(tond->nl_ncp, tond->nl_cred);
2734 }
2735 fromnd->nl_flags |= NLC_NCPISLOCKED;
2736
2737 /*
2738 * make sure the parent directories linkages are the same
2739 */
2740 if (fncpd != fromnd->nl_ncp->nc_parent ||
2741 tncpd != tond->nl_ncp->nc_parent) {
2742 cache_drop(fncpd);
2743 cache_drop(tncpd);
2744 return (ENOENT);
2745 }
2746
2747 /*
2748 * Both the source and target must be within the same filesystem and
2749 * in the same filesystem as their parent directories within the
2750 * namecache topology.
aa40af33
MD
2751 *
2752 * NOTE: fromnd's nc_mount or nc_vp could be NULL.
fad57d0e
MD
2753 */
2754 mp = fncpd->nc_mount;
2755 if (mp != tncpd->nc_mount || mp != fromnd->nl_ncp->nc_mount ||
2756 mp != tond->nl_ncp->nc_mount) {
2757 cache_drop(fncpd);
2758 cache_drop(tncpd);
2759 return (EXDEV);
2760 }
2761
2762 /*
2763 * If the target exists and either the source or target is a directory,
2764 * then both must be directories.
aa40af33
MD
2765 *
2766 * Due to relocking of the source, fromnd->nl_ncp->nc_vp might have
2767 * become NULL.
fad57d0e
MD
2768 */
2769 if (tond->nl_ncp->nc_vp) {
aa40af33
MD
2770 if (fromnd->nl_ncp->nc_vp == NULL) {
2771 error = ENOENT;
2772 } else if (fromnd->nl_ncp->nc_vp->v_type == VDIR) {
fad57d0e
MD
2773 if (tond->nl_ncp->nc_vp->v_type != VDIR)
2774 error = ENOTDIR;
2775 } else if (tond->nl_ncp->nc_vp->v_type == VDIR) {
984263bc 2776 error = EISDIR;
984263bc
MD
2777 }
2778 }
fad57d0e 2779
984263bc 2780 /*
fad57d0e
MD
2781 * You cannot rename a source into itself or a subdirectory of itself.
2782 * We check this by travsersing the target directory upwards looking
2783 * for a match against the source.
984263bc 2784 */
fad57d0e
MD
2785 if (error == 0) {
2786 for (ncp = tncpd; ncp; ncp = ncp->nc_parent) {
2787 if (fromnd->nl_ncp == ncp) {
2788 error = EINVAL;
2789 break;
2790 }
984263bc 2791 }
fad57d0e
MD
2792 }
2793
2794 cache_drop(fncpd);
2795 cache_drop(tncpd);
bc8fd9e0
MD
2796
2797 /*
2798 * Even though the namespaces are different, they may still represent
2799 * hardlinks to the same file. The filesystem might have a hard time
2800 * with this so we issue a NREMOVE of the source instead of a NRENAME
2801 * when we detect the situation.
2802 */
2803 if (error == 0) {
2804 if (fromnd->nl_ncp->nc_vp == tond->nl_ncp->nc_vp) {
2805 error = VOP_NREMOVE(fromnd->nl_ncp, fromnd->nl_cred);
2806 } else {
2807 error = VOP_NRENAME(fromnd->nl_ncp, tond->nl_ncp,
2808 tond->nl_cred);
2809 }
2810 }
984263bc
MD
2811 return (error);
2812}
2813
2814/*
9697c509 2815 * rename_args(char *from, char *to)
41c20dac 2816 *
9697c509
DRJ
2817 * Rename files. Source and destination must either both be directories,
2818 * or both not be directories. If target is a directory, it must be empty.
984263bc 2819 */
984263bc 2820int
753fd850 2821sys_rename(struct rename_args *uap)
9697c509 2822{
fad57d0e 2823 struct nlookupdata fromnd, tond;
9697c509
DRJ
2824 int error;
2825
fad57d0e
MD
2826 error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
2827 if (error == 0) {
2828 error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
2829 if (error == 0)
2830 error = kern_rename(&fromnd, &tond);
2831 nlookup_done(&tond);
2832 }
2833 nlookup_done(&fromnd);
9697c509
DRJ
2834 return (error);
2835}
2836
2837int
fad57d0e 2838kern_mkdir(struct nlookupdata *nd, int mode)
984263bc 2839{
dadab5e9
MD
2840 struct thread *td = curthread;
2841 struct proc *p = td->td_proc;
fad57d0e 2842 struct namecache *ncp;
41c20dac 2843 struct vnode *vp;
984263bc
MD
2844 struct vattr vattr;
2845 int error;
984263bc
MD
2846
2847 bwillwrite();
fad57d0e
MD
2848 nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE;
2849 if ((error = nlookup(nd)) != 0)
984263bc 2850 return (error);
fad57d0e
MD
2851
2852 ncp = nd->nl_ncp;
2853 if (ncp->nc_vp)
984263bc 2854 return (EEXIST);
fad57d0e 2855
984263bc
MD
2856 VATTR_NULL(&vattr);
2857 vattr.va_type = VDIR;
9697c509 2858 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
fad57d0e
MD
2859
2860 vp = NULL;
2861 error = VOP_NMKDIR(ncp, &vp, p->p_ucred, &vattr);
9697c509 2862 if (error == 0)
fad57d0e 2863 vput(vp);
984263bc
MD
2864 return (error);
2865}
2866
2867/*
9697c509 2868 * mkdir_args(char *path, int mode)
41c20dac 2869 *
9697c509 2870 * Make a directory file.
984263bc 2871 */
984263bc
MD
2872/* ARGSUSED */
2873int
753fd850 2874sys_mkdir(struct mkdir_args *uap)
9697c509 2875{
fad57d0e 2876 struct nlookupdata nd;
9697c509
DRJ
2877 int error;
2878
fad57d0e
MD
2879 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2880 if (error == 0)
2881 error = kern_mkdir(&nd, uap->mode);
2882 nlookup_done(&nd);
9697c509
DRJ
2883 return (error);
2884}
2885
2886int
fad57d0e 2887kern_rmdir(struct nlookupdata *nd)
984263bc 2888{
fad57d0e 2889 struct namecache *ncp;
984263bc 2890 int error;
984263bc
MD
2891
2892 bwillwrite();
fad57d0e
MD
2893 nd->nl_flags |= NLC_DELETE;
2894 if ((error = nlookup(nd)) != 0)
984263bc 2895 return (error);
fad57d0e
MD
2896
2897 ncp = nd->nl_ncp;
2898 error = VOP_NRMDIR(ncp, nd->nl_cred);
984263bc
MD
2899 return (error);
2900}
2901
984263bc 2902/*
9697c509 2903 * rmdir_args(char *path)
41c20dac 2904 *
9697c509 2905 * Remove a directory file.
984263bc 2906 */
9697c509 2907/* ARGSUSED */
984263bc 2908int
753fd850 2909sys_rmdir(struct rmdir_args *uap)
984263bc 2910{
fad57d0e 2911 struct nlookupdata nd;
9697c509
DRJ
2912 int error;
2913
fad57d0e
MD
2914 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2915 if (error == 0)
2916 error = kern_rmdir(&nd);
2917 nlookup_done(&nd);
984263bc
MD
2918 return (error);
2919}
984263bc 2920
984263bc 2921int
53b02c48
JS
2922kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res,
2923 enum uio_seg direction)
984263bc 2924{
dadab5e9
MD
2925 struct thread *td = curthread;
2926 struct proc *p = td->td_proc;
984263bc
MD
2927 struct vnode *vp;
2928 struct file *fp;
2929 struct uio auio;
2930 struct iovec aiov;
2931 long loff;
2932 int error, eofflag;
2933
5b287bba 2934 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc 2935 return (error);
5b287bba
MD
2936 if ((fp->f_flag & FREAD) == 0) {
2937 error = EBADF;
2938 goto done;
2939 }
984263bc
MD
2940 vp = (struct vnode *)fp->f_data;
2941unionread:
5b287bba
MD
2942 if (vp->v_type != VDIR) {
2943 error = EINVAL;
2944 goto done;
2945 }
9697c509
DRJ
2946 aiov.iov_base = buf;
2947 aiov.iov_len = count;
984263bc
MD
2948 auio.uio_iov = &aiov;
2949 auio.uio_iovcnt = 1;
2950 auio.uio_rw = UIO_READ;
53b02c48 2951 auio.uio_segflg = direction;
dadab5e9 2952 auio.uio_td = td;
9697c509 2953 auio.uio_resid = count;
ca466bae
MD
2954 /* vn_lock(vp, LK_SHARED | LK_RETRY); */
2955 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc
MD
2956 loff = auio.uio_offset = fp->f_offset;
2957 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
2958 fp->f_offset = auio.uio_offset;
ca466bae 2959 VOP_UNLOCK(vp, 0);
984263bc 2960 if (error)
5b287bba 2961 goto done;
9697c509 2962 if (count == auio.uio_resid) {
984263bc 2963 if (union_dircheckp) {
dadab5e9 2964 error = union_dircheckp(td, &vp, fp);
984263bc
MD
2965 if (error == -1)
2966 goto unionread;
2967 if (error)
5b287bba 2968 goto done;
984263bc
MD
2969 }
2970 if ((vp->v_flag & VROOT) &&
2971 (vp->v_mount->mnt_flag & MNT_UNION)) {
2972 struct vnode *tvp = vp;
2973 vp = vp->v_mount->mnt_vnodecovered;
597aea93 2974 vref(vp);
fbb4eeab 2975 fp->f_data = vp;
984263bc
MD
2976 fp->f_offset = 0;
2977 vrele(tvp);
2978 goto unionread;
2979 }
2980 }
9697c509
DRJ
2981 if (basep) {
2982 *basep = loff;
984263bc 2983 }
9697c509 2984 *res = count - auio.uio_resid;
5b287bba
MD
2985done:
2986 fdrop(fp);
9697c509
DRJ
2987 return (error);
2988}
2989
2990/*
2991 * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
2992 *
2993 * Read a block of directory entries in a file system independent format.
2994 */
2995int
753fd850 2996sys_getdirentries(struct getdirentries_args *uap)
9697c509
DRJ
2997{
2998 long base;
2999 int error;
3000
3001 error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base,
53b02c48 3002 &uap->sysmsg_result, UIO_USERSPACE);
9697c509
DRJ
3003
3004 if (error == 0)
3005 error = copyout(&base, uap->basep, sizeof(*uap->basep));
984263bc
MD
3006 return (error);
3007}
41c20dac
MD
3008
3009/*
3010 * getdents_args(int fd, char *buf, size_t count)
3011 */
984263bc 3012int
753fd850 3013sys_getdents(struct getdents_args *uap)
984263bc 3014{
9697c509
DRJ
3015 int error;
3016
3017 error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL,
53b02c48 3018 &uap->sysmsg_result, UIO_USERSPACE);
41c20dac 3019
9697c509 3020 return (error);
984263bc
MD
3021}
3022
3023/*
41c20dac
MD
3024 * umask(int newmask)
3025 *
984263bc
MD
3026 * Set the mode mask for creation of filesystem nodes.
3027 *
3028 * MP SAFE
3029 */
984263bc 3030int
753fd850 3031sys_umask(struct umask_args *uap)
984263bc 3032{
dadab5e9
MD
3033 struct thread *td = curthread;
3034 struct proc *p = td->td_proc;
41c20dac 3035 struct filedesc *fdp;
984263bc
MD
3036
3037 fdp = p->p_fd;
c7114eea 3038 uap->sysmsg_result = fdp->fd_cmask;
ab2eb4eb 3039 fdp->fd_cmask = uap->newmask & ALLPERMS;
984263bc
MD
3040 return (0);
3041}
3042
3043/*
41c20dac
MD
3044 * revoke(char *path)
3045 *
984263bc
MD
3046 * Void all references to file by ripping underlying filesystem
3047 * away from vnode.
3048 */
984263bc
MD
3049/* ARGSUSED */
3050int
753fd850 3051sys_revoke(struct revoke_args *uap)
984263bc 3052{
fad57d0e 3053 struct nlookupdata nd;
984263bc 3054 struct vattr vattr;
fad57d0e
MD
3055 struct vnode *vp;
3056 struct ucred *cred;
984263bc 3057 int error;
984263bc 3058
fad57d0e 3059 vp = NULL;
ab2eb4eb 3060 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
3061 if (error == 0)
3062 error = nlookup(&nd);
3063 if (error == 0)
3064 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
3065 cred = crhold(nd.nl_cred);
3066 nlookup_done(&nd);
3067 if (error == 0) {
3068 if (vp->v_type != VCHR && vp->v_type != VBLK)
3069 error = EINVAL;
3070 if (error == 0)
87de5057 3071 error = VOP_GETATTR(vp, &vattr);
fad57d0e
MD
3072 if (error == 0 && cred->cr_uid != vattr.va_uid)
3073 error = suser_cred(cred, PRISON_ROOT);
3074 if (error == 0 && count_udev(vp->v_udev) > 0) {
3075 if ((error = vx_lock(vp)) == 0) {
3076 VOP_REVOKE(vp, REVOKEALL);
3077 vx_unlock(vp);
3078 }
5fd012e0 3079 }
fad57d0e 3080 vrele(vp);
5fd012e0 3081 }
ac3d60cb
MD
3082 if (cred)
3083 crfree(cred);
984263bc
MD
3084 return (error);
3085}
3086
984263bc 3087/*
41c20dac
MD
3088 * getfh_args(char *fname, fhandle_t *fhp)
3089 *
984263bc
MD
3090 * Get (NFS) file handle
3091 */
984263bc 3092int
753fd850 3093sys_getfh(struct getfh_args *uap)
984263bc 3094{
dadab5e9 3095 struct thread *td = curthread;
fad57d0e 3096 struct nlookupdata nd;
984263bc 3097 fhandle_t fh;
41c20dac 3098 struct vnode *vp;
984263bc
MD
3099 int error;
3100
3101 /*
3102 * Must be super user
3103 */
fad57d0e 3104 if ((error = suser(td)) != 0)
984263bc 3105 return (error);
fad57d0e
MD
3106
3107 vp = NULL;
3108 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
3109 if (error == 0)
3110 error = nlookup(&nd);
3111 if (error == 0)
3112 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3113 nlookup_done(&nd);
3114 if (error == 0) {
3115 bzero(&fh, sizeof(fh));
3116 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3117 error = VFS_VPTOFH(vp, &fh.fh_fid);
3118 vput(vp);
3119 if (error == 0)
3120 error = copyout(&fh, uap->fhp, sizeof(fh));
3121 }
984263bc
MD
3122 return (error);
3123}
3124
3125/*
41c20dac
MD
3126 * fhopen_args(const struct fhandle *u_fhp, int flags)
3127 *
984263bc
MD
3128 * syscall for the rpc.lockd to use to translate a NFS file handle into
3129 * an open descriptor.
3130 *
3131 * warning: do not remove the suser() call or this becomes one giant
3132 * security hole.
3133 */
984263bc 3134int
753fd850 3135sys_fhopen(struct fhopen_args *uap)
984263bc 3136{
dadab5e9
MD
3137 struct thread *td = curthread;
3138 struct proc *p = td->td_proc;
984263bc
MD
3139 struct mount *mp;
3140 struct vnode *vp;
3141 struct fhandle fhp;
3142 struct vattr vat;
3143 struct vattr *vap = &vat;
3144 struct flock lf;
984263bc
MD
3145 int fmode, mode, error, type;
3146 struct file *nfp;
fad57d0e 3147 struct file *fp;
984263bc
MD
3148 int indx;
3149
3150 /*
3151 * Must be super user
3152 */
dadab5e9 3153 error = suser(td);
984263bc
MD
3154 if (error)
3155 return (error);
3156
ab2eb4eb 3157 fmode = FFLAGS(uap->flags);
984263bc
MD
3158 /* why not allow a non-read/write open for our lockd? */
3159 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3160 return (EINVAL);
ab2eb4eb 3161 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
984263bc
MD
3162 if (error)
3163 return(error);
3164 /* find the mount point */
3165 mp = vfs_getvfs(&fhp.fh_fsid);
3166 if (mp == NULL)
3167 return (ESTALE);
3168 /* now give me my vnode, it gets returned to me locked */
3169 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
3170 if (error)
3171 return (error);
3172 /*
3173 * from now on we have to make sure not
3174 * to forget about the vnode
3175 * any error that causes an abort must vput(vp)
3176 * just set error = err and 'goto bad;'.
3177 */
3178
3179 /*
3180 * from vn_open
3181 */
3182 if (vp->v_type == VLNK) {
3183 error = EMLINK;
3184 goto bad;
3185 }
3186 if (vp->v_type == VSOCK) {
3187 error = EOPNOTSUPP;
3188 goto bad;
3189 }
3190 mode = 0;
3191 if (fmode & (FWRITE | O_TRUNC)) {
3192 if (vp->v_type == VDIR) {
3193 error = EISDIR;
3194 goto bad;
3195 }
3196 error = vn_writechk(vp);
3197 if (error)
3198 goto bad;
3199 mode |= VWRITE;
3200 }
3201 if (fmode & FREAD)
3202 mode |= VREAD;
3203 if (mode) {
87de5057 3204 error = VOP_ACCESS(vp, mode, p->p_ucred);
984263bc
MD
3205 if (error)
3206 goto bad;
3207 }
3208 if (fmode & O_TRUNC) {
ca466bae
MD
3209 VOP_UNLOCK(vp, 0); /* XXX */
3210 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
984263bc
MD
3211 VATTR_NULL(vap);
3212 vap->va_size = 0;
87de5057 3213 error = VOP_SETATTR(vp, vap, p->p_ucred);
984263bc
MD
3214 if (error)
3215 goto bad;
3216 }
fad57d0e 3217
984263bc 3218 /*
fad57d0e
MD
3219 * VOP_OPEN needs the file pointer so it can potentially override
3220 * it.
3221 *
3222 * WARNING! no f_ncp will be associated when fhopen()ing a directory.
3223 * XXX
984263bc 3224 */
259b8ea0 3225 if ((error = falloc(p, &nfp, &indx)) != 0)
fad57d0e
MD
3226 goto bad;
3227 fp = nfp;
3228
87de5057 3229 error = VOP_OPEN(vp, fmode, p->p_ucred, fp);
fad57d0e
MD
3230 if (error) {
3231 /*
3232 * setting f_ops this way prevents VOP_CLOSE from being
3233 * called or fdrop() releasing the vp from v_data. Since
3234 * the VOP_OPEN failed we don't want to VOP_CLOSE.
3235 */
3236 fp->f_ops = &badfileops;
3237 fp->f_data = NULL;
259b8ea0 3238 goto bad_drop;
984263bc 3239 }
fad57d0e
MD
3240
3241 /*
8ddc6004
MD
3242 * The fp is given its own reference, we still have our ref and lock.
3243 *
7540ab49 3244 * Assert that all regular files must be created with a VM object.
984263bc 3245 */
7540ab49
MD
3246 if (vp->v_type == VREG && vp->v_object == NULL) {
3247 printf("fhopen: regular file did not have VM object: %p\n", vp);
259b8ea0 3248 goto bad_drop;
fad57d0e 3249 }
984263bc 3250
fad57d0e 3251 /*
259b8ea0 3252 * The open was successful. Handle any locking requirements.
fad57d0e 3253 */
984263bc
MD
3254 if (fmode & (O_EXLOCK | O_SHLOCK)) {