Add kernel-layer support for chflags checks, remove (most) from the VFS layer.
[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 $
91fdc001 40 * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.135 2008/11/11 00:55:49 pavalos 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>
895c1f85 61#include <sys/priv.h>
9658f145 62#include <sys/jail.h>
dadab5e9 63#include <sys/namei.h>
21739618 64#include <sys/nlookup.h>
984263bc
MD
65#include <sys/dirent.h>
66#include <sys/extattr.h>
8fa76237 67#include <sys/spinlock.h>
8f6f8622 68#include <sys/kern_syscall.h>
70aac194 69#include <sys/objcache.h>
984263bc 70#include <sys/sysctl.h>
408357d8
MD
71
72#include <sys/buf2.h>
28623bf9
MD
73#include <sys/file2.h>
74#include <sys/spinlock2.h>
75
984263bc
MD
76#include <vm/vm.h>
77#include <vm/vm_object.h>
984263bc
MD
78#include <vm/vm_page.h>
79
28623bf9
MD
80#include <machine/limits.h>
81#include <machine/stdarg.h>
82
83#include <vfs/union/union.h>
dadab5e9 84
28623bf9 85static void mount_warning(struct mount *mp, const char *ctl, ...);
d3c546e6 86static int mount_path(struct proc *p, struct mount *mp, char **rb, char **fb);
e24b948e 87static int checkvp_chdir (struct vnode *vn, struct thread *td);
77480a97 88static void checkdirs (struct nchandle *old_nch, struct nchandle *new_nch);
402ed7e1 89static int chroot_refuse_vdir_fds (struct filedesc *fdp);
75ffff0d 90static int chroot_visible_mnt(struct mount *mp, struct proc *p);
402ed7e1
RG
91static int getutimes (const struct timeval *, struct timespec *);
92static int setfown (struct vnode *, uid_t, gid_t);
93static int setfmode (struct vnode *, int);
94static int setfflags (struct vnode *, int);
3a907475
MD
95static int setutimes (struct vnode *, struct vattr *,
96 const struct timespec *, int);
984263bc
MD
97static int usermount = 0; /* if 1, non-root can mount fs. */
98
402ed7e1 99int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
984263bc
MD
100
101SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
102
103/*
104 * Virtual File System System Calls
105 */
106
107/*
108 * Mount a file system.
109 */
41c20dac
MD
110/*
111 * mount_args(char *type, char *path, int flags, caddr_t data)
112 */
984263bc
MD
113/* ARGSUSED */
114int
753fd850 115sys_mount(struct mount_args *uap)
984263bc 116{
dadab5e9
MD
117 struct thread *td = curthread;
118 struct proc *p = td->td_proc;
984263bc 119 struct vnode *vp;
28623bf9 120 struct nchandle nch;
984263bc
MD
121 struct mount *mp;
122 struct vfsconf *vfsp;
123 int error, flag = 0, flag2 = 0;
1d505369 124 int hasmount;
984263bc 125 struct vattr va;
21739618 126 struct nlookupdata nd;
984263bc 127 char fstypename[MFSNAMELEN];
acde96db 128 struct ucred *cred = p->p_ucred;
984263bc 129
cd0d3e17 130 KKASSERT(p);
57962f80 131 if (jailed(cred))
84c3a710 132 return (EPERM);
895c1f85 133 if (usermount == 0 && (error = priv_check(td, PRIV_ROOT)))
984263bc
MD
134 return (error);
135 /*
136 * Do not allow NFS export by non-root users.
137 */
ab2eb4eb 138 if (uap->flags & MNT_EXPORTED) {
895c1f85 139 error = priv_check(td, PRIV_ROOT);
984263bc
MD
140 if (error)
141 return (error);
142 }
143 /*
144 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
145 */
895c1f85 146 if (priv_check(td, PRIV_ROOT))
ab2eb4eb 147 uap->flags |= MNT_NOSUID | MNT_NODEV;
21739618 148
984263bc 149 /*
28623bf9 150 * Lookup the requested path and extract the nch and vnode.
984263bc 151 */
21739618
MD
152 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
153 if (error == 0) {
154 if ((error = nlookup(&nd)) == 0) {
28623bf9 155 if (nd.nl_nch.ncp->nc_vp == NULL)
21739618
MD
156 error = ENOENT;
157 }
158 }
159 if (error) {
160 nlookup_done(&nd);
984263bc 161 return (error);
21739618
MD
162 }
163
164 /*
165 * Extract the locked+refd ncp and cleanup the nd structure
166 */
28623bf9
MD
167 nch = nd.nl_nch;
168 cache_zero(&nd.nl_nch);
21739618
MD
169 nlookup_done(&nd);
170
28623bf9 171 if ((nch.ncp->nc_flag & NCF_ISMOUNTPT) && cache_findmount(&nch))
1d505369
MD
172 hasmount = 1;
173 else
174 hasmount = 0;
175
176
21739618 177 /*
28623bf9 178 * now we have the locked ref'd nch and unreferenced vnode.
21739618 179 */
28623bf9 180 vp = nch.ncp->nc_vp;
87de5057 181 if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
28623bf9 182 cache_put(&nch);
21739618
MD
183 return (error);
184 }
28623bf9 185 cache_unlock(&nch);
21739618
MD
186
187 /*
28623bf9 188 * Now we have an unlocked ref'd nch and a locked ref'd vp
21739618 189 */
ab2eb4eb 190 if (uap->flags & MNT_UPDATE) {
67863d04 191 if ((vp->v_flag & (VROOT|VPFSROOT)) == 0) {
28623bf9 192 cache_drop(&nch);
984263bc
MD
193 vput(vp);
194 return (EINVAL);
195 }
196 mp = vp->v_mount;
197 flag = mp->mnt_flag;
198 flag2 = mp->mnt_kern_flag;
199 /*
200 * We only allow the filesystem to be reloaded if it
201 * is currently mounted read-only.
202 */
ab2eb4eb 203 if ((uap->flags & MNT_RELOAD) &&
984263bc 204 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
28623bf9 205 cache_drop(&nch);
984263bc
MD
206 vput(vp);
207 return (EOPNOTSUPP); /* Needs translation */
208 }
209 /*
210 * Only root, or the user that did the original mount is
211 * permitted to update it.
212 */
acde96db 213 if (mp->mnt_stat.f_owner != cred->cr_uid &&
895c1f85 214 (error = priv_check(td, PRIV_ROOT))) {
28623bf9 215 cache_drop(&nch);
984263bc
MD
216 vput(vp);
217 return (error);
218 }
f9642f56 219 if (vfs_busy(mp, LK_NOWAIT)) {
28623bf9 220 cache_drop(&nch);
984263bc
MD
221 vput(vp);
222 return (EBUSY);
223 }
1d505369 224 if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
28623bf9 225 cache_drop(&nch);
f9642f56 226 vfs_unbusy(mp);
984263bc
MD
227 vput(vp);
228 return (EBUSY);
229 }
230 vp->v_flag |= VMOUNT;
984263bc 231 mp->mnt_flag |=
ab2eb4eb 232 uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
a11aaa81 233 vn_unlock(vp);
984263bc
MD
234 goto update;
235 }
236 /*
237 * If the user is not root, ensure that they own the directory
238 * onto which we are attempting to mount.
239 */
87de5057 240 if ((error = VOP_GETATTR(vp, &va)) ||
895c1f85 241 (va.va_uid != cred->cr_uid && (error = priv_check(td, PRIV_ROOT)))) {
28623bf9 242 cache_drop(&nch);
984263bc
MD
243 vput(vp);
244 return (error);
245 }
87de5057 246 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
28623bf9 247 cache_drop(&nch);
984263bc
MD
248 vput(vp);
249 return (error);
250 }
251 if (vp->v_type != VDIR) {
28623bf9 252 cache_drop(&nch);
984263bc
MD
253 vput(vp);
254 return (ENOTDIR);
255 }
5dc91765
MD
256 if (vp->v_mount->mnt_kern_flag & MNTK_NOSTKMNT) {
257 cache_drop(&nch);
258 vput(vp);
259 return (EPERM);
260 }
ab2eb4eb 261 if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
28623bf9 262 cache_drop(&nch);
984263bc
MD
263 vput(vp);
264 return (error);
265 }
2613053d 266 vfsp = vfsconf_find_by_name(fstypename);
984263bc
MD
267 if (vfsp == NULL) {
268 linker_file_t lf;
269
270 /* Only load modules for root (very important!) */
895c1f85 271 if ((error = priv_check(td, PRIV_ROOT)) != 0) {
28623bf9 272 cache_drop(&nch);
984263bc
MD
273 vput(vp);
274 return error;
275 }
276 error = linker_load_file(fstypename, &lf);
277 if (error || lf == NULL) {
28623bf9 278 cache_drop(&nch);
984263bc
MD
279 vput(vp);
280 if (lf == NULL)
281 error = ENODEV;
282 return error;
283 }
284 lf->userrefs++;
285 /* lookup again, see if the VFS was loaded */
2613053d 286 vfsp = vfsconf_find_by_name(fstypename);
984263bc
MD
287 if (vfsp == NULL) {
288 lf->userrefs--;
289 linker_file_unload(lf);
28623bf9 290 cache_drop(&nch);
984263bc
MD
291 vput(vp);
292 return (ENODEV);
293 }
294 }
1d505369 295 if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
28623bf9 296 cache_drop(&nch);
984263bc
MD
297 vput(vp);
298 return (EBUSY);
299 }
300 vp->v_flag |= VMOUNT;
984263bc
MD
301
302 /*
303 * Allocate and initialize the filesystem.
304 */
efda3bd0 305 mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
984263bc
MD
306 TAILQ_INIT(&mp->mnt_nvnodelist);
307 TAILQ_INIT(&mp->mnt_reservedvnlist);
2281065e 308 TAILQ_INIT(&mp->mnt_jlist);
984263bc 309 mp->mnt_nvnodelistsize = 0;
ab6f251b 310 lockinit(&mp->mnt_lock, "vfslock", 0, 0);
f9642f56 311 vfs_busy(mp, LK_NOWAIT);
984263bc
MD
312 mp->mnt_op = vfsp->vfc_vfsops;
313 mp->mnt_vfc = vfsp;
314 vfsp->vfc_refcount++;
315 mp->mnt_stat.f_type = vfsp->vfc_typenum;
316 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
317 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
acde96db 318 mp->mnt_stat.f_owner = cred->cr_uid;
984263bc 319 mp->mnt_iosize_max = DFLTPHYS;
a11aaa81 320 vn_unlock(vp);
984263bc
MD
321update:
322 /*
323 * Set the mount level flags.
324 */
ab2eb4eb 325 if (uap->flags & MNT_RDONLY)
984263bc
MD
326 mp->mnt_flag |= MNT_RDONLY;
327 else if (mp->mnt_flag & MNT_RDONLY)
328 mp->mnt_kern_flag |= MNTK_WANTRDWR;
329 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
330 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
331 MNT_NOSYMFOLLOW | MNT_IGNORE |
332 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
ab2eb4eb 333 mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
984263bc
MD
334 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
335 MNT_NOSYMFOLLOW | MNT_IGNORE |
336 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
337 /*
338 * Mount the filesystem.
339 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
fad57d0e 340 * get.
984263bc 341 */
acde96db 342 error = VFS_MOUNT(mp, uap->path, uap->data, cred);
984263bc
MD
343 if (mp->mnt_flag & MNT_UPDATE) {
344 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
345 mp->mnt_flag &= ~MNT_RDONLY;
346 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
347 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
348 if (error) {
349 mp->mnt_flag = flag;
350 mp->mnt_kern_flag = flag2;
351 }
f9642f56 352 vfs_unbusy(mp);
984263bc 353 vp->v_flag &= ~VMOUNT;
984263bc 354 vrele(vp);
28623bf9 355 cache_drop(&nch);
984263bc
MD
356 return (error);
357 }
ca466bae 358 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc 359 /*
21739618 360 * Put the new filesystem on the mount list after root. The mount
28623bf9
MD
361 * point gets its own mnt_ncmountpt (unless the VFS already set one
362 * up) which represents the root of the mount. The lookup code
363 * detects the mount point going forward and checks the root of
364 * the mount going backwards.
8e005a45
MD
365 *
366 * It is not necessary to invalidate or purge the vnode underneath
367 * because elements under the mount will be given their own glue
368 * namecache record.
984263bc 369 */
984263bc 370 if (!error) {
28623bf9
MD
371 if (mp->mnt_ncmountpt.ncp == NULL) {
372 /*
373 * allocate, then unlock, but leave the ref intact
374 */
375 cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
376 cache_unlock(&mp->mnt_ncmountpt);
377 }
378 mp->mnt_ncmounton = nch; /* inherits ref */
379 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
380
381 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
984263bc 382 vp->v_flag &= ~VMOUNT;
861905fb 383 mountlist_insert(mp, MNTINS_LAST);
a11aaa81 384 vn_unlock(vp);
c0721f5f 385 checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
41a01a4d 386 error = vfs_allocate_syncvnode(mp);
f9642f56 387 vfs_unbusy(mp);
1d505369
MD
388 error = VFS_START(mp, 0);
389 vrele(vp);
984263bc 390 } else {
66a1ddf5
MD
391 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
392 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
393 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
394 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
395 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
984263bc 396 vp->v_flag &= ~VMOUNT;
984263bc 397 mp->mnt_vfc->vfc_refcount--;
f9642f56 398 vfs_unbusy(mp);
efda3bd0 399 kfree(mp, M_MOUNT);
28623bf9 400 cache_drop(&nch);
984263bc
MD
401 vput(vp);
402 }
403 return (error);
404}
405
406/*
407 * Scan all active processes to see if any of them have a current
408 * or root directory onto which the new filesystem has just been
409 * mounted. If so, replace them with the new mount point.
21739618
MD
410 *
411 * The passed ncp is ref'd and locked (from the mount code) and
412 * must be associated with the vnode representing the root of the
413 * mount point.
984263bc 414 */
8fa76237 415struct checkdirs_info {
77480a97
MD
416 struct nchandle old_nch;
417 struct nchandle new_nch;
418 struct vnode *old_vp;
419 struct vnode *new_vp;
8fa76237
MD
420};
421
422static int checkdirs_callback(struct proc *p, void *data);
423
984263bc 424static void
77480a97 425checkdirs(struct nchandle *old_nch, struct nchandle *new_nch)
984263bc 426{
8fa76237 427 struct checkdirs_info info;
77480a97 428 struct vnode *olddp;
984263bc 429 struct vnode *newdp;
690a3127 430 struct mount *mp;
984263bc 431
77480a97
MD
432 /*
433 * If the old mount point's vnode has a usecount of 1, it is not
434 * being held as a descriptor anywhere.
435 */
436 olddp = old_nch->ncp->nc_vp;
3c37c940 437 if (olddp == NULL || olddp->v_sysref.refcnt == 1)
984263bc 438 return;
77480a97
MD
439
440 /*
441 * Force the root vnode of the new mount point to be resolved
442 * so we can update any matching processes.
443 */
444 mp = new_nch->mount;
690a3127 445 if (VFS_ROOT(mp, &newdp))
984263bc 446 panic("mount: lost mount");
77480a97
MD
447 cache_setunresolved(new_nch);
448 cache_setvp(new_nch, newdp);
21739618 449
77480a97
MD
450 /*
451 * Special handling of the root node
452 */
21739618
MD
453 if (rootvnode == olddp) {
454 vref(newdp);
77480a97 455 vfs_cache_setroot(newdp, cache_hold(new_nch));
21739618
MD
456 }
457
77480a97
MD
458 /*
459 * Pass newdp separately so the callback does not have to access
460 * it via new_nch->ncp->nc_vp.
461 */
462 info.old_nch = *old_nch;
463 info.new_nch = *new_nch;
464 info.new_vp = newdp;
8fa76237
MD
465 allproc_scan(checkdirs_callback, &info);
466 vput(newdp);
467}
468
469/*
470 * NOTE: callback is not MP safe because the scanned process's filedesc
471 * structure can be ripped out from under us, amoung other things.
472 */
473static int
474checkdirs_callback(struct proc *p, void *data)
475{
476 struct checkdirs_info *info = data;
477 struct filedesc *fdp;
28623bf9
MD
478 struct nchandle ncdrop1;
479 struct nchandle ncdrop2;
8fa76237
MD
480 struct vnode *vprele1;
481 struct vnode *vprele2;
482
483 if ((fdp = p->p_fd) != NULL) {
28623bf9
MD
484 cache_zero(&ncdrop1);
485 cache_zero(&ncdrop2);
8fa76237
MD
486 vprele1 = NULL;
487 vprele2 = NULL;
488
489 /*
490 * MPUNSAFE - XXX fdp can be pulled out from under a
491 * foreign process.
492 *
493 * A shared filedesc is ok, we don't have to copy it
494 * because we are making this change globally.
495 */
496 spin_lock_wr(&fdp->fd_spin);
77480a97
MD
497 if (fdp->fd_ncdir.mount == info->old_nch.mount &&
498 fdp->fd_ncdir.ncp == info->old_nch.ncp) {
8fa76237 499 vprele1 = fdp->fd_cdir;
77480a97
MD
500 vref(info->new_vp);
501 fdp->fd_cdir = info->new_vp;
8fa76237 502 ncdrop1 = fdp->fd_ncdir;
77480a97 503 cache_copy(&info->new_nch, &fdp->fd_ncdir);
984263bc 504 }
77480a97
MD
505 if (fdp->fd_nrdir.mount == info->old_nch.mount &&
506 fdp->fd_nrdir.ncp == info->old_nch.ncp) {
8fa76237 507 vprele2 = fdp->fd_rdir;
77480a97
MD
508 vref(info->new_vp);
509 fdp->fd_rdir = info->new_vp;
8fa76237 510 ncdrop2 = fdp->fd_nrdir;
77480a97 511 cache_copy(&info->new_nch, &fdp->fd_nrdir);
984263bc 512 }
8fa76237 513 spin_unlock_wr(&fdp->fd_spin);
28623bf9
MD
514 if (ncdrop1.ncp)
515 cache_drop(&ncdrop1);
516 if (ncdrop2.ncp)
517 cache_drop(&ncdrop2);
8fa76237
MD
518 if (vprele1)
519 vrele(vprele1);
520 if (vprele2)
521 vrele(vprele2);
984263bc 522 }
8fa76237 523 return(0);
984263bc
MD
524}
525
526/*
527 * Unmount a file system.
528 *
529 * Note: unmount takes a path to the vnode mounted on as argument,
530 * not special file (as before).
531 */
41c20dac
MD
532/*
533 * umount_args(char *path, int flags)
534 */
984263bc
MD
535/* ARGSUSED */
536int
753fd850 537sys_unmount(struct unmount_args *uap)
984263bc 538{
dadab5e9
MD
539 struct thread *td = curthread;
540 struct proc *p = td->td_proc;
d2293868 541 struct mount *mp = NULL;
984263bc 542 int error;
fad57d0e 543 struct nlookupdata nd;
984263bc 544
dadab5e9 545 KKASSERT(p);
84c3a710
HP
546 if (p->p_ucred->cr_prison != NULL)
547 return (EPERM);
895c1f85 548 if (usermount == 0 && (error = priv_check(td, PRIV_ROOT)))
84c3a710
HP
549 return (error);
550
ab2eb4eb 551 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
552 if (error == 0)
553 error = nlookup(&nd);
fad57d0e 554 if (error)
d2293868 555 goto out;
fad57d0e 556
28623bf9 557 mp = nd.nl_nch.mount;
984263bc
MD
558
559 /*
560 * Only root, or the user that did the original mount is
561 * permitted to unmount this filesystem.
562 */
563 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
895c1f85 564 (error = priv_check(td, PRIV_ROOT)))
d2293868 565 goto out;
984263bc
MD
566
567 /*
568 * Don't allow unmounting the root file system.
569 */
570 if (mp->mnt_flag & MNT_ROOTFS) {
d2293868
MD
571 error = EINVAL;
572 goto out;
984263bc
MD
573 }
574
575 /*
576 * Must be the root of the filesystem
577 */
28623bf9 578 if (nd.nl_nch.ncp != mp->mnt_ncmountpt.ncp) {
d2293868
MD
579 error = EINVAL;
580 goto out;
984263bc 581 }
d2293868
MD
582
583out:
584 nlookup_done(&nd);
585 if (error)
586 return (error);
acde96db 587 return (dounmount(mp, uap->flags));
984263bc
MD
588}
589
590/*
591 * Do the actual file system unmount.
592 */
861905fb
MD
593static int
594dounmount_interlock(struct mount *mp)
595{
596 if (mp->mnt_kern_flag & MNTK_UNMOUNT)
597 return (EBUSY);
598 mp->mnt_kern_flag |= MNTK_UNMOUNT;
599 return(0);
600}
601
984263bc 602int
acde96db 603dounmount(struct mount *mp, int flags)
984263bc 604{
dcf26f79 605 struct namecache *ncp;
28623bf9 606 struct nchandle nch;
2ec4b00d 607 struct vnode *vp;
984263bc
MD
608 int error;
609 int async_flag;
2df9419c 610 int lflags;
28623bf9 611 int freeok = 1;
984263bc 612
861905fb
MD
613 /*
614 * Exclusive access for unmounting purposes
615 */
616 if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
617 return (error);
618
619 /*
620 * Allow filesystems to detect that a forced unmount is in progress.
621 */
984263bc
MD
622 if (flags & MNT_FORCE)
623 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
2df9419c 624 lflags = LK_EXCLUSIVE | ((flags & MNT_FORCE) ? 0 : LK_NOWAIT);
df4f70a6 625 error = lockmgr(&mp->mnt_lock, lflags);
984263bc
MD
626 if (error) {
627 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
628 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 629 wakeup(mp);
984263bc
MD
630 return (error);
631 }
632
633 if (mp->mnt_flag & MNT_EXPUBLIC)
634 vfs_setpublicfs(NULL, NULL, NULL);
635
636 vfs_msync(mp, MNT_WAIT);
637 async_flag = mp->mnt_flag & MNT_ASYNC;
638 mp->mnt_flag &=~ MNT_ASYNC;
dcf26f79
MD
639
640 /*
28623bf9
MD
641 * If this filesystem isn't aliasing other filesystems,
642 * try to invalidate any remaining namecache entries and
643 * check the count afterwords.
dcf26f79 644 */
28623bf9
MD
645 if ((mp->mnt_kern_flag & MNTK_NCALIASED) == 0) {
646 cache_lock(&mp->mnt_ncmountpt);
647 cache_inval(&mp->mnt_ncmountpt, CINV_DESTROY|CINV_CHILDREN);
648 cache_unlock(&mp->mnt_ncmountpt);
649
650 if ((ncp = mp->mnt_ncmountpt.ncp) != NULL &&
651 (ncp->nc_refs != 1 || TAILQ_FIRST(&ncp->nc_list))) {
dcf26f79
MD
652
653 if ((flags & MNT_FORCE) == 0) {
654 error = EBUSY;
28623bf9
MD
655 mount_warning(mp, "Cannot unmount: "
656 "%d namecache "
657 "references still "
658 "present",
659 ncp->nc_refs - 1);
dcf26f79 660 } else {
28623bf9
MD
661 mount_warning(mp, "Forced unmount: "
662 "%d namecache "
663 "references still "
664 "present",
665 ncp->nc_refs - 1);
666 freeok = 0;
dcf26f79
MD
667 }
668 }
669 }
670
28623bf9
MD
671 /*
672 * nchandle records ref the mount structure. Expect a count of 1
673 * (our mount->mnt_ncmountpt).
674 */
675 if (mp->mnt_refs != 1) {
676 if ((flags & MNT_FORCE) == 0) {
677 mount_warning(mp, "Cannot unmount: "
678 "%d process references still "
679 "present", mp->mnt_refs);
680 error = EBUSY;
681 } else {
682 mount_warning(mp, "Forced unmount: "
683 "%d process references still "
684 "present", mp->mnt_refs);
685 freeok = 0;
686 }
687 }
688
2ec4b00d
MD
689 /*
690 * Decomission our special mnt_syncer vnode. This also stops
691 * the vnlru code. If we are unable to unmount we recommission
692 * the vnode.
693 */
dcf26f79 694 if (error == 0) {
2ec4b00d
MD
695 if ((vp = mp->mnt_syncer) != NULL) {
696 mp->mnt_syncer = NULL;
697 vrele(vp);
698 }
dcf26f79
MD
699 if (((mp->mnt_flag & MNT_RDONLY) ||
700 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) ||
701 (flags & MNT_FORCE)) {
702 error = VFS_UNMOUNT(mp, flags);
703 }
704 }
984263bc 705 if (error) {
41a01a4d
MD
706 if (mp->mnt_syncer == NULL)
707 vfs_allocate_syncvnode(mp);
984263bc
MD
708 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
709 mp->mnt_flag |= async_flag;
df4f70a6 710 lockmgr(&mp->mnt_lock, LK_RELEASE);
984263bc 711 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 712 wakeup(mp);
984263bc
MD
713 return (error);
714 }
432b8263
MD
715 /*
716 * Clean up any journals still associated with the mount after
717 * filesystem activity has ceased.
718 */
719 journal_remove_all_journals(mp,
720 ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));
721
861905fb 722 mountlist_remove(mp);
0961aa92
MD
723
724 /*
725 * Remove any installed vnode ops here so the individual VFSs don't
726 * have to.
727 */
66a1ddf5
MD
728 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
729 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
730 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
731 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
732 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
0961aa92 733
28623bf9
MD
734 if (mp->mnt_ncmountpt.ncp != NULL) {
735 nch = mp->mnt_ncmountpt;
736 cache_zero(&mp->mnt_ncmountpt);
737 cache_clrmountpt(&nch);
738 cache_drop(&nch);
739 }
740 if (mp->mnt_ncmounton.ncp != NULL) {
741 nch = mp->mnt_ncmounton;
742 cache_zero(&mp->mnt_ncmounton);
743 cache_clrmountpt(&nch);
744 cache_drop(&nch);
984263bc 745 }
1d505369 746
984263bc
MD
747 mp->mnt_vfc->vfc_refcount--;
748 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
749 panic("unmount: dangling vnode");
df4f70a6 750 lockmgr(&mp->mnt_lock, LK_RELEASE);
984263bc 751 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 752 wakeup(mp);
28623bf9
MD
753 if (freeok)
754 kfree(mp, M_MOUNT);
984263bc
MD
755 return (0);
756}
757
28623bf9
MD
758static
759void
760mount_warning(struct mount *mp, const char *ctl, ...)
761{
762 char *ptr;
763 char *buf;
764 __va_list va;
765
766 __va_start(va, ctl);
767 if (cache_fullpath(NULL, &mp->mnt_ncmounton, &ptr, &buf) == 0) {
6ea70f76 768 kprintf("unmount(%s): ", ptr);
379210cb 769 kvprintf(ctl, va);
6ea70f76 770 kprintf("\n");
28623bf9
MD
771 kfree(buf, M_TEMP);
772 } else {
56bcacad
MD
773 kprintf("unmount(%p", mp);
774 if (mp->mnt_ncmounton.ncp && mp->mnt_ncmounton.ncp->nc_name)
775 kprintf(",%s", mp->mnt_ncmounton.ncp->nc_name);
776 kprintf("): ");
379210cb 777 kvprintf(ctl, va);
6ea70f76 778 kprintf("\n");
28623bf9
MD
779 }
780 __va_end(va);
781}
782
984263bc 783/*
d3c546e6
MD
784 * Shim cache_fullpath() to handle the case where a process is chrooted into
785 * a subdirectory of a mount. In this case if the root mount matches the
786 * process root directory's mount we have to specify the process's root
787 * directory instead of the mount point, because the mount point might
788 * be above the root directory.
789 */
790static
791int
792mount_path(struct proc *p, struct mount *mp, char **rb, char **fb)
793{
794 struct nchandle *nch;
d3c546e6
MD
795
796 if (p && p->p_fd->fd_nrdir.mount == mp)
797 nch = &p->p_fd->fd_nrdir;
798 else
799 nch = &mp->mnt_ncmountpt;
800 return(cache_fullpath(p, nch, rb, fb));
801}
802
803/*
984263bc
MD
804 * Sync each mounted filesystem.
805 */
984263bc
MD
806
807#ifdef DEBUG
808static int syncprt = 0;
809SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
6bdbb368 810#endif /* DEBUG */
984263bc 811
861905fb
MD
812static int sync_callback(struct mount *mp, void *data);
813
984263bc
MD
814/* ARGSUSED */
815int
753fd850 816sys_sync(struct sync_args *uap)
984263bc 817{
861905fb 818 mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
6bdbb368 819#ifdef DEBUG
861905fb
MD
820 /*
821 * print out buffer pool stat information on each sync() call.
822 */
984263bc
MD
823 if (syncprt)
824 vfs_bufstats();
6bdbb368 825#endif /* DEBUG */
984263bc
MD
826 return (0);
827}
828
861905fb
MD
829static
830int
831sync_callback(struct mount *mp, void *data __unused)
832{
833 int asyncflag;
834
835 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
836 asyncflag = mp->mnt_flag & MNT_ASYNC;
837 mp->mnt_flag &= ~MNT_ASYNC;
838 vfs_msync(mp, MNT_NOWAIT);
87de5057 839 VFS_SYNC(mp, MNT_NOWAIT);
861905fb
MD
840 mp->mnt_flag |= asyncflag;
841 }
842 return(0);
843}
844
984263bc
MD
845/* XXX PRISON: could be per prison flag */
846static int prison_quotas;
847#if 0
848SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
849#endif
850
851/*
41c20dac
MD
852 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
853 *
984263bc
MD
854 * Change filesystem quotas.
855 */
984263bc
MD
856/* ARGSUSED */
857int
753fd850 858sys_quotactl(struct quotactl_args *uap)
984263bc 859{
fad57d0e
MD
860 struct nlookupdata nd;
861 struct thread *td;
862 struct proc *p;
41c20dac 863 struct mount *mp;
984263bc 864 int error;
984263bc 865
fad57d0e
MD
866 td = curthread;
867 p = td->td_proc;
41c20dac 868 if (p->p_ucred->cr_prison && !prison_quotas)
984263bc 869 return (EPERM);
fad57d0e 870
ab2eb4eb 871 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
872 if (error == 0)
873 error = nlookup(&nd);
874 if (error == 0) {
28623bf9 875 mp = nd.nl_nch.mount;
ab2eb4eb 876 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
acde96db 877 uap->arg, nd.nl_cred);
fad57d0e
MD
878 }
879 nlookup_done(&nd);
880 return (error);
984263bc
MD
881}
882
949ecb9b 883/*
2281065e 884 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
949ecb9b
MD
885 * void *buf, int buflen)
886 *
887 * This function operates on a mount point and executes the specified
888 * operation using the specified control data, and possibly returns data.
889 *
890 * The actual number of bytes stored in the result buffer is returned, 0
891 * if none, otherwise an error is returned.
892 */
893/* ARGSUSED */
894int
753fd850 895sys_mountctl(struct mountctl_args *uap)
949ecb9b
MD
896{
897 struct thread *td = curthread;
898 struct proc *p = td->td_proc;
2281065e 899 struct file *fp;
949ecb9b
MD
900 void *ctl = NULL;
901 void *buf = NULL;
902 char *path = NULL;
903 int error;
904
905 /*
906 * Sanity and permissions checks. We must be root.
907 */
908 KKASSERT(p);
909 if (p->p_ucred->cr_prison != NULL)
910 return (EPERM);
895c1f85 911 if ((error = priv_check(td, PRIV_ROOT)) != 0)
949ecb9b
MD
912 return (error);
913
914 /*
915 * Argument length checks
916 */
39b13188 917 if (uap->ctllen < 0 || uap->ctllen > 1024)
949ecb9b 918 return (EINVAL);
39b13188 919 if (uap->buflen < 0 || uap->buflen > 16 * 1024)
949ecb9b
MD
920 return (EINVAL);
921 if (uap->path == NULL)
922 return (EINVAL);
923
924 /*
925 * Allocate the necessary buffers and copyin data
926 */
70aac194 927 path = objcache_get(namei_oc, M_WAITOK);
949ecb9b
MD
928 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
929 if (error)
930 goto done;
931
932 if (uap->ctllen) {
efda3bd0 933 ctl = kmalloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
949ecb9b
MD
934 error = copyin(uap->ctl, ctl, uap->ctllen);
935 if (error)
936 goto done;
937 }
938 if (uap->buflen)
efda3bd0 939 buf = kmalloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
949ecb9b
MD
940
941 /*
2281065e
MD
942 * Validate the descriptor
943 */
ecda6326
MD
944 if (uap->fd >= 0) {
945 fp = holdfp(p->p_fd, uap->fd, -1);
946 if (fp == NULL) {
947 error = EBADF;
948 goto done;
949 }
950 } else {
951 fp = NULL;
2281065e 952 }
2281065e
MD
953
954 /*
949ecb9b
MD
955 * Execute the internal kernel function and clean up.
956 */
2281065e
MD
957 error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
958 if (fp)
9f87144f 959 fdrop(fp);
949ecb9b
MD
960 if (error == 0 && uap->sysmsg_result > 0)
961 error = copyout(buf, uap->buf, uap->sysmsg_result);
962done:
963 if (path)
70aac194 964 objcache_put(namei_oc, path);
949ecb9b 965 if (ctl)
efda3bd0 966 kfree(ctl, M_TEMP);
949ecb9b 967 if (buf)
efda3bd0 968 kfree(buf, M_TEMP);
949ecb9b
MD
969 return (error);
970}
971
972/*
973 * Execute a mount control operation by resolving the path to a mount point
974 * and calling vop_mountctl().
67863d04
MD
975 *
976 * Use the mount point from the nch instead of the vnode so nullfs mounts
977 * can properly spike the VOP.
949ecb9b
MD
978 */
979int
2281065e
MD
980kern_mountctl(const char *path, int op, struct file *fp,
981 const void *ctl, int ctllen,
949ecb9b
MD
982 void *buf, int buflen, int *res)
983{
949ecb9b
MD
984 struct vnode *vp;
985 struct mount *mp;
986 struct nlookupdata nd;
987 int error;
988
989 *res = 0;
990 vp = NULL;
991 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
992 if (error == 0)
993 error = nlookup(&nd);
994 if (error == 0)
28623bf9 995 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
67863d04 996 mp = nd.nl_nch.mount;
949ecb9b
MD
997 nlookup_done(&nd);
998 if (error)
999 return (error);
1000
949ecb9b
MD
1001 /*
1002 * Must be the root of the filesystem
1003 */
67863d04 1004 if ((vp->v_flag & (VROOT|VPFSROOT)) == 0) {
949ecb9b
MD
1005 vput(vp);
1006 return (EINVAL);
1007 }
2281065e 1008 error = vop_mountctl(mp->mnt_vn_use_ops, op, fp, ctl, ctllen,
949ecb9b
MD
1009 buf, buflen, res);
1010 vput(vp);
1011 return (error);
1012}
1013
984263bc 1014int
fad57d0e 1015kern_statfs(struct nlookupdata *nd, struct statfs *buf)
984263bc 1016{
dadab5e9 1017 struct thread *td = curthread;
75ffff0d 1018 struct proc *p = td->td_proc;
41c20dac
MD
1019 struct mount *mp;
1020 struct statfs *sp;
75ffff0d 1021 char *fullpath, *freepath;
984263bc 1022 int error;
984263bc 1023
fad57d0e 1024 if ((error = nlookup(nd)) != 0)
984263bc 1025 return (error);
28623bf9 1026 mp = nd->nl_nch.mount;
984263bc 1027 sp = &mp->mnt_stat;
acde96db 1028 if ((error = VFS_STATFS(mp, sp, nd->nl_cred)) != 0)
984263bc 1029 return (error);
75ffff0d 1030
d3c546e6 1031 error = mount_path(p, mp, &fullpath, &freepath);
75ffff0d
JS
1032 if (error)
1033 return(error);
1034 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1035 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
efda3bd0 1036 kfree(freepath, M_TEMP);
75ffff0d 1037
984263bc 1038 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509
DRJ
1039 bcopy(sp, buf, sizeof(*buf));
1040 /* Only root should have access to the fsid's. */
895c1f85 1041 if (priv_check(td, PRIV_ROOT))
9697c509
DRJ
1042 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
1043 return (0);
984263bc
MD
1044}
1045
1046/*
9697c509 1047 * statfs_args(char *path, struct statfs *buf)
41c20dac 1048 *
984263bc
MD
1049 * Get filesystem statistics.
1050 */
984263bc 1051int
753fd850 1052sys_statfs(struct statfs_args *uap)
9697c509 1053{
fad57d0e 1054 struct nlookupdata nd;
9697c509
DRJ
1055 struct statfs buf;
1056 int error;
1057
fad57d0e
MD
1058 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1059 if (error == 0)
1060 error = kern_statfs(&nd, &buf);
1061 nlookup_done(&nd);
9697c509
DRJ
1062 if (error == 0)
1063 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1064 return (error);
1065}
1066
1067int
1068kern_fstatfs(int fd, struct statfs *buf)
984263bc 1069{
dadab5e9
MD
1070 struct thread *td = curthread;
1071 struct proc *p = td->td_proc;
984263bc
MD
1072 struct file *fp;
1073 struct mount *mp;
1fd87d54 1074 struct statfs *sp;
75ffff0d 1075 char *fullpath, *freepath;
984263bc 1076 int error;
984263bc 1077
dadab5e9 1078 KKASSERT(p);
5b287bba 1079 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc
MD
1080 return (error);
1081 mp = ((struct vnode *)fp->f_data)->v_mount;
5b287bba
MD
1082 if (mp == NULL) {
1083 error = EBADF;
1084 goto done;
1085 }
1086 if (fp->f_cred == NULL) {
1087 error = EINVAL;
1088 goto done;
1089 }
984263bc 1090 sp = &mp->mnt_stat;
5b287bba
MD
1091 if ((error = VFS_STATFS(mp, sp, fp->f_cred)) != 0)
1092 goto done;
75ffff0d 1093
d3c546e6 1094 if ((error = mount_path(p, mp, &fullpath, &freepath)) != 0)
5b287bba 1095 goto done;
75ffff0d
JS
1096 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1097 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
efda3bd0 1098 kfree(freepath, M_TEMP);
75ffff0d 1099
984263bc 1100 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509 1101 bcopy(sp, buf, sizeof(*buf));
75ffff0d 1102
9697c509 1103 /* Only root should have access to the fsid's. */
895c1f85 1104 if (priv_check(td, PRIV_ROOT))
9697c509 1105 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
5b287bba
MD
1106 error = 0;
1107done:
1108 fdrop(fp);
1109 return (error);
9697c509
DRJ
1110}
1111
1112/*
1113 * fstatfs_args(int fd, struct statfs *buf)
1114 *
1115 * Get filesystem statistics.
1116 */
1117int
753fd850 1118sys_fstatfs(struct fstatfs_args *uap)
9697c509
DRJ
1119{
1120 struct statfs buf;
1121 int error;
1122
1123 error = kern_fstatfs(uap->fd, &buf);
1124
1125 if (error == 0)
1126 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1127 return (error);
984263bc
MD
1128}
1129
d9fad06e
MD
1130int
1131kern_statvfs(struct nlookupdata *nd, struct statvfs *buf)
1132{
1133 struct mount *mp;
1134 struct statvfs *sp;
1135 int error;
1136
1137 if ((error = nlookup(nd)) != 0)
1138 return (error);
1139 mp = nd->nl_nch.mount;
1140 sp = &mp->mnt_vstat;
1141 if ((error = VFS_STATVFS(mp, sp, nd->nl_cred)) != 0)
1142 return (error);
1143
1144 sp->f_flag = 0;
1145 if (mp->mnt_flag & MNT_RDONLY)
1146 sp->f_flag |= ST_RDONLY;
1147 if (mp->mnt_flag & MNT_NOSUID)
1148 sp->f_flag |= ST_NOSUID;
1149 bcopy(sp, buf, sizeof(*buf));
1150 return (0);
1151}
1152
1153/*
1154 * statfs_args(char *path, struct statfs *buf)
1155 *
1156 * Get filesystem statistics.
1157 */
1158int
1159sys_statvfs(struct statvfs_args *uap)
1160{
1161 struct nlookupdata nd;
1162 struct statvfs buf;
1163 int error;
1164
1165 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1166 if (error == 0)
1167 error = kern_statvfs(&nd, &buf);
1168 nlookup_done(&nd);
1169 if (error == 0)
1170 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1171 return (error);
1172}
1173
1174int
1175kern_fstatvfs(int fd, struct statvfs *buf)
1176{
1177 struct thread *td = curthread;
1178 struct proc *p = td->td_proc;
1179 struct file *fp;
1180 struct mount *mp;
1181 struct statvfs *sp;
1182 int error;
1183
1184 KKASSERT(p);
1185 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
1186 return (error);
1187 mp = ((struct vnode *)fp->f_data)->v_mount;
1188 if (mp == NULL) {
1189 error = EBADF;
1190 goto done;
1191 }
1192 if (fp->f_cred == NULL) {
1193 error = EINVAL;
1194 goto done;
1195 }
1196 sp = &mp->mnt_vstat;
1197 if ((error = VFS_STATVFS(mp, sp, fp->f_cred)) != 0)
1198 goto done;
1199
1200 sp->f_flag = 0;
1201 if (mp->mnt_flag & MNT_RDONLY)
1202 sp->f_flag |= ST_RDONLY;
1203 if (mp->mnt_flag & MNT_NOSUID)
1204 sp->f_flag |= ST_NOSUID;
1205
1206 bcopy(sp, buf, sizeof(*buf));
1207 error = 0;
1208done:
1209 fdrop(fp);
1210 return (error);
1211}
1212
1213/*
1214 * fstatfs_args(int fd, struct statfs *buf)
1215 *
1216 * Get filesystem statistics.
1217 */
1218int
1219sys_fstatvfs(struct fstatvfs_args *uap)
1220{
1221 struct statvfs buf;
1222 int error;
1223
1224 error = kern_fstatvfs(uap->fd, &buf);
1225
1226 if (error == 0)
1227 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1228 return (error);
1229}
1230
984263bc 1231/*
41c20dac
MD
1232 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
1233 *
984263bc
MD
1234 * Get statistics on all filesystems.
1235 */
861905fb
MD
1236
1237struct getfsstat_info {
1238 struct statfs *sfsp;
1239 long count;
1240 long maxcount;
1241 int error;
1242 int flags;
861905fb
MD
1243 struct proc *p;
1244};
1245
1246static int getfsstat_callback(struct mount *, void *);
1247
41c20dac 1248/* ARGSUSED */
984263bc 1249int
753fd850 1250sys_getfsstat(struct getfsstat_args *uap)
984263bc 1251{
dadab5e9 1252 struct thread *td = curthread;
75ffff0d 1253 struct proc *p = td->td_proc;
861905fb 1254 struct getfsstat_info info;
75ffff0d 1255
861905fb 1256 bzero(&info, sizeof(info));
861905fb
MD
1257
1258 info.maxcount = uap->bufsize / sizeof(struct statfs);
1259 info.sfsp = uap->buf;
1260 info.count = 0;
1261 info.flags = uap->flags;
861905fb
MD
1262 info.p = p;
1263
1264 mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
1265 if (info.sfsp && info.count > info.maxcount)
1266 uap->sysmsg_result = info.maxcount;
75ffff0d 1267 else
861905fb
MD
1268 uap->sysmsg_result = info.count;
1269 return (info.error);
1270}
1271
1272static int
1273getfsstat_callback(struct mount *mp, void *data)
1274{
1275 struct getfsstat_info *info = data;
1276 struct statfs *sp;
1277 char *freepath;
1278 char *fullpath;
1279 int error;
1280
1281 if (info->sfsp && info->count < info->maxcount) {
28623bf9 1282 if (info->p && !chroot_visible_mnt(mp, info->p))
861905fb
MD
1283 return(0);
1284 sp = &mp->mnt_stat;
1285
1286 /*
1287 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1288 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1289 * overrides MNT_WAIT.
1290 */
1291 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1292 (info->flags & MNT_WAIT)) &&
acde96db 1293 (error = VFS_STATFS(mp, sp, info->p->p_ucred))) {
861905fb
MD
1294 return(0);
1295 }
1296 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1297
d3c546e6 1298 error = mount_path(info->p, mp, &fullpath, &freepath);
861905fb
MD
1299 if (error) {
1300 info->error = error;
1301 return(-1);
984263bc 1302 }
861905fb
MD
1303 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1304 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
efda3bd0 1305 kfree(freepath, M_TEMP);
861905fb
MD
1306
1307 error = copyout(sp, info->sfsp, sizeof(*sp));
1308 if (error) {
1309 info->error = error;
1310 return (-1);
984263bc 1311 }
861905fb 1312 ++info->sfsp;
984263bc 1313 }
861905fb
MD
1314 info->count++;
1315 return(0);
984263bc
MD
1316}
1317
1318/*
00fe9d48
MD
1319 * getvfsstat_args(struct statfs *buf, struct statvfs *vbuf,
1320 long bufsize, int flags)
1321 *
1322 * Get statistics on all filesystems.
1323 */
1324
1325struct getvfsstat_info {
1326 struct statfs *sfsp;
1327 struct statvfs *vsfsp;
1328 long count;
1329 long maxcount;
1330 int error;
1331 int flags;
1332 struct proc *p;
1333};
1334
1335static int getvfsstat_callback(struct mount *, void *);
1336
1337/* ARGSUSED */
1338int
1339sys_getvfsstat(struct getvfsstat_args *uap)
1340{
1341 struct thread *td = curthread;
1342 struct proc *p = td->td_proc;
1343 struct getvfsstat_info info;
1344
1345 bzero(&info, sizeof(info));
1346
1347 info.maxcount = uap->vbufsize / sizeof(struct statvfs);
1348 info.sfsp = uap->buf;
1349 info.vsfsp = uap->vbuf;
1350 info.count = 0;
1351 info.flags = uap->flags;
1352 info.p = p;
1353
1354 mountlist_scan(getvfsstat_callback, &info, MNTSCAN_FORWARD);
1355 if (info.vsfsp && info.count > info.maxcount)
1356 uap->sysmsg_result = info.maxcount;
1357 else
1358 uap->sysmsg_result = info.count;
1359 return (info.error);
1360}
1361
1362static int
1363getvfsstat_callback(struct mount *mp, void *data)
1364{
1365 struct getvfsstat_info *info = data;
1366 struct statfs *sp;
1367 struct statvfs *vsp;
1368 char *freepath;
1369 char *fullpath;
1370 int error;
1371
1372 if (info->vsfsp && info->count < info->maxcount) {
1373 if (info->p && !chroot_visible_mnt(mp, info->p))
1374 return(0);
1375 sp = &mp->mnt_stat;
1376 vsp = &mp->mnt_vstat;
1377
1378 /*
1379 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1380 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1381 * overrides MNT_WAIT.
1382 */
1383 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1384 (info->flags & MNT_WAIT)) &&
1385 (error = VFS_STATFS(mp, sp, info->p->p_ucred))) {
1386 return(0);
1387 }
1388 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1389
1390 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1391 (info->flags & MNT_WAIT)) &&
1392 (error = VFS_STATVFS(mp, vsp, info->p->p_ucred))) {
1393 return(0);
1394 }
1395 vsp->f_flag = 0;
1396 if (mp->mnt_flag & MNT_RDONLY)
1397 vsp->f_flag |= ST_RDONLY;
1398 if (mp->mnt_flag & MNT_NOSUID)
1399 vsp->f_flag |= ST_NOSUID;
1400
1401 error = mount_path(info->p, mp, &fullpath, &freepath);
1402 if (error) {
1403 info->error = error;
1404 return(-1);
1405 }
1406 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1407 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1408 kfree(freepath, M_TEMP);
1409
1410 error = copyout(sp, info->sfsp, sizeof(*sp));
1411 if (error == 0)
1412 error = copyout(vsp, info->vsfsp, sizeof(*vsp));
1413 if (error) {
1414 info->error = error;
1415 return (-1);
1416 }
1417 ++info->sfsp;
1418 ++info->vsfsp;
1419 }
1420 info->count++;
1421 return(0);
1422}
1423
1424
1425/*
41c20dac
MD
1426 * fchdir_args(int fd)
1427 *
984263bc
MD
1428 * Change current working directory to a given file descriptor.
1429 */
984263bc
MD
1430/* ARGSUSED */
1431int
753fd850 1432sys_fchdir(struct fchdir_args *uap)
984263bc 1433{
dadab5e9
MD
1434 struct thread *td = curthread;
1435 struct proc *p = td->td_proc;
41c20dac 1436 struct filedesc *fdp = p->p_fd;
21739618 1437 struct vnode *vp, *ovp;
984263bc
MD
1438 struct mount *mp;
1439 struct file *fp;
28623bf9 1440 struct nchandle nch, onch, tnch;
984263bc
MD
1441 int error;
1442
5b287bba 1443 if ((error = holdvnode(fdp, uap->fd, &fp)) != 0)
984263bc
MD
1444 return (error);
1445 vp = (struct vnode *)fp->f_data;
597aea93 1446 vref(vp);
ca466bae 1447 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
28623bf9 1448 if (vp->v_type != VDIR || fp->f_nchandle.ncp == NULL)
984263bc
MD
1449 error = ENOTDIR;
1450 else
87de5057 1451 error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
984263bc
MD
1452 if (error) {
1453 vput(vp);
5b287bba 1454 fdrop(fp);
984263bc
MD
1455 return (error);
1456 }
28623bf9 1457 cache_copy(&fp->f_nchandle, &nch);
1d505369
MD
1458
1459 /*
1460 * If the ncp has become a mount point, traverse through
1461 * the mount point.
1462 */
1463
28623bf9
MD
1464 while (!error && (nch.ncp->nc_flag & NCF_ISMOUNTPT) &&
1465 (mp = cache_findmount(&nch)) != NULL
1d505369 1466 ) {
28623bf9 1467 error = nlookup_mp(mp, &tnch);
21739618 1468 if (error == 0) {
28623bf9 1469 cache_unlock(&tnch); /* leave ref intact */
21739618 1470 vput(vp);
28623bf9 1471 vp = tnch.ncp->nc_vp;
87de5057 1472 error = vget(vp, LK_SHARED);
21739618 1473 KKASSERT(error == 0);
28623bf9
MD
1474 cache_drop(&nch);
1475 nch = tnch;
21739618
MD
1476 }
1477 }
1478 if (error == 0) {
1479 ovp = fdp->fd_cdir;
28623bf9 1480 onch = fdp->fd_ncdir;
a11aaa81 1481 vn_unlock(vp); /* leave ref intact */
21739618 1482 fdp->fd_cdir = vp;
28623bf9
MD
1483 fdp->fd_ncdir = nch;
1484 cache_drop(&onch);
21739618
MD
1485 vrele(ovp);
1486 } else {
28623bf9 1487 cache_drop(&nch);
21739618
MD
1488 vput(vp);
1489 }
5b287bba 1490 fdrop(fp);
21739618 1491 return (error);
984263bc
MD
1492}
1493
9697c509 1494int
21739618 1495kern_chdir(struct nlookupdata *nd)
9697c509
DRJ
1496{
1497 struct thread *td = curthread;
1498 struct proc *p = td->td_proc;
1499 struct filedesc *fdp = p->p_fd;
21739618 1500 struct vnode *vp, *ovp;
28623bf9 1501 struct nchandle onch;
9697c509
DRJ
1502 int error;
1503
21739618 1504 if ((error = nlookup(nd)) != 0)
9697c509 1505 return (error);
28623bf9 1506 if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
21739618 1507 return (ENOENT);
87de5057 1508 if ((error = vget(vp, LK_SHARED)) != 0)
21739618
MD
1509 return (error);
1510
1511 error = checkvp_chdir(vp, td);
a11aaa81 1512 vn_unlock(vp);
21739618
MD
1513 if (error == 0) {
1514 ovp = fdp->fd_cdir;
28623bf9
MD
1515 onch = fdp->fd_ncdir;
1516 cache_unlock(&nd->nl_nch); /* leave reference intact */
1517 fdp->fd_ncdir = nd->nl_nch;
21739618 1518 fdp->fd_cdir = vp;
28623bf9 1519 cache_drop(&onch);
21739618 1520 vrele(ovp);
28623bf9 1521 cache_zero(&nd->nl_nch);
21739618
MD
1522 } else {
1523 vrele(vp);
e24b948e 1524 }
e24b948e 1525 return (error);
9697c509
DRJ
1526}
1527
984263bc 1528/*
41c20dac
MD
1529 * chdir_args(char *path)
1530 *
984263bc
MD
1531 * Change current working directory (``.'').
1532 */
984263bc 1533int
753fd850 1534sys_chdir(struct chdir_args *uap)
984263bc 1535{
21739618 1536 struct nlookupdata nd;
9697c509 1537 int error;
984263bc 1538
21739618 1539 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e 1540 if (error == 0)
21739618 1541 error = kern_chdir(&nd);
fad57d0e 1542 nlookup_done(&nd);
9697c509 1543 return (error);
984263bc
MD
1544}
1545
1546/*
1547 * Helper function for raised chroot(2) security function: Refuse if
1548 * any filedescriptors are open directories.
1549 */
1550static int
c972a82f 1551chroot_refuse_vdir_fds(struct filedesc *fdp)
984263bc
MD
1552{
1553 struct vnode *vp;
1554 struct file *fp;
1555 int error;
1556 int fd;
1557
1558 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
5b287bba 1559 if ((error = holdvnode(fdp, fd, &fp)) != 0)
984263bc
MD
1560 continue;
1561 vp = (struct vnode *)fp->f_data;
5b287bba
MD
1562 if (vp->v_type != VDIR) {
1563 fdrop(fp);
984263bc 1564 continue;
5b287bba
MD
1565 }
1566 fdrop(fp);
984263bc
MD
1567 return(EPERM);
1568 }
1569 return (0);
1570}
1571
1572/*
1573 * This sysctl determines if we will allow a process to chroot(2) if it
1574 * has a directory open:
1575 * 0: disallowed for all processes.
1576 * 1: allowed for processes that were not already chroot(2)'ed.
1577 * 2: allowed for all processes.
1578 */
1579
1580static int chroot_allow_open_directories = 1;
1581
1582SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1583 &chroot_allow_open_directories, 0, "");
1584
1585/*
21739618
MD
1586 * chroot to the specified namecache entry. We obtain the vp from the
1587 * namecache data. The passed ncp must be locked and referenced and will
1588 * remain locked and referenced on return.
e24b948e 1589 */
53dd6631 1590int
28623bf9 1591kern_chroot(struct nchandle *nch)
e24b948e
MD
1592{
1593 struct thread *td = curthread;
1594 struct proc *p = td->td_proc;
1595 struct filedesc *fdp = p->p_fd;
21739618 1596 struct vnode *vp;
e24b948e
MD
1597 int error;
1598
1599 /*
1600 * Only root can chroot
1601 */
895c1f85 1602 if ((error = priv_check_cred(p->p_ucred, PRIV_ROOT, PRISON_ROOT)) != 0)
e24b948e
MD
1603 return (error);
1604
1605 /*
1606 * Disallow open directory descriptors (fchdir() breakouts).
1607 */
1608 if (chroot_allow_open_directories == 0 ||
1609 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1610 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1611 return (error);
1612 }
28623bf9 1613 if ((vp = nch->ncp->nc_vp) == NULL)
21739618
MD
1614 return (ENOENT);
1615
87de5057 1616 if ((error = vget(vp, LK_SHARED)) != 0)
21739618 1617 return (error);
e24b948e
MD
1618
1619 /*
1620 * Check the validity of vp as a directory to change to and
1621 * associate it with rdir/jdir.
1622 */
21739618 1623 error = checkvp_chdir(vp, td);
a11aaa81 1624 vn_unlock(vp); /* leave reference intact */
21739618 1625 if (error == 0) {
e24b948e 1626 vrele(fdp->fd_rdir);
21739618 1627 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */
28623bf9
MD
1628 cache_drop(&fdp->fd_nrdir);
1629 cache_copy(nch, &fdp->fd_nrdir);
e24b948e
MD
1630 if (fdp->fd_jdir == NULL) {
1631 fdp->fd_jdir = vp;
597aea93 1632 vref(fdp->fd_jdir);
28623bf9 1633 cache_copy(nch, &fdp->fd_njdir);
e24b948e 1634 }
21739618
MD
1635 } else {
1636 vrele(vp);
e24b948e
MD
1637 }
1638 return (error);
1639}
1640
1641/*
41c20dac
MD
1642 * chroot_args(char *path)
1643 *
984263bc
MD
1644 * Change notion of root (``/'') directory.
1645 */
984263bc
MD
1646/* ARGSUSED */
1647int
753fd850 1648sys_chroot(struct chroot_args *uap)
984263bc 1649{
dadab5e9 1650 struct thread *td = curthread;
21739618 1651 struct nlookupdata nd;
e24b948e 1652 int error;
984263bc 1653
e24b948e 1654 KKASSERT(td->td_proc);
21739618 1655 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
53dd6631
JS
1656 if (error) {
1657 nlookup_done(&nd);
1658 return(error);
1659 }
3a907475 1660 nd.nl_flags |= NLC_EXEC;
53dd6631
JS
1661 error = nlookup(&nd);
1662 if (error == 0)
28623bf9 1663 error = kern_chroot(&nd.nl_nch);
fad57d0e 1664 nlookup_done(&nd);
53dd6631 1665 return(error);
984263bc
MD
1666}
1667
1668/*
e24b948e
MD
1669 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1670 * determine whether it is legal to chdir to the vnode. The vnode's state
1671 * is not changed by this call.
984263bc 1672 */
e24b948e
MD
1673int
1674checkvp_chdir(struct vnode *vp, struct thread *td)
984263bc 1675{
984263bc
MD
1676 int error;
1677
984263bc
MD
1678 if (vp->v_type != VDIR)
1679 error = ENOTDIR;
1680 else
87de5057 1681 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred);
984263bc
MD
1682 return (error);
1683}
1684
984263bc 1685int
fad57d0e 1686kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
984263bc 1687{
dadab5e9
MD
1688 struct thread *td = curthread;
1689 struct proc *p = td->td_proc;
fbfe4e7d 1690 struct lwp *lp = td->td_lwp;
41c20dac 1691 struct filedesc *fdp = p->p_fd;
9697c509 1692 int cmode, flags;
984263bc 1693 struct file *nfp;
fad57d0e
MD
1694 struct file *fp;
1695 struct vnode *vp;
984263bc
MD
1696 int type, indx, error;
1697 struct flock lf;
984263bc 1698
984263bc
MD
1699 if ((oflags & O_ACCMODE) == O_ACCMODE)
1700 return (EINVAL);
1701 flags = FFLAGS(oflags);
fad57d0e 1702 error = falloc(p, &nfp, NULL);
984263bc
MD
1703 if (error)
1704 return (error);
1705 fp = nfp;
3a907475 1706 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
fad57d0e 1707
984263bc 1708 /*
fad57d0e
MD
1709 * XXX p_dupfd is a real mess. It allows a device to return a
1710 * file descriptor to be duplicated rather then doing the open
1711 * itself.
984263bc 1712 */
fbfe4e7d 1713 lp->lwp_dupfd = -1;
21739618
MD
1714
1715 /*
fad57d0e
MD
1716 * Call vn_open() to do the lookup and assign the vnode to the
1717 * file pointer. vn_open() does not change the ref count on fp
1718 * and the vnode, on success, will be inherited by the file pointer
1719 * and unlocked.
21739618 1720 */
fad57d0e
MD
1721 nd->nl_flags |= NLC_LOCKVP;
1722 error = vn_open(nd, fp, flags, cmode);
1723 nlookup_done(nd);
984263bc
MD
1724 if (error) {
1725 /*
984263bc
MD
1726 * handle special fdopen() case. bleh. dupfdopen() is
1727 * responsible for dropping the old contents of ofiles[indx]
1728 * if it succeeds.
fad57d0e 1729 *
259b8ea0
MD
1730 * Note that fsetfd() will add a ref to fp which represents
1731 * the fd_files[] assignment. We must still drop our
1732 * reference.
984263bc 1733 */
fbfe4e7d 1734 if ((error == ENODEV || error == ENXIO) && lp->lwp_dupfd >= 0) {
259b8ea0
MD
1735 if (fdalloc(p, 0, &indx) == 0) {
1736 error = dupfdopen(p, indx, lp->lwp_dupfd, flags, error);
fad57d0e
MD
1737 if (error == 0) {
1738 *res = indx;
9f87144f 1739 fdrop(fp); /* our ref */
fad57d0e
MD
1740 return (0);
1741 }
259b8ea0 1742 fsetfd(p, NULL, indx);
fad57d0e 1743 }
984263bc 1744 }
9f87144f 1745 fdrop(fp); /* our ref */
984263bc
MD
1746 if (error == ERESTART)
1747 error = EINTR;
1748 return (error);
1749 }
fad57d0e
MD
1750
1751 /*
1752 * ref the vnode for ourselves so it can't be ripped out from under
1753 * is. XXX need an ND flag to request that the vnode be returned
1754 * anyway.
259b8ea0
MD
1755 *
1756 * Reserve a file descriptor but do not assign it until the open
1757 * succeeds.
fad57d0e
MD
1758 */
1759 vp = (struct vnode *)fp->f_data;
1760 vref(vp);
259b8ea0 1761 if ((error = fdalloc(p, 0, &indx)) != 0) {
9f87144f 1762 fdrop(fp);
fad57d0e
MD
1763 vrele(vp);
1764 return (error);
1765 }
1766
1767 /*
1768 * If no error occurs the vp will have been assigned to the file
1769 * pointer.
1770 */
fbfe4e7d 1771 lp->lwp_dupfd = 0;
984263bc 1772
984263bc
MD
1773 if (flags & (O_EXLOCK | O_SHLOCK)) {
1774 lf.l_whence = SEEK_SET;
1775 lf.l_start = 0;
1776 lf.l_len = 0;
1777 if (flags & O_EXLOCK)
1778 lf.l_type = F_WRLCK;
1779 else
1780 lf.l_type = F_RDLCK;
71c18fe3
MD
1781 if (flags & FNONBLOCK)
1782 type = 0;
1783 else
1784 type = F_WAIT;
fad57d0e 1785
984263bc
MD
1786 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1787 /*
259b8ea0
MD
1788 * lock request failed. Clean up the reserved
1789 * descriptor.
984263bc 1790 */
fad57d0e 1791 vrele(vp);
259b8ea0 1792 fsetfd(p, NULL, indx);
9f87144f 1793 fdrop(fp);
984263bc
MD
1794 return (error);
1795 }
984263bc
MD
1796 fp->f_flag |= FHASLOCK;
1797 }
7540ab49
MD
1798#if 0
1799 /*
1800 * Assert that all regular file vnodes were created with a object.
1801 */
1802 KASSERT(vp->v_type != VREG || vp->v_object != NULL,
1803 ("open: regular file has no backing object after vn_open"));
1804#endif
984263bc 1805
fad57d0e 1806 vrele(vp);
21739618
MD
1807
1808 /*
984263bc
MD
1809 * release our private reference, leaving the one associated with the
1810 * descriptor table intact.
1811 */
259b8ea0 1812 fsetfd(p, fp, indx);
9f87144f 1813 fdrop(fp);
9697c509 1814 *res = indx;
984263bc
MD
1815 return (0);
1816}
1817
984263bc 1818/*
9697c509 1819 * open_args(char *path, int flags, int mode)
41c20dac 1820 *
9697c509
DRJ
1821 * Check permissions, allocate an open file structure,
1822 * and call the device open routine if any.
984263bc 1823 */
984263bc 1824int
753fd850 1825sys_open(struct open_args *uap)
984263bc 1826{
fad57d0e 1827 struct nlookupdata nd;
9697c509
DRJ
1828 int error;
1829
fad57d0e
MD
1830 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1831 if (error == 0) {
1832 error = kern_open(&nd, uap->flags,
1833 uap->mode, &uap->sysmsg_result);
1834 }
1835 nlookup_done(&nd);
9697c509 1836 return (error);
984263bc 1837}
984263bc 1838
984263bc 1839int
0e9b9130 1840kern_mknod(struct nlookupdata *nd, int mode, int rmajor, int rminor)
984263bc 1841{
dadab5e9
MD
1842 struct thread *td = curthread;
1843 struct proc *p = td->td_proc;
41c20dac 1844 struct vnode *vp;
984263bc
MD
1845 struct vattr vattr;
1846 int error;
1847 int whiteout = 0;
984263bc 1848
dadab5e9
MD
1849 KKASSERT(p);
1850
9697c509 1851 switch (mode & S_IFMT) {
984263bc
MD
1852 case S_IFCHR:
1853 case S_IFBLK:
895c1f85 1854 error = priv_check(td, PRIV_ROOT);
984263bc
MD
1855 break;
1856 default:
895c1f85 1857 error = priv_check_cred(p->p_ucred, PRIV_ROOT, PRISON_ROOT);
984263bc
MD
1858 break;
1859 }
1860 if (error)
1861 return (error);
fad57d0e 1862
c4df9635 1863 bwillinode(1);
5312fa43 1864 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 1865 if ((error = nlookup(nd)) != 0)
984263bc 1866 return (error);
28623bf9 1867 if (nd->nl_nch.ncp->nc_vp)
fad57d0e 1868 return (EEXIST);
28623bf9 1869 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 1870 return (error);
984263bc 1871
fad57d0e
MD
1872 VATTR_NULL(&vattr);
1873 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
0e9b9130
MD
1874 vattr.va_rmajor = rmajor;
1875 vattr.va_rminor = rminor;
fad57d0e
MD
1876 whiteout = 0;
1877
1878 switch (mode & S_IFMT) {
1879 case S_IFMT: /* used by badsect to flag bad sectors */
1880 vattr.va_type = VBAD;
1881 break;
1882 case S_IFCHR:
1883 vattr.va_type = VCHR;
1884 break;
1885 case S_IFBLK:
1886 vattr.va_type = VBLK;
1887 break;
1888 case S_IFWHT:
1889 whiteout = 1;
1890 break;
bed9e372
MD
1891 case S_IFDIR:
1892 /* special directories support for HAMMER */
1893 vattr.va_type = VDIR;
1894 break;
fad57d0e
MD
1895 default:
1896 error = EINVAL;
1897 break;
984263bc 1898 }
9697c509 1899 if (error == 0) {
fad57d0e 1900 if (whiteout) {
5312fa43
MD
1901 error = VOP_NWHITEOUT(&nd->nl_nch, nd->nl_dvp,
1902 nd->nl_cred, NAMEI_CREATE);
fad57d0e
MD
1903 } else {
1904 vp = NULL;
5312fa43
MD
1905 error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp,
1906 &vp, nd->nl_cred, &vattr);
984263bc 1907 if (error == 0)
fad57d0e 1908 vput(vp);
984263bc 1909 }
984263bc 1910 }
9697c509
DRJ
1911 return (error);
1912}
1913
1914/*
1915 * mknod_args(char *path, int mode, int dev)
1916 *
1917 * Create a special file.
1918 */
1919int
753fd850 1920sys_mknod(struct mknod_args *uap)
9697c509 1921{
fad57d0e 1922 struct nlookupdata nd;
9697c509
DRJ
1923 int error;
1924
fad57d0e 1925 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
0e9b9130
MD
1926 if (error == 0) {
1927 error = kern_mknod(&nd, uap->mode,
1928 umajor(uap->dev), uminor(uap->dev));
1929 }
fad57d0e 1930 nlookup_done(&nd);
984263bc
MD
1931 return (error);
1932}
1933
984263bc 1934int
fad57d0e 1935kern_mkfifo(struct nlookupdata *nd, int mode)
984263bc 1936{
dadab5e9
MD
1937 struct thread *td = curthread;
1938 struct proc *p = td->td_proc;
984263bc 1939 struct vattr vattr;
fad57d0e 1940 struct vnode *vp;
984263bc 1941 int error;
984263bc 1942
c4df9635 1943 bwillinode(1);
fad57d0e 1944
5312fa43 1945 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 1946 if ((error = nlookup(nd)) != 0)
984263bc 1947 return (error);
28623bf9 1948 if (nd->nl_nch.ncp->nc_vp)
984263bc 1949 return (EEXIST);
28623bf9 1950 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 1951 return (error);
fad57d0e 1952
984263bc
MD
1953 VATTR_NULL(&vattr);
1954 vattr.va_type = VFIFO;
136178b3 1955 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
fad57d0e 1956 vp = NULL;
5312fa43 1957 error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp, &vp, nd->nl_cred, &vattr);
984263bc 1958 if (error == 0)
fad57d0e 1959 vput(vp);
136178b3
DRJ
1960 return (error);
1961}
1962
1963/*
1964 * mkfifo_args(char *path, int mode)
1965 *
1966 * Create a named pipe.
1967 */
1968int
753fd850 1969sys_mkfifo(struct mkfifo_args *uap)
136178b3 1970{
fad57d0e 1971 struct nlookupdata nd;
136178b3
DRJ
1972 int error;
1973
fad57d0e
MD
1974 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1975 if (error == 0)
1976 error = kern_mkfifo(&nd, uap->mode);
1977 nlookup_done(&nd);
984263bc
MD
1978 return (error);
1979}
1980
8c40f81e 1981static int hardlink_check_uid = 0;
43a0f7ae 1982SYSCTL_INT(_security, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
8c40f81e
MD
1983 &hardlink_check_uid, 0,
1984 "Unprivileged processes cannot create hard links to files owned by other "
1985 "users");
1986static int hardlink_check_gid = 0;
43a0f7ae 1987SYSCTL_INT(_security, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
8c40f81e
MD
1988 &hardlink_check_gid, 0,
1989 "Unprivileged processes cannot create hard links to files owned by other "
1990 "groups");
1991
1992static int
1993can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
1994{
1995 struct vattr va;
1996 int error;
1997
1998 /*
1999 * Shortcut if disabled
2000 */
2001 if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
2002 return (0);
2003
2004 /*
2005 * root cred can always hardlink
2006 */
895c1f85 2007 if (priv_check_cred(cred, PRIV_ROOT, PRISON_ROOT) == 0)
8c40f81e
MD
2008 return (0);
2009
2010 /*
2011 * Otherwise only if the originating file is owned by the
2012 * same user or group. Note that any group is allowed if
2013 * the file is owned by the caller.
2014 */
87de5057 2015 error = VOP_GETATTR(vp, &va);
8c40f81e
MD
2016 if (error != 0)
2017 return (error);
2018
2019 if (hardlink_check_uid) {
2020 if (cred->cr_uid != va.va_uid)
2021 return (EPERM);
2022 }
2023
2024 if (hardlink_check_gid) {
2025 if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
2026 return (EPERM);
2027 }
2028
2029 return (0);
2030}
2031
984263bc 2032int
fad57d0e 2033kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
984263bc 2034{
dadab5e9 2035 struct thread *td = curthread;
41c20dac 2036 struct vnode *vp;
984263bc
MD
2037 int error;
2038
fad57d0e
MD
2039 /*
2040 * Lookup the source and obtained a locked vnode.
2041 *
3a907475
MD
2042 * You may only hardlink a file which you have write permission
2043 * on or which you own.
2044 *
fad57d0e
MD
2045 * XXX relookup on vget failure / race ?
2046 */
c4df9635 2047 bwillinode(1);
3a907475 2048 nd->nl_flags |= NLC_WRITE | NLC_OWN | NLC_HLINK;
fad57d0e 2049 if ((error = nlookup(nd)) != 0)
984263bc 2050 return (error);
28623bf9 2051 vp = nd->nl_nch.ncp->nc_vp;
fad57d0e 2052 KKASSERT(vp != NULL);
984263bc 2053 if (vp->v_type == VDIR)
fad57d0e 2054 return (EPERM); /* POSIX */
28623bf9 2055 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2056 return (error);
87de5057 2057 if ((error = vget(vp, LK_EXCLUSIVE)) != 0)
fad57d0e
MD
2058 return (error);
2059
2060 /*
2061 * Unlock the source so we can lookup the target without deadlocking
2062 * (XXX vp is locked already, possible other deadlock?). The target
2063 * must not exist.
2064 */
2065 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
2066 nd->nl_flags &= ~NLC_NCPISLOCKED;
28623bf9 2067 cache_unlock(&nd->nl_nch);
fad57d0e 2068
5312fa43 2069 linknd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e
MD
2070 if ((error = nlookup(linknd)) != 0) {
2071 vput(vp);
2072 return (error);
984263bc 2073 }
28623bf9 2074 if (linknd->nl_nch.ncp->nc_vp) {
fad57d0e
MD
2075 vput(vp);
2076 return (EEXIST);
2077 }
2078
2079 /*
2080 * Finally run the new API VOP.
2081 */
8c40f81e 2082 error = can_hardlink(vp, td, td->td_proc->p_ucred);
5312fa43
MD
2083 if (error == 0) {
2084 error = VOP_NLINK(&linknd->nl_nch, linknd->nl_dvp,
2085 vp, linknd->nl_cred);
2086 }
fad57d0e 2087 vput(vp);
984263bc
MD
2088 return (error);
2089}
2090
2091/*
9697c509 2092 * link_args(char *path, char *link)
41c20dac 2093 *
9697c509 2094 * Make a hard file link.
984263bc 2095 */
984263bc 2096int
753fd850 2097sys_link(struct link_args *uap)
9697c509 2098{
fad57d0e 2099 struct nlookupdata nd, linknd;
9697c509
DRJ
2100 int error;
2101
fad57d0e
MD
2102 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2103 if (error == 0) {
2104 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
2105 if (error == 0)
2106 error = kern_link(&nd, &linknd);
2107 nlookup_done(&linknd);
2108 }
2109 nlookup_done(&nd);
9697c509
DRJ
2110 return (error);
2111}
2112
2113int
fad57d0e 2114kern_symlink(struct nlookupdata *nd, char *path, int mode)
984263bc
MD
2115{
2116 struct vattr vattr;
fad57d0e 2117 struct vnode *vp;
dff430ab 2118 struct vnode *dvp;
984263bc 2119 int error;
984263bc 2120
c4df9635 2121 bwillinode(1);
5312fa43 2122 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 2123 if ((error = nlookup(nd)) != 0)
9697c509 2124 return (error);
28623bf9 2125 if (nd->nl_nch.ncp->nc_vp)
9697c509 2126 return (EEXIST);
28623bf9 2127 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2128 return (error);
5312fa43 2129 dvp = nd->nl_dvp;
984263bc 2130 VATTR_NULL(&vattr);
fad57d0e 2131 vattr.va_mode = mode;
dff430ab 2132 error = VOP_NSYMLINK(&nd->nl_nch, dvp, &vp, nd->nl_cred, &vattr, path);
984263bc 2133 if (error == 0)
fad57d0e 2134 vput(vp);
9697c509
DRJ
2135 return (error);
2136}
2137
2138/*
2139 * symlink(char *path, char *link)
2140 *
2141 * Make a symbolic link.
2142 */
2143int
753fd850 2144sys_symlink(struct symlink_args *uap)
9697c509
DRJ
2145{
2146 struct thread *td = curthread;
fad57d0e 2147 struct nlookupdata nd;
9697c509
DRJ
2148 char *path;
2149 int error;
fad57d0e 2150 int mode;
9697c509 2151
70aac194 2152 path = objcache_get(namei_oc, M_WAITOK);
9697c509 2153 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
e24b948e 2154 if (error == 0) {
fad57d0e
MD
2155 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
2156 if (error == 0) {
2157 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
2158 error = kern_symlink(&nd, path, mode);
2159 }
2160 nlookup_done(&nd);
e24b948e 2161 }
70aac194 2162 objcache_put(namei_oc, path);
984263bc
MD
2163 return (error);
2164}
2165
2166/*
41c20dac
MD
2167 * undelete_args(char *path)
2168 *
984263bc
MD
2169 * Delete a whiteout from the filesystem.
2170 */
2171/* ARGSUSED */
2172int
753fd850 2173sys_undelete(struct undelete_args *uap)
984263bc 2174{
fad57d0e 2175 struct nlookupdata nd;
984263bc 2176 int error;
984263bc 2177
ab2eb4eb 2178 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
c4df9635 2179 bwillinode(1);
5312fa43 2180 nd.nl_flags |= NLC_DELETE | NLC_REFDVP;
fad57d0e
MD
2181 if (error == 0)
2182 error = nlookup(&nd);
b8997912 2183 if (error == 0)
28623bf9 2184 error = ncp_writechk(&nd.nl_nch);
dff430ab 2185 if (error == 0) {
5312fa43
MD
2186 error = VOP_NWHITEOUT(&nd.nl_nch, nd.nl_dvp, nd.nl_cred,
2187 NAMEI_DELETE);
dff430ab 2188 }
fad57d0e 2189 nlookup_done(&nd);
984263bc
MD
2190 return (error);
2191}
2192
984263bc 2193int
fad57d0e 2194kern_unlink(struct nlookupdata *nd)
984263bc 2195{
984263bc 2196 int error;
984263bc 2197
c4df9635 2198 bwillinode(1);
5312fa43 2199 nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
fad57d0e 2200 if ((error = nlookup(nd)) != 0)
984263bc 2201 return (error);
28623bf9 2202 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2203 return (error);
5312fa43 2204 error = VOP_NREMOVE(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
984263bc
MD
2205 return (error);
2206}
2207
2208/*
9697c509 2209 * unlink_args(char *path)
41c20dac 2210 *
9697c509 2211 * Delete a name from the filesystem.
984263bc 2212 */
984263bc 2213int
753fd850 2214sys_unlink(struct unlink_args *uap)
9697c509 2215{
fad57d0e 2216 struct nlookupdata nd;
9697c509
DRJ
2217 int error;
2218
fad57d0e
MD
2219 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2220 if (error == 0)
2221 error = kern_unlink(&nd);
2222 nlookup_done(&nd);
9697c509
DRJ
2223 return (error);
2224}
2225
2226int
5a3fe67d 2227kern_lseek(int fd, off_t offset, int whence, off_t *res)
984263bc 2228{
dadab5e9
MD
2229 struct thread *td = curthread;
2230 struct proc *p = td->td_proc;
41c20dac 2231 struct file *fp;
116edfbf 2232 struct vnode *vp;
984263bc 2233 struct vattr vattr;
116edfbf 2234 off_t new_offset;
984263bc
MD
2235 int error;
2236
228b401d
MD
2237 fp = holdfp(p->p_fd, fd, -1);
2238 if (fp == NULL)
984263bc 2239 return (EBADF);
228b401d
MD
2240 if (fp->f_type != DTYPE_VNODE) {
2241 error = ESPIPE;
2242 goto done;
2243 }
116edfbf 2244 vp = (struct vnode *)fp->f_data;
228b401d 2245
9697c509 2246 switch (whence) {
984263bc 2247 case L_INCR:
116edfbf 2248 new_offset = fp->f_offset + offset;
228b401d 2249 error = 0;
984263bc
MD
2250 break;
2251 case L_XTND:
116edfbf
MD
2252 error = VOP_GETATTR(vp, &vattr);
2253 new_offset = offset + vattr.va_size;
984263bc
MD
2254 break;
2255 case L_SET:
116edfbf 2256 new_offset = offset;
228b401d 2257 error = 0;
984263bc
MD
2258 break;
2259 default:
116edfbf 2260 new_offset = 0;
228b401d
MD
2261 error = EINVAL;
2262 break;
984263bc 2263 }
116edfbf
MD
2264
2265 /*
2266 * Validate the seek position. Negative offsets are not allowed
2267 * for regular files, block specials, or directories.
2268 */
2269 if (error == 0) {
2270 if (new_offset < 0 &&
2271 (vp->v_type == VREG || vp->v_type == VDIR ||
2272 vp->v_type == VCHR || vp->v_type == VBLK)) {
2273 error = EINVAL;
b06b3b87
MD
2274 } else {
2275 fp->f_offset = new_offset;
116edfbf 2276 }
116edfbf 2277 }
9697c509 2278 *res = fp->f_offset;
228b401d
MD
2279done:
2280 fdrop(fp);
2281 return (error);
984263bc
MD
2282}
2283
984263bc 2284/*
9697c509 2285 * lseek_args(int fd, int pad, off_t offset, int whence)
41c20dac 2286 *
9697c509 2287 * Reposition read/write file offset.
984263bc 2288 */
984263bc 2289int
753fd850 2290sys_lseek(struct lseek_args *uap)
984263bc 2291{
984263bc
MD
2292 int error;
2293
9697c509 2294 error = kern_lseek(uap->fd, uap->offset, uap->whence,
116edfbf 2295 &uap->sysmsg_offset);
9697c509 2296
984263bc
MD
2297 return (error);
2298}
984263bc 2299
984263bc 2300int
fad57d0e 2301kern_access(struct nlookupdata *nd, int aflags)
984263bc 2302{
41c20dac 2303 struct vnode *vp;
984263bc 2304 int error, flags;
984263bc 2305
fad57d0e
MD
2306 if ((error = nlookup(nd)) != 0)
2307 return (error);
806dcf9a 2308retry:
28623bf9 2309 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 2310 if (error)
fad57d0e 2311 return (error);
984263bc
MD
2312
2313 /* Flags == 0 means only check for existence. */
9697c509 2314 if (aflags) {
984263bc 2315 flags = 0;
9697c509 2316 if (aflags & R_OK)
984263bc 2317 flags |= VREAD;
9697c509 2318 if (aflags & W_OK)
984263bc 2319 flags |= VWRITE;
9697c509 2320 if (aflags & X_OK)
984263bc 2321 flags |= VEXEC;
468bb1f9 2322 if ((flags & VWRITE) == 0 ||
28623bf9 2323 (error = vn_writechk(vp, &nd->nl_nch)) == 0)
87de5057 2324 error = VOP_ACCESS(vp, flags, nd->nl_cred);
806dcf9a
MD
2325
2326 /*
2327 * If the file handle is stale we have to re-resolve the
2328 * entry. This is a hack at the moment.
2329 */
2330 if (error == ESTALE) {
5e1a76ea 2331 vput(vp);
28623bf9
MD
2332 cache_setunresolved(&nd->nl_nch);
2333 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
806dcf9a 2334 if (error == 0) {
806dcf9a
MD
2335 vp = NULL;
2336 goto retry;
2337 }
5e1a76ea 2338 return(error);
806dcf9a 2339 }
984263bc 2340 }
984263bc 2341 vput(vp);
984263bc
MD
2342 return (error);
2343}
2344
984263bc 2345/*
9697c509 2346 * access_args(char *path, int flags)
41c20dac 2347 *
9697c509 2348 * Check access permissions.
984263bc 2349 */
984263bc 2350int
753fd850 2351sys_access(struct access_args *uap)
984263bc 2352{
fad57d0e 2353 struct nlookupdata nd;
9697c509 2354 int error;
984263bc 2355
fad57d0e
MD
2356 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2357 if (error == 0)
2358 error = kern_access(&nd, uap->flags);
2359 nlookup_done(&nd);
984263bc
MD
2360 return (error);
2361}
2362
984263bc 2363int
21739618 2364kern_stat(struct nlookupdata *nd, struct stat *st)
984263bc 2365{
984263bc 2366 int error;
21739618
MD
2367 struct vnode *vp;
2368 thread_t td;
984263bc 2369
21739618 2370 if ((error = nlookup(nd)) != 0)
984263bc 2371 return (error);
806dcf9a 2372again:
28623bf9 2373 if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
21739618
MD
2374 return (ENOENT);
2375
2376 td = curthread;
87de5057 2377 if ((error = vget(vp, LK_SHARED)) != 0)
21739618 2378 return (error);
87de5057 2379 error = vn_stat(vp, st, nd->nl_cred);
806dcf9a
MD
2380
2381 /*
2382 * If the file handle is stale we have to re-resolve the entry. This
2383 * is a hack at the moment.
2384 */
2385 if (error == ESTALE) {
5dc91765 2386 vput(vp);
28623bf9
MD
2387 cache_setunresolved(&nd->nl_nch);
2388 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
5dc91765 2389 if (error == 0)
806dcf9a 2390 goto again;
5dc91765
MD
2391 } else {
2392 vput(vp);
806dcf9a 2393 }
984263bc
MD
2394 return (error);
2395}
2396
2397/*
41c20dac
MD
2398 * stat_args(char *path, struct stat *ub)
2399 *
984263bc
MD
2400 * Get file status; this version follows links.
2401 */
984263bc 2402int
753fd850 2403sys_stat(struct stat_args *uap)
984263bc 2404{
21739618 2405 struct nlookupdata nd;
9697c509
DRJ
2406 struct stat st;
2407 int error;
984263bc 2408
21739618
MD
2409 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2410 if (error == 0) {
2411 error = kern_stat(&nd, &st);
2412 if (error == 0)
2413 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 2414 }
fad57d0e 2415 nlookup_done(&nd);
984263bc
MD
2416 return (error);
2417}
2418
2419/*
41c20dac
MD
2420 * lstat_args(char *path, struct stat *ub)
2421 *
984263bc
MD
2422 * Get file status; this version does not follow links.
2423 */
984263bc 2424int
753fd850 2425sys_lstat(struct lstat_args *uap)
984263bc 2426{
21739618 2427 struct nlookupdata nd;
9697c509
DRJ
2428 struct stat st;
2429 int error;
984263bc 2430
21739618
MD
2431 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2432 if (error == 0) {
2433 error = kern_stat(&nd, &st);
2434 if (error == 0)
2435 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 2436 }
fad57d0e 2437 nlookup_done(&nd);
984263bc
MD
2438 return (error);
2439}
2440
984263bc 2441/*
41c20dac
MD
2442 * pathconf_Args(char *path, int name)
2443 *
984263bc
MD
2444 * Get configurable pathname variables.
2445 */
984263bc
MD
2446/* ARGSUSED */
2447int
753fd850 2448sys_pathconf(struct pathconf_args *uap)
984263bc 2449{
fad57d0e
MD
2450 struct nlookupdata nd;
2451 struct vnode *vp;
984263bc 2452 int error;
984263bc 2453
fad57d0e 2454 vp = NULL;
ab2eb4eb 2455 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2456 if (error == 0)
2457 error = nlookup(&nd);
2458 if (error == 0)
28623bf9 2459 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
fad57d0e
MD
2460 nlookup_done(&nd);
2461 if (error == 0) {
a385a915 2462 error = VOP_PATHCONF(vp, uap->name, &uap->sysmsg_reg);
fad57d0e
MD
2463 vput(vp);
2464 }
984263bc
MD
2465 return (error);
2466}
2467
bbab27f6
DRJ
2468/*
2469 * XXX: daver
2470 * kern_readlink isn't properly split yet. There is a copyin burried
2471 * in VOP_READLINK().
2472 */
984263bc 2473int
fad57d0e 2474kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
984263bc 2475{
dadab5e9
MD
2476 struct thread *td = curthread;
2477 struct proc *p = td->td_proc;
41c20dac 2478 struct vnode *vp;
984263bc
MD
2479 struct iovec aiov;
2480 struct uio auio;
2481 int error;
984263bc 2482
fad57d0e
MD
2483 if ((error = nlookup(nd)) != 0)
2484 return (error);
28623bf9 2485 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 2486 if (error)
984263bc 2487 return (error);
fad57d0e 2488 if (vp->v_type != VLNK) {
984263bc 2489 error = EINVAL;
fad57d0e 2490 } else {
9697c509
DRJ
2491 aiov.iov_base = buf;
2492 aiov.iov_len = count;
984263bc
MD
2493 auio.uio_iov = &aiov;
2494 auio.uio_iovcnt = 1;
2495 auio.uio_offset = 0;
2496 auio.uio_rw = UIO_READ;
2497 auio.uio_segflg = UIO_USERSPACE;
dadab5e9 2498 auio.uio_td = td;
9697c509 2499 auio.uio_resid = count;
984263bc
MD
2500 error = VOP_READLINK(vp, &auio, p->p_ucred);
2501 }
2502 vput(vp);
9697c509
DRJ
2503 *res = count - auio.uio_resid;
2504 return (error);
2505}
2506
2507/*
2508 * readlink_args(char *path, char *buf, int count)
2509 *
2510 * Return target name of a symbolic link.
2511 */
2512int
753fd850 2513sys_readlink(struct readlink_args *uap)
9697c509 2514{
fad57d0e 2515 struct nlookupdata nd;
9697c509
DRJ
2516 int error;
2517
fad57d0e
MD
2518 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2519 if (error == 0) {
2520 error = kern_readlink(&nd, uap->buf, uap->count,
2521 &uap->sysmsg_result);
2522 }
2523 nlookup_done(&nd);
984263bc
MD
2524 return (error);
2525}
2526
2527static int
41c20dac 2528setfflags(struct vnode *vp, int flags)
984263bc 2529{
dadab5e9
MD
2530 struct thread *td = curthread;
2531 struct proc *p = td->td_proc;
984263bc
MD
2532 int error;
2533 struct vattr vattr;
2534
2535 /*
2536 * Prevent non-root users from setting flags on devices. When
2537 * a device is reused, users can retain ownership of the device
2538 * if they are allowed to set flags and programs assume that
2539 * chown can't fail when done as root.
2540 */
2541 if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
895c1f85 2542 ((error = priv_check_cred(p->p_ucred, PRIV_ROOT, PRISON_ROOT)) != 0))
984263bc
MD
2543 return (error);
2544
186fccd6
MD
2545 /*
2546 * note: vget is required for any operation that might mod the vnode
2547 * so VINACTIVE is properly cleared.
2548 */
87de5057 2549 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2550 VATTR_NULL(&vattr);
2551 vattr.va_flags = flags;
87de5057 2552 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2553 vput(vp);
2554 }
984263bc
MD
2555 return (error);
2556}
2557
2558/*
41c20dac
MD
2559 * chflags(char *path, int flags)
2560 *
984263bc
MD
2561 * Change flags of a file given a path name.
2562 */
984263bc
MD
2563/* ARGSUSED */
2564int
753fd850 2565sys_chflags(struct chflags_args *uap)
984263bc 2566{
fad57d0e
MD
2567 struct nlookupdata nd;
2568 struct vnode *vp;
984263bc 2569 int error;
984263bc 2570
fad57d0e 2571 vp = NULL;
ab2eb4eb 2572 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2573 if (error == 0)
2574 error = nlookup(&nd);
2575 if (error == 0)
28623bf9 2576 error = ncp_writechk(&nd.nl_nch);
72310cfb 2577 if (error == 0)
28623bf9 2578 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
fad57d0e
MD
2579 nlookup_done(&nd);
2580 if (error == 0) {
ab2eb4eb 2581 error = setfflags(vp, uap->flags);
fad57d0e
MD
2582 vrele(vp);
2583 }
2584 return (error);
984263bc
MD
2585}
2586
2587/*
91fdc001
PA
2588 * lchflags(char *path, int flags)
2589 *
2590 * Change flags of a file given a path name, but don't follow symlinks.
2591 */
2592/* ARGSUSED */
2593int
2594sys_lchflags(struct lchflags_args *uap)
2595{
2596 struct nlookupdata nd;
2597 struct vnode *vp;
2598 int error;
2599
2600 vp = NULL;
2601 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
91fdc001
PA
2602 if (error == 0)
2603 error = nlookup(&nd);
2604 if (error == 0)
2605 error = ncp_writechk(&nd.nl_nch);
2606 if (error == 0)
2607 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
2608 nlookup_done(&nd);
2609 if (error == 0) {
2610 error = setfflags(vp, uap->flags);
2611 vrele(vp);
2612 }
2613 return (error);
2614}
2615
2616/*
41c20dac
MD
2617 * fchflags_args(int fd, int flags)
2618 *
984263bc
MD
2619 * Change flags of a file given a file descriptor.
2620 */
984263bc
MD
2621/* ARGSUSED */
2622int
753fd850 2623sys_fchflags(struct fchflags_args *uap)
984263bc 2624{
dadab5e9
MD
2625 struct thread *td = curthread;
2626 struct proc *p = td->td_proc;
984263bc
MD
2627 struct file *fp;
2628 int error;
2629
5b287bba 2630 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2631 return (error);
28623bf9
MD
2632 if (fp->f_nchandle.ncp)
2633 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
2634 if (error == 0)
2635 error = setfflags((struct vnode *) fp->f_data, uap->flags);
5b287bba
MD
2636 fdrop(fp);
2637 return (error);
984263bc
MD
2638}
2639
2640static int
41c20dac 2641setfmode(struct vnode *vp, int mode)
984263bc 2642{
dadab5e9
MD
2643 struct thread *td = curthread;
2644 struct proc *p = td->td_proc;
984263bc
MD
2645 int error;
2646 struct vattr vattr;
2647
186fccd6
MD
2648 /*
2649 * note: vget is required for any operation that might mod the vnode
2650 * so VINACTIVE is properly cleared.
2651 */
87de5057 2652 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2653 VATTR_NULL(&vattr);
2654 vattr.va_mode = mode & ALLPERMS;
87de5057 2655 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2656 vput(vp);
2657 }
984263bc
MD
2658 return error;
2659}
2660
9697c509 2661int
fad57d0e 2662kern_chmod(struct nlookupdata *nd, int mode)
9697c509 2663{
fad57d0e 2664 struct vnode *vp;
9697c509
DRJ
2665 int error;
2666
fad57d0e 2667 if ((error = nlookup(nd)) != 0)
9697c509 2668 return (error);
28623bf9 2669 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 2670 return (error);
28623bf9 2671 if ((error = ncp_writechk(&nd->nl_nch)) == 0)
72310cfb 2672 error = setfmode(vp, mode);
fad57d0e
MD
2673 vrele(vp);
2674 return (error);
9697c509
DRJ
2675}
2676
984263bc 2677/*
41c20dac
MD
2678 * chmod_args(char *path, int mode)
2679 *
984263bc
MD
2680 * Change mode of a file given path name.
2681 */
984263bc
MD
2682/* ARGSUSED */
2683int
753fd850 2684sys_chmod(struct chmod_args *uap)
984263bc 2685{
fad57d0e 2686 struct nlookupdata nd;
9697c509 2687 int error;
984263bc 2688
fad57d0e
MD
2689 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2690 if (error == 0)
2691 error = kern_chmod(&nd, uap->mode);
2692 nlookup_done(&nd);
9697c509 2693 return (error);
984263bc
MD
2694}
2695
2696/*
41c20dac
MD
2697 * lchmod_args(char *path, int mode)
2698 *
984263bc
MD
2699 * Change mode of a file given path name (don't follow links.)
2700 */
984263bc
MD
2701/* ARGSUSED */
2702int
753fd850 2703sys_lchmod(struct lchmod_args *uap)
984263bc 2704{
fad57d0e 2705 struct nlookupdata nd;
984263bc 2706 int error;
984263bc 2707
fad57d0e
MD
2708 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2709 if (error == 0)
2710 error = kern_chmod(&nd, uap->mode);
2711 nlookup_done(&nd);
2712 return (error);
984263bc
MD
2713}
2714
2715/*
41c20dac
MD
2716 * fchmod_args(int fd, int mode)
2717 *
984263bc
MD
2718 * Change mode of a file given a file descriptor.
2719 */
984263bc
MD
2720/* ARGSUSED */
2721int
753fd850 2722sys_fchmod(struct fchmod_args *uap)
984263bc 2723{
dadab5e9
MD
2724 struct thread *td = curthread;
2725 struct proc *p = td->td_proc;
984263bc
MD
2726 struct file *fp;
2727 int error;
2728
5b287bba 2729 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2730 return (error);
28623bf9
MD
2731 if (fp->f_nchandle.ncp)
2732 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
2733 if (error == 0)
2734 error = setfmode((struct vnode *)fp->f_data, uap->mode);
5b287bba
MD
2735 fdrop(fp);
2736 return (error);
984263bc
MD
2737}
2738
2739static int
41c20dac 2740setfown(struct vnode *vp, uid_t uid, gid_t gid)
984263bc 2741{
dadab5e9
MD
2742 struct thread *td = curthread;
2743 struct proc *p = td->td_proc;
984263bc
MD
2744 int error;
2745 struct vattr vattr;
2746
186fccd6
MD
2747 /*
2748 * note: vget is required for any operation that might mod the vnode
2749 * so VINACTIVE is properly cleared.
2750 */
87de5057 2751 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2752 VATTR_NULL(&vattr);
2753 vattr.va_uid = uid;
2754 vattr.va_gid = gid;
87de5057 2755 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2756 vput(vp);
2757 }
984263bc
MD
2758 return error;
2759}
2760
9697c509 2761int
fad57d0e 2762kern_chown(struct nlookupdata *nd, int uid, int gid)
9697c509 2763{
fad57d0e 2764 struct vnode *vp;
9697c509
DRJ
2765 int error;
2766
fad57d0e 2767 if ((error = nlookup(nd)) != 0)
9697c509 2768 return (error);
28623bf9 2769 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 2770 return (error);
28623bf9 2771 if ((error = ncp_writechk(&nd->nl_nch)) == 0)
72310cfb 2772 error = setfown(vp, uid, gid);
fad57d0e 2773 vrele(vp);
9697c509
DRJ
2774 return (error);
2775}
2776
984263bc 2777/*
41c20dac
MD
2778 * chown(char *path, int uid, int gid)
2779 *
984263bc
MD
2780 * Set ownership given a path name.
2781 */
984263bc 2782int
753fd850 2783sys_chown(struct chown_args *uap)
984263bc 2784{
fad57d0e 2785 struct nlookupdata nd;
9697c509
DRJ
2786 int error;
2787
fad57d0e
MD
2788 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2789 if (error == 0)
2790 error = kern_chown(&nd, uap->uid, uap->gid);
2791 nlookup_done(&nd);
984263bc
MD
2792 return (error);
2793}
2794
2795/*
41c20dac
MD
2796 * lchown_args(char *path, int uid, int gid)
2797 *
984263bc
MD
2798 * Set ownership given a path name, do not cross symlinks.
2799 */
984263bc 2800int
753fd850 2801sys_lchown(struct lchown_args *uap)
984263bc 2802{
fad57d0e 2803 struct nlookupdata nd;
984263bc 2804 int error;
9697c509 2805
fad57d0e
MD
2806 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2807 if (error == 0)
2808 error = kern_chown(&nd, uap->uid, uap->gid);
2809 nlookup_done(&nd);
984263bc
MD
2810 return (error);
2811}
2812
2813/*
41c20dac
MD
2814 * fchown_args(int fd, int uid, int gid)
2815 *
984263bc
MD
2816 * Set ownership given a file descriptor.
2817 */
984263bc
MD
2818/* ARGSUSED */
2819int
753fd850 2820sys_fchown(struct fchown_args *uap)
984263bc 2821{
dadab5e9
MD
2822 struct thread *td = curthread;
2823 struct proc *p = td->td_proc;
984263bc
MD
2824 struct file *fp;
2825 int error;
2826
5b287bba 2827 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2828 return (error);
28623bf9
MD
2829 if (fp->f_nchandle.ncp)
2830 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
2831 if (error == 0)
2832 error = setfown((struct vnode *)fp->f_data, uap->uid, uap->gid);
5b287bba
MD
2833 fdrop(fp);
2834 return (error);
984263bc
MD
2835}
2836
2837static int
9697c509 2838getutimes(const struct timeval *tvp, struct timespec *tsp)
984263bc
MD
2839{
2840 struct timeval tv[2];
984263bc 2841
9697c509 2842 if (tvp == NULL) {
984263bc
MD
2843 microtime(&tv[0]);
2844 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2845 tsp[1] = tsp[0];
2846 } else {
9697c509
DRJ
2847 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2848 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
984263bc
MD
2849 }
2850 return 0;
2851}
2852
2853static int
3a907475
MD
2854setutimes(struct vnode *vp, struct vattr *vattr,
2855 const struct timespec *ts, int nullflag)
984263bc 2856{
dadab5e9
MD
2857 struct thread *td = curthread;
2858 struct proc *p = td->td_proc;
984263bc 2859 int error;
984263bc 2860
3a907475
MD
2861 VATTR_NULL(vattr);
2862 vattr->va_atime = ts[0];
2863 vattr->va_mtime = ts[1];
2864 if (nullflag)
2865 vattr->va_vaflags |= VA_UTIMES_NULL;
2866 error = VOP_SETATTR(vp, vattr, p->p_ucred);
2867
984263bc
MD
2868 return error;
2869}
2870
9697c509 2871int
fad57d0e 2872kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
9697c509
DRJ
2873{
2874 struct timespec ts[2];
fad57d0e 2875 struct vnode *vp;
3a907475 2876 struct vattr vattr;
9697c509
DRJ
2877 int error;
2878
fad57d0e 2879 if ((error = getutimes(tptr, ts)) != 0)
9697c509 2880 return (error);
3a907475
MD
2881
2882 /*
2883 * NOTE: utimes() succeeds for the owner even if the file
2884 * is not user-writable.
2885 */
2886 nd->nl_flags |= NLC_OWN | NLC_WRITE;
2887
fad57d0e 2888 if ((error = nlookup(nd)) != 0)
9697c509 2889 return (error);
28623bf9 2890 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2891 return (error);
28623bf9 2892 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 2893 return (error);
ee89633d
MD
2894
2895 /*
3a907475
MD
2896 * note: vget is required for any operation that might mod the vnode
2897 * so VINACTIVE is properly cleared.
ee89633d 2898 */
3a907475
MD
2899 if ((error = vn_writechk(vp, &nd->nl_nch)) == 0) {
2900 error = vget(vp, LK_EXCLUSIVE);
2901 if (error == 0) {
2902 error = setutimes(vp, &vattr, ts, (tptr == NULL));
2903 vput(vp);
2904 }
918101d3 2905 }
fad57d0e 2906 vrele(vp);
9697c509
DRJ
2907 return (error);
2908}
2909
984263bc 2910/*
41c20dac
MD
2911 * utimes_args(char *path, struct timeval *tptr)
2912 *
984263bc
MD
2913 * Set the access and modification times of a file.
2914 */
984263bc 2915int
753fd850 2916sys_utimes(struct utimes_args *uap)
984263bc 2917{
9697c509 2918 struct timeval tv[2];
fad57d0e 2919 struct nlookupdata nd;
9697c509
DRJ
2920 int error;
2921
2922 if (uap->tptr) {
2923 error = copyin(uap->tptr, tv, sizeof(tv));
2924 if (error)
2925 return (error);
2926 }
fad57d0e
MD
2927 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2928 if (error == 0)
2929 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2930 nlookup_done(&nd);
984263bc
MD
2931 return (error);
2932}
2933
2934/*
41c20dac
MD
2935 * lutimes_args(char *path, struct timeval *tptr)
2936 *
984263bc
MD
2937 * Set the access and modification times of a file.
2938 */
984263bc 2939int
753fd850 2940sys_lutimes(struct lutimes_args *uap)
984263bc 2941{
9697c509 2942 struct timeval tv[2];
fad57d0e 2943 struct nlookupdata nd;
9697c509
DRJ
2944 int error;
2945
2946 if (uap->tptr) {
2947 error = copyin(uap->tptr, tv, sizeof(tv));
2948 if (error)
2949 return (error);
2950 }
fad57d0e
MD
2951 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2952 if (error == 0)
2953 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2954 nlookup_done(&nd);
984263bc
MD
2955 return (error);
2956}
2957
3a907475
MD
2958/*
2959 * Set utimes on a file descriptor. The creds used to open the
2960 * file are used to determine whether the operation is allowed
2961 * or not.
2962 */
984263bc 2963int
9697c509 2964kern_futimes(int fd, struct timeval *tptr)
984263bc 2965{
dadab5e9
MD
2966 struct thread *td = curthread;
2967 struct proc *p = td->td_proc;
984263bc
MD
2968 struct timespec ts[2];
2969 struct file *fp;
3a907475
MD
2970 struct vnode *vp;
2971 struct vattr vattr;
984263bc
MD
2972 int error;
2973
9697c509
DRJ
2974 error = getutimes(tptr, ts);
2975 if (error)
984263bc 2976 return (error);
5b287bba 2977 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc 2978 return (error);
28623bf9
MD
2979 if (fp->f_nchandle.ncp)
2980 error = ncp_writechk(&fp->f_nchandle);
3a907475
MD
2981 if (error == 0) {
2982 vp = fp->f_data;
2983 error = vget(vp, LK_EXCLUSIVE);
2984 if (error == 0) {
2985 error = VOP_GETATTR(vp, &vattr);
2986 if (error == 0) {
2987 error = naccess_va(&vattr, NLC_OWN | NLC_WRITE,
2988 fp->f_cred);
2989 }
2990 if (error == 0) {
2991 error = setutimes(vp, &vattr, ts,
2992 (tptr == NULL));
2993 }
2994 vput(vp);
2995 }
2996 }
5b287bba 2997 fdrop(fp);
9697c509 2998 return (error);
984263bc
MD
2999}
3000
3001/*
9697c509 3002 * futimes_args(int fd, struct timeval *tptr)
41c20dac 3003 *
9697c509 3004 * Set the access and modification times of a file.
984263bc 3005 */
984263bc 3006int
753fd850 3007sys_futimes(struct futimes_args *uap)
9697c509
DRJ
3008{
3009 struct timeval tv[2];
3010 int error;
3011
3012 if (uap->tptr) {
3013 error = copyin(uap->tptr, tv, sizeof(tv));
3014 if (error)
3015 return (error);
3016 }
3017
b651c485 3018 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
9697c509
DRJ
3019
3020 return (error);
3021}
3022
3023int
fad57d0e 3024kern_truncate(struct nlookupdata *nd, off_t length)
984263bc 3025{
41c20dac 3026 struct vnode *vp;
984263bc
MD
3027 struct vattr vattr;
3028 int error;
984263bc 3029
9697c509 3030 if (length < 0)
984263bc 3031 return(EINVAL);
3a907475 3032 nd->nl_flags |= NLC_WRITE | NLC_TRUNCATE;
fad57d0e 3033 if ((error = nlookup(nd)) != 0)
984263bc 3034 return (error);
28623bf9 3035 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 3036 return (error);
28623bf9 3037 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 3038 return (error);
ca466bae 3039 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) != 0) {
fad57d0e
MD
3040 vrele(vp);
3041 return (error);
3042 }
3043 if (vp->v_type == VDIR) {
984263bc 3044 error = EISDIR;
3a907475 3045 } else if ((error = vn_writechk(vp, &nd->nl_nch)) == 0) {
984263bc 3046 VATTR_NULL(&vattr);
9697c509 3047 vattr.va_size = length;
87de5057 3048 error = VOP_SETATTR(vp, &vattr, nd->nl_cred);
984263bc
MD
3049 }
3050 vput(vp);
3051 return (error);
3052}
3053
9697c509
DRJ
3054/*
3055 * truncate(char *path, int pad, off_t length)
3056 *
3057 * Truncate a file given its path name.
3058 */
3059int
753fd850 3060sys_truncate(struct truncate_args *uap)
9697c509 3061{
fad57d0e 3062 struct nlookupdata nd;
9697c509
DRJ
3063 int error;
3064
fad57d0e
MD
3065 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3066 if (error == 0)
3067 error = kern_truncate(&nd, uap->length);
3068 nlookup_done(&nd);
9697c509
DRJ
3069 return error;
3070}
3071
984263bc 3072int
8f6f8622 3073kern_ftruncate(int fd, off_t length)
984263bc 3074{
dadab5e9
MD
3075 struct thread *td = curthread;
3076 struct proc *p = td->td_proc;
984263bc
MD
3077 struct vattr vattr;
3078 struct vnode *vp;
3079 struct file *fp;
3080 int error;
3081
8f6f8622 3082 if (length < 0)
984263bc 3083 return(EINVAL);
5b287bba 3084 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc 3085 return (error);
28623bf9
MD
3086 if (fp->f_nchandle.ncp) {
3087 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
3088 if (error)
3089 goto done;
3090 }
5b287bba
MD
3091 if ((fp->f_flag & FWRITE) == 0) {
3092 error = EINVAL;
3093 goto done;
3094 }
3a907475
MD
3095 if (fp->f_flag & FAPPENDONLY) { /* inode was set s/uapnd */
3096 error = EINVAL;
3097 goto done;
3098 }
984263bc 3099 vp = (struct vnode *)fp->f_data;
ca466bae 3100 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
5b287bba 3101 if (vp->v_type == VDIR) {
984263bc 3102 error = EISDIR;
468bb1f9 3103 } else if ((error = vn_writechk(vp, NULL)) == 0) {
984263bc 3104 VATTR_NULL(&vattr);
8f6f8622 3105 vattr.va_size = length;
87de5057 3106 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
984263bc 3107 }
a11aaa81 3108 vn_unlock(vp);
5b287bba
MD
3109done:
3110 fdrop(fp);
984263bc
MD
3111 return (error);
3112}
3113
984263bc 3114/*
8f6f8622 3115 * ftruncate_args(int fd, int pad, off_t length)
41c20dac 3116 *
8f6f8622 3117 * Truncate a file given a file descriptor.
984263bc 3118 */
984263bc 3119int
753fd850 3120sys_ftruncate(struct ftruncate_args *uap)
984263bc 3121{
8f6f8622 3122 int error;
984263bc 3123
8f6f8622
DRJ
3124 error = kern_ftruncate(uap->fd, uap->length);
3125
3126 return (error);
984263bc
MD
3127}
3128
984263bc 3129/*
41c20dac
MD
3130 * fsync(int fd)
3131 *
984263bc
MD
3132 * Sync an open file.
3133 */
984263bc
MD
3134/* ARGSUSED */
3135int
753fd850 3136sys_fsync(struct fsync_args *uap)
984263bc 3137{
dadab5e9
MD
3138 struct thread *td = curthread;
3139 struct proc *p = td->td_proc;
41c20dac 3140 struct vnode *vp;
984263bc
MD
3141 struct file *fp;
3142 vm_object_t obj;
3143 int error;
3144
5b287bba 3145 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc
MD
3146 return (error);
3147 vp = (struct vnode *)fp->f_data;
ca466bae 3148 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
7540ab49 3149 if ((obj = vp->v_object) != NULL)
984263bc 3150 vm_object_page_clean(obj, 0, 0, 0);
408357d8
MD
3151 if ((error = VOP_FSYNC(vp, MNT_WAIT)) == 0 && vp->v_mount)
3152 error = buf_fsync(vp);
a11aaa81 3153 vn_unlock(vp);
5b287bba 3154 fdrop(fp);
984263bc
MD
3155 return (error);
3156}
3157
984263bc 3158int
fad57d0e 3159kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
984263bc 3160{
28623bf9
MD
3161 struct nchandle fnchd;
3162 struct nchandle tnchd;
fad57d0e 3163 struct namecache *ncp;
dff430ab
MD
3164 struct vnode *fdvp;
3165 struct vnode *tdvp;
fad57d0e 3166 struct mount *mp;
984263bc
MD
3167 int error;
3168
c4df9635 3169 bwillinode(1);
3a907475 3170 fromnd->nl_flags |= NLC_REFDVP | NLC_RENAME_SRC;
fad57d0e 3171 if ((error = nlookup(fromnd)) != 0)
984263bc 3172 return (error);
28623bf9 3173 if ((fnchd.ncp = fromnd->nl_nch.ncp->nc_parent) == NULL)
fad57d0e 3174 return (ENOENT);
28623bf9
MD
3175 fnchd.mount = fromnd->nl_nch.mount;
3176 cache_hold(&fnchd);
fad57d0e
MD
3177
3178 /*
28623bf9 3179 * unlock the source nch so we can lookup the target nch without
fad57d0e
MD
3180 * deadlocking. The target may or may not exist so we do not check
3181 * for a target vp like kern_mkdir() and other creation functions do.
3182 *
3183 * The source and target directories are ref'd and rechecked after
3184 * everything is relocked to determine if the source or target file
3185 * has been renamed.
3186 */
3187 KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
3188 fromnd->nl_flags &= ~NLC_NCPISLOCKED;
28623bf9 3189 cache_unlock(&fromnd->nl_nch);
fad57d0e 3190
3a907475 3191 tond->nl_flags |= NLC_RENAME_DST | NLC_REFDVP;
fad57d0e 3192 if ((error = nlookup(tond)) != 0) {
28623bf9 3193 cache_drop(&fnchd);
fad57d0e
MD
3194 return (error);
3195 }
28623bf9
MD
3196 if ((tnchd.ncp = tond->nl_nch.ncp->nc_parent) == NULL) {
3197 cache_drop(&fnchd);
fad57d0e
MD
3198 return (ENOENT);
3199 }
28623bf9
MD
3200 tnchd.mount = tond->nl_nch.mount;
3201 cache_hold(&tnchd);
fad57d0e
MD
3202
3203 /*
3204 * If the source and target are the same there is nothing to do
3205 */
28623bf9
MD
3206 if (fromnd->nl_nch.ncp == tond->nl_nch.ncp) {
3207 cache_drop(&fnchd);
3208 cache_drop(&tnchd);
fad57d0e
MD
3209 return (0);
3210 }
3211
3212 /*
1d505369
MD
3213 * Mount points cannot be renamed or overwritten
3214 */
28623bf9
MD
3215 if ((fromnd->nl_nch.ncp->nc_flag | tond->nl_nch.ncp->nc_flag) &
3216 NCF_ISMOUNTPT
1d505369 3217 ) {
28623bf9
MD
3218 cache_drop(&fnchd);
3219 cache_drop(&tnchd);
1d505369
MD
3220 return (EINVAL);
3221 }
3222
3223 /*
aa40af33
MD
3224 * relock the source ncp. NOTE AFTER RELOCKING: the source ncp
3225 * may have become invalid while it was unlocked, nc_vp and nc_mount
3226 * could be NULL.
fad57d0e 3227 */