TRIM support
[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 $
40 */
41
984263bc
MD
42#include <sys/param.h>
43#include <sys/systm.h>
44#include <sys/buf.h>
16df2c74 45#include <sys/conf.h>
984263bc
MD
46#include <sys/sysent.h>
47#include <sys/malloc.h>
48#include <sys/mount.h>
432b8263 49#include <sys/mountctl.h>
984263bc 50#include <sys/sysproto.h>
984263bc
MD
51#include <sys/filedesc.h>
52#include <sys/kernel.h>
53#include <sys/fcntl.h>
54#include <sys/file.h>
55#include <sys/linker.h>
56#include <sys/stat.h>
57#include <sys/unistd.h>
58#include <sys/vnode.h>
59#include <sys/proc.h>
895c1f85 60#include <sys/priv.h>
9658f145 61#include <sys/jail.h>
dadab5e9 62#include <sys/namei.h>
21739618 63#include <sys/nlookup.h>
984263bc
MD
64#include <sys/dirent.h>
65#include <sys/extattr.h>
8fa76237 66#include <sys/spinlock.h>
8f6f8622 67#include <sys/kern_syscall.h>
70aac194 68#include <sys/objcache.h>
984263bc 69#include <sys/sysctl.h>
408357d8
MD
70
71#include <sys/buf2.h>
28623bf9
MD
72#include <sys/file2.h>
73#include <sys/spinlock2.h>
684a93c4 74#include <sys/mplock2.h>
28623bf9 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
c6b20585
SW
85static void mount_warning(struct mount *mp, const char *ctl, ...)
86 __printflike(2, 3);
d3c546e6 87static int mount_path(struct proc *p, struct mount *mp, char **rb, char **fb);
e24b948e 88static int checkvp_chdir (struct vnode *vn, struct thread *td);
77480a97 89static void checkdirs (struct nchandle *old_nch, struct nchandle *new_nch);
402ed7e1 90static int chroot_refuse_vdir_fds (struct filedesc *fdp);
75ffff0d 91static int chroot_visible_mnt(struct mount *mp, struct proc *p);
402ed7e1
RG
92static int getutimes (const struct timeval *, struct timespec *);
93static int setfown (struct vnode *, uid_t, gid_t);
94static int setfmode (struct vnode *, int);
95static int setfflags (struct vnode *, int);
3a907475
MD
96static int setutimes (struct vnode *, struct vattr *,
97 const struct timespec *, int);
984263bc
MD
98static int usermount = 0; /* if 1, non-root can mount fs. */
99
402ed7e1 100int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
984263bc 101
093e85dc
SG
102SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
103 "Allow non-root users to mount filesystems");
984263bc
MD
104
105/*
106 * Virtual File System System Calls
107 */
108
109/*
110 * Mount a file system.
3919ced0 111 *
41c20dac 112 * mount_args(char *type, char *path, int flags, caddr_t data)
3919ced0
MD
113 *
114 * MPALMOSTSAFE
41c20dac 115 */
984263bc 116int
753fd850 117sys_mount(struct mount_args *uap)
984263bc 118{
dadab5e9 119 struct thread *td = curthread;
984263bc 120 struct vnode *vp;
28623bf9 121 struct nchandle nch;
8b02b69a 122 struct mount *mp, *nullmp;
984263bc
MD
123 struct vfsconf *vfsp;
124 int error, flag = 0, flag2 = 0;
1d505369 125 int hasmount;
984263bc 126 struct vattr va;
21739618 127 struct nlookupdata nd;
984263bc 128 char fstypename[MFSNAMELEN];
3919ced0 129 struct ucred *cred;
984263bc 130
3919ced0 131 get_mplock();
9910d07b 132 cred = td->td_ucred;
3919ced0
MD
133 if (jailed(cred)) {
134 error = EPERM;
135 goto done;
136 }
895c1f85 137 if (usermount == 0 && (error = priv_check(td, PRIV_ROOT)))
3919ced0
MD
138 goto done;
139
984263bc
MD
140 /*
141 * Do not allow NFS export by non-root users.
142 */
ab2eb4eb 143 if (uap->flags & MNT_EXPORTED) {
895c1f85 144 error = priv_check(td, PRIV_ROOT);
984263bc 145 if (error)
3919ced0 146 goto done;
984263bc
MD
147 }
148 /*
149 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
150 */
895c1f85 151 if (priv_check(td, PRIV_ROOT))
ab2eb4eb 152 uap->flags |= MNT_NOSUID | MNT_NODEV;
21739618 153
984263bc 154 /*
28623bf9 155 * Lookup the requested path and extract the nch and vnode.
984263bc 156 */
21739618
MD
157 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
158 if (error == 0) {
159 if ((error = nlookup(&nd)) == 0) {
28623bf9 160 if (nd.nl_nch.ncp->nc_vp == NULL)
21739618
MD
161 error = ENOENT;
162 }
163 }
164 if (error) {
165 nlookup_done(&nd);
3919ced0 166 goto done;
21739618
MD
167 }
168
169 /*
8b02b69a
SK
170 * If the target filesystem is resolved via a nullfs mount, then
171 * nd.nl_nch.mount will be pointing to the nullfs mount structure
172 * instead of the target file system. We need it in case we are
173 * doing an update.
174 */
175 nullmp = nd.nl_nch.mount;
176
177 /*
21739618
MD
178 * Extract the locked+refd ncp and cleanup the nd structure
179 */
28623bf9
MD
180 nch = nd.nl_nch;
181 cache_zero(&nd.nl_nch);
21739618
MD
182 nlookup_done(&nd);
183
28623bf9 184 if ((nch.ncp->nc_flag & NCF_ISMOUNTPT) && cache_findmount(&nch))
1d505369
MD
185 hasmount = 1;
186 else
187 hasmount = 0;
188
189
21739618 190 /*
28623bf9 191 * now we have the locked ref'd nch and unreferenced vnode.
21739618 192 */
28623bf9 193 vp = nch.ncp->nc_vp;
87de5057 194 if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
28623bf9 195 cache_put(&nch);
3919ced0 196 goto done;
21739618 197 }
28623bf9 198 cache_unlock(&nch);
21739618
MD
199
200 /*
8b02b69a
SK
201 * Extract the file system type. We need to know this early, to take
202 * appropriate actions if we are dealing with a nullfs.
203 */
204 if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
205 cache_drop(&nch);
206 vput(vp);
3919ced0 207 goto done;
8b02b69a
SK
208 }
209
210 /*
28623bf9 211 * Now we have an unlocked ref'd nch and a locked ref'd vp
21739618 212 */
ab2eb4eb 213 if (uap->flags & MNT_UPDATE) {
67863d04 214 if ((vp->v_flag & (VROOT|VPFSROOT)) == 0) {
28623bf9 215 cache_drop(&nch);
984263bc 216 vput(vp);
3919ced0
MD
217 error = EINVAL;
218 goto done;
984263bc 219 }
8b02b69a
SK
220
221 if (strncmp(fstypename, "null", 5) == 0) {
222 KKASSERT(nullmp);
223 mp = nullmp;
224 } else {
225 mp = vp->v_mount;
226 }
227
984263bc
MD
228 flag = mp->mnt_flag;
229 flag2 = mp->mnt_kern_flag;
230 /*
231 * We only allow the filesystem to be reloaded if it
232 * is currently mounted read-only.
233 */
ab2eb4eb 234 if ((uap->flags & MNT_RELOAD) &&
984263bc 235 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
28623bf9 236 cache_drop(&nch);
984263bc 237 vput(vp);
3919ced0
MD
238 error = EOPNOTSUPP; /* Needs translation */
239 goto done;
984263bc
MD
240 }
241 /*
242 * Only root, or the user that did the original mount is
243 * permitted to update it.
244 */
acde96db 245 if (mp->mnt_stat.f_owner != cred->cr_uid &&
895c1f85 246 (error = priv_check(td, PRIV_ROOT))) {
28623bf9 247 cache_drop(&nch);
984263bc 248 vput(vp);
3919ced0 249 goto done;
984263bc 250 }
f9642f56 251 if (vfs_busy(mp, LK_NOWAIT)) {
28623bf9 252 cache_drop(&nch);
984263bc 253 vput(vp);
3919ced0
MD
254 error = EBUSY;
255 goto done;
984263bc 256 }
1d505369 257 if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
28623bf9 258 cache_drop(&nch);
f9642f56 259 vfs_unbusy(mp);
984263bc 260 vput(vp);
3919ced0
MD
261 error = EBUSY;
262 goto done;
984263bc 263 }
2247fe02 264 vsetflags(vp, VMOUNT);
984263bc 265 mp->mnt_flag |=
ab2eb4eb 266 uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
a11aaa81 267 vn_unlock(vp);
984263bc
MD
268 goto update;
269 }
270 /*
271 * If the user is not root, ensure that they own the directory
272 * onto which we are attempting to mount.
273 */
87de5057 274 if ((error = VOP_GETATTR(vp, &va)) ||
895c1f85 275 (va.va_uid != cred->cr_uid && (error = priv_check(td, PRIV_ROOT)))) {
28623bf9 276 cache_drop(&nch);
984263bc 277 vput(vp);
3919ced0 278 goto done;
984263bc 279 }
87de5057 280 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
28623bf9 281 cache_drop(&nch);
984263bc 282 vput(vp);
3919ced0 283 goto done;
984263bc
MD
284 }
285 if (vp->v_type != VDIR) {
28623bf9 286 cache_drop(&nch);
984263bc 287 vput(vp);
3919ced0
MD
288 error = ENOTDIR;
289 goto done;
984263bc 290 }
5dc91765
MD
291 if (vp->v_mount->mnt_kern_flag & MNTK_NOSTKMNT) {
292 cache_drop(&nch);
293 vput(vp);
3919ced0
MD
294 error = EPERM;
295 goto done;
5dc91765 296 }
2613053d 297 vfsp = vfsconf_find_by_name(fstypename);
984263bc
MD
298 if (vfsp == NULL) {
299 linker_file_t lf;
300
301 /* Only load modules for root (very important!) */
895c1f85 302 if ((error = priv_check(td, PRIV_ROOT)) != 0) {
28623bf9 303 cache_drop(&nch);
984263bc 304 vput(vp);
3919ced0 305 goto done;
984263bc 306 }
1c0e3286 307 error = linker_load_file(fstypename, &lf);
984263bc 308 if (error || lf == NULL) {
28623bf9 309 cache_drop(&nch);
984263bc
MD
310 vput(vp);
311 if (lf == NULL)
312 error = ENODEV;
3919ced0 313 goto done;
984263bc
MD
314 }
315 lf->userrefs++;
316 /* lookup again, see if the VFS was loaded */
2613053d 317 vfsp = vfsconf_find_by_name(fstypename);
984263bc
MD
318 if (vfsp == NULL) {
319 lf->userrefs--;
320 linker_file_unload(lf);
28623bf9 321 cache_drop(&nch);
984263bc 322 vput(vp);
3919ced0
MD
323 error = ENODEV;
324 goto done;
984263bc
MD
325 }
326 }
1d505369 327 if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
28623bf9 328 cache_drop(&nch);
984263bc 329 vput(vp);
3919ced0
MD
330 error = EBUSY;
331 goto done;
984263bc 332 }
2247fe02 333 vsetflags(vp, VMOUNT);
984263bc
MD
334
335 /*
336 * Allocate and initialize the filesystem.
337 */
efda3bd0 338 mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
aac0aabd 339 mount_init(mp);
f9642f56 340 vfs_busy(mp, LK_NOWAIT);
984263bc
MD
341 mp->mnt_op = vfsp->vfc_vfsops;
342 mp->mnt_vfc = vfsp;
343 vfsp->vfc_refcount++;
344 mp->mnt_stat.f_type = vfsp->vfc_typenum;
345 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
346 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
acde96db 347 mp->mnt_stat.f_owner = cred->cr_uid;
a11aaa81 348 vn_unlock(vp);
984263bc
MD
349update:
350 /*
351 * Set the mount level flags.
352 */
ab2eb4eb 353 if (uap->flags & MNT_RDONLY)
984263bc
MD
354 mp->mnt_flag |= MNT_RDONLY;
355 else if (mp->mnt_flag & MNT_RDONLY)
356 mp->mnt_kern_flag |= MNTK_WANTRDWR;
357 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
358 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
e0fb398b 359 MNT_NOSYMFOLLOW | MNT_IGNORE | MNT_TRIM |
984263bc 360 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
ab2eb4eb 361 mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
984263bc 362 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
e0fb398b 363 MNT_NOSYMFOLLOW | MNT_IGNORE | MNT_TRIM |
984263bc
MD
364 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
365 /*
366 * Mount the filesystem.
367 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
fad57d0e 368 * get.
984263bc 369 */
acde96db 370 error = VFS_MOUNT(mp, uap->path, uap->data, cred);
984263bc
MD
371 if (mp->mnt_flag & MNT_UPDATE) {
372 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
373 mp->mnt_flag &= ~MNT_RDONLY;
374 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
375 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
376 if (error) {
377 mp->mnt_flag = flag;
378 mp->mnt_kern_flag = flag2;
379 }
f9642f56 380 vfs_unbusy(mp);
2247fe02 381 vclrflags(vp, VMOUNT);
984263bc 382 vrele(vp);
28623bf9 383 cache_drop(&nch);
3919ced0 384 goto done;
984263bc 385 }
ca466bae 386 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc 387 /*
21739618 388 * Put the new filesystem on the mount list after root. The mount
28623bf9
MD
389 * point gets its own mnt_ncmountpt (unless the VFS already set one
390 * up) which represents the root of the mount. The lookup code
391 * detects the mount point going forward and checks the root of
392 * the mount going backwards.
8e005a45
MD
393 *
394 * It is not necessary to invalidate or purge the vnode underneath
395 * because elements under the mount will be given their own glue
396 * namecache record.
984263bc 397 */
984263bc 398 if (!error) {
28623bf9
MD
399 if (mp->mnt_ncmountpt.ncp == NULL) {
400 /*
401 * allocate, then unlock, but leave the ref intact
402 */
403 cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
404 cache_unlock(&mp->mnt_ncmountpt);
405 }
406 mp->mnt_ncmounton = nch; /* inherits ref */
407 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
408
409 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
2247fe02 410 vclrflags(vp, VMOUNT);
861905fb 411 mountlist_insert(mp, MNTINS_LAST);
a11aaa81 412 vn_unlock(vp);
c0721f5f 413 checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
41a01a4d 414 error = vfs_allocate_syncvnode(mp);
f9642f56 415 vfs_unbusy(mp);
1d505369
MD
416 error = VFS_START(mp, 0);
417 vrele(vp);
984263bc 418 } else {
66a1ddf5
MD
419 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
420 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
421 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
422 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
423 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
2247fe02 424 vclrflags(vp, VMOUNT);
984263bc 425 mp->mnt_vfc->vfc_refcount--;
f9642f56 426 vfs_unbusy(mp);
efda3bd0 427 kfree(mp, M_MOUNT);
28623bf9 428 cache_drop(&nch);
984263bc
MD
429 vput(vp);
430 }
3919ced0
MD
431done:
432 rel_mplock();
984263bc
MD
433 return (error);
434}
435
436/*
437 * Scan all active processes to see if any of them have a current
438 * or root directory onto which the new filesystem has just been
439 * mounted. If so, replace them with the new mount point.
21739618
MD
440 *
441 * The passed ncp is ref'd and locked (from the mount code) and
442 * must be associated with the vnode representing the root of the
443 * mount point.
984263bc 444 */
8fa76237 445struct checkdirs_info {
77480a97
MD
446 struct nchandle old_nch;
447 struct nchandle new_nch;
448 struct vnode *old_vp;
449 struct vnode *new_vp;
8fa76237
MD
450};
451
452static int checkdirs_callback(struct proc *p, void *data);
453
984263bc 454static void
77480a97 455checkdirs(struct nchandle *old_nch, struct nchandle *new_nch)
984263bc 456{
8fa76237 457 struct checkdirs_info info;
77480a97 458 struct vnode *olddp;
984263bc 459 struct vnode *newdp;
690a3127 460 struct mount *mp;
984263bc 461
77480a97
MD
462 /*
463 * If the old mount point's vnode has a usecount of 1, it is not
464 * being held as a descriptor anywhere.
465 */
466 olddp = old_nch->ncp->nc_vp;
3c37c940 467 if (olddp == NULL || olddp->v_sysref.refcnt == 1)
984263bc 468 return;
77480a97
MD
469
470 /*
471 * Force the root vnode of the new mount point to be resolved
472 * so we can update any matching processes.
473 */
474 mp = new_nch->mount;
690a3127 475 if (VFS_ROOT(mp, &newdp))
984263bc 476 panic("mount: lost mount");
77480a97
MD
477 cache_setunresolved(new_nch);
478 cache_setvp(new_nch, newdp);
21739618 479
77480a97
MD
480 /*
481 * Special handling of the root node
482 */
21739618
MD
483 if (rootvnode == olddp) {
484 vref(newdp);
77480a97 485 vfs_cache_setroot(newdp, cache_hold(new_nch));
21739618
MD
486 }
487
77480a97
MD
488 /*
489 * Pass newdp separately so the callback does not have to access
490 * it via new_nch->ncp->nc_vp.
491 */
492 info.old_nch = *old_nch;
493 info.new_nch = *new_nch;
494 info.new_vp = newdp;
8fa76237
MD
495 allproc_scan(checkdirs_callback, &info);
496 vput(newdp);
497}
498
499/*
500 * NOTE: callback is not MP safe because the scanned process's filedesc
501 * structure can be ripped out from under us, amoung other things.
502 */
503static int
504checkdirs_callback(struct proc *p, void *data)
505{
506 struct checkdirs_info *info = data;
507 struct filedesc *fdp;
28623bf9
MD
508 struct nchandle ncdrop1;
509 struct nchandle ncdrop2;
8fa76237
MD
510 struct vnode *vprele1;
511 struct vnode *vprele2;
512
513 if ((fdp = p->p_fd) != NULL) {
28623bf9
MD
514 cache_zero(&ncdrop1);
515 cache_zero(&ncdrop2);
8fa76237
MD
516 vprele1 = NULL;
517 vprele2 = NULL;
518
519 /*
520 * MPUNSAFE - XXX fdp can be pulled out from under a
521 * foreign process.
522 *
523 * A shared filedesc is ok, we don't have to copy it
524 * because we are making this change globally.
525 */
287a8577 526 spin_lock(&fdp->fd_spin);
77480a97
MD
527 if (fdp->fd_ncdir.mount == info->old_nch.mount &&
528 fdp->fd_ncdir.ncp == info->old_nch.ncp) {
8fa76237 529 vprele1 = fdp->fd_cdir;
77480a97
MD
530 vref(info->new_vp);
531 fdp->fd_cdir = info->new_vp;
8fa76237 532 ncdrop1 = fdp->fd_ncdir;
77480a97 533 cache_copy(&info->new_nch, &fdp->fd_ncdir);
984263bc 534 }
77480a97
MD
535 if (fdp->fd_nrdir.mount == info->old_nch.mount &&
536 fdp->fd_nrdir.ncp == info->old_nch.ncp) {
8fa76237 537 vprele2 = fdp->fd_rdir;
77480a97
MD
538 vref(info->new_vp);
539 fdp->fd_rdir = info->new_vp;
8fa76237 540 ncdrop2 = fdp->fd_nrdir;
77480a97 541 cache_copy(&info->new_nch, &fdp->fd_nrdir);
984263bc 542 }
287a8577 543 spin_unlock(&fdp->fd_spin);
28623bf9
MD
544 if (ncdrop1.ncp)
545 cache_drop(&ncdrop1);
546 if (ncdrop2.ncp)
547 cache_drop(&ncdrop2);
8fa76237
MD
548 if (vprele1)
549 vrele(vprele1);
550 if (vprele2)
551 vrele(vprele2);
984263bc 552 }
8fa76237 553 return(0);
984263bc
MD
554}
555
556/*
557 * Unmount a file system.
558 *
559 * Note: unmount takes a path to the vnode mounted on as argument,
560 * not special file (as before).
3919ced0 561 *
41c20dac 562 * umount_args(char *path, int flags)
3919ced0
MD
563 *
564 * MPALMOSTSAFE
41c20dac 565 */
984263bc 566int
753fd850 567sys_unmount(struct unmount_args *uap)
984263bc 568{
dadab5e9 569 struct thread *td = curthread;
f64b567c 570 struct proc *p __debugvar = td->td_proc;
d2293868 571 struct mount *mp = NULL;
fad57d0e 572 struct nlookupdata nd;
3919ced0 573 int error;
984263bc 574
dadab5e9 575 KKASSERT(p);
3919ced0 576 get_mplock();
9910d07b 577 if (td->td_ucred->cr_prison != NULL) {
3919ced0
MD
578 error = EPERM;
579 goto done;
580 }
895c1f85 581 if (usermount == 0 && (error = priv_check(td, PRIV_ROOT)))
3919ced0 582 goto done;
84c3a710 583
ab2eb4eb 584 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
585 if (error == 0)
586 error = nlookup(&nd);
fad57d0e 587 if (error)
d2293868 588 goto out;
fad57d0e 589
28623bf9 590 mp = nd.nl_nch.mount;
984263bc
MD
591
592 /*
593 * Only root, or the user that did the original mount is
594 * permitted to unmount this filesystem.
595 */
9910d07b 596 if ((mp->mnt_stat.f_owner != td->td_ucred->cr_uid) &&
895c1f85 597 (error = priv_check(td, PRIV_ROOT)))
d2293868 598 goto out;
984263bc
MD
599
600 /*
601 * Don't allow unmounting the root file system.
602 */
603 if (mp->mnt_flag & MNT_ROOTFS) {
d2293868
MD
604 error = EINVAL;
605 goto out;
984263bc
MD
606 }
607
608 /*
609 * Must be the root of the filesystem
610 */
28623bf9 611 if (nd.nl_nch.ncp != mp->mnt_ncmountpt.ncp) {
d2293868
MD
612 error = EINVAL;
613 goto out;
984263bc 614 }
d2293868
MD
615
616out:
617 nlookup_done(&nd);
3919ced0
MD
618 if (error == 0)
619 error = dounmount(mp, uap->flags);
620done:
621 rel_mplock();
622 return (error);
984263bc
MD
623}
624
625/*
626 * Do the actual file system unmount.
627 */
861905fb
MD
628static int
629dounmount_interlock(struct mount *mp)
630{
631 if (mp->mnt_kern_flag & MNTK_UNMOUNT)
632 return (EBUSY);
633 mp->mnt_kern_flag |= MNTK_UNMOUNT;
634 return(0);
635}
636
32b7c8fd
AH
637static int
638unmount_allproc_cb(struct proc *p, void *arg)
639{
640 struct mount *mp;
641
642 if (p->p_textnch.ncp == NULL)
643 return 0;
644
645 mp = (struct mount *)arg;
646 if (p->p_textnch.mount == mp)
647 cache_drop(&p->p_textnch);
648
649 return 0;
650}
651
984263bc 652int
acde96db 653dounmount(struct mount *mp, int flags)
984263bc 654{
dcf26f79 655 struct namecache *ncp;
28623bf9 656 struct nchandle nch;
2ec4b00d 657 struct vnode *vp;
984263bc
MD
658 int error;
659 int async_flag;
2df9419c 660 int lflags;
28623bf9 661 int freeok = 1;
984263bc 662
861905fb
MD
663 /*
664 * Exclusive access for unmounting purposes
665 */
666 if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
667 return (error);
668
669 /*
670 * Allow filesystems to detect that a forced unmount is in progress.
671 */
984263bc
MD
672 if (flags & MNT_FORCE)
673 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
2df9419c 674 lflags = LK_EXCLUSIVE | ((flags & MNT_FORCE) ? 0 : LK_NOWAIT);
df4f70a6 675 error = lockmgr(&mp->mnt_lock, lflags);
984263bc
MD
676 if (error) {
677 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
678 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 679 wakeup(mp);
984263bc
MD
680 return (error);
681 }
682
683 if (mp->mnt_flag & MNT_EXPUBLIC)
684 vfs_setpublicfs(NULL, NULL, NULL);
685
686 vfs_msync(mp, MNT_WAIT);
687 async_flag = mp->mnt_flag & MNT_ASYNC;
688 mp->mnt_flag &=~ MNT_ASYNC;
dcf26f79
MD
689
690 /*
28623bf9
MD
691 * If this filesystem isn't aliasing other filesystems,
692 * try to invalidate any remaining namecache entries and
693 * check the count afterwords.
dcf26f79 694 */
28623bf9
MD
695 if ((mp->mnt_kern_flag & MNTK_NCALIASED) == 0) {
696 cache_lock(&mp->mnt_ncmountpt);
697 cache_inval(&mp->mnt_ncmountpt, CINV_DESTROY|CINV_CHILDREN);
698 cache_unlock(&mp->mnt_ncmountpt);
699
700 if ((ncp = mp->mnt_ncmountpt.ncp) != NULL &&
701 (ncp->nc_refs != 1 || TAILQ_FIRST(&ncp->nc_list))) {
32b7c8fd
AH
702 allproc_scan(&unmount_allproc_cb, mp);
703 }
704
705 if ((ncp = mp->mnt_ncmountpt.ncp) != NULL &&
706 (ncp->nc_refs != 1 || TAILQ_FIRST(&ncp->nc_list))) {
dcf26f79
MD
707
708 if ((flags & MNT_FORCE) == 0) {
709 error = EBUSY;
28623bf9
MD
710 mount_warning(mp, "Cannot unmount: "
711 "%d namecache "
712 "references still "
713 "present",
714 ncp->nc_refs - 1);
dcf26f79 715 } else {
28623bf9
MD
716 mount_warning(mp, "Forced unmount: "
717 "%d namecache "
718 "references still "
719 "present",
720 ncp->nc_refs - 1);
721 freeok = 0;
dcf26f79
MD
722 }
723 }
724 }
725
28623bf9
MD
726 /*
727 * nchandle records ref the mount structure. Expect a count of 1
728 * (our mount->mnt_ncmountpt).
729 */
730 if (mp->mnt_refs != 1) {
731 if ((flags & MNT_FORCE) == 0) {
732 mount_warning(mp, "Cannot unmount: "
733 "%d process references still "
734 "present", mp->mnt_refs);
735 error = EBUSY;
736 } else {
737 mount_warning(mp, "Forced unmount: "
738 "%d process references still "
739 "present", mp->mnt_refs);
740 freeok = 0;
741 }
742 }
743
2ec4b00d
MD
744 /*
745 * Decomission our special mnt_syncer vnode. This also stops
746 * the vnlru code. If we are unable to unmount we recommission
747 * the vnode.
748 */
dcf26f79 749 if (error == 0) {
2ec4b00d
MD
750 if ((vp = mp->mnt_syncer) != NULL) {
751 mp->mnt_syncer = NULL;
752 vrele(vp);
753 }
dcf26f79
MD
754 if (((mp->mnt_flag & MNT_RDONLY) ||
755 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) ||
756 (flags & MNT_FORCE)) {
757 error = VFS_UNMOUNT(mp, flags);
758 }
759 }
984263bc 760 if (error) {
41a01a4d
MD
761 if (mp->mnt_syncer == NULL)
762 vfs_allocate_syncvnode(mp);
984263bc
MD
763 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
764 mp->mnt_flag |= async_flag;
df4f70a6 765 lockmgr(&mp->mnt_lock, LK_RELEASE);
984263bc 766 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 767 wakeup(mp);
984263bc
MD
768 return (error);
769 }
432b8263
MD
770 /*
771 * Clean up any journals still associated with the mount after
772 * filesystem activity has ceased.
773 */
774 journal_remove_all_journals(mp,
775 ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));
776
861905fb 777 mountlist_remove(mp);
0961aa92
MD
778
779 /*
780 * Remove any installed vnode ops here so the individual VFSs don't
781 * have to.
782 */
66a1ddf5
MD
783 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
784 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
785 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
786 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
787 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
0961aa92 788
28623bf9
MD
789 if (mp->mnt_ncmountpt.ncp != NULL) {
790 nch = mp->mnt_ncmountpt;
791 cache_zero(&mp->mnt_ncmountpt);
792 cache_clrmountpt(&nch);
793 cache_drop(&nch);
794 }
795 if (mp->mnt_ncmounton.ncp != NULL) {
796 nch = mp->mnt_ncmounton;
797 cache_zero(&mp->mnt_ncmounton);
798 cache_clrmountpt(&nch);
799 cache_drop(&nch);
984263bc 800 }
1d505369 801
984263bc
MD
802 mp->mnt_vfc->vfc_refcount--;
803 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
804 panic("unmount: dangling vnode");
df4f70a6 805 lockmgr(&mp->mnt_lock, LK_RELEASE);
984263bc 806 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 807 wakeup(mp);
28623bf9
MD
808 if (freeok)
809 kfree(mp, M_MOUNT);
984263bc
MD
810 return (0);
811}
812
28623bf9
MD
813static
814void
815mount_warning(struct mount *mp, const char *ctl, ...)
816{
817 char *ptr;
818 char *buf;
819 __va_list va;
820
821 __va_start(va, ctl);
5b4cfb7e 822 if (cache_fullpath(NULL, &mp->mnt_ncmounton, &ptr, &buf, 0) == 0) {
6ea70f76 823 kprintf("unmount(%s): ", ptr);
379210cb 824 kvprintf(ctl, va);
6ea70f76 825 kprintf("\n");
28623bf9
MD
826 kfree(buf, M_TEMP);
827 } else {
56bcacad
MD
828 kprintf("unmount(%p", mp);
829 if (mp->mnt_ncmounton.ncp && mp->mnt_ncmounton.ncp->nc_name)
830 kprintf(",%s", mp->mnt_ncmounton.ncp->nc_name);
831 kprintf("): ");
379210cb 832 kvprintf(ctl, va);
6ea70f76 833 kprintf("\n");
28623bf9
MD
834 }
835 __va_end(va);
836}
837
984263bc 838/*
d3c546e6
MD
839 * Shim cache_fullpath() to handle the case where a process is chrooted into
840 * a subdirectory of a mount. In this case if the root mount matches the
841 * process root directory's mount we have to specify the process's root
842 * directory instead of the mount point, because the mount point might
843 * be above the root directory.
844 */
845static
846int
847mount_path(struct proc *p, struct mount *mp, char **rb, char **fb)
848{
849 struct nchandle *nch;
d3c546e6
MD
850
851 if (p && p->p_fd->fd_nrdir.mount == mp)
852 nch = &p->p_fd->fd_nrdir;
853 else
854 nch = &mp->mnt_ncmountpt;
5b4cfb7e 855 return(cache_fullpath(p, nch, rb, fb, 0));
d3c546e6
MD
856}
857
858/*
984263bc
MD
859 * Sync each mounted filesystem.
860 */
984263bc
MD
861
862#ifdef DEBUG
863static int syncprt = 0;
864SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
6bdbb368 865#endif /* DEBUG */
984263bc 866
861905fb
MD
867static int sync_callback(struct mount *mp, void *data);
868
984263bc 869int
753fd850 870sys_sync(struct sync_args *uap)
984263bc 871{
861905fb 872 mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
6bdbb368 873#ifdef DEBUG
861905fb
MD
874 /*
875 * print out buffer pool stat information on each sync() call.
876 */
984263bc
MD
877 if (syncprt)
878 vfs_bufstats();
6bdbb368 879#endif /* DEBUG */
984263bc
MD
880 return (0);
881}
882
861905fb
MD
883static
884int
885sync_callback(struct mount *mp, void *data __unused)
886{
887 int asyncflag;
888
889 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
890 asyncflag = mp->mnt_flag & MNT_ASYNC;
891 mp->mnt_flag &= ~MNT_ASYNC;
892 vfs_msync(mp, MNT_NOWAIT);
28271622 893 VFS_SYNC(mp, MNT_NOWAIT | MNT_LAZY);
861905fb
MD
894 mp->mnt_flag |= asyncflag;
895 }
896 return(0);
897}
898
984263bc
MD
899/* XXX PRISON: could be per prison flag */
900static int prison_quotas;
901#if 0
902SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
903#endif
904
905/*
41c20dac
MD
906 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
907 *
984263bc 908 * Change filesystem quotas.
3919ced0
MD
909 *
910 * MPALMOSTSAFE
984263bc 911 */
984263bc 912int
753fd850 913sys_quotactl(struct quotactl_args *uap)
984263bc 914{
fad57d0e
MD
915 struct nlookupdata nd;
916 struct thread *td;
41c20dac 917 struct mount *mp;
984263bc 918 int error;
984263bc 919
3919ced0 920 get_mplock();
fad57d0e 921 td = curthread;
9910d07b 922 if (td->td_ucred->cr_prison && !prison_quotas) {
3919ced0
MD
923 error = EPERM;
924 goto done;
925 }
fad57d0e 926
ab2eb4eb 927 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
928 if (error == 0)
929 error = nlookup(&nd);
930 if (error == 0) {
28623bf9 931 mp = nd.nl_nch.mount;
ab2eb4eb 932 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
acde96db 933 uap->arg, nd.nl_cred);
fad57d0e
MD
934 }
935 nlookup_done(&nd);
3919ced0
MD
936done:
937 rel_mplock();
fad57d0e 938 return (error);
984263bc
MD
939}
940
949ecb9b 941/*
2281065e 942 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
949ecb9b
MD
943 * void *buf, int buflen)
944 *
945 * This function operates on a mount point and executes the specified
946 * operation using the specified control data, and possibly returns data.
947 *
948 * The actual number of bytes stored in the result buffer is returned, 0
949 * if none, otherwise an error is returned.
3919ced0
MD
950 *
951 * MPALMOSTSAFE
949ecb9b 952 */
949ecb9b 953int
753fd850 954sys_mountctl(struct mountctl_args *uap)
949ecb9b
MD
955{
956 struct thread *td = curthread;
957 struct proc *p = td->td_proc;
2281065e 958 struct file *fp;
949ecb9b
MD
959 void *ctl = NULL;
960 void *buf = NULL;
961 char *path = NULL;
962 int error;
963
964 /*
965 * Sanity and permissions checks. We must be root.
966 */
967 KKASSERT(p);
9910d07b 968 if (td->td_ucred->cr_prison != NULL)
949ecb9b 969 return (EPERM);
dad088a5
MD
970 if ((uap->op != MOUNTCTL_MOUNTFLAGS) &&
971 (error = priv_check(td, PRIV_ROOT)) != 0)
949ecb9b
MD
972 return (error);
973
974 /*
975 * Argument length checks
976 */
39b13188 977 if (uap->ctllen < 0 || uap->ctllen > 1024)
949ecb9b 978 return (EINVAL);
39b13188 979 if (uap->buflen < 0 || uap->buflen > 16 * 1024)
949ecb9b
MD
980 return (EINVAL);
981 if (uap->path == NULL)
982 return (EINVAL);
983
984 /*
985 * Allocate the necessary buffers and copyin data
986 */
70aac194 987 path = objcache_get(namei_oc, M_WAITOK);
949ecb9b
MD
988 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
989 if (error)
990 goto done;
991
992 if (uap->ctllen) {
efda3bd0 993 ctl = kmalloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
949ecb9b
MD
994 error = copyin(uap->ctl, ctl, uap->ctllen);
995 if (error)
996 goto done;
997 }
998 if (uap->buflen)
efda3bd0 999 buf = kmalloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
949ecb9b
MD
1000
1001 /*
2281065e
MD
1002 * Validate the descriptor
1003 */
ecda6326
MD
1004 if (uap->fd >= 0) {
1005 fp = holdfp(p->p_fd, uap->fd, -1);
1006 if (fp == NULL) {
1007 error = EBADF;
1008 goto done;
1009 }
1010 } else {
1011 fp = NULL;
2281065e 1012 }
2281065e
MD
1013
1014 /*
949ecb9b
MD
1015 * Execute the internal kernel function and clean up.
1016 */
3919ced0 1017 get_mplock();
2281065e 1018 error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
3919ced0 1019 rel_mplock();
2281065e 1020 if (fp)
9f87144f 1021 fdrop(fp);
949ecb9b
MD
1022 if (error == 0 && uap->sysmsg_result > 0)
1023 error = copyout(buf, uap->buf, uap->sysmsg_result);
1024done:
1025 if (path)
70aac194 1026 objcache_put(namei_oc, path);
949ecb9b 1027 if (ctl)
efda3bd0 1028 kfree(ctl, M_TEMP);
949ecb9b 1029 if (buf)
efda3bd0 1030 kfree(buf, M_TEMP);
949ecb9b
MD
1031 return (error);
1032}
1033
1034/*
1035 * Execute a mount control operation by resolving the path to a mount point
1036 * and calling vop_mountctl().
67863d04
MD
1037 *
1038 * Use the mount point from the nch instead of the vnode so nullfs mounts
1039 * can properly spike the VOP.
949ecb9b
MD
1040 */
1041int
2281065e
MD
1042kern_mountctl(const char *path, int op, struct file *fp,
1043 const void *ctl, int ctllen,
949ecb9b
MD
1044 void *buf, int buflen, int *res)
1045{
949ecb9b
MD
1046 struct vnode *vp;
1047 struct mount *mp;
1048 struct nlookupdata nd;
1049 int error;
1050
1051 *res = 0;
1052 vp = NULL;
1053 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
1054 if (error == 0)
1055 error = nlookup(&nd);
1056 if (error == 0)
28623bf9 1057 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
67863d04 1058 mp = nd.nl_nch.mount;
949ecb9b
MD
1059 nlookup_done(&nd);
1060 if (error)
1061 return (error);
d246c910 1062 vn_unlock(vp);
949ecb9b 1063
949ecb9b
MD
1064 /*
1065 * Must be the root of the filesystem
1066 */
67863d04 1067 if ((vp->v_flag & (VROOT|VPFSROOT)) == 0) {
d246c910 1068 vrele(vp);
949ecb9b
MD
1069 return (EINVAL);
1070 }
aac0aabd
MD
1071 error = vop_mountctl(mp->mnt_vn_use_ops, vp, op, fp, ctl, ctllen,
1072 buf, buflen, res);
d246c910 1073 vrele(vp);
949ecb9b
MD
1074 return (error);
1075}
1076
984263bc 1077int
fad57d0e 1078kern_statfs(struct nlookupdata *nd, struct statfs *buf)
984263bc 1079{
dadab5e9 1080 struct thread *td = curthread;
75ffff0d 1081 struct proc *p = td->td_proc;
41c20dac
MD
1082 struct mount *mp;
1083 struct statfs *sp;
75ffff0d 1084 char *fullpath, *freepath;
984263bc 1085 int error;
984263bc 1086
fad57d0e 1087 if ((error = nlookup(nd)) != 0)
984263bc 1088 return (error);
28623bf9 1089 mp = nd->nl_nch.mount;
984263bc 1090 sp = &mp->mnt_stat;
acde96db 1091 if ((error = VFS_STATFS(mp, sp, nd->nl_cred)) != 0)
984263bc 1092 return (error);
75ffff0d 1093
d3c546e6 1094 error = mount_path(p, mp, &fullpath, &freepath);
75ffff0d
JS
1095 if (error)
1096 return(error);
1097 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1098 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
efda3bd0 1099 kfree(freepath, M_TEMP);
75ffff0d 1100
984263bc 1101 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509
DRJ
1102 bcopy(sp, buf, sizeof(*buf));
1103 /* Only root should have access to the fsid's. */
895c1f85 1104 if (priv_check(td, PRIV_ROOT))
9697c509
DRJ
1105 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
1106 return (0);
984263bc
MD
1107}
1108
1109/*
9697c509 1110 * statfs_args(char *path, struct statfs *buf)
41c20dac 1111 *
984263bc
MD
1112 * Get filesystem statistics.
1113 */
984263bc 1114int
753fd850 1115sys_statfs(struct statfs_args *uap)
9697c509 1116{
fad57d0e 1117 struct nlookupdata nd;
9697c509
DRJ
1118 struct statfs buf;
1119 int error;
1120
fad57d0e
MD
1121 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1122 if (error == 0)
1123 error = kern_statfs(&nd, &buf);
1124 nlookup_done(&nd);
9697c509
DRJ
1125 if (error == 0)
1126 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1127 return (error);
1128}
1129
1130int
1131kern_fstatfs(int fd, struct statfs *buf)
984263bc 1132{
dadab5e9
MD
1133 struct thread *td = curthread;
1134 struct proc *p = td->td_proc;
984263bc
MD
1135 struct file *fp;
1136 struct mount *mp;
1fd87d54 1137 struct statfs *sp;
75ffff0d 1138 char *fullpath, *freepath;
984263bc 1139 int error;
984263bc 1140
dadab5e9 1141 KKASSERT(p);
5b287bba 1142 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc 1143 return (error);
92734e3d
MD
1144
1145 /*
1146 * Try to use mount info from any overlays rather than the
1147 * mount info for the underlying vnode, otherwise we will
1148 * fail when operating on null-mounted paths inside a chroot.
1149 */
1150 if ((mp = fp->f_nchandle.mount) == NULL)
1151 mp = ((struct vnode *)fp->f_data)->v_mount;
5b287bba
MD
1152 if (mp == NULL) {
1153 error = EBADF;
1154 goto done;
1155 }
1156 if (fp->f_cred == NULL) {
1157 error = EINVAL;
1158 goto done;
1159 }
984263bc 1160 sp = &mp->mnt_stat;
5b287bba
MD
1161 if ((error = VFS_STATFS(mp, sp, fp->f_cred)) != 0)
1162 goto done;
75ffff0d 1163
d3c546e6 1164 if ((error = mount_path(p, mp, &fullpath, &freepath)) != 0)
5b287bba 1165 goto done;
75ffff0d
JS
1166 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1167 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
efda3bd0 1168 kfree(freepath, M_TEMP);
75ffff0d 1169
984263bc 1170 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509 1171 bcopy(sp, buf, sizeof(*buf));
75ffff0d 1172
9697c509 1173 /* Only root should have access to the fsid's. */
895c1f85 1174 if (priv_check(td, PRIV_ROOT))
9697c509 1175 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
5b287bba
MD
1176 error = 0;
1177done:
1178 fdrop(fp);
1179 return (error);
9697c509
DRJ
1180}
1181
1182/*
1183 * fstatfs_args(int fd, struct statfs *buf)
1184 *
1185 * Get filesystem statistics.
1186 */
1187int
753fd850 1188sys_fstatfs(struct fstatfs_args *uap)
9697c509
DRJ
1189{
1190 struct statfs buf;
1191 int error;
1192
1193 error = kern_fstatfs(uap->fd, &buf);
1194
1195 if (error == 0)
1196 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1197 return (error);
984263bc
MD
1198}
1199
d9fad06e
MD
1200int
1201kern_statvfs(struct nlookupdata *nd, struct statvfs *buf)
1202{
1203 struct mount *mp;
1204 struct statvfs *sp;
1205 int error;
1206
1207 if ((error = nlookup(nd)) != 0)
1208 return (error);
1209 mp = nd->nl_nch.mount;
1210 sp = &mp->mnt_vstat;
1211 if ((error = VFS_STATVFS(mp, sp, nd->nl_cred)) != 0)
1212 return (error);
1213
1214 sp->f_flag = 0;
1215 if (mp->mnt_flag & MNT_RDONLY)
1216 sp->f_flag |= ST_RDONLY;
1217 if (mp->mnt_flag & MNT_NOSUID)
1218 sp->f_flag |= ST_NOSUID;
1219 bcopy(sp, buf, sizeof(*buf));
1220 return (0);
1221}
1222
1223/*
1224 * statfs_args(char *path, struct statfs *buf)
1225 *
1226 * Get filesystem statistics.
1227 */
1228int
1229sys_statvfs(struct statvfs_args *uap)
1230{
1231 struct nlookupdata nd;
1232 struct statvfs buf;
1233 int error;
1234
1235 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1236 if (error == 0)
1237 error = kern_statvfs(&nd, &buf);
1238 nlookup_done(&nd);
1239 if (error == 0)
1240 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1241 return (error);
1242}
1243
1244int
1245kern_fstatvfs(int fd, struct statvfs *buf)
1246{
1247 struct thread *td = curthread;
1248 struct proc *p = td->td_proc;
1249 struct file *fp;
1250 struct mount *mp;
1251 struct statvfs *sp;
1252 int error;
1253
1254 KKASSERT(p);
1255 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
1256 return (error);
92734e3d
MD
1257 if ((mp = fp->f_nchandle.mount) == NULL)
1258 mp = ((struct vnode *)fp->f_data)->v_mount;
d9fad06e
MD
1259 if (mp == NULL) {
1260 error = EBADF;
1261 goto done;
1262 }
1263 if (fp->f_cred == NULL) {
1264 error = EINVAL;
1265 goto done;
1266 }
1267 sp = &mp->mnt_vstat;
1268 if ((error = VFS_STATVFS(mp, sp, fp->f_cred)) != 0)
1269 goto done;
1270
1271 sp->f_flag = 0;
1272 if (mp->mnt_flag & MNT_RDONLY)
1273 sp->f_flag |= ST_RDONLY;
1274 if (mp->mnt_flag & MNT_NOSUID)
1275 sp->f_flag |= ST_NOSUID;
1276
1277 bcopy(sp, buf, sizeof(*buf));
1278 error = 0;
1279done:
1280 fdrop(fp);
1281 return (error);
1282}
1283
1284/*
1285 * fstatfs_args(int fd, struct statfs *buf)
1286 *
1287 * Get filesystem statistics.
1288 */
1289int
1290sys_fstatvfs(struct fstatvfs_args *uap)
1291{
1292 struct statvfs buf;
1293 int error;
1294
1295 error = kern_fstatvfs(uap->fd, &buf);
1296
1297 if (error == 0)
1298 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1299 return (error);
1300}
1301
984263bc 1302/*
41c20dac
MD
1303 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
1304 *
984263bc
MD
1305 * Get statistics on all filesystems.
1306 */
861905fb
MD
1307
1308struct getfsstat_info {
1309 struct statfs *sfsp;
1310 long count;
1311 long maxcount;
1312 int error;
1313 int flags;
9910d07b 1314 struct thread *td;
861905fb
MD
1315};
1316
1317static int getfsstat_callback(struct mount *, void *);
1318
984263bc 1319int
753fd850 1320sys_getfsstat(struct getfsstat_args *uap)
984263bc 1321{
dadab5e9 1322 struct thread *td = curthread;
861905fb 1323 struct getfsstat_info info;
75ffff0d 1324
861905fb 1325 bzero(&info, sizeof(info));
861905fb
MD
1326
1327 info.maxcount = uap->bufsize / sizeof(struct statfs);
1328 info.sfsp = uap->buf;
1329 info.count = 0;
1330 info.flags = uap->flags;
9910d07b 1331 info.td = td;
861905fb
MD
1332
1333 mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
1334 if (info.sfsp && info.count > info.maxcount)
1335 uap->sysmsg_result = info.maxcount;
75ffff0d 1336 else
861905fb
MD
1337 uap->sysmsg_result = info.count;
1338 return (info.error);
1339}
1340
1341static int
1342getfsstat_callback(struct mount *mp, void *data)
1343{
1344 struct getfsstat_info *info = data;
1345 struct statfs *sp;
1346 char *freepath;
1347 char *fullpath;
1348 int error;
1349
1350 if (info->sfsp && info->count < info->maxcount) {
9910d07b
MD
1351 if (info->td->td_proc &&
1352 !chroot_visible_mnt(mp, info->td->td_proc)) {
861905fb 1353 return(0);
9910d07b 1354 }
861905fb
MD
1355 sp = &mp->mnt_stat;
1356
1357 /*
1358 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1359 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1360 * overrides MNT_WAIT.
1361 */
1362 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1363 (info->flags & MNT_WAIT)) &&
9910d07b 1364 (error = VFS_STATFS(mp, sp, info->td->td_ucred))) {
861905fb
MD
1365 return(0);
1366 }
1367 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1368
9910d07b 1369 error = mount_path(info->td->td_proc, mp, &fullpath, &freepath);
861905fb
MD
1370 if (error) {
1371 info->error = error;
1372 return(-1);
984263bc 1373 }
861905fb
MD
1374 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1375 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
efda3bd0 1376 kfree(freepath, M_TEMP);
861905fb
MD
1377
1378 error = copyout(sp, info->sfsp, sizeof(*sp));
1379 if (error) {
1380 info->error = error;
1381 return (-1);
984263bc 1382 }
861905fb 1383 ++info->sfsp;
984263bc 1384 }
861905fb
MD
1385 info->count++;
1386 return(0);
984263bc
MD
1387}
1388
1389/*
00fe9d48
MD
1390 * getvfsstat_args(struct statfs *buf, struct statvfs *vbuf,
1391 long bufsize, int flags)
1392 *
1393 * Get statistics on all filesystems.
1394 */
1395
1396struct getvfsstat_info {
1397 struct statfs *sfsp;
1398 struct statvfs *vsfsp;
1399 long count;
1400 long maxcount;
1401 int error;
1402 int flags;
9910d07b 1403 struct thread *td;
00fe9d48
MD
1404};
1405
1406static int getvfsstat_callback(struct mount *, void *);
1407
00fe9d48
MD
1408int
1409sys_getvfsstat(struct getvfsstat_args *uap)
1410{
1411 struct thread *td = curthread;
00fe9d48
MD
1412 struct getvfsstat_info info;
1413
1414 bzero(&info, sizeof(info));
1415
1416 info.maxcount = uap->vbufsize / sizeof(struct statvfs);
1417 info.sfsp = uap->buf;
1418 info.vsfsp = uap->vbuf;
1419 info.count = 0;
1420 info.flags = uap->flags;
9910d07b 1421 info.td = td;
00fe9d48
MD
1422
1423 mountlist_scan(getvfsstat_callback, &info, MNTSCAN_FORWARD);
1424 if (info.vsfsp && info.count > info.maxcount)
1425 uap->sysmsg_result = info.maxcount;
1426 else
1427 uap->sysmsg_result = info.count;
1428 return (info.error);
1429}
1430
1431static int
1432getvfsstat_callback(struct mount *mp, void *data)
1433{
1434 struct getvfsstat_info *info = data;
1435 struct statfs *sp;
1436 struct statvfs *vsp;
1437 char *freepath;
1438 char *fullpath;
1439 int error;
1440
1441 if (info->vsfsp && info->count < info->maxcount) {
9910d07b
MD
1442 if (info->td->td_proc &&
1443 !chroot_visible_mnt(mp, info->td->td_proc)) {
00fe9d48 1444 return(0);
9910d07b 1445 }
00fe9d48
MD
1446 sp = &mp->mnt_stat;
1447 vsp = &mp->mnt_vstat;
1448
1449 /*
1450 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1451 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1452 * overrides MNT_WAIT.
1453 */
1454 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1455 (info->flags & MNT_WAIT)) &&
9910d07b 1456 (error = VFS_STATFS(mp, sp, info->td->td_ucred))) {
00fe9d48
MD
1457 return(0);
1458 }
1459 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1460
1461 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1462 (info->flags & MNT_WAIT)) &&
9910d07b 1463 (error = VFS_STATVFS(mp, vsp, info->td->td_ucred))) {
00fe9d48
MD
1464 return(0);
1465 }
1466 vsp->f_flag = 0;
1467 if (mp->mnt_flag & MNT_RDONLY)
1468 vsp->f_flag |= ST_RDONLY;
1469 if (mp->mnt_flag & MNT_NOSUID)
1470 vsp->f_flag |= ST_NOSUID;
1471
9910d07b 1472 error = mount_path(info->td->td_proc, mp, &fullpath, &freepath);
00fe9d48
MD
1473 if (error) {
1474 info->error = error;
1475 return(-1);
1476 }
1477 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1478 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1479 kfree(freepath, M_TEMP);
1480
1481 error = copyout(sp, info->sfsp, sizeof(*sp));
1482 if (error == 0)
1483 error = copyout(vsp, info->vsfsp, sizeof(*vsp));
1484 if (error) {
1485 info->error = error;
1486 return (-1);
1487 }
1488 ++info->sfsp;
1489 ++info->vsfsp;
1490 }
1491 info->count++;
1492 return(0);
1493}
1494
1495
1496/*
41c20dac
MD
1497 * fchdir_args(int fd)
1498 *
984263bc
MD
1499 * Change current working directory to a given file descriptor.
1500 */
984263bc 1501int
753fd850 1502sys_fchdir(struct fchdir_args *uap)
984263bc 1503{
dadab5e9
MD
1504 struct thread *td = curthread;
1505 struct proc *p = td->td_proc;
41c20dac 1506 struct filedesc *fdp = p->p_fd;
21739618 1507 struct vnode *vp, *ovp;
984263bc
MD
1508 struct mount *mp;
1509 struct file *fp;
28623bf9 1510 struct nchandle nch, onch, tnch;
984263bc
MD
1511 int error;
1512
5b287bba 1513 if ((error = holdvnode(fdp, uap->fd, &fp)) != 0)
984263bc 1514 return (error);
0ac7b714 1515 lwkt_gettoken(&p->p_token);
984263bc 1516 vp = (struct vnode *)fp->f_data;
597aea93 1517 vref(vp);
ca466bae 1518 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
609c6f34 1519 if (fp->f_nchandle.ncp == NULL)
984263bc
MD
1520 error = ENOTDIR;
1521 else
609c6f34 1522 error = checkvp_chdir(vp, td);
984263bc
MD
1523 if (error) {
1524 vput(vp);
3919ced0 1525 goto done;
984263bc 1526 }
28623bf9 1527 cache_copy(&fp->f_nchandle, &nch);
1d505369
MD
1528
1529 /*
1530 * If the ncp has become a mount point, traverse through
1531 * the mount point.
1532 */
1533
28623bf9
MD
1534 while (!error && (nch.ncp->nc_flag & NCF_ISMOUNTPT) &&
1535 (mp = cache_findmount(&nch)) != NULL
1d505369 1536 ) {
28623bf9 1537 error = nlookup_mp(mp, &tnch);
21739618 1538 if (error == 0) {
28623bf9 1539 cache_unlock(&tnch); /* leave ref intact */
21739618 1540 vput(vp);
28623bf9 1541 vp = tnch.ncp->nc_vp;
87de5057 1542 error = vget(vp, LK_SHARED);
21739618 1543 KKASSERT(error == 0);
28623bf9
MD
1544 cache_drop(&nch);
1545 nch = tnch;
21739618
MD
1546 }
1547 }
1548 if (error == 0) {
1549 ovp = fdp->fd_cdir;
28623bf9 1550 onch = fdp->fd_ncdir;
a11aaa81 1551 vn_unlock(vp); /* leave ref intact */
21739618 1552 fdp->fd_cdir = vp;
28623bf9
MD
1553 fdp->fd_ncdir = nch;
1554 cache_drop(&onch);
21739618
MD
1555 vrele(ovp);
1556 } else {
28623bf9 1557 cache_drop(&nch);
21739618
MD
1558 vput(vp);
1559 }
5b287bba 1560 fdrop(fp);
3919ced0 1561done:
0ac7b714 1562 lwkt_reltoken(&p->p_token);
21739618 1563 return (error);
984263bc
MD
1564}
1565
9697c509 1566int
21739618 1567kern_chdir(struct nlookupdata *nd)
9697c509
DRJ
1568{
1569 struct thread *td = curthread;
1570 struct proc *p = td->td_proc;
1571 struct filedesc *fdp = p->p_fd;
21739618 1572 struct vnode *vp, *ovp;
28623bf9 1573 struct nchandle onch;
9697c509
DRJ
1574 int error;
1575
21739618 1576 if ((error = nlookup(nd)) != 0)
9697c509 1577 return (error);
28623bf9 1578 if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
21739618 1579 return (ENOENT);
87de5057 1580 if ((error = vget(vp, LK_SHARED)) != 0)
21739618
MD
1581 return (error);
1582
0ac7b714 1583 lwkt_gettoken(&p->p_token);
21739618 1584 error = checkvp_chdir(vp, td);
a11aaa81 1585 vn_unlock(vp);
21739618
MD
1586 if (error == 0) {
1587 ovp = fdp->fd_cdir;
28623bf9
MD
1588 onch = fdp->fd_ncdir;
1589 cache_unlock(&nd->nl_nch); /* leave reference intact */
1590 fdp->fd_ncdir = nd->nl_nch;
21739618 1591 fdp->fd_cdir = vp;
28623bf9 1592 cache_drop(&onch);
21739618 1593 vrele(ovp);
28623bf9 1594 cache_zero(&nd->nl_nch);
21739618
MD
1595 } else {
1596 vrele(vp);
e24b948e 1597 }
0ac7b714 1598 lwkt_reltoken(&p->p_token);
e24b948e 1599 return (error);
9697c509
DRJ
1600}
1601
984263bc 1602/*
41c20dac
MD
1603 * chdir_args(char *path)
1604 *
984263bc
MD
1605 * Change current working directory (``.'').
1606 */
984263bc 1607int
753fd850 1608sys_chdir(struct chdir_args *uap)
984263bc 1609{
21739618 1610 struct nlookupdata nd;
9697c509 1611 int error;
984263bc 1612
21739618 1613 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e 1614 if (error == 0)
21739618 1615 error = kern_chdir(&nd);
fad57d0e 1616 nlookup_done(&nd);
9697c509 1617 return (error);
984263bc
MD
1618}
1619
1620/*
1621 * Helper function for raised chroot(2) security function: Refuse if
1622 * any filedescriptors are open directories.
1623 */
1624static int
c972a82f 1625chroot_refuse_vdir_fds(struct filedesc *fdp)
984263bc
MD
1626{
1627 struct vnode *vp;
1628 struct file *fp;
1629 int error;
1630 int fd;
1631
1632 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
5b287bba 1633 if ((error = holdvnode(fdp, fd, &fp)) != 0)
984263bc
MD
1634 continue;
1635 vp = (struct vnode *)fp->f_data;
5b287bba
MD
1636 if (vp->v_type != VDIR) {
1637 fdrop(fp);
984263bc 1638 continue;
5b287bba
MD
1639 }
1640 fdrop(fp);
984263bc
MD
1641 return(EPERM);
1642 }
1643 return (0);
1644}
1645
1646/*
1647 * This sysctl determines if we will allow a process to chroot(2) if it
1648 * has a directory open:
1649 * 0: disallowed for all processes.
1650 * 1: allowed for processes that were not already chroot(2)'ed.
1651 * 2: allowed for all processes.
1652 */
1653
1654static int chroot_allow_open_directories = 1;
1655
1656SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1657 &chroot_allow_open_directories, 0, "");
1658
1659/*
21739618
MD
1660 * chroot to the specified namecache entry. We obtain the vp from the
1661 * namecache data. The passed ncp must be locked and referenced and will
1662 * remain locked and referenced on return.
e24b948e 1663 */
53dd6631 1664int
28623bf9 1665kern_chroot(struct nchandle *nch)
e24b948e
MD
1666{
1667 struct thread *td = curthread;
1668 struct proc *p = td->td_proc;
1669 struct filedesc *fdp = p->p_fd;
21739618 1670 struct vnode *vp;
e24b948e
MD
1671 int error;
1672
1673 /*
ca3cd02d 1674 * Only privileged user can chroot
e24b948e 1675 */
9910d07b 1676 error = priv_check_cred(td->td_ucred, PRIV_VFS_CHROOT, 0);
3a591c90 1677 if (error)
e24b948e
MD
1678 return (error);
1679
1680 /*
1681 * Disallow open directory descriptors (fchdir() breakouts).
1682 */
1683 if (chroot_allow_open_directories == 0 ||
1684 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1685 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1686 return (error);
1687 }
28623bf9 1688 if ((vp = nch->ncp->nc_vp) == NULL)
21739618
MD
1689 return (ENOENT);
1690
87de5057 1691 if ((error = vget(vp, LK_SHARED)) != 0)
21739618 1692 return (error);
e24b948e
MD
1693
1694 /*
1695 * Check the validity of vp as a directory to change to and
1696 * associate it with rdir/jdir.
1697 */
21739618 1698 error = checkvp_chdir(vp, td);
a11aaa81 1699 vn_unlock(vp); /* leave reference intact */
21739618 1700 if (error == 0) {
e24b948e 1701 vrele(fdp->fd_rdir);
21739618 1702 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */
28623bf9
MD
1703 cache_drop(&fdp->fd_nrdir);
1704 cache_copy(nch, &fdp->fd_nrdir);
e24b948e
MD
1705 if (fdp->fd_jdir == NULL) {
1706 fdp->fd_jdir = vp;
597aea93 1707 vref(fdp->fd_jdir);
28623bf9 1708 cache_copy(nch, &fdp->fd_njdir);
e24b948e 1709 }
21739618
MD
1710 } else {
1711 vrele(vp);
e24b948e
MD
1712 }
1713 return (error);
1714}
1715
1716/*
41c20dac
MD
1717 * chroot_args(char *path)
1718 *
984263bc
MD
1719 * Change notion of root (``/'') directory.
1720 */
984263bc 1721int
753fd850 1722sys_chroot(struct chroot_args *uap)
984263bc 1723{
f64b567c 1724 struct thread *td __debugvar = curthread;
21739618 1725 struct nlookupdata nd;
e24b948e 1726 int error;
984263bc 1727
e24b948e 1728 KKASSERT(td->td_proc);
21739618 1729 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3919ced0
MD
1730 if (error == 0) {
1731 nd.nl_flags |= NLC_EXEC;
1732 error = nlookup(&nd);
1733 if (error == 0)
1734 error = kern_chroot(&nd.nl_nch);
53dd6631 1735 }
fad57d0e 1736 nlookup_done(&nd);
53dd6631 1737 return(error);
984263bc
MD
1738}
1739
72f12c5b
AH
1740int
1741sys_chroot_kernel(struct chroot_kernel_args *uap)
1742{
1743 struct thread *td = curthread;
1744 struct nlookupdata nd;
1745 struct nchandle *nch;
1746 struct vnode *vp;
1747 int error;
1748
72f12c5b
AH
1749 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1750 if (error)
1751 goto error_nond;
1752
1753 error = nlookup(&nd);
1754 if (error)
1755 goto error_out;
1756
1757 nch = &nd.nl_nch;
1758
1759 error = priv_check_cred(td->td_ucred, PRIV_VFS_CHROOT, 0);
1760 if (error)
1761 goto error_out;
1762
1763 if ((vp = nch->ncp->nc_vp) == NULL) {
1764 error = ENOENT;
1765 goto error_out;
1766 }
1767
1768 if ((error = cache_vref(nch, nd.nl_cred, &vp)) != 0)
1769 goto error_out;
1770
1771 kprintf("chroot_kernel: set new rootnch/rootvnode to %s\n", uap->path);
0ac7b714 1772 get_mplock();
72f12c5b 1773 vfs_cache_setroot(vp, cache_hold(nch));
0ac7b714 1774 rel_mplock();
72f12c5b
AH
1775
1776error_out:
1777 nlookup_done(&nd);
1778error_nond:
72f12c5b
AH
1779 return(error);
1780}
1781
984263bc 1782/*
e24b948e
MD
1783 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1784 * determine whether it is legal to chdir to the vnode. The vnode's state
1785 * is not changed by this call.
984263bc 1786 */
e24b948e
MD
1787int
1788checkvp_chdir(struct vnode *vp, struct thread *td)
984263bc 1789{
984263bc
MD
1790 int error;
1791
984263bc
MD
1792 if (vp->v_type != VDIR)
1793 error = ENOTDIR;
1794 else
9910d07b 1795 error = VOP_EACCESS(vp, VEXEC, td->td_ucred);
984263bc
MD
1796 return (error);
1797}
1798
984263bc 1799int
fad57d0e 1800kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
984263bc 1801{
dadab5e9
MD
1802 struct thread *td = curthread;
1803 struct proc *p = td->td_proc;
fbfe4e7d 1804 struct lwp *lp = td->td_lwp;
41c20dac 1805 struct filedesc *fdp = p->p_fd;
9697c509 1806 int cmode, flags;
984263bc 1807 struct file *nfp;
fad57d0e
MD
1808 struct file *fp;
1809 struct vnode *vp;
984263bc
MD
1810 int type, indx, error;
1811 struct flock lf;
984263bc 1812
984263bc
MD
1813 if ((oflags & O_ACCMODE) == O_ACCMODE)
1814 return (EINVAL);
1815 flags = FFLAGS(oflags);
f3a2d8c4 1816 error = falloc(lp, &nfp, NULL);
984263bc
MD
1817 if (error)
1818 return (error);
1819 fp = nfp;
3a907475 1820 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
fad57d0e 1821
984263bc 1822 /*
fad57d0e
MD
1823 * XXX p_dupfd is a real mess. It allows a device to return a
1824 * file descriptor to be duplicated rather then doing the open
1825 * itself.
984263bc 1826 */
fbfe4e7d 1827 lp->lwp_dupfd = -1;
21739618
MD
1828
1829 /*
fad57d0e
MD
1830 * Call vn_open() to do the lookup and assign the vnode to the
1831 * file pointer. vn_open() does not change the ref count on fp
1832 * and the vnode, on success, will be inherited by the file pointer
1833 * and unlocked.
21739618 1834 */
fad57d0e
MD
1835 nd->nl_flags |= NLC_LOCKVP;
1836 error = vn_open(nd, fp, flags, cmode);
1837 nlookup_done(nd);
984263bc
MD
1838 if (error) {
1839 /*
984263bc
MD
1840 * handle special fdopen() case. bleh. dupfdopen() is
1841 * responsible for dropping the old contents of ofiles[indx]
1842 * if it succeeds.
fad57d0e 1843 *
259b8ea0
MD
1844 * Note that fsetfd() will add a ref to fp which represents
1845 * the fd_files[] assignment. We must still drop our
1846 * reference.
984263bc 1847 */
fbfe4e7d 1848 if ((error == ENODEV || error == ENXIO) && lp->lwp_dupfd >= 0) {
259b8ea0 1849 if (fdalloc(p, 0, &indx) == 0) {
f3a2d8c4 1850 error = dupfdopen(fdp, indx, lp->lwp_dupfd, flags, error);
fad57d0e
MD
1851 if (error == 0) {
1852 *res = indx;
9f87144f 1853 fdrop(fp); /* our ref */
fad57d0e
MD
1854 return (0);
1855 }
f3a2d8c4 1856 fsetfd(fdp, NULL, indx);
fad57d0e 1857 }
984263bc 1858 }
9f87144f 1859 fdrop(fp); /* our ref */
984263bc
MD
1860 if (error == ERESTART)
1861 error = EINTR;
1862 return (error);
1863 }
fad57d0e
MD
1864
1865 /*
1866 * ref the vnode for ourselves so it can't be ripped out from under
1867 * is. XXX need an ND flag to request that the vnode be returned
1868 * anyway.
259b8ea0
MD
1869 *
1870 * Reserve a file descriptor but do not assign it until the open
1871 * succeeds.
fad57d0e
MD
1872 */
1873 vp = (struct vnode *)fp->f_data;
1874 vref(vp);
259b8ea0 1875 if ((error = fdalloc(p, 0, &indx)) != 0) {
9f87144f 1876 fdrop(fp);
fad57d0e
MD
1877 vrele(vp);
1878 return (error);
1879 }
1880
1881 /*
1882 * If no error occurs the vp will have been assigned to the file
1883 * pointer.
1884 */
fbfe4e7d 1885 lp->lwp_dupfd = 0;
984263bc 1886
984263bc
MD
1887 if (flags & (O_EXLOCK | O_SHLOCK)) {
1888 lf.l_whence = SEEK_SET;
1889 lf.l_start = 0;
1890 lf.l_len = 0;
1891 if (flags & O_EXLOCK)
1892 lf.l_type = F_WRLCK;
1893 else
1894 lf.l_type = F_RDLCK;
71c18fe3
MD
1895 if (flags & FNONBLOCK)
1896 type = 0;
1897 else
1898 type = F_WAIT;
fad57d0e 1899
984263bc
MD
1900 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1901 /*
259b8ea0
MD
1902 * lock request failed. Clean up the reserved
1903 * descriptor.
984263bc 1904 */
fad57d0e 1905 vrele(vp);
f3a2d8c4 1906 fsetfd(fdp, NULL, indx);
9f87144f 1907 fdrop(fp);
984263bc
MD
1908 return (error);
1909 }
984263bc
MD
1910 fp->f_flag |= FHASLOCK;
1911 }
7540ab49
MD
1912#if 0
1913 /*
1914 * Assert that all regular file vnodes were created with a object.
1915 */
1916 KASSERT(vp->v_type != VREG || vp->v_object != NULL,
1917 ("open: regular file has no backing object after vn_open"));
1918#endif
984263bc 1919
fad57d0e 1920 vrele(vp);
21739618
MD
1921
1922 /*
984263bc
MD
1923 * release our private reference, leaving the one associated with the
1924 * descriptor table intact.
1925 */
f3a2d8c4 1926 fsetfd(fdp, fp, indx);
9f87144f 1927 fdrop(fp);
9697c509 1928 *res = indx;
984263bc
MD
1929 return (0);
1930}
1931
984263bc 1932/*
9697c509 1933 * open_args(char *path, int flags, int mode)
41c20dac 1934 *
9697c509
DRJ
1935 * Check permissions, allocate an open file structure,
1936 * and call the device open routine if any.
984263bc 1937 */
984263bc 1938int
753fd850 1939sys_open(struct open_args *uap)
984263bc 1940{
fad57d0e 1941 struct nlookupdata nd;
9697c509
DRJ
1942 int error;
1943
85dbab7f 1944 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
fad57d0e
MD
1945 if (error == 0) {
1946 error = kern_open(&nd, uap->flags,
1947 uap->mode, &uap->sysmsg_result);
1948 }
1949 nlookup_done(&nd);
9697c509 1950 return (error);
984263bc 1951}
984263bc 1952
83a11774
NT
1953/*
1954 * openat_args(int fd, char *path, int flags, int mode)
1955 */
1956int
1957sys_openat(struct openat_args *uap)
1958{
83a11774
NT
1959 struct nlookupdata nd;
1960 int error;
1db695af 1961 struct file *fp;
83a11774 1962
1db695af
NT
1963 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
1964 if (error == 0) {
1965 error = kern_open(&nd, uap->flags, uap->mode,
1966 &uap->sysmsg_result);
83a11774 1967 }
1db695af 1968 nlookup_done_at(&nd, fp);
83a11774
NT
1969 return (error);
1970}
1971
984263bc 1972int
0e9b9130 1973kern_mknod(struct nlookupdata *nd, int mode, int rmajor, int rminor)
984263bc 1974{
dadab5e9
MD
1975 struct thread *td = curthread;
1976 struct proc *p = td->td_proc;
41c20dac 1977 struct vnode *vp;
984263bc
MD
1978 struct vattr vattr;
1979 int error;
1980 int whiteout = 0;
984263bc 1981
dadab5e9
MD
1982 KKASSERT(p);
1983
fad57d0e
MD
1984 VATTR_NULL(&vattr);
1985 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
0e9b9130
MD
1986 vattr.va_rmajor = rmajor;
1987 vattr.va_rminor = rminor;
fad57d0e
MD
1988
1989 switch (mode & S_IFMT) {
1990 case S_IFMT: /* used by badsect to flag bad sectors */
9910d07b 1991 error = priv_check_cred(td->td_ucred, PRIV_VFS_MKNOD_BAD, 0);
fad57d0e
MD
1992 vattr.va_type = VBAD;
1993 break;
1994 case S_IFCHR:
44a36778 1995 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
fad57d0e
MD
1996 vattr.va_type = VCHR;
1997 break;
1998 case S_IFBLK:
44a36778 1999 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
fad57d0e
MD
2000 vattr.va_type = VBLK;
2001 break;
2002 case S_IFWHT:
9910d07b 2003 error = priv_check_cred(td->td_ucred, PRIV_VFS_MKNOD_WHT, 0);
fad57d0e
MD
2004 whiteout = 1;
2005 break;
44a36778 2006 case S_IFDIR: /* special directories support for HAMMER */
9910d07b 2007 error = priv_check_cred(td->td_ucred, PRIV_VFS_MKNOD_DIR, 0);
bed9e372
MD
2008 vattr.va_type = VDIR;
2009 break;
fad57d0e
MD
2010 default:
2011 error = EINVAL;
2012 break;
984263bc 2013 }
44a36778
MN
2014
2015 if (error)
2016 return (error);
2017
2018 bwillinode(1);
2019 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
2020 if ((error = nlookup(nd)) != 0)
2021 return (error);
2022 if (nd->nl_nch.ncp->nc_vp)
2023 return (EEXIST);
2024 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2025 return (error);
2026
2027 if (whiteout) {
2028 error = VOP_NWHITEOUT(&nd->nl_nch, nd->nl_dvp,
2029 nd->nl_cred, NAMEI_CREATE);
2030 } else {
2031 vp = NULL;
2032 error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp,
2033 &vp, nd->nl_cred, &vattr);
2034 if (error == 0)
2035 vput(vp);
984263bc 2036 }
9697c509
DRJ
2037 return (error);
2038}
2039
2040/*
2041 * mknod_args(char *path, int mode, int dev)
2042 *
2043 * Create a special file.
2044 */
2045int
753fd850 2046sys_mknod(struct mknod_args *uap)
9697c509 2047{
fad57d0e 2048 struct nlookupdata nd;
9697c509
DRJ
2049 int error;
2050
fad57d0e 2051 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
0e9b9130
MD
2052 if (error == 0) {
2053 error = kern_mknod(&nd, uap->mode,
2054 umajor(uap->dev), uminor(uap->dev));
2055 }
fad57d0e 2056 nlookup_done(&nd);
984263bc
MD
2057 return (error);
2058}
2059
6d98f93c
SW
2060/*
2061 * mknodat_args(int fd, char *path, mode_t mode, dev_t dev)
2062 *
2063 * Create a special file. The path is relative to the directory associated
2064 * with fd.
6d98f93c
SW
2065 */
2066int
2067sys_mknodat(struct mknodat_args *uap)
2068{
2069 struct nlookupdata nd;
2070 struct file *fp;
2071 int error;
2072
6d98f93c
SW
2073 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
2074 if (error == 0) {
2075 error = kern_mknod(&nd, uap->mode,
2076 umajor(uap->dev), uminor(uap->dev));
2077 }
2078 nlookup_done_at(&nd, fp);
6d98f93c
SW
2079 return (error);
2080}
2081
984263bc 2082int
fad57d0e 2083kern_mkfifo(struct nlookupdata *nd, int mode)
984263bc 2084{
dadab5e9
MD
2085 struct thread *td = curthread;
2086 struct proc *p = td->td_proc;
984263bc 2087 struct vattr vattr;
fad57d0e 2088 struct vnode *vp;
984263bc 2089 int error;
984263bc 2090
c4df9635 2091 bwillinode(1);
fad57d0e 2092
5312fa43 2093 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 2094 if ((error = nlookup(nd)) != 0)
984263bc 2095 return (error);
28623bf9 2096 if (nd->nl_nch.ncp->nc_vp)
984263bc 2097 return (EEXIST);
28623bf9 2098 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2099 return (error);
fad57d0e 2100
984263bc
MD
2101 VATTR_NULL(&vattr);
2102 vattr.va_type = VFIFO;
136178b3 2103 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
fad57d0e 2104 vp = NULL;
5312fa43 2105 error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp, &vp, nd->nl_cred, &vattr);
984263bc 2106 if (error == 0)
fad57d0e 2107 vput(vp);
136178b3
DRJ
2108 return (error);
2109}
2110
2111/*
2112 * mkfifo_args(char *path, int mode)
2113 *
2114 * Create a named pipe.
2115 */
2116int
753fd850 2117sys_mkfifo(struct mkfifo_args *uap)
136178b3 2118{
fad57d0e 2119 struct nlookupdata nd;
136178b3
DRJ
2120 int error;
2121
fad57d0e
MD
2122 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2123 if (error == 0)
2124 error = kern_mkfifo(&nd, uap->mode);
2125 nlookup_done(&nd);
984263bc
MD
2126 return (error);
2127}
2128
6d98f93c
SW
2129/*
2130 * mkfifoat_args(int fd, char *path, mode_t mode)
2131 *
2132 * Create a named pipe. The path is relative to the directory associated
2133 * with fd.
6d98f93c
SW
2134 */
2135int
2136sys_mkfifoat(struct mkfifoat_args *uap)
2137{
2138 struct nlookupdata nd;
2139 struct file *fp;
2140 int error;
2141
6d98f93c
SW
2142 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
2143 if (error == 0)
2144 error = kern_mkfifo(&nd, uap->mode);
2145 nlookup_done_at(&nd, fp);
6d98f93c
SW
2146 return (error);
2147}
2148
8c40f81e 2149static int hardlink_check_uid = 0;
43a0f7ae 2150SYSCTL_INT(_security, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
8c40f81e
MD
2151 &hardlink_check_uid, 0,
2152 "Unprivileged processes cannot create hard links to files owned by other "
2153 "users");
2154static int hardlink_check_gid = 0;
43a0f7ae 2155SYSCTL_INT(_security, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
8c40f81e
MD
2156 &hardlink_check_gid, 0,
2157 "Unprivileged processes cannot create hard links to files owned by other "
2158 "groups");
2159
2160static int
2161can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
2162{
2163 struct vattr va;
2164 int error;
2165
2166 /*
2167 * Shortcut if disabled
2168 */
2169 if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
2170 return (0);
2171
2172 /*
f6df0641 2173 * Privileged user can always hardlink
8c40f81e 2174 */
f6df0641 2175 if (priv_check_cred(cred, PRIV_VFS_LINK, 0) == 0)
8c40f81e
MD
2176 return (0);
2177
2178 /*
2179 * Otherwise only if the originating file is owned by the
2180 * same user or group. Note that any group is allowed if
2181 * the file is owned by the caller.
2182 */
87de5057 2183 error = VOP_GETATTR(vp, &va);
8c40f81e
MD
2184 if (error != 0)
2185 return (error);
2186
2187 if (hardlink_check_uid) {
2188 if (cred->cr_uid != va.va_uid)
2189 return (EPERM);
2190 }
2191
2192 if (hardlink_check_gid) {
2193 if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
2194 return (EPERM);
2195 }
2196
2197 return (0);
2198}
2199
984263bc 2200int
fad57d0e 2201kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
984263bc 2202{
dadab5e9 2203 struct thread *td = curthread;
41c20dac 2204 struct vnode *vp;
984263bc
MD
2205 int error;
2206
fad57d0e
MD
2207 /*
2208 * Lookup the source and obtained a locked vnode.
2209 *
3a907475
MD
2210 * You may only hardlink a file which you have write permission
2211 * on or which you own.
2212 *
fad57d0e
MD
2213 * XXX relookup on vget failure / race ?
2214 */
c4df9635 2215 bwillinode(1);
3a907475 2216 nd->nl_flags |= NLC_WRITE | NLC_OWN | NLC_HLINK;
fad57d0e 2217 if ((error = nlookup(nd)) != 0)
984263bc 2218 return (error);
28623bf9 2219 vp = nd->nl_nch.ncp->nc_vp;
fad57d0e 2220 KKASSERT(vp != NULL);
984263bc 2221 if (vp->v_type == VDIR)
fad57d0e 2222 return (EPERM); /* POSIX */
28623bf9 2223 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2224 return (error);
87de5057 2225 if ((error = vget(vp, LK_EXCLUSIVE)) != 0)
fad57d0e
MD
2226 return (error);
2227
2228 /*
2229 * Unlock the source so we can lookup the target without deadlocking
2230 * (XXX vp is locked already, possible other deadlock?). The target
2231 * must not exist.
2232 */
2233 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
2234 nd->nl_flags &= ~NLC_NCPISLOCKED;
28623bf9 2235 cache_unlock(&nd->nl_nch);
1cd65690 2236 vn_unlock(vp);
fad57d0e 2237
5312fa43 2238 linknd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 2239 if ((error = nlookup(linknd)) != 0) {
1cd65690 2240 vrele(vp);
fad57d0e 2241 return (error);
984263bc 2242 }
28623bf9 2243 if (linknd->nl_nch.ncp->nc_vp) {
1cd65690 2244 vrele(vp);
fad57d0e
MD
2245 return (EEXIST);
2246 }
1cd65690
MD
2247 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) != 0) {
2248 vrele(vp);
2249 return (error);
2250 }
fad57d0e
MD
2251
2252 /*
2253 * Finally run the new API VOP.
2254 */
9910d07b 2255 error = can_hardlink(vp, td, td->td_ucred);
5312fa43
MD
2256 if (error == 0) {
2257 error = VOP_NLINK(&linknd->nl_nch, linknd->nl_dvp,
2258 vp, linknd->nl_cred);
2259 }
fad57d0e 2260 vput(vp);
984263bc
MD
2261 return (error);
2262}
2263
2264/*
9697c509 2265 * link_args(char *path, char *link)
41c20dac 2266 *
9697c509 2267 * Make a hard file link.
984263bc 2268 */
984263bc 2269int
753fd850 2270sys_link(struct link_args *uap)
9697c509 2271{
fad57d0e 2272 struct nlookupdata nd, linknd;
9697c509
DRJ
2273 int error;
2274
fad57d0e
MD
2275 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2276 if (error == 0) {
2277 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
2278 if (error == 0)
2279 error = kern_link(&nd, &linknd);
2280 nlookup_done(&linknd);
2281 }
2282 nlookup_done(&nd);
9697c509
DRJ
2283 return (error);
2284}
2285
2286int
fad57d0e 2287kern_symlink(struct nlookupdata *nd, char *path, int mode)
984263bc
MD
2288{
2289 struct vattr vattr;
fad57d0e 2290 struct vnode *vp;
dff430ab 2291 struct vnode *dvp;
984263bc 2292 int error;
984263bc 2293
c4df9635 2294 bwillinode(1);
5312fa43 2295 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 2296 if ((error = nlookup(nd)) != 0)
9697c509 2297 return (error);
28623bf9 2298 if (nd->nl_nch.ncp->nc_vp)
9697c509 2299 return (EEXIST);
28623bf9 2300 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2301 return (error);
5312fa43 2302 dvp = nd->nl_dvp;
984263bc 2303 VATTR_NULL(&vattr);
fad57d0e 2304 vattr.va_mode = mode;
dff430ab 2305 error = VOP_NSYMLINK(&nd->nl_nch, dvp, &vp, nd->nl_cred, &vattr, path);
984263bc 2306 if (error == 0)
fad57d0e 2307 vput(vp);
9697c509
DRJ
2308 return (error);
2309}
2310
2311/*
2312 * symlink(char *path, char *link)
2313 *
2314 * Make a symbolic link.
2315 */
2316int
753fd850 2317sys_symlink(struct symlink_args *uap)
9697c509
DRJ
2318{
2319 struct thread *td = curthread;
fad57d0e 2320 struct nlookupdata nd;
9697c509
DRJ
2321 char *path;
2322 int error;
fad57d0e 2323 int mode;
9697c509 2324
70aac194 2325 path = objcache_get(namei_oc, M_WAITOK);
9697c509 2326 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
e24b948e 2327 if (error == 0) {
fad57d0e
MD
2328 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
2329 if (error == 0) {
2330 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
2331 error = kern_symlink(&nd, path, mode);
2332 }
2333 nlookup_done(&nd);
e24b948e 2334 }
70aac194 2335 objcache_put(namei_oc, path);
984263bc
MD
2336 return (error);
2337}
2338
2339/*
6d98f93c
SW
2340 * symlinkat_args(char *path1, int fd, char *path2)
2341 *
2342 * Make a symbolic link. The path2 argument is relative to the directory
2343 * associated with fd.
6d98f93c
SW
2344 */
2345int
2346sys_symlinkat(struct symlinkat_args *uap)
2347{
2348 struct thread *td = curthread;
2349 struct nlookupdata nd;
2350 struct file *fp;
2351 char *path1;
2352 int error;
2353 int mode;
2354
2355 path1 = objcache_get(namei_oc, M_WAITOK);
2356 error = copyinstr(uap->path1, path1, MAXPATHLEN, NULL);
2357 if (error == 0) {
6d98f93c
SW
2358 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path2,
2359 UIO_USERSPACE, 0);
2360 if (error == 0) {
2361 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
2362 error = kern_symlink(&nd, path1, mode);
2363 }
2364 nlookup_done_at(&nd, fp);
6d98f93c
SW
2365 }
2366 objcache_put(namei_oc, path1);
2367 return (error);
2368}
2369
2370/*
41c20dac
MD
2371 * undelete_args(char *path)
2372 *
984263bc
MD
2373 * Delete a whiteout from the filesystem.
2374 */
984263bc 2375int
753fd850 2376sys_undelete(struct undelete_args *uap)
984263bc 2377{
fad57d0e 2378 struct nlookupdata nd;
984263bc 2379 int error;
984263bc 2380
ab2eb4eb 2381 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
c4df9635 2382 bwillinode(1);
5312fa43 2383 nd.nl_flags |= NLC_DELETE | NLC_REFDVP;
fad57d0e
MD
2384 if (error == 0)
2385 error = nlookup(&nd);
b8997912 2386 if (error == 0)
28623bf9 2387 error = ncp_writechk(&nd.nl_nch);
dff430ab 2388 if (error == 0) {
5312fa43
MD
2389 error = VOP_NWHITEOUT(&nd.nl_nch, nd.nl_dvp, nd.nl_cred,
2390 NAMEI_DELETE);
dff430ab 2391 }
fad57d0e 2392 nlookup_done(&nd);
984263bc
MD
2393 return (error);
2394}
2395
984263bc 2396int
fad57d0e 2397kern_unlink(struct nlookupdata *nd)
984263bc 2398{
984263bc 2399 int error;
984263bc 2400
c4df9635 2401 bwillinode(1);
5312fa43 2402 nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
fad57d0e 2403 if ((error = nlookup(nd)) != 0)
984263bc 2404 return (error);
28623bf9 2405 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2406 return (error);
5312fa43 2407 error = VOP_NREMOVE(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
984263bc
MD
2408 return (error);
2409}
2410
2411/*
9697c509 2412 * unlink_args(char *path)
41c20dac 2413 *
9697c509 2414 * Delete a name from the filesystem.
984263bc 2415 */
984263bc 2416int
753fd850 2417sys_unlink(struct unlink_args *uap)
9697c509 2418{
fad57d0e 2419 struct nlookupdata nd;
9697c509
DRJ
2420 int error;
2421
fad57d0e
MD
2422 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2423 if (error == 0)
2424 error = kern_unlink(&nd);
2425 nlookup_done(&nd);
9697c509
DRJ
2426 return (error);
2427}
2428
3a6d8629
NT
2429
2430/*
2431 * unlinkat_args(int fd, char *path, int flags)
2432 *
2433 * Delete the file or directory entry pointed to by fd/path.
2434 */
2435int
2436sys_unlinkat(struct unlinkat_args *uap)
2437{
2438 struct nlookupdata nd;
2439 struct file *fp;
2440 int error;
2441
2442 if (uap->flags & ~AT_REMOVEDIR)
2443 return (EINVAL);
2444
2445 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
2446 if (error == 0) {
2447 if (uap->flags & AT_REMOVEDIR)
2448 error = kern_rmdir(&nd);
2449 else
2450 error = kern_unlink(&nd);
2451 }
2452 nlookup_done_at(&nd, fp);
2453 return (error);
2454}
2455
9697c509 2456int
5a3fe67d 2457kern_lseek(int fd, off_t offset, int whence, off_t *res)
984263bc 2458{
dadab5e9
MD
2459 struct thread *td = curthread;
2460 struct proc *p = td->td_proc;
41c20dac 2461 struct file *fp;
116edfbf 2462 struct vnode *vp;
984263bc 2463 struct vattr vattr;
116edfbf 2464 off_t new_offset;
984263bc
MD
2465 int error;
2466
228b401d
MD
2467 fp = holdfp(p->p_fd, fd, -1);
2468 if (fp == NULL)
984263bc 2469 return (EBADF);
228b401d
MD
2470 if (fp->f_type != DTYPE_VNODE) {
2471 error = ESPIPE;
2472 goto done;
2473 }
116edfbf 2474 vp = (struct vnode *)fp->f_data;
228b401d 2475
9697c509 2476 switch (whence) {
984263bc 2477 case L_INCR:
287a8577 2478 spin_lock(&fp->f_spin);
116edfbf 2479 new_offset = fp->f_offset + offset;
228b401d 2480 error = 0;
984263bc
MD
2481 break;
2482 case L_XTND:
116edfbf 2483 error = VOP_GETATTR(vp, &vattr);
287a8577 2484 spin_lock(&fp->f_spin);
116edfbf 2485 new_offset = offset + vattr.va_size;
984263bc
MD
2486 break;
2487 case L_SET:
116edfbf 2488 new_offset = offset;
228b401d 2489 error = 0;
287a8577 2490 spin_lock(&fp->f_spin);
984263bc
MD
2491 break;
2492 default:
116edfbf 2493 new_offset = 0;
228b401d 2494 error = EINVAL;
287a8577 2495 spin_lock(&fp->f_spin);
228b401d 2496 break;
984263bc 2497 }
116edfbf
MD
2498
2499 /*
2500 * Validate the seek position. Negative offsets are not allowed
58675cb0
MD
2501 * for regular files or directories.
2502 *
2503 * Normally we would also not want to allow negative offsets for
2504 * character and block-special devices. However kvm addresses
2505 * on 64 bit architectures might appear to be negative and must
2506 * be allowed.
116edfbf
MD
2507 */
2508 if (error == 0) {
2509 if (new_offset < 0 &&
58675cb0 2510 (vp->v_type == VREG || vp->v_type == VDIR)) {
116edfbf 2511 error = EINVAL;
b06b3b87
MD
2512 } else {
2513 fp->f_offset = new_offset;
116edfbf 2514 }
116edfbf 2515 }
9697c509 2516 *res = fp->f_offset;
287a8577 2517 spin_unlock(&fp->f_spin);
228b401d
MD
2518done:
2519 fdrop(fp);
2520 return (error);
984263bc
MD
2521}
2522
984263bc 2523/*
9697c509 2524 * lseek_args(int fd, int pad, off_t offset, int whence)
41c20dac 2525 *
9697c509 2526 * Reposition read/write file offset.
984263bc 2527 */
984263bc 2528int
753fd850 2529sys_lseek(struct lseek_args *uap)
984263bc 2530{
984263bc
MD
2531 int error;
2532
9697c509 2533 error = kern_lseek(uap->fd, uap->offset, uap->whence,
116edfbf 2534 &uap->sysmsg_offset);
9697c509 2535
984263bc
MD
2536 return (error);
2537}
984263bc 2538
18cf460b
NT
2539/*
2540 * Check if current process can access given file. amode is a bitmask of *_OK
2541 * access bits. flags is a bitmask of AT_* flags.
2542 */
984263bc 2543int
18cf460b 2544kern_access(struct nlookupdata *nd, int amode, int flags)
984263bc 2545{
41c20dac 2546 struct vnode *vp;
18cf460b 2547 int error, mode;
984263bc 2548
18cf460b
NT
2549 if (flags & ~AT_EACCESS)
2550 return (EINVAL);
fad57d0e
MD
2551 if ((error = nlookup(nd)) != 0)
2552 return (error);
806dcf9a 2553retry:
28623bf9 2554 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 2555 if (error)
fad57d0e 2556 return (error);
984263bc
MD
2557
2558 /* Flags == 0 means only check for existence. */
18cf460b
NT
2559 if (amode) {
2560 mode = 0;
2561 if (amode & R_OK)
2562 mode |= VREAD;
2563 if (amode & W_OK)
2564 mode |= VWRITE;
2565 if (amode & X_OK)
2566 mode |= VEXEC;
2567 if ((mode & VWRITE) == 0 ||
28623bf9 2568 (error = vn_writechk(vp, &nd->nl_nch)) == 0)
18cf460b 2569 error = VOP_ACCESS_FLAGS(vp, mode, flags, nd->nl_cred);
806dcf9a
MD
2570
2571 /*
2572 * If the file handle is stale we have to re-resolve the
2573 * entry. This is a hack at the moment.
2574 */
2575 if (error == ESTALE) {
5e1a76ea 2576 vput(vp);
28623bf9
MD
2577 cache_setunresolved(&nd->nl_nch);
2578 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
806dcf9a 2579 if (error == 0) {
806dcf9a
MD
2580 vp = NULL;
2581 goto retry;
2582 }
5e1a76ea 2583 return(error);
806dcf9a 2584 }
984263bc 2585 }
984263bc 2586 vput(vp);
984263bc
MD
2587 return (error);
2588}
2589
984263bc 2590/*
9697c509 2591 * access_args(char *path, int flags)
41c20dac 2592 *
9697c509 2593 * Check access permissions.
984263bc 2594 */
984263bc 2595int
753fd850 2596sys_access(struct access_args *uap)
984263bc 2597{
fad57d0e 2598 struct nlookupdata nd;
9697c509 2599 int error;
984263bc 2600
fad57d0e
MD
2601 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2602 if (error == 0)
18cf460b 2603 error = kern_access(&nd, uap->flags, 0);
fad57d0e 2604 nlookup_done(&nd);
984263bc
MD
2605 return (error);
2606}
2607
18cf460b
NT
2608
2609/*
2610 * faccessat_args(int fd, char *path, int amode, int flags)
2611 *
2612 * Check access permissions.
2613 */
2614int
2615sys_faccessat(struct faccessat_args *uap)
2616{
2617 struct nlookupdata nd;
2618 struct file *fp;
2619 int error;
2620
2621 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE,
2622 NLC_FOLLOW);
2623 if (error == 0)
2624 error = kern_access(&nd, uap->amode, uap->flags);
2625 nlookup_done_at(&nd, fp);
2626 return (error);
2627}
2628
2629
984263bc 2630int
21739618 2631kern_stat(struct nlookupdata *nd, struct stat *st)
984263bc 2632{
984263bc 2633 int error;
21739618 2634 struct vnode *vp;
984263bc 2635
21739618 2636 if ((error = nlookup(nd)) != 0)
984263bc 2637 return (error);
806dcf9a 2638again:
28623bf9 2639 if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
21739618
MD
2640 return (ENOENT);
2641
87de5057 2642 if ((error = vget(vp, LK_SHARED)) != 0)
21739618 2643 return (error);
87de5057 2644 error = vn_stat(vp, st, nd->nl_cred);
806dcf9a
MD
2645
2646 /*
2647 * If the file handle is stale we have to re-resolve the entry. This
2648 * is a hack at the moment.
2649 */
2650 if (error == ESTALE) {
5dc91765 2651 vput(vp);
28623bf9
MD
2652 cache_setunresolved(&nd->nl_nch);
2653 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
5dc91765 2654 if (error == 0)
806dcf9a 2655 goto again;
5dc91765
MD
2656 } else {
2657 vput(vp);
806dcf9a 2658 }
984263bc
MD
2659 return (error);
2660}
2661
2662/*
41c20dac
MD
2663 * stat_args(char *path, struct stat *ub)
2664 *
984263bc
MD
2665 * Get file status; this version follows links.
2666 */
984263bc 2667int
753fd850 2668sys_stat(struct stat_args *uap)
984263bc 2669{
21739618 2670 struct nlookupdata nd;
9697c509
DRJ
2671 struct stat st;
2672 int error;
984263bc 2673
21739618
MD
2674 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2675 if (error == 0) {
2676 error = kern_stat(&nd, &st);
2677 if (error == 0)
2678 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 2679 }
fad57d0e 2680 nlookup_done(&nd);
984263bc
MD
2681 return (error);
2682}
2683
2684/*
41c20dac
MD
2685 * lstat_args(char *path, struct stat *ub)
2686 *
984263bc
MD
2687 * Get file status; this version does not follow links.
2688 */
984263bc 2689int
753fd850 2690sys_lstat(struct lstat_args *uap)
984263bc 2691{
21739618 2692 struct nlookupdata nd;
9697c509
DRJ
2693 struct stat st;
2694 int error;
984263bc 2695
21739618
MD
2696 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2697 if (error == 0) {
2698 error = kern_stat(&nd, &st);
2699 if (error == 0)
2700 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 2701 }
fad57d0e 2702 nlookup_done(&nd);
984263bc
MD
2703 return (error);
2704}
2705
984263bc 2706/*
e101a3f4
NT
2707 * fstatat_args(int fd, char *path, struct stat *sb, int flags)
2708 *
2709 * Get status of file pointed to by fd/path.
2710 */
2711int
2712sys_fstatat(struct fstatat_args *uap)
2713{
2714 struct nlookupdata nd;
2715 struct stat st;
2716 int error;
2717 int flags;
2718 struct file *fp;
2719
dbeafe1e 2720 if (uap->flags & ~AT_SYMLINK_NOFOLLOW)
e101a3f4
NT
2721 return (EINVAL);
2722
2723 flags = (uap->flags & AT_SYMLINK_NOFOLLOW) ? 0 : NLC_FOLLOW;
2724
2725 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path,
2726 UIO_USERSPACE, flags);
2727 if (error == 0) {
2728 error = kern_stat(&nd, &st);
2729 if (error == 0)
2730 error = copyout(&st, uap->sb, sizeof(*uap->sb));
2731 }
2732 nlookup_done_at(&nd, fp);
2733 return (error);
2734}
2735
2736/*
41c20dac
MD
2737 * pathconf_Args(char *path, int name)
2738 *
984263bc
MD
2739 * Get configurable pathname variables.
2740 */
984263bc 2741int
753fd850 2742sys_pathconf(struct pathconf_args *uap)
984263bc 2743{
fad57d0e
MD
2744 struct nlookupdata nd;
2745 struct vnode *vp;
984263bc 2746 int error;
984263bc 2747
fad57d0e 2748 vp = NULL;
ab2eb4eb 2749 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2750 if (error == 0)
2751 error = nlookup(&nd);
2752 if (error == 0)
28623bf9 2753 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
fad57d0e
MD
2754 nlookup_done(&nd);
2755 if (error == 0) {
a385a915 2756 error = VOP_PATHCONF(vp, uap->name, &uap->sysmsg_reg);
fad57d0e
MD
2757 vput(vp);
2758 }
984263bc
MD
2759 return (error);
2760}
2761
bbab27f6
DRJ
2762/*
2763 * XXX: daver
2764 * kern_readlink isn't properly split yet. There is a copyin burried
2765 * in VOP_READLINK().
2766 */
984263bc 2767int
fad57d0e 2768kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
984263bc 2769{
dadab5e9 2770 struct thread *td = curthread;
41c20dac 2771 struct vnode *vp;
984263bc
MD
2772 struct iovec aiov;
2773 struct uio auio;
2774 int error;
984263bc 2775
fad57d0e
MD
2776 if ((error = nlookup(nd)) != 0)
2777 return (error);
28623bf9 2778 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 2779 if (error)
984263bc 2780 return (error);
fad57d0e 2781 if (vp->v_type != VLNK) {
984263bc 2782 error = EINVAL;
fad57d0e 2783 } else {
9697c509
DRJ
2784 aiov.iov_base = buf;
2785 aiov.iov_len = count;
984263bc
MD
2786 auio.uio_iov = &aiov;
2787 auio.uio_iovcnt = 1;
2788 auio.uio_offset = 0;
2789 auio.uio_rw = UIO_READ;
2790 auio.uio_segflg = UIO_USERSPACE;
dadab5e9 2791 auio.uio_td = td;
9697c509 2792 auio.uio_resid = count;
9910d07b 2793 error = VOP_READLINK(vp, &auio, td->td_ucred);
984263bc
MD
2794 }
2795 vput(vp);
9697c509
DRJ
2796 *res = count - auio.uio_resid;
2797 return (error);
2798}
2799
2800/*
2801 * readlink_args(char *path, char *buf, int count)
2802 *
2803 * Return target name of a symbolic link.
2804 */
2805int
753fd850 2806sys_readlink(struct readlink_args *uap)
9697c509 2807{
fad57d0e 2808 struct nlookupdata nd;
9697c509
DRJ
2809 int error;
2810
fad57d0e
MD
2811 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2812 if (error == 0) {
2813 error = kern_readlink(&nd, uap->buf, uap->count,
2814 &uap->sysmsg_result);
2815 }
2816 nlookup_done(&nd);
984263bc
MD
2817 return (error);
2818}
2819
6d98f93c
SW
2820/*
2821 * readlinkat_args(int fd, char *path, char *buf, size_t bufsize)
2822 *
2823 * Return target name of a symbolic link. The path is relative to the
2824 * directory associated with fd.
6d98f93c
SW
2825 */
2826int
2827sys_readlinkat(struct readlinkat_args *uap)
2828{
2829 struct nlookupdata nd;
2830 struct file *fp;
2831 int error;
2832
6d98f93c
SW
2833 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path, UIO_USERSPACE, 0);
2834 if (error == 0) {
2835 error = kern_readlink(&nd, uap->buf, uap->bufsize,
2836 &uap->sysmsg_result);
2837 }
2838 nlookup_done_at(&nd, fp);
6d98f93c
SW
2839 return (error);
2840}
2841
984263bc 2842static int
41c20dac 2843setfflags(struct vnode *vp, int flags)
984263bc 2844{
dadab5e9 2845 struct thread *td = curthread;
984263bc
MD
2846 int error;
2847 struct vattr vattr;
2848
2849 /*
2850 * Prevent non-root users from setting flags on devices. When
2851 * a device is reused, users can retain ownership of the device
2852 * if they are allowed to set flags and programs assume that
2853 * chown can't fail when done as root.
2854 */
2855 if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
9910d07b 2856 ((error = priv_check_cred(td->td_ucred, PRIV_VFS_CHFLAGS_DEV, 0)) != 0))
984263bc
MD
2857 return (error);
2858
186fccd6
MD
2859 /*
2860 * note: vget is required for any operation that might mod the vnode
2861 * so VINACTIVE is properly cleared.
2862 */
87de5057 2863 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2864 VATTR_NULL(&vattr);
2865 vattr.va_flags = flags;
9910d07b 2866 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
186fccd6
MD
2867 vput(vp);
2868 }
984263bc
MD
2869 return (error);
2870}
2871
2872/*
41c20dac
MD
2873 * chflags(char *path, int flags)
2874 *
984263bc
MD
2875 * Change flags of a file given a path name.
2876 */
984263bc 2877int
753fd850 2878sys_chflags(struct chflags_args *uap)
984263bc 2879{
fad57d0e
MD
2880 struct nlookupdata nd;
2881 struct vnode *vp;
984263bc 2882 int error;
984263bc 2883
fad57d0e 2884 vp = NULL;
ab2eb4eb 2885 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2886 if (error == 0)
2887 error = nlookup(&nd);
2888 if (error == 0)
28623bf9 2889 error = ncp_writechk(&nd.nl_nch);
72310cfb 2890 if (error == 0)
28623bf9 2891 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
fad57d0e
MD
2892 nlookup_done(&nd);
2893 if (error == 0) {
ab2eb4eb 2894 error = setfflags(vp, uap->flags);
fad57d0e
MD
2895 vrele(vp);
2896 }
2897 return (error);
984263bc
MD
2898}
2899
2900/*
91fdc001
PA
2901 * lchflags(char *path, int flags)
2902 *
2903 * Change flags of a file given a path name, but don't follow symlinks.
2904 */
91fdc001
PA
2905int
2906sys_lchflags(struct lchflags_args *uap)
2907{
2908 struct nlookupdata nd;
2909 struct vnode *vp;
2910 int error;
2911
2912 vp = NULL;
2913 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
91fdc001
PA
2914 if (error == 0)
2915 error = nlookup(&nd);
2916 if (error == 0)
2917 error = ncp_writechk(&nd.nl_nch);
2918 if (error == 0)
2919 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
2920 nlookup_done(&nd);
2921 if (error == 0) {
2922 error = setfflags(vp, uap->flags);
2923 vrele(vp);
2924 }
2925 return (error);
2926}
2927
2928/*
41c20dac
MD
2929 * fchflags_args(int fd, int flags)
2930 *
984263bc
MD
2931 * Change flags of a file given a file descriptor.
2932 */
984263bc 2933int
753fd850 2934sys_fchflags(struct fchflags_args *uap)
984263bc 2935{
dadab5e9
MD
2936 struct thread *td = curthread;
2937 struct proc *p = td->td_proc;
984263bc
MD
2938 struct file *fp;
2939 int error;
2940
5b287bba 2941 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2942 return (error);
28623bf9
MD
2943 if (fp->f_nchandle.ncp)
2944 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
2945 if (error == 0)
2946 error = setfflags((struct vnode *) fp->f_data, uap->flags);
5b287bba
MD
2947 fdrop(fp);
2948 return (error);
984263bc
MD
2949}
2950
2951static int
41c20dac 2952setfmode(struct vnode *vp, int mode)
984263bc 2953{
dadab5e9 2954 struct thread *td = curthread;
984263bc
MD
2955 int error;
2956 struct vattr vattr;
2957
186fccd6
MD
2958 /*
2959 * note: vget is required for any operation that might mod the vnode
2960 * so VINACTIVE is properly cleared.
2961 */
87de5057 2962 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2963 VATTR_NULL(&vattr);
2964 vattr.va_mode = mode & ALLPERMS;
9910d07b 2965 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
186fccd6
MD
2966 vput(vp);
2967 }
984263bc
MD
2968 return error;
2969}
2970
9697c509 2971int
fad57d0e 2972kern_chmod(struct nlookupdata *nd, int mode)
9697c509 2973{
fad57d0e 2974 struct vnode *vp;
9697c509
DRJ
2975 int error;
2976
fad57d0e 2977 if ((error = nlookup(nd)) != 0)
9697c509 2978 return (error);
28623bf9 2979 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 2980 return (error);
28623bf9 2981 if ((error = ncp_writechk(&nd->nl_nch)) == 0)
72310cfb 2982 error = setfmode(vp, mode);
fad57d0e
MD
2983 vrele(vp);
2984 return (error);
9697c509
DRJ
2985}
2986
984263bc 2987/*
41c20dac
MD
2988 * chmod_args(char *path, int mode)
2989 *
984263bc
MD
2990 * Change mode of a file given path name.
2991 */
984263bc 2992int
753fd850 2993sys_chmod(struct chmod_args *uap)
984263bc 2994{
fad57d0e 2995 struct nlookupdata nd;
9697c509 2996 int error;
984263bc 2997
fad57d0e
MD
2998 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2999 if (error == 0)
3000 error = kern_chmod(&nd, uap->mode);
3001 nlookup_done(&nd);
9697c509 3002 return (error);
984263bc
MD
3003}
3004
3005/*
41c20dac
MD
3006 * lchmod_args(char *path, int mode)
3007 *
984263bc
MD
3008 * Change mode of a file given path name (don't follow links.)
3009 */
984263bc 3010int
753fd850 3011sys_lchmod(struct lchmod_args *uap)
984263bc 3012{
fad57d0e 3013 struct nlookupdata nd;
984263bc 3014 int error;
984263bc 3015
fad57d0e
MD
3016 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3017 if (error == 0)
3018 error = kern_chmod(&nd, uap->mode);
3019 nlookup_done(&nd);
3020 return (error);
984263bc
MD
3021}
3022
3023/*
41c20dac
MD
3024 * fchmod_args(int fd, int mode)
3025 *
984263bc
MD
3026 * Change mode of a file given a file descriptor.
3027 */
984263bc 3028int
753fd850 3029sys_fchmod(struct fchmod_args *uap)
984263bc 3030{
dadab5e9
MD
3031 struct thread *td = curthread;
3032 struct proc *p = td->td_proc;
984263bc
MD
3033 struct file *fp;
3034 int error;
3035
5b287bba 3036 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 3037 return (error);
28623bf9
MD
3038 if (fp->f_nchandle.ncp)
3039 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
3040 if (error == 0)
3041 error = setfmode((struct vnode *)fp->f_data, uap->mode);
5b287bba
MD
3042 fdrop(fp);
3043 return (error);
984263bc
MD
3044}
3045
8cde3c79
NT
3046/*
3047 * fchmodat_args(char *path, int mode)
3048 *
3049 * Change mode of a file pointed to by fd/path.
3050 */
3051int
3052sys_fchmodat(struct fchmodat_args *uap)
3053{
3054 struct nlookupdata nd;
3055 struct file *fp;
3056 int error;
3057 int flags;
3058
dbeafe1e 3059 if (uap->flags & ~AT_SYMLINK_NOFOLLOW)
8cde3c79
NT
3060 return (EINVAL);
3061 flags = (uap->flags & AT_SYMLINK_NOFOLLOW) ? 0 : NLC_FOLLOW;
3062
3063 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path,
3064 UIO_USERSPACE, flags);
3065 if (error == 0)
3066 error = kern_chmod(&nd, uap->mode);
3067 nlookup_done_at(&nd, fp);
3068 return (error);
3069}
3070
984263bc 3071static int
41c20dac 3072setfown(struct vnode *vp, uid_t uid, gid_t gid)
984263bc 3073{
dadab5e9 3074 struct thread *td = curthread;
984263bc
MD
3075 int error;
3076 struct vattr vattr;
3077
186fccd6
MD
3078 /*
3079 * note: vget is required for any operation that might mod the vnode
3080 * so VINACTIVE is properly cleared.
3081 */
87de5057 3082 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
3083 VATTR_NULL(&vattr);
3084 vattr.va_uid = uid;
3085 vattr.va_gid = gid;
9910d07b 3086 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
186fccd6
MD
3087 vput(vp);
3088 }
984263bc
MD
3089 return error;
3090}
3091
9697c509 3092int
fad57d0e 3093kern_chown(struct nlookupdata *nd, int uid, int gid)
9697c509 3094{
fad57d0e 3095 struct vnode *vp;
9697c509
DRJ
3096 int error;
3097
fad57d0e 3098 if ((error = nlookup(nd)) != 0)
9697c509 3099 return (error);
28623bf9 3100 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 3101 return (error);
28623bf9 3102 if ((error = ncp_writechk(&nd->nl_nch)) == 0)
72310cfb 3103 error = setfown(vp, uid, gid);
fad57d0e 3104 vrele(vp);
9697c509
DRJ
3105 return (error);
3106}
3107
984263bc 3108/*
41c20dac
MD
3109 * chown(char *path, int uid, int gid)
3110 *
984263bc
MD
3111 * Set ownership given a path name.
3112 */
984263bc 3113int
753fd850 3114sys_chown(struct chown_args *uap)
984263bc 3115{
fad57d0e 3116 struct nlookupdata nd;
9697c509
DRJ
3117 int error;
3118
fad57d0e
MD
3119 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3120 if (error == 0)
3121 error = kern_chown(&nd, uap->uid, uap->gid);
3122 nlookup_done(&nd);
984263bc
MD
3123 return (error);
3124}
3125
3126/*
41c20dac
MD
3127 * lchown_args(char *path, int uid, int gid)
3128 *
984263bc
MD
3129 * Set ownership given a path name, do not cross symlinks.
3130 */
984263bc 3131int
753fd850 3132sys_lchown(struct lchown_args *uap)
984263bc 3133{
fad57d0e 3134 struct nlookupdata nd;
984263bc 3135 int error;
9697c509 3136
fad57d0e
MD
3137 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3138 if (error == 0)
3139 error = kern_chown(&nd, uap->uid, uap->gid);
3140 nlookup_done(&nd);
984263bc
MD
3141 return (error);
3142}
3143
3144/*
41c20dac
MD
3145 * fchown_args(int fd, int uid, int gid)
3146 *
984263bc
MD
3147 * Set ownership given a file descriptor.
3148 */
984263bc 3149int
753fd850 3150sys_fchown(struct fchown_args *uap)
984263bc 3151{
dadab5e9
MD
3152 struct thread *td = curthread;
3153 struct proc *p = td->td_proc;
984263bc
MD
3154 struct file *fp;
3155 int error;
3156
5b287bba 3157 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 3158 return (error);
28623bf9
MD
3159 if (fp->f_nchandle.ncp)
3160 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
3161 if (error == 0)
3162 error = setfown((struct vnode *)fp->f_data, uap->uid, uap->gid);
5b287bba
MD
3163 fdrop(fp);
3164 return (error);
984263bc
MD
3165}
3166
32013176
NT
3167/*
3168 * fchownat(int fd, char *path, int uid, int gid, int flags)
3169 *
3170 * Set ownership of file pointed to by fd/path.
3171 */
3172int
3173sys_fchownat(struct fchownat_args *uap)
3174{
3175 struct nlookupdata nd;
3176 struct file *fp;
3177 int error;
3178 int flags;
3179
dbeafe1e 3180 if (uap->flags & ~AT_SYMLINK_NOFOLLOW)
32013176
NT
3181 return (EINVAL);
3182 flags = (uap->flags & AT_SYMLINK_NOFOLLOW) ? 0 : NLC_FOLLOW;
3183
3184 error = nlookup_init_at(&nd, &fp, uap->fd, uap->path,
3185 UIO_USERSPACE, flags);
3186 if (error == 0)
3187 error = kern_chown(&nd, uap->uid, uap->gid);
3188 nlookup_done_at(&nd, fp);
3189 return (error);
3190}
3191
3192
984263bc 3193static int
9697c509 3194getutimes(const struct timeval *tvp, struct timespec *tsp)
984263bc
MD
3195{
3196 struct timeval tv[2];
984263bc 3197
9697c509 3198 if (tvp == NULL) {
984263bc
MD
3199 microtime(&tv[0]);
3200 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
3201 tsp[1] = tsp[0];
3202 } else {
9697c509
DRJ
3203 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3204 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
984263bc
MD
3205 }
3206 return 0;
3207}
3208
3209static int
3a907475
MD
3210setutimes(struct vnode *vp, struct vattr *vattr,
3211 const struct timespec *ts, int nullflag)
984263bc 3212{
dadab5e9 3213 struct thread *td = curthread;
984263bc 3214 int error;
984263bc 3215
3a907475
MD
3216 VATTR_NULL(vattr);
3217 vattr->va_atime = ts[0];
3218 vattr->va_mtime = ts[1];
3219 if (nullflag)
3220 vattr->va_vaflags |= VA_UTIMES_NULL;
9910d07b 3221 error = VOP_SETATTR(vp, vattr, td->td_ucred);
3a907475 3222
984263bc
MD
3223 return error;
3224}
3225
9697c509 3226int
fad57d0e 3227kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
9697c509
DRJ
3228{
3229 struct timespec ts[2];
fad57d0e 3230 struct vnode *vp;
3a907475 3231 struct vattr vattr;
9697c509
DRJ
3232 int error;
3233
fad57d0e 3234 if ((error = getutimes(tptr, ts)) != 0)
9697c509 3235 return (error);
3a907475
MD
3236
3237 /*
3238 * NOTE: utimes() succeeds for the owner even if the file
3239 * is not user-writable.
3240 */
3241 nd->nl_flags |= NLC_OWN | NLC_WRITE;
3242
fad57d0e 3243 if ((error = nlookup(nd)) != 0)
9697c509 3244 return (error);
28623bf9 3245 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 3246 return (error);
28623bf9 3247 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 3248 return (error);
ee89633d
MD
3249
3250 /*
3a907475
MD
3251 * note: vget is required for any operation that might mod the vnode
3252 * so VINACTIVE is properly cleared.
ee89633d 3253 */