- Add devq which stores mbuf dequeued from ifnet.if_snd during ifnet.if_sart.
[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 $
5312fa43 40 * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.126 2008/05/09 17:52:17 dillon Exp $
984263bc
MD
41 */
42
984263bc
MD
43#include <sys/param.h>
44#include <sys/systm.h>
45#include <sys/buf.h>
16df2c74 46#include <sys/conf.h>
984263bc
MD
47#include <sys/sysent.h>
48#include <sys/malloc.h>
49#include <sys/mount.h>
432b8263 50#include <sys/mountctl.h>
984263bc 51#include <sys/sysproto.h>
984263bc
MD
52#include <sys/filedesc.h>
53#include <sys/kernel.h>
54#include <sys/fcntl.h>
55#include <sys/file.h>
56#include <sys/linker.h>
57#include <sys/stat.h>
58#include <sys/unistd.h>
59#include <sys/vnode.h>
60#include <sys/proc.h>
dadab5e9 61#include <sys/namei.h>
21739618 62#include <sys/nlookup.h>
984263bc
MD
63#include <sys/dirent.h>
64#include <sys/extattr.h>
8fa76237 65#include <sys/spinlock.h>
8f6f8622 66#include <sys/kern_syscall.h>
70aac194 67#include <sys/objcache.h>
984263bc 68#include <sys/sysctl.h>
408357d8
MD
69
70#include <sys/buf2.h>
28623bf9
MD
71#include <sys/file2.h>
72#include <sys/spinlock2.h>
73
984263bc
MD
74#include <vm/vm.h>
75#include <vm/vm_object.h>
984263bc
MD
76#include <vm/vm_page.h>
77
28623bf9
MD
78#include <machine/limits.h>
79#include <machine/stdarg.h>
80
81#include <vfs/union/union.h>
dadab5e9 82
28623bf9 83static void mount_warning(struct mount *mp, const char *ctl, ...);
d3c546e6 84static int mount_path(struct proc *p, struct mount *mp, char **rb, char **fb);
e24b948e 85static int checkvp_chdir (struct vnode *vn, struct thread *td);
77480a97 86static void checkdirs (struct nchandle *old_nch, struct nchandle *new_nch);
402ed7e1 87static int chroot_refuse_vdir_fds (struct filedesc *fdp);
75ffff0d 88static int chroot_visible_mnt(struct mount *mp, struct proc *p);
402ed7e1
RG
89static int getutimes (const struct timeval *, struct timespec *);
90static int setfown (struct vnode *, uid_t, gid_t);
91static int setfmode (struct vnode *, int);
92static int setfflags (struct vnode *, int);
93static int setutimes (struct vnode *, const struct timespec *, int);
984263bc
MD
94static int usermount = 0; /* if 1, non-root can mount fs. */
95
402ed7e1 96int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
984263bc
MD
97
98SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
99
100/*
101 * Virtual File System System Calls
102 */
103
104/*
105 * Mount a file system.
106 */
41c20dac
MD
107/*
108 * mount_args(char *type, char *path, int flags, caddr_t data)
109 */
984263bc
MD
110/* ARGSUSED */
111int
753fd850 112sys_mount(struct mount_args *uap)
984263bc 113{
dadab5e9
MD
114 struct thread *td = curthread;
115 struct proc *p = td->td_proc;
984263bc 116 struct vnode *vp;
28623bf9 117 struct nchandle nch;
984263bc
MD
118 struct mount *mp;
119 struct vfsconf *vfsp;
120 int error, flag = 0, flag2 = 0;
1d505369 121 int hasmount;
984263bc 122 struct vattr va;
21739618 123 struct nlookupdata nd;
984263bc 124 char fstypename[MFSNAMELEN];
acde96db 125 struct ucred *cred = p->p_ucred;
984263bc 126
cd0d3e17 127 KKASSERT(p);
acde96db 128 if (cred->cr_prison != NULL)
84c3a710 129 return (EPERM);
dadab5e9 130 if (usermount == 0 && (error = suser(td)))
984263bc
MD
131 return (error);
132 /*
133 * Do not allow NFS export by non-root users.
134 */
ab2eb4eb 135 if (uap->flags & MNT_EXPORTED) {
dadab5e9 136 error = suser(td);
984263bc
MD
137 if (error)
138 return (error);
139 }
140 /*
141 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
142 */
dadab5e9 143 if (suser(td))
ab2eb4eb 144 uap->flags |= MNT_NOSUID | MNT_NODEV;
21739618 145
984263bc 146 /*
28623bf9 147 * Lookup the requested path and extract the nch and vnode.
984263bc 148 */
21739618
MD
149 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
150 if (error == 0) {
151 if ((error = nlookup(&nd)) == 0) {
28623bf9 152 if (nd.nl_nch.ncp->nc_vp == NULL)
21739618
MD
153 error = ENOENT;
154 }
155 }
156 if (error) {
157 nlookup_done(&nd);
984263bc 158 return (error);
21739618
MD
159 }
160
161 /*
162 * Extract the locked+refd ncp and cleanup the nd structure
163 */
28623bf9
MD
164 nch = nd.nl_nch;
165 cache_zero(&nd.nl_nch);
21739618
MD
166 nlookup_done(&nd);
167
28623bf9 168 if ((nch.ncp->nc_flag & NCF_ISMOUNTPT) && cache_findmount(&nch))
1d505369
MD
169 hasmount = 1;
170 else
171 hasmount = 0;
172
173
21739618 174 /*
28623bf9 175 * now we have the locked ref'd nch and unreferenced vnode.
21739618 176 */
28623bf9 177 vp = nch.ncp->nc_vp;
87de5057 178 if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
28623bf9 179 cache_put(&nch);
21739618
MD
180 return (error);
181 }
28623bf9 182 cache_unlock(&nch);
21739618
MD
183
184 /*
28623bf9 185 * Now we have an unlocked ref'd nch and a locked ref'd vp
21739618 186 */
ab2eb4eb 187 if (uap->flags & MNT_UPDATE) {
984263bc 188 if ((vp->v_flag & VROOT) == 0) {
28623bf9 189 cache_drop(&nch);
984263bc
MD
190 vput(vp);
191 return (EINVAL);
192 }
193 mp = vp->v_mount;
194 flag = mp->mnt_flag;
195 flag2 = mp->mnt_kern_flag;
196 /*
197 * We only allow the filesystem to be reloaded if it
198 * is currently mounted read-only.
199 */
ab2eb4eb 200 if ((uap->flags & MNT_RELOAD) &&
984263bc 201 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
28623bf9 202 cache_drop(&nch);
984263bc
MD
203 vput(vp);
204 return (EOPNOTSUPP); /* Needs translation */
205 }
206 /*
207 * Only root, or the user that did the original mount is
208 * permitted to update it.
209 */
acde96db 210 if (mp->mnt_stat.f_owner != cred->cr_uid &&
dadab5e9 211 (error = suser(td))) {
28623bf9 212 cache_drop(&nch);
984263bc
MD
213 vput(vp);
214 return (error);
215 }
f9642f56 216 if (vfs_busy(mp, LK_NOWAIT)) {
28623bf9 217 cache_drop(&nch);
984263bc
MD
218 vput(vp);
219 return (EBUSY);
220 }
1d505369 221 if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
28623bf9 222 cache_drop(&nch);
f9642f56 223 vfs_unbusy(mp);
984263bc
MD
224 vput(vp);
225 return (EBUSY);
226 }
227 vp->v_flag |= VMOUNT;
984263bc 228 mp->mnt_flag |=
ab2eb4eb 229 uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
a11aaa81 230 vn_unlock(vp);
984263bc
MD
231 goto update;
232 }
233 /*
234 * If the user is not root, ensure that they own the directory
235 * onto which we are attempting to mount.
236 */
87de5057 237 if ((error = VOP_GETATTR(vp, &va)) ||
acde96db 238 (va.va_uid != cred->cr_uid && (error = suser(td)))) {
28623bf9 239 cache_drop(&nch);
984263bc
MD
240 vput(vp);
241 return (error);
242 }
87de5057 243 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
28623bf9 244 cache_drop(&nch);
984263bc
MD
245 vput(vp);
246 return (error);
247 }
248 if (vp->v_type != VDIR) {
28623bf9 249 cache_drop(&nch);
984263bc
MD
250 vput(vp);
251 return (ENOTDIR);
252 }
5dc91765
MD
253 if (vp->v_mount->mnt_kern_flag & MNTK_NOSTKMNT) {
254 cache_drop(&nch);
255 vput(vp);
256 return (EPERM);
257 }
ab2eb4eb 258 if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
28623bf9 259 cache_drop(&nch);
984263bc
MD
260 vput(vp);
261 return (error);
262 }
21739618 263 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
984263bc
MD
264 if (!strcmp(vfsp->vfc_name, fstypename))
265 break;
21739618 266 }
984263bc
MD
267 if (vfsp == NULL) {
268 linker_file_t lf;
269
270 /* Only load modules for root (very important!) */
dadab5e9 271 if ((error = suser(td)) != 0) {
28623bf9 272 cache_drop(&nch);
984263bc
MD
273 vput(vp);
274 return error;
275 }
276 error = linker_load_file(fstypename, &lf);
277 if (error || lf == NULL) {
28623bf9 278 cache_drop(&nch);
984263bc
MD
279 vput(vp);
280 if (lf == NULL)
281 error = ENODEV;
282 return error;
283 }
284 lf->userrefs++;
285 /* lookup again, see if the VFS was loaded */
21739618 286 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
984263bc
MD
287 if (!strcmp(vfsp->vfc_name, fstypename))
288 break;
21739618 289 }
984263bc
MD
290 if (vfsp == NULL) {
291 lf->userrefs--;
292 linker_file_unload(lf);
28623bf9 293 cache_drop(&nch);
984263bc
MD
294 vput(vp);
295 return (ENODEV);
296 }
297 }
1d505369 298 if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
28623bf9 299 cache_drop(&nch);
984263bc
MD
300 vput(vp);
301 return (EBUSY);
302 }
303 vp->v_flag |= VMOUNT;
984263bc
MD
304
305 /*
306 * Allocate and initialize the filesystem.
307 */
efda3bd0 308 mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
984263bc
MD
309 TAILQ_INIT(&mp->mnt_nvnodelist);
310 TAILQ_INIT(&mp->mnt_reservedvnlist);
2281065e 311 TAILQ_INIT(&mp->mnt_jlist);
984263bc 312 mp->mnt_nvnodelistsize = 0;
ab6f251b 313 lockinit(&mp->mnt_lock, "vfslock", 0, 0);
f9642f56 314 vfs_busy(mp, LK_NOWAIT);
984263bc
MD
315 mp->mnt_op = vfsp->vfc_vfsops;
316 mp->mnt_vfc = vfsp;
317 vfsp->vfc_refcount++;
318 mp->mnt_stat.f_type = vfsp->vfc_typenum;
319 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
320 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
acde96db 321 mp->mnt_stat.f_owner = cred->cr_uid;
984263bc 322 mp->mnt_iosize_max = DFLTPHYS;
a11aaa81 323 vn_unlock(vp);
984263bc
MD
324update:
325 /*
326 * Set the mount level flags.
327 */
ab2eb4eb 328 if (uap->flags & MNT_RDONLY)
984263bc
MD
329 mp->mnt_flag |= MNT_RDONLY;
330 else if (mp->mnt_flag & MNT_RDONLY)
331 mp->mnt_kern_flag |= MNTK_WANTRDWR;
332 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
333 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
334 MNT_NOSYMFOLLOW | MNT_IGNORE |
335 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
ab2eb4eb 336 mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
984263bc
MD
337 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
338 MNT_NOSYMFOLLOW | MNT_IGNORE |
339 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
340 /*
341 * Mount the filesystem.
342 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
fad57d0e 343 * get.
984263bc 344 */
acde96db 345 error = VFS_MOUNT(mp, uap->path, uap->data, cred);
984263bc
MD
346 if (mp->mnt_flag & MNT_UPDATE) {
347 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
348 mp->mnt_flag &= ~MNT_RDONLY;
349 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
350 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
351 if (error) {
352 mp->mnt_flag = flag;
353 mp->mnt_kern_flag = flag2;
354 }
f9642f56 355 vfs_unbusy(mp);
984263bc 356 vp->v_flag &= ~VMOUNT;
984263bc 357 vrele(vp);
28623bf9 358 cache_drop(&nch);
984263bc
MD
359 return (error);
360 }
ca466bae 361 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
984263bc 362 /*
21739618 363 * Put the new filesystem on the mount list after root. The mount
28623bf9
MD
364 * point gets its own mnt_ncmountpt (unless the VFS already set one
365 * up) which represents the root of the mount. The lookup code
366 * detects the mount point going forward and checks the root of
367 * the mount going backwards.
8e005a45
MD
368 *
369 * It is not necessary to invalidate or purge the vnode underneath
370 * because elements under the mount will be given their own glue
371 * namecache record.
984263bc 372 */
984263bc 373 if (!error) {
28623bf9
MD
374 if (mp->mnt_ncmountpt.ncp == NULL) {
375 /*
376 * allocate, then unlock, but leave the ref intact
377 */
378 cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
379 cache_unlock(&mp->mnt_ncmountpt);
380 }
381 mp->mnt_ncmounton = nch; /* inherits ref */
382 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
383
384 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
984263bc 385 vp->v_flag &= ~VMOUNT;
861905fb 386 mountlist_insert(mp, MNTINS_LAST);
77480a97 387 checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
a11aaa81 388 vn_unlock(vp);
41a01a4d 389 error = vfs_allocate_syncvnode(mp);
f9642f56 390 vfs_unbusy(mp);
1d505369
MD
391 error = VFS_START(mp, 0);
392 vrele(vp);
984263bc 393 } else {
66a1ddf5
MD
394 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
395 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
396 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
397 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
398 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
984263bc 399 vp->v_flag &= ~VMOUNT;
984263bc 400 mp->mnt_vfc->vfc_refcount--;
f9642f56 401 vfs_unbusy(mp);
efda3bd0 402 kfree(mp, M_MOUNT);
28623bf9 403 cache_drop(&nch);
984263bc
MD
404 vput(vp);
405 }
406 return (error);
407}
408
409/*
410 * Scan all active processes to see if any of them have a current
411 * or root directory onto which the new filesystem has just been
412 * mounted. If so, replace them with the new mount point.
21739618
MD
413 *
414 * The passed ncp is ref'd and locked (from the mount code) and
415 * must be associated with the vnode representing the root of the
416 * mount point.
984263bc 417 */
8fa76237 418struct checkdirs_info {
77480a97
MD
419 struct nchandle old_nch;
420 struct nchandle new_nch;
421 struct vnode *old_vp;
422 struct vnode *new_vp;
8fa76237
MD
423};
424
425static int checkdirs_callback(struct proc *p, void *data);
426
984263bc 427static void
77480a97 428checkdirs(struct nchandle *old_nch, struct nchandle *new_nch)
984263bc 429{
8fa76237 430 struct checkdirs_info info;
77480a97 431 struct vnode *olddp;
984263bc 432 struct vnode *newdp;
690a3127 433 struct mount *mp;
984263bc 434
77480a97
MD
435 /*
436 * If the old mount point's vnode has a usecount of 1, it is not
437 * being held as a descriptor anywhere.
438 */
439 olddp = old_nch->ncp->nc_vp;
3c37c940 440 if (olddp == NULL || olddp->v_sysref.refcnt == 1)
984263bc 441 return;
77480a97
MD
442
443 /*
444 * Force the root vnode of the new mount point to be resolved
445 * so we can update any matching processes.
446 */
447 mp = new_nch->mount;
690a3127 448 if (VFS_ROOT(mp, &newdp))
984263bc 449 panic("mount: lost mount");
77480a97
MD
450 cache_setunresolved(new_nch);
451 cache_setvp(new_nch, newdp);
21739618 452
77480a97
MD
453 /*
454 * Special handling of the root node
455 */
21739618
MD
456 if (rootvnode == olddp) {
457 vref(newdp);
77480a97 458 vfs_cache_setroot(newdp, cache_hold(new_nch));
21739618
MD
459 }
460
77480a97
MD
461 /*
462 * Pass newdp separately so the callback does not have to access
463 * it via new_nch->ncp->nc_vp.
464 */
465 info.old_nch = *old_nch;
466 info.new_nch = *new_nch;
467 info.new_vp = newdp;
8fa76237
MD
468 allproc_scan(checkdirs_callback, &info);
469 vput(newdp);
470}
471
472/*
473 * NOTE: callback is not MP safe because the scanned process's filedesc
474 * structure can be ripped out from under us, amoung other things.
475 */
476static int
477checkdirs_callback(struct proc *p, void *data)
478{
479 struct checkdirs_info *info = data;
480 struct filedesc *fdp;
28623bf9
MD
481 struct nchandle ncdrop1;
482 struct nchandle ncdrop2;
8fa76237
MD
483 struct vnode *vprele1;
484 struct vnode *vprele2;
485
486 if ((fdp = p->p_fd) != NULL) {
28623bf9
MD
487 cache_zero(&ncdrop1);
488 cache_zero(&ncdrop2);
8fa76237
MD
489 vprele1 = NULL;
490 vprele2 = NULL;
491
492 /*
493 * MPUNSAFE - XXX fdp can be pulled out from under a
494 * foreign process.
495 *
496 * A shared filedesc is ok, we don't have to copy it
497 * because we are making this change globally.
498 */
499 spin_lock_wr(&fdp->fd_spin);
77480a97
MD
500 if (fdp->fd_ncdir.mount == info->old_nch.mount &&
501 fdp->fd_ncdir.ncp == info->old_nch.ncp) {
8fa76237 502 vprele1 = fdp->fd_cdir;
77480a97
MD
503 vref(info->new_vp);
504 fdp->fd_cdir = info->new_vp;
8fa76237 505 ncdrop1 = fdp->fd_ncdir;
77480a97 506 cache_copy(&info->new_nch, &fdp->fd_ncdir);
984263bc 507 }
77480a97
MD
508 if (fdp->fd_nrdir.mount == info->old_nch.mount &&
509 fdp->fd_nrdir.ncp == info->old_nch.ncp) {
8fa76237 510 vprele2 = fdp->fd_rdir;
77480a97
MD
511 vref(info->new_vp);
512 fdp->fd_rdir = info->new_vp;
8fa76237 513 ncdrop2 = fdp->fd_nrdir;
77480a97 514 cache_copy(&info->new_nch, &fdp->fd_nrdir);
984263bc 515 }
8fa76237 516 spin_unlock_wr(&fdp->fd_spin);
28623bf9
MD
517 if (ncdrop1.ncp)
518 cache_drop(&ncdrop1);
519 if (ncdrop2.ncp)
520 cache_drop(&ncdrop2);
8fa76237
MD
521 if (vprele1)
522 vrele(vprele1);
523 if (vprele2)
524 vrele(vprele2);
984263bc 525 }
8fa76237 526 return(0);
984263bc
MD
527}
528
529/*
530 * Unmount a file system.
531 *
532 * Note: unmount takes a path to the vnode mounted on as argument,
533 * not special file (as before).
534 */
41c20dac
MD
535/*
536 * umount_args(char *path, int flags)
537 */
984263bc
MD
538/* ARGSUSED */
539int
753fd850 540sys_unmount(struct unmount_args *uap)
984263bc 541{
dadab5e9
MD
542 struct thread *td = curthread;
543 struct proc *p = td->td_proc;
d2293868 544 struct mount *mp = NULL;
984263bc 545 int error;
fad57d0e 546 struct nlookupdata nd;
984263bc 547
dadab5e9 548 KKASSERT(p);
84c3a710
HP
549 if (p->p_ucred->cr_prison != NULL)
550 return (EPERM);
551 if (usermount == 0 && (error = suser(td)))
552 return (error);
553
ab2eb4eb 554 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
555 if (error == 0)
556 error = nlookup(&nd);
fad57d0e 557 if (error)
d2293868 558 goto out;
fad57d0e 559
28623bf9 560 mp = nd.nl_nch.mount;
984263bc
MD
561
562 /*
563 * Only root, or the user that did the original mount is
564 * permitted to unmount this filesystem.
565 */
566 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
d2293868
MD
567 (error = suser(td)))
568 goto out;
984263bc
MD
569
570 /*
571 * Don't allow unmounting the root file system.
572 */
573 if (mp->mnt_flag & MNT_ROOTFS) {
d2293868
MD
574 error = EINVAL;
575 goto out;
984263bc
MD
576 }
577
578 /*
579 * Must be the root of the filesystem
580 */
28623bf9 581 if (nd.nl_nch.ncp != mp->mnt_ncmountpt.ncp) {
d2293868
MD
582 error = EINVAL;
583 goto out;
984263bc 584 }
d2293868
MD
585
586out:
587 nlookup_done(&nd);
588 if (error)
589 return (error);
acde96db 590 return (dounmount(mp, uap->flags));
984263bc
MD
591}
592
593/*
594 * Do the actual file system unmount.
595 */
861905fb
MD
596static int
597dounmount_interlock(struct mount *mp)
598{
599 if (mp->mnt_kern_flag & MNTK_UNMOUNT)
600 return (EBUSY);
601 mp->mnt_kern_flag |= MNTK_UNMOUNT;
602 return(0);
603}
604
984263bc 605int
acde96db 606dounmount(struct mount *mp, int flags)
984263bc 607{
dcf26f79 608 struct namecache *ncp;
28623bf9 609 struct nchandle nch;
984263bc
MD
610 int error;
611 int async_flag;
2df9419c 612 int lflags;
28623bf9 613 int freeok = 1;
984263bc 614
861905fb
MD
615 /*
616 * Exclusive access for unmounting purposes
617 */
618 if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
619 return (error);
620
621 /*
622 * Allow filesystems to detect that a forced unmount is in progress.
623 */
984263bc
MD
624 if (flags & MNT_FORCE)
625 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
2df9419c 626 lflags = LK_EXCLUSIVE | ((flags & MNT_FORCE) ? 0 : LK_NOWAIT);
df4f70a6 627 error = lockmgr(&mp->mnt_lock, lflags);
984263bc
MD
628 if (error) {
629 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
630 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 631 wakeup(mp);
984263bc
MD
632 return (error);
633 }
634
635 if (mp->mnt_flag & MNT_EXPUBLIC)
636 vfs_setpublicfs(NULL, NULL, NULL);
637
638 vfs_msync(mp, MNT_WAIT);
639 async_flag = mp->mnt_flag & MNT_ASYNC;
640 mp->mnt_flag &=~ MNT_ASYNC;
dcf26f79
MD
641
642 /*
28623bf9
MD
643 * If this filesystem isn't aliasing other filesystems,
644 * try to invalidate any remaining namecache entries and
645 * check the count afterwords.
dcf26f79 646 */
28623bf9
MD
647 if ((mp->mnt_kern_flag & MNTK_NCALIASED) == 0) {
648 cache_lock(&mp->mnt_ncmountpt);
649 cache_inval(&mp->mnt_ncmountpt, CINV_DESTROY|CINV_CHILDREN);
650 cache_unlock(&mp->mnt_ncmountpt);
651
652 if ((ncp = mp->mnt_ncmountpt.ncp) != NULL &&
653 (ncp->nc_refs != 1 || TAILQ_FIRST(&ncp->nc_list))) {
dcf26f79
MD
654
655 if ((flags & MNT_FORCE) == 0) {
656 error = EBUSY;
28623bf9
MD
657 mount_warning(mp, "Cannot unmount: "
658 "%d namecache "
659 "references still "
660 "present",
661 ncp->nc_refs - 1);
dcf26f79 662 } else {
28623bf9
MD
663 mount_warning(mp, "Forced unmount: "
664 "%d namecache "
665 "references still "
666 "present",
667 ncp->nc_refs - 1);
668 freeok = 0;
dcf26f79
MD
669 }
670 }
671 }
672
28623bf9
MD
673 /*
674 * nchandle records ref the mount structure. Expect a count of 1
675 * (our mount->mnt_ncmountpt).
676 */
677 if (mp->mnt_refs != 1) {
678 if ((flags & MNT_FORCE) == 0) {
679 mount_warning(mp, "Cannot unmount: "
680 "%d process references still "
681 "present", mp->mnt_refs);
682 error = EBUSY;
683 } else {
684 mount_warning(mp, "Forced unmount: "
685 "%d process references still "
686 "present", mp->mnt_refs);
687 freeok = 0;
688 }
689 }
690
dcf26f79
MD
691 if (error == 0) {
692 if (mp->mnt_syncer != NULL)
693 vrele(mp->mnt_syncer);
694 if (((mp->mnt_flag & MNT_RDONLY) ||
695 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) ||
696 (flags & MNT_FORCE)) {
697 error = VFS_UNMOUNT(mp, flags);
698 }
699 }
984263bc 700 if (error) {
41a01a4d
MD
701 if (mp->mnt_syncer == NULL)
702 vfs_allocate_syncvnode(mp);
984263bc
MD
703 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
704 mp->mnt_flag |= async_flag;
df4f70a6 705 lockmgr(&mp->mnt_lock, LK_RELEASE);
984263bc 706 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 707 wakeup(mp);
984263bc
MD
708 return (error);
709 }
432b8263
MD
710 /*
711 * Clean up any journals still associated with the mount after
712 * filesystem activity has ceased.
713 */
714 journal_remove_all_journals(mp,
715 ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));
716
861905fb 717 mountlist_remove(mp);
0961aa92
MD
718
719 /*
720 * Remove any installed vnode ops here so the individual VFSs don't
721 * have to.
722 */
66a1ddf5
MD
723 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
724 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
725 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
726 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
727 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
0961aa92 728
28623bf9
MD
729 if (mp->mnt_ncmountpt.ncp != NULL) {
730 nch = mp->mnt_ncmountpt;
731 cache_zero(&mp->mnt_ncmountpt);
732 cache_clrmountpt(&nch);
733 cache_drop(&nch);
734 }
735 if (mp->mnt_ncmounton.ncp != NULL) {
736 nch = mp->mnt_ncmounton;
737 cache_zero(&mp->mnt_ncmounton);
738 cache_clrmountpt(&nch);
739 cache_drop(&nch);
984263bc 740 }
1d505369 741
984263bc
MD
742 mp->mnt_vfc->vfc_refcount--;
743 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
744 panic("unmount: dangling vnode");
df4f70a6 745 lockmgr(&mp->mnt_lock, LK_RELEASE);
984263bc 746 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 747 wakeup(mp);
28623bf9
MD
748 if (freeok)
749 kfree(mp, M_MOUNT);
984263bc
MD
750 return (0);
751}
752
28623bf9
MD
753static
754void
755mount_warning(struct mount *mp, const char *ctl, ...)
756{
757 char *ptr;
758 char *buf;
759 __va_list va;
760
761 __va_start(va, ctl);
762 if (cache_fullpath(NULL, &mp->mnt_ncmounton, &ptr, &buf) == 0) {
6ea70f76 763 kprintf("unmount(%s): ", ptr);
379210cb 764 kvprintf(ctl, va);
6ea70f76 765 kprintf("\n");
28623bf9
MD
766 kfree(buf, M_TEMP);
767 } else {
56bcacad
MD
768 kprintf("unmount(%p", mp);
769 if (mp->mnt_ncmounton.ncp && mp->mnt_ncmounton.ncp->nc_name)
770 kprintf(",%s", mp->mnt_ncmounton.ncp->nc_name);
771 kprintf("): ");
379210cb 772 kvprintf(ctl, va);
6ea70f76 773 kprintf("\n");
28623bf9
MD
774 }
775 __va_end(va);
776}
777
d3c546e6
MD
778/*
779 * Shim cache_fullpath() to handle the case where a process is chrooted into
780 * a subdirectory of a mount. In this case if the root mount matches the
781 * process root directory's mount we have to specify the process's root
782 * directory instead of the mount point, because the mount point might
783 * be above the root directory.
784 */
785static
786int
787mount_path(struct proc *p, struct mount *mp, char **rb, char **fb)
788{
789 struct nchandle *nch;
d3c546e6
MD
790
791 if (p && p->p_fd->fd_nrdir.mount == mp)
792 nch = &p->p_fd->fd_nrdir;
793 else
794 nch = &mp->mnt_ncmountpt;
795 return(cache_fullpath(p, nch, rb, fb));
796}
797
984263bc
MD
798/*
799 * Sync each mounted filesystem.
800 */
984263bc
MD
801
802#ifdef DEBUG
803static int syncprt = 0;
804SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
6bdbb368 805#endif /* DEBUG */
984263bc 806
861905fb
MD
807static int sync_callback(struct mount *mp, void *data);
808
984263bc
MD
809/* ARGSUSED */
810int
753fd850 811sys_sync(struct sync_args *uap)
984263bc 812{
861905fb 813 mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
6bdbb368 814#ifdef DEBUG
861905fb
MD
815 /*
816 * print out buffer pool stat information on each sync() call.
817 */
984263bc
MD
818 if (syncprt)
819 vfs_bufstats();
6bdbb368 820#endif /* DEBUG */
984263bc
MD
821 return (0);
822}
823
861905fb
MD
824static
825int
826sync_callback(struct mount *mp, void *data __unused)
827{
828 int asyncflag;
829
830 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
831 asyncflag = mp->mnt_flag & MNT_ASYNC;
832 mp->mnt_flag &= ~MNT_ASYNC;
833 vfs_msync(mp, MNT_NOWAIT);
87de5057 834 VFS_SYNC(mp, MNT_NOWAIT);
861905fb
MD
835 mp->mnt_flag |= asyncflag;
836 }
837 return(0);
838}
839
984263bc
MD
840/* XXX PRISON: could be per prison flag */
841static int prison_quotas;
842#if 0
843SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
844#endif
845
846/*
41c20dac
MD
847 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
848 *
984263bc
MD
849 * Change filesystem quotas.
850 */
984263bc
MD
851/* ARGSUSED */
852int
753fd850 853sys_quotactl(struct quotactl_args *uap)
984263bc 854{
fad57d0e
MD
855 struct nlookupdata nd;
856 struct thread *td;
857 struct proc *p;
41c20dac 858 struct mount *mp;
984263bc 859 int error;
984263bc 860
fad57d0e
MD
861 td = curthread;
862 p = td->td_proc;
41c20dac 863 if (p->p_ucred->cr_prison && !prison_quotas)
984263bc 864 return (EPERM);
fad57d0e 865
ab2eb4eb 866 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
867 if (error == 0)
868 error = nlookup(&nd);
869 if (error == 0) {
28623bf9 870 mp = nd.nl_nch.mount;
ab2eb4eb 871 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
acde96db 872 uap->arg, nd.nl_cred);
fad57d0e
MD
873 }
874 nlookup_done(&nd);
875 return (error);
984263bc
MD
876}
877
949ecb9b 878/*
2281065e 879 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
949ecb9b
MD
880 * void *buf, int buflen)
881 *
882 * This function operates on a mount point and executes the specified
883 * operation using the specified control data, and possibly returns data.
884 *
885 * The actual number of bytes stored in the result buffer is returned, 0
886 * if none, otherwise an error is returned.
887 */
888/* ARGSUSED */
889int
753fd850 890sys_mountctl(struct mountctl_args *uap)
949ecb9b
MD
891{
892 struct thread *td = curthread;
893 struct proc *p = td->td_proc;
2281065e 894 struct file *fp;
949ecb9b
MD
895 void *ctl = NULL;
896 void *buf = NULL;
897 char *path = NULL;
898 int error;
899
900 /*
901 * Sanity and permissions checks. We must be root.
902 */
903 KKASSERT(p);
904 if (p->p_ucred->cr_prison != NULL)
905 return (EPERM);
906 if ((error = suser(td)) != 0)
907 return (error);
908
909 /*
910 * Argument length checks
911 */
39b13188 912 if (uap->ctllen < 0 || uap->ctllen > 1024)
949ecb9b 913 return (EINVAL);
39b13188 914 if (uap->buflen < 0 || uap->buflen > 16 * 1024)
949ecb9b
MD
915 return (EINVAL);
916 if (uap->path == NULL)
917 return (EINVAL);
918
919 /*
920 * Allocate the necessary buffers and copyin data
921 */
70aac194 922 path = objcache_get(namei_oc, M_WAITOK);
949ecb9b
MD
923 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
924 if (error)
925 goto done;
926
927 if (uap->ctllen) {
efda3bd0 928 ctl = kmalloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
949ecb9b
MD
929 error = copyin(uap->ctl, ctl, uap->ctllen);
930 if (error)
931 goto done;
932 }
933 if (uap->buflen)
efda3bd0 934 buf = kmalloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
949ecb9b 935
2281065e
MD
936 /*
937 * Validate the descriptor
938 */
ecda6326
MD
939 if (uap->fd >= 0) {
940 fp = holdfp(p->p_fd, uap->fd, -1);
941 if (fp == NULL) {
942 error = EBADF;
943 goto done;
944 }
945 } else {
946 fp = NULL;
2281065e 947 }
2281065e 948
949ecb9b
MD
949 /*
950 * Execute the internal kernel function and clean up.
951 */
2281065e
MD
952 error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
953 if (fp)
9f87144f 954 fdrop(fp);
949ecb9b
MD
955 if (error == 0 && uap->sysmsg_result > 0)
956 error = copyout(buf, uap->buf, uap->sysmsg_result);
957done:
958 if (path)
70aac194 959 objcache_put(namei_oc, path);
949ecb9b 960 if (ctl)
efda3bd0 961 kfree(ctl, M_TEMP);
949ecb9b 962 if (buf)
efda3bd0 963 kfree(buf, M_TEMP);
949ecb9b
MD
964 return (error);
965}
966
967/*
968 * Execute a mount control operation by resolving the path to a mount point
969 * and calling vop_mountctl().
970 */
971int
2281065e
MD
972kern_mountctl(const char *path, int op, struct file *fp,
973 const void *ctl, int ctllen,
949ecb9b
MD
974 void *buf, int buflen, int *res)
975{
949ecb9b
MD
976 struct vnode *vp;
977 struct mount *mp;
978 struct nlookupdata nd;
979 int error;
980
981 *res = 0;
982 vp = NULL;
983 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
984 if (error == 0)
985 error = nlookup(&nd);
986 if (error == 0)
28623bf9 987 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
949ecb9b
MD
988 nlookup_done(&nd);
989 if (error)
990 return (error);
991
992 mp = vp->v_mount;
993
994 /*
995 * Must be the root of the filesystem
996 */
997 if ((vp->v_flag & VROOT) == 0) {
998 vput(vp);
999 return (EINVAL);
1000 }
2281065e 1001 error = vop_mountctl(mp->mnt_vn_use_ops, op, fp, ctl, ctllen,
949ecb9b
MD
1002 buf, buflen, res);
1003 vput(vp);
1004 return (error);
1005}
1006
984263bc 1007int
fad57d0e 1008kern_statfs(struct nlookupdata *nd, struct statfs *buf)
984263bc 1009{
dadab5e9 1010 struct thread *td = curthread;
75ffff0d 1011 struct proc *p = td->td_proc;
41c20dac
MD
1012 struct mount *mp;
1013 struct statfs *sp;
75ffff0d 1014 char *fullpath, *freepath;
984263bc 1015 int error;
984263bc 1016
fad57d0e 1017 if ((error = nlookup(nd)) != 0)
984263bc 1018 return (error);
28623bf9 1019 mp = nd->nl_nch.mount;
984263bc 1020 sp = &mp->mnt_stat;
acde96db 1021 if ((error = VFS_STATFS(mp, sp, nd->nl_cred)) != 0)
984263bc 1022 return (error);
75ffff0d 1023
d3c546e6 1024 error = mount_path(p, mp, &fullpath, &freepath);
75ffff0d
JS
1025 if (error)
1026 return(error);
1027 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1028 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
efda3bd0 1029 kfree(freepath, M_TEMP);
75ffff0d 1030
984263bc 1031 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509
DRJ
1032 bcopy(sp, buf, sizeof(*buf));
1033 /* Only root should have access to the fsid's. */
1034 if (suser(td))
1035 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
1036 return (0);
984263bc
MD
1037}
1038
1039/*
9697c509 1040 * statfs_args(char *path, struct statfs *buf)
41c20dac 1041 *
984263bc
MD
1042 * Get filesystem statistics.
1043 */
984263bc 1044int
753fd850 1045sys_statfs(struct statfs_args *uap)
9697c509 1046{
fad57d0e 1047 struct nlookupdata nd;
9697c509
DRJ
1048 struct statfs buf;
1049 int error;
1050
fad57d0e
MD
1051 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1052 if (error == 0)
1053 error = kern_statfs(&nd, &buf);
1054 nlookup_done(&nd);
9697c509
DRJ
1055 if (error == 0)
1056 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1057 return (error);
1058}
1059
1060int
1061kern_fstatfs(int fd, struct statfs *buf)
984263bc 1062{
dadab5e9
MD
1063 struct thread *td = curthread;
1064 struct proc *p = td->td_proc;
984263bc
MD
1065 struct file *fp;
1066 struct mount *mp;
1fd87d54 1067 struct statfs *sp;
75ffff0d 1068 char *fullpath, *freepath;
984263bc 1069 int error;
984263bc 1070
dadab5e9 1071 KKASSERT(p);
5b287bba 1072 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc
MD
1073 return (error);
1074 mp = ((struct vnode *)fp->f_data)->v_mount;
5b287bba
MD
1075 if (mp == NULL) {
1076 error = EBADF;
1077 goto done;
1078 }
1079 if (fp->f_cred == NULL) {
1080 error = EINVAL;
1081 goto done;
1082 }
984263bc 1083 sp = &mp->mnt_stat;
5b287bba
MD
1084 if ((error = VFS_STATFS(mp, sp, fp->f_cred)) != 0)
1085 goto done;
75ffff0d 1086
d3c546e6 1087 if ((error = mount_path(p, mp, &fullpath, &freepath)) != 0)
5b287bba 1088 goto done;
75ffff0d
JS
1089 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1090 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
efda3bd0 1091 kfree(freepath, M_TEMP);
75ffff0d 1092
984263bc 1093 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509 1094 bcopy(sp, buf, sizeof(*buf));
75ffff0d 1095
9697c509
DRJ
1096 /* Only root should have access to the fsid's. */
1097 if (suser(td))
1098 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
5b287bba
MD
1099 error = 0;
1100done:
1101 fdrop(fp);
1102 return (error);
9697c509
DRJ
1103}
1104
1105/*
1106 * fstatfs_args(int fd, struct statfs *buf)
1107 *
1108 * Get filesystem statistics.
1109 */
1110int
753fd850 1111sys_fstatfs(struct fstatfs_args *uap)
9697c509
DRJ
1112{
1113 struct statfs buf;
1114 int error;
1115
1116 error = kern_fstatfs(uap->fd, &buf);
1117
1118 if (error == 0)
1119 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1120 return (error);
984263bc
MD
1121}
1122
1123/*
41c20dac
MD
1124 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
1125 *
984263bc
MD
1126 * Get statistics on all filesystems.
1127 */
861905fb
MD
1128
1129struct getfsstat_info {
1130 struct statfs *sfsp;
1131 long count;
1132 long maxcount;
1133 int error;
1134 int flags;
861905fb
MD
1135 struct proc *p;
1136};
1137
1138static int getfsstat_callback(struct mount *, void *);
1139
41c20dac 1140/* ARGSUSED */
984263bc 1141int
753fd850 1142sys_getfsstat(struct getfsstat_args *uap)
984263bc 1143{
dadab5e9 1144 struct thread *td = curthread;
75ffff0d 1145 struct proc *p = td->td_proc;
861905fb 1146 struct getfsstat_info info;
75ffff0d 1147
861905fb 1148 bzero(&info, sizeof(info));
861905fb
MD
1149
1150 info.maxcount = uap->bufsize / sizeof(struct statfs);
1151 info.sfsp = uap->buf;
1152 info.count = 0;
1153 info.flags = uap->flags;
861905fb
MD
1154 info.p = p;
1155
1156 mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
1157 if (info.sfsp && info.count > info.maxcount)
1158 uap->sysmsg_result = info.maxcount;
75ffff0d 1159 else
861905fb
MD
1160 uap->sysmsg_result = info.count;
1161 return (info.error);
1162}
1163
1164static int
1165getfsstat_callback(struct mount *mp, void *data)
1166{
1167 struct getfsstat_info *info = data;
1168 struct statfs *sp;
1169 char *freepath;
1170 char *fullpath;
1171 int error;
1172
1173 if (info->sfsp && info->count < info->maxcount) {
28623bf9 1174 if (info->p && !chroot_visible_mnt(mp, info->p))
861905fb
MD
1175 return(0);
1176 sp = &mp->mnt_stat;
1177
1178 /*
1179 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1180 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1181 * overrides MNT_WAIT.
1182 */
1183 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1184 (info->flags & MNT_WAIT)) &&
acde96db 1185 (error = VFS_STATFS(mp, sp, info->p->p_ucred))) {
861905fb
MD
1186 return(0);
1187 }
1188 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1189
d3c546e6 1190 error = mount_path(info->p, mp, &fullpath, &freepath);
861905fb
MD
1191 if (error) {
1192 info->error = error;
1193 return(-1);
984263bc 1194 }
861905fb
MD
1195 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1196 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
efda3bd0 1197 kfree(freepath, M_TEMP);
861905fb
MD
1198
1199 error = copyout(sp, info->sfsp, sizeof(*sp));
1200 if (error) {
1201 info->error = error;
1202 return (-1);
984263bc 1203 }
861905fb 1204 ++info->sfsp;
984263bc 1205 }
861905fb
MD
1206 info->count++;
1207 return(0);
984263bc
MD
1208}
1209
1210/*
41c20dac
MD
1211 * fchdir_args(int fd)
1212 *
984263bc
MD
1213 * Change current working directory to a given file descriptor.
1214 */
984263bc
MD
1215/* ARGSUSED */
1216int
753fd850 1217sys_fchdir(struct fchdir_args *uap)
984263bc 1218{
dadab5e9
MD
1219 struct thread *td = curthread;
1220 struct proc *p = td->td_proc;
41c20dac 1221 struct filedesc *fdp = p->p_fd;
21739618 1222 struct vnode *vp, *ovp;
984263bc
MD
1223 struct mount *mp;
1224 struct file *fp;
28623bf9 1225 struct nchandle nch, onch, tnch;
984263bc
MD
1226 int error;
1227
5b287bba 1228 if ((error = holdvnode(fdp, uap->fd, &fp)) != 0)
984263bc
MD
1229 return (error);
1230 vp = (struct vnode *)fp->f_data;
597aea93 1231 vref(vp);
ca466bae 1232 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
28623bf9 1233 if (vp->v_type != VDIR || fp->f_nchandle.ncp == NULL)
984263bc
MD
1234 error = ENOTDIR;
1235 else
87de5057 1236 error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
984263bc
MD
1237 if (error) {
1238 vput(vp);
5b287bba 1239 fdrop(fp);
984263bc
MD
1240 return (error);
1241 }
28623bf9 1242 cache_copy(&fp->f_nchandle, &nch);
1d505369
MD
1243
1244 /*
1245 * If the ncp has become a mount point, traverse through
1246 * the mount point.
1247 */
1248
28623bf9
MD
1249 while (!error && (nch.ncp->nc_flag & NCF_ISMOUNTPT) &&
1250 (mp = cache_findmount(&nch)) != NULL
1d505369 1251 ) {
28623bf9 1252 error = nlookup_mp(mp, &tnch);
21739618 1253 if (error == 0) {
28623bf9 1254 cache_unlock(&tnch); /* leave ref intact */
21739618 1255 vput(vp);
28623bf9 1256 vp = tnch.ncp->nc_vp;
87de5057 1257 error = vget(vp, LK_SHARED);
21739618 1258 KKASSERT(error == 0);
28623bf9
MD
1259 cache_drop(&nch);
1260 nch = tnch;
21739618
MD
1261 }
1262 }
1263 if (error == 0) {
1264 ovp = fdp->fd_cdir;
28623bf9 1265 onch = fdp->fd_ncdir;
a11aaa81 1266 vn_unlock(vp); /* leave ref intact */
21739618 1267 fdp->fd_cdir = vp;
28623bf9
MD
1268 fdp->fd_ncdir = nch;
1269 cache_drop(&onch);
21739618
MD
1270 vrele(ovp);
1271 } else {
28623bf9 1272 cache_drop(&nch);
21739618
MD
1273 vput(vp);
1274 }
5b287bba 1275 fdrop(fp);
21739618 1276 return (error);
984263bc
MD
1277}
1278
9697c509 1279int
21739618 1280kern_chdir(struct nlookupdata *nd)
9697c509
DRJ
1281{
1282 struct thread *td = curthread;
1283 struct proc *p = td->td_proc;
1284 struct filedesc *fdp = p->p_fd;
21739618 1285 struct vnode *vp, *ovp;
28623bf9 1286 struct nchandle onch;
9697c509
DRJ
1287 int error;
1288
21739618 1289 if ((error = nlookup(nd)) != 0)
9697c509 1290 return (error);
28623bf9 1291 if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
21739618 1292 return (ENOENT);
87de5057 1293 if ((error = vget(vp, LK_SHARED)) != 0)
21739618
MD
1294 return (error);
1295
1296 error = checkvp_chdir(vp, td);
a11aaa81 1297 vn_unlock(vp);
21739618
MD
1298 if (error == 0) {
1299 ovp = fdp->fd_cdir;
28623bf9
MD
1300 onch = fdp->fd_ncdir;
1301 cache_unlock(&nd->nl_nch); /* leave reference intact */
1302 fdp->fd_ncdir = nd->nl_nch;
21739618 1303 fdp->fd_cdir = vp;
28623bf9 1304 cache_drop(&onch);
21739618 1305 vrele(ovp);
28623bf9 1306 cache_zero(&nd->nl_nch);
21739618
MD
1307 } else {
1308 vrele(vp);
e24b948e 1309 }
e24b948e 1310 return (error);
9697c509
DRJ
1311}
1312
984263bc 1313/*
41c20dac
MD
1314 * chdir_args(char *path)
1315 *
984263bc
MD
1316 * Change current working directory (``.'').
1317 */
984263bc 1318int
753fd850 1319sys_chdir(struct chdir_args *uap)
984263bc 1320{
21739618 1321 struct nlookupdata nd;
9697c509 1322 int error;
984263bc 1323
21739618 1324 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e 1325 if (error == 0)
21739618 1326 error = kern_chdir(&nd);
fad57d0e 1327 nlookup_done(&nd);
9697c509 1328 return (error);
984263bc
MD
1329}
1330
1331/*
1332 * Helper function for raised chroot(2) security function: Refuse if
1333 * any filedescriptors are open directories.
1334 */
1335static int
c972a82f 1336chroot_refuse_vdir_fds(struct filedesc *fdp)
984263bc
MD
1337{
1338 struct vnode *vp;
1339 struct file *fp;
1340 int error;
1341 int fd;
1342
1343 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
5b287bba 1344 if ((error = holdvnode(fdp, fd, &fp)) != 0)
984263bc
MD
1345 continue;
1346 vp = (struct vnode *)fp->f_data;
5b287bba
MD
1347 if (vp->v_type != VDIR) {
1348 fdrop(fp);
984263bc 1349 continue;
5b287bba
MD
1350 }
1351 fdrop(fp);
984263bc
MD
1352 return(EPERM);
1353 }
1354 return (0);
1355}
1356
1357/*
1358 * This sysctl determines if we will allow a process to chroot(2) if it
1359 * has a directory open:
1360 * 0: disallowed for all processes.
1361 * 1: allowed for processes that were not already chroot(2)'ed.
1362 * 2: allowed for all processes.
1363 */
1364
1365static int chroot_allow_open_directories = 1;
1366
1367SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1368 &chroot_allow_open_directories, 0, "");
1369
e24b948e 1370/*
21739618
MD
1371 * chroot to the specified namecache entry. We obtain the vp from the
1372 * namecache data. The passed ncp must be locked and referenced and will
1373 * remain locked and referenced on return.
e24b948e 1374 */
53dd6631 1375int
28623bf9 1376kern_chroot(struct nchandle *nch)
e24b948e
MD
1377{
1378 struct thread *td = curthread;
1379 struct proc *p = td->td_proc;
1380 struct filedesc *fdp = p->p_fd;
21739618 1381 struct vnode *vp;
e24b948e
MD
1382 int error;
1383
1384 /*
1385 * Only root can chroot
1386 */
1387 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
1388 return (error);
1389
1390 /*
1391 * Disallow open directory descriptors (fchdir() breakouts).
1392 */
1393 if (chroot_allow_open_directories == 0 ||
1394 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1395 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1396 return (error);
1397 }
28623bf9 1398 if ((vp = nch->ncp->nc_vp) == NULL)
21739618
MD
1399 return (ENOENT);
1400
87de5057 1401 if ((error = vget(vp, LK_SHARED)) != 0)
21739618 1402 return (error);
e24b948e
MD
1403
1404 /*
1405 * Check the validity of vp as a directory to change to and
1406 * associate it with rdir/jdir.
1407 */
21739618 1408 error = checkvp_chdir(vp, td);
a11aaa81 1409 vn_unlock(vp); /* leave reference intact */
21739618 1410 if (error == 0) {
e24b948e 1411 vrele(fdp->fd_rdir);
21739618 1412 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */
28623bf9
MD
1413 cache_drop(&fdp->fd_nrdir);
1414 cache_copy(nch, &fdp->fd_nrdir);
e24b948e
MD
1415 if (fdp->fd_jdir == NULL) {
1416 fdp->fd_jdir = vp;
597aea93 1417 vref(fdp->fd_jdir);
28623bf9 1418 cache_copy(nch, &fdp->fd_njdir);
e24b948e 1419 }
21739618
MD
1420 } else {
1421 vrele(vp);
e24b948e
MD
1422 }
1423 return (error);
1424}
1425
984263bc 1426/*
41c20dac
MD
1427 * chroot_args(char *path)
1428 *
984263bc
MD
1429 * Change notion of root (``/'') directory.
1430 */
984263bc
MD
1431/* ARGSUSED */
1432int
753fd850 1433sys_chroot(struct chroot_args *uap)
984263bc 1434{
dadab5e9 1435 struct thread *td = curthread;
21739618 1436 struct nlookupdata nd;
e24b948e 1437 int error;
984263bc 1438
e24b948e 1439 KKASSERT(td->td_proc);
21739618 1440 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
53dd6631
JS
1441 if (error) {
1442 nlookup_done(&nd);
1443 return(error);
1444 }
1445 error = nlookup(&nd);
1446 if (error == 0)
28623bf9 1447 error = kern_chroot(&nd.nl_nch);
fad57d0e 1448 nlookup_done(&nd);
53dd6631 1449 return(error);
984263bc
MD
1450}
1451
1452/*
e24b948e
MD
1453 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1454 * determine whether it is legal to chdir to the vnode. The vnode's state
1455 * is not changed by this call.
984263bc 1456 */
e24b948e
MD
1457int
1458checkvp_chdir(struct vnode *vp, struct thread *td)
984263bc 1459{
984263bc
MD
1460 int error;
1461
984263bc
MD
1462 if (vp->v_type != VDIR)
1463 error = ENOTDIR;
1464 else
87de5057 1465 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred);
984263bc
MD
1466 return (error);
1467}
1468
984263bc 1469int
fad57d0e 1470kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
984263bc 1471{
dadab5e9
MD
1472 struct thread *td = curthread;
1473 struct proc *p = td->td_proc;
fbfe4e7d 1474 struct lwp *lp = td->td_lwp;
41c20dac 1475 struct filedesc *fdp = p->p_fd;
9697c509 1476 int cmode, flags;
984263bc 1477 struct file *nfp;
fad57d0e
MD
1478 struct file *fp;
1479 struct vnode *vp;
984263bc
MD
1480 int type, indx, error;
1481 struct flock lf;
984263bc 1482
984263bc
MD
1483 if ((oflags & O_ACCMODE) == O_ACCMODE)
1484 return (EINVAL);
1485 flags = FFLAGS(oflags);
fad57d0e 1486 error = falloc(p, &nfp, NULL);
984263bc
MD
1487 if (error)
1488 return (error);
1489 fp = nfp;
9697c509 1490 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
fad57d0e 1491
984263bc 1492 /*
fad57d0e
MD
1493 * XXX p_dupfd is a real mess. It allows a device to return a
1494 * file descriptor to be duplicated rather then doing the open
1495 * itself.
984263bc 1496 */
fbfe4e7d 1497 lp->lwp_dupfd = -1;
21739618
MD
1498
1499 /*
fad57d0e
MD
1500 * Call vn_open() to do the lookup and assign the vnode to the
1501 * file pointer. vn_open() does not change the ref count on fp
1502 * and the vnode, on success, will be inherited by the file pointer
1503 * and unlocked.
21739618 1504 */
fad57d0e
MD
1505 nd->nl_flags |= NLC_LOCKVP;
1506 error = vn_open(nd, fp, flags, cmode);
1507 nlookup_done(nd);
984263bc 1508 if (error) {
984263bc
MD
1509 /*
1510 * handle special fdopen() case. bleh. dupfdopen() is
1511 * responsible for dropping the old contents of ofiles[indx]
1512 * if it succeeds.
fad57d0e 1513 *
259b8ea0
MD
1514 * Note that fsetfd() will add a ref to fp which represents
1515 * the fd_files[] assignment. We must still drop our
1516 * reference.
984263bc 1517 */
fbfe4e7d 1518 if ((error == ENODEV || error == ENXIO) && lp->lwp_dupfd >= 0) {
259b8ea0
MD
1519 if (fdalloc(p, 0, &indx) == 0) {
1520 error = dupfdopen(p, indx, lp->lwp_dupfd, flags, error);
fad57d0e
MD
1521 if (error == 0) {
1522 *res = indx;
9f87144f 1523 fdrop(fp); /* our ref */
fad57d0e
MD
1524 return (0);
1525 }
259b8ea0 1526 fsetfd(p, NULL, indx);
fad57d0e 1527 }
984263bc 1528 }
9f87144f 1529 fdrop(fp); /* our ref */
984263bc
MD
1530 if (error == ERESTART)
1531 error = EINTR;
1532 return (error);
1533 }
fad57d0e
MD
1534
1535 /*
1536 * ref the vnode for ourselves so it can't be ripped out from under
1537 * is. XXX need an ND flag to request that the vnode be returned
1538 * anyway.
259b8ea0
MD
1539 *
1540 * Reserve a file descriptor but do not assign it until the open
1541 * succeeds.
fad57d0e
MD
1542 */
1543 vp = (struct vnode *)fp->f_data;
1544 vref(vp);
259b8ea0 1545 if ((error = fdalloc(p, 0, &indx)) != 0) {
9f87144f 1546 fdrop(fp);
fad57d0e
MD
1547 vrele(vp);
1548 return (error);
1549 }
1550
1551 /*
1552 * If no error occurs the vp will have been assigned to the file
1553 * pointer.
1554 */
fbfe4e7d 1555 lp->lwp_dupfd = 0;
984263bc 1556
984263bc
MD
1557 if (flags & (O_EXLOCK | O_SHLOCK)) {
1558 lf.l_whence = SEEK_SET;
1559 lf.l_start = 0;
1560 lf.l_len = 0;
1561 if (flags & O_EXLOCK)
1562 lf.l_type = F_WRLCK;
1563 else
1564 lf.l_type = F_RDLCK;
71c18fe3
MD
1565 if (flags & FNONBLOCK)
1566 type = 0;
1567 else
1568 type = F_WAIT;
fad57d0e 1569
984263bc
MD
1570 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1571 /*
259b8ea0
MD
1572 * lock request failed. Clean up the reserved
1573 * descriptor.
984263bc 1574 */
fad57d0e 1575 vrele(vp);
259b8ea0 1576 fsetfd(p, NULL, indx);
9f87144f 1577 fdrop(fp);
984263bc
MD
1578 return (error);
1579 }
984263bc
MD
1580 fp->f_flag |= FHASLOCK;
1581 }
7540ab49
MD
1582#if 0
1583 /*
1584 * Assert that all regular file vnodes were created with a object.
1585 */
1586 KASSERT(vp->v_type != VREG || vp->v_object != NULL,
1587 ("open: regular file has no backing object after vn_open"));
1588#endif
984263bc 1589
fad57d0e 1590 vrele(vp);
21739618 1591
984263bc
MD
1592 /*
1593 * release our private reference, leaving the one associated with the
1594 * descriptor table intact.
1595 */
259b8ea0 1596 fsetfd(p, fp, indx);
9f87144f 1597 fdrop(fp);
9697c509 1598 *res = indx;
984263bc
MD
1599 return (0);
1600}
1601
984263bc 1602/*
9697c509 1603 * open_args(char *path, int flags, int mode)
41c20dac 1604 *
9697c509
DRJ
1605 * Check permissions, allocate an open file structure,
1606 * and call the device open routine if any.
984263bc 1607 */
984263bc 1608int
753fd850 1609sys_open(struct open_args *uap)
984263bc 1610{
fad57d0e 1611 struct nlookupdata nd;
9697c509
DRJ
1612 int error;
1613
fad57d0e
MD
1614 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1615 if (error == 0) {
1616 error = kern_open(&nd, uap->flags,
1617 uap->mode, &uap->sysmsg_result);
1618 }
1619 nlookup_done(&nd);
9697c509 1620 return (error);
984263bc 1621}
984263bc 1622
984263bc 1623int
0e9b9130 1624kern_mknod(struct nlookupdata *nd, int mode, int rmajor, int rminor)
984263bc 1625{
dadab5e9
MD
1626 struct thread *td = curthread;
1627 struct proc *p = td->td_proc;
41c20dac 1628 struct vnode *vp;
984263bc
MD
1629 struct vattr vattr;
1630 int error;
1631 int whiteout = 0;
984263bc 1632
dadab5e9
MD
1633 KKASSERT(p);
1634
9697c509 1635 switch (mode & S_IFMT) {
984263bc
MD
1636 case S_IFCHR:
1637 case S_IFBLK:
dadab5e9 1638 error = suser(td);
984263bc
MD
1639 break;
1640 default:
dadab5e9 1641 error = suser_cred(p->p_ucred, PRISON_ROOT);
984263bc
MD
1642 break;
1643 }
1644 if (error)
1645 return (error);
fad57d0e 1646
984263bc 1647 bwillwrite();
5312fa43 1648 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 1649 if ((error = nlookup(nd)) != 0)
984263bc 1650 return (error);
28623bf9 1651 if (nd->nl_nch.ncp->nc_vp)
fad57d0e 1652 return (EEXIST);
28623bf9 1653 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 1654 return (error);
984263bc 1655
fad57d0e
MD
1656 VATTR_NULL(&vattr);
1657 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
0e9b9130
MD
1658 vattr.va_rmajor = rmajor;
1659 vattr.va_rminor = rminor;
fad57d0e
MD
1660 whiteout = 0;
1661
1662 switch (mode & S_IFMT) {
1663 case S_IFMT: /* used by badsect to flag bad sectors */
1664 vattr.va_type = VBAD;
1665 break;
1666 case S_IFCHR:
1667 vattr.va_type = VCHR;
1668 break;
1669 case S_IFBLK:
1670 vattr.va_type = VBLK;
1671 break;
1672 case S_IFWHT:
1673 whiteout = 1;
1674 break;
1675 default:
1676 error = EINVAL;
1677 break;
984263bc 1678 }
9697c509 1679 if (error == 0) {
fad57d0e 1680 if (whiteout) {
5312fa43
MD
1681 error = VOP_NWHITEOUT(&nd->nl_nch, nd->nl_dvp,
1682 nd->nl_cred, NAMEI_CREATE);
fad57d0e
MD
1683 } else {
1684 vp = NULL;
5312fa43
MD
1685 error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp,
1686 &vp, nd->nl_cred, &vattr);
984263bc 1687 if (error == 0)
fad57d0e 1688 vput(vp);
984263bc 1689 }
984263bc 1690 }
9697c509
DRJ
1691 return (error);
1692}
1693
1694/*
1695 * mknod_args(char *path, int mode, int dev)
1696 *
1697 * Create a special file.
1698 */
1699int
753fd850 1700sys_mknod(struct mknod_args *uap)
9697c509 1701{
fad57d0e 1702 struct nlookupdata nd;
9697c509
DRJ
1703 int error;
1704
fad57d0e 1705 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
0e9b9130
MD
1706 if (error == 0) {
1707 error = kern_mknod(&nd, uap->mode,
1708 umajor(uap->dev), uminor(uap->dev));
1709 }
fad57d0e 1710 nlookup_done(&nd);
984263bc
MD
1711 return (error);
1712}
1713
984263bc 1714int
fad57d0e 1715kern_mkfifo(struct nlookupdata *nd, int mode)
984263bc 1716{
dadab5e9
MD
1717 struct thread *td = curthread;
1718 struct proc *p = td->td_proc;
984263bc 1719 struct vattr vattr;
fad57d0e 1720 struct vnode *vp;
984263bc 1721 int error;
984263bc
MD
1722
1723 bwillwrite();
fad57d0e 1724
5312fa43 1725 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 1726 if ((error = nlookup(nd)) != 0)
984263bc 1727 return (error);
28623bf9 1728 if (nd->nl_nch.ncp->nc_vp)
984263bc 1729 return (EEXIST);
28623bf9 1730 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 1731 return (error);
fad57d0e 1732
984263bc
MD
1733 VATTR_NULL(&vattr);
1734 vattr.va_type = VFIFO;
136178b3 1735 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
fad57d0e 1736 vp = NULL;
5312fa43 1737 error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp, &vp, nd->nl_cred, &vattr);
984263bc 1738 if (error == 0)
fad57d0e 1739 vput(vp);
136178b3
DRJ
1740 return (error);
1741}
1742
1743/*
1744 * mkfifo_args(char *path, int mode)
1745 *
1746 * Create a named pipe.
1747 */
1748int
753fd850 1749sys_mkfifo(struct mkfifo_args *uap)
136178b3 1750{
fad57d0e 1751 struct nlookupdata nd;
136178b3
DRJ
1752 int error;
1753
fad57d0e
MD
1754 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1755 if (error == 0)
1756 error = kern_mkfifo(&nd, uap->mode);
1757 nlookup_done(&nd);
984263bc
MD
1758 return (error);
1759}
1760
8c40f81e 1761static int hardlink_check_uid = 0;
43a0f7ae 1762SYSCTL_INT(_security, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
8c40f81e
MD
1763 &hardlink_check_uid, 0,
1764 "Unprivileged processes cannot create hard links to files owned by other "
1765 "users");
1766static int hardlink_check_gid = 0;
43a0f7ae 1767SYSCTL_INT(_security, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
8c40f81e
MD
1768 &hardlink_check_gid, 0,
1769 "Unprivileged processes cannot create hard links to files owned by other "
1770 "groups");
1771
1772static int
1773can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
1774{
1775 struct vattr va;
1776 int error;
1777
1778 /*
1779 * Shortcut if disabled
1780 */
1781 if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
1782 return (0);
1783
1784 /*
1785 * root cred can always hardlink
1786 */
1787 if (suser_cred(cred, PRISON_ROOT) == 0)
1788 return (0);
1789
1790 /*
1791 * Otherwise only if the originating file is owned by the
1792 * same user or group. Note that any group is allowed if
1793 * the file is owned by the caller.
1794 */
87de5057 1795 error = VOP_GETATTR(vp, &va);
8c40f81e
MD
1796 if (error != 0)
1797 return (error);
1798
1799 if (hardlink_check_uid) {
1800 if (cred->cr_uid != va.va_uid)
1801 return (EPERM);
1802 }
1803
1804 if (hardlink_check_gid) {
1805 if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
1806 return (EPERM);
1807 }
1808
1809 return (0);
1810}
1811
984263bc 1812int
fad57d0e 1813kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
984263bc 1814{
dadab5e9 1815 struct thread *td = curthread;
41c20dac 1816 struct vnode *vp;
984263bc
MD
1817 int error;
1818
fad57d0e
MD
1819 /*
1820 * Lookup the source and obtained a locked vnode.
1821 *
1822 * XXX relookup on vget failure / race ?
1823 */
984263bc 1824 bwillwrite();
fad57d0e 1825 if ((error = nlookup(nd)) != 0)
984263bc 1826 return (error);
28623bf9 1827 vp = nd->nl_nch.ncp->nc_vp;
fad57d0e 1828 KKASSERT(vp != NULL);
984263bc 1829 if (vp->v_type == VDIR)
fad57d0e 1830 return (EPERM); /* POSIX */
28623bf9 1831 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 1832 return (error);
87de5057 1833 if ((error = vget(vp, LK_EXCLUSIVE)) != 0)
fad57d0e
MD
1834 return (error);
1835
1836 /*
1837 * Unlock the source so we can lookup the target without deadlocking
1838 * (XXX vp is locked already, possible other deadlock?). The target
1839 * must not exist.
1840 */
1841 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
1842 nd->nl_flags &= ~NLC_NCPISLOCKED;
28623bf9 1843 cache_unlock(&nd->nl_nch);
fad57d0e 1844
5312fa43 1845 linknd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e
MD
1846 if ((error = nlookup(linknd)) != 0) {
1847 vput(vp);
1848 return (error);
984263bc 1849 }
28623bf9 1850 if (linknd->nl_nch.ncp->nc_vp) {
fad57d0e
MD
1851 vput(vp);
1852 return (EEXIST);
1853 }
1854
1855 /*
1856 * Finally run the new API VOP.
1857 */
8c40f81e 1858 error = can_hardlink(vp, td, td->td_proc->p_ucred);
5312fa43
MD
1859 if (error == 0) {
1860 error = VOP_NLINK(&linknd->nl_nch, linknd->nl_dvp,
1861 vp, linknd->nl_cred);
1862 }
fad57d0e 1863 vput(vp);
984263bc
MD
1864 return (error);
1865}
1866
1867/*
9697c509 1868 * link_args(char *path, char *link)
41c20dac 1869 *
9697c509 1870 * Make a hard file link.
984263bc 1871 */
984263bc 1872int
753fd850 1873sys_link(struct link_args *uap)
9697c509 1874{
fad57d0e 1875 struct nlookupdata nd, linknd;
9697c509
DRJ
1876 int error;
1877
fad57d0e
MD
1878 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1879 if (error == 0) {
1880 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
1881 if (error == 0)
1882 error = kern_link(&nd, &linknd);
1883 nlookup_done(&linknd);
1884 }
1885 nlookup_done(&nd);
9697c509
DRJ
1886 return (error);
1887}
1888
1889int
fad57d0e 1890kern_symlink(struct nlookupdata *nd, char *path, int mode)
984263bc
MD
1891{
1892 struct vattr vattr;
fad57d0e 1893 struct vnode *vp;
dff430ab 1894 struct vnode *dvp;
984263bc 1895 int error;
984263bc 1896
984263bc 1897 bwillwrite();
5312fa43 1898 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 1899 if ((error = nlookup(nd)) != 0)
9697c509 1900 return (error);
28623bf9 1901 if (nd->nl_nch.ncp->nc_vp)
9697c509 1902 return (EEXIST);
28623bf9 1903 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 1904 return (error);
5312fa43 1905 dvp = nd->nl_dvp;
984263bc 1906 VATTR_NULL(&vattr);
fad57d0e 1907 vattr.va_mode = mode;
dff430ab 1908 error = VOP_NSYMLINK(&nd->nl_nch, dvp, &vp, nd->nl_cred, &vattr, path);
984263bc 1909 if (error == 0)
fad57d0e 1910 vput(vp);
9697c509
DRJ
1911 return (error);
1912}
1913
1914/*
1915 * symlink(char *path, char *link)
1916 *
1917 * Make a symbolic link.
1918 */
1919int
753fd850 1920sys_symlink(struct symlink_args *uap)
9697c509
DRJ
1921{
1922 struct thread *td = curthread;
fad57d0e 1923 struct nlookupdata nd;
9697c509
DRJ
1924 char *path;
1925 int error;
fad57d0e 1926 int mode;
9697c509 1927
70aac194 1928 path = objcache_get(namei_oc, M_WAITOK);
9697c509 1929 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
e24b948e 1930 if (error == 0) {
fad57d0e
MD
1931 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
1932 if (error == 0) {
1933 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
1934 error = kern_symlink(&nd, path, mode);
1935 }
1936 nlookup_done(&nd);
e24b948e 1937 }
70aac194 1938 objcache_put(namei_oc, path);
984263bc
MD
1939 return (error);
1940}
1941
1942/*
41c20dac
MD
1943 * undelete_args(char *path)
1944 *
984263bc
MD
1945 * Delete a whiteout from the filesystem.
1946 */
1947/* ARGSUSED */
1948int
753fd850 1949sys_undelete(struct undelete_args *uap)
984263bc 1950{
fad57d0e 1951 struct nlookupdata nd;
984263bc 1952 int error;
984263bc 1953
ab2eb4eb 1954 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
984263bc 1955 bwillwrite();
5312fa43 1956 nd.nl_flags |= NLC_DELETE | NLC_REFDVP;
fad57d0e
MD
1957 if (error == 0)
1958 error = nlookup(&nd);
72310cfb 1959 if (error == 0)
28623bf9 1960 error = ncp_writechk(&nd.nl_nch);
dff430ab 1961 if (error == 0) {
5312fa43
MD
1962 error = VOP_NWHITEOUT(&nd.nl_nch, nd.nl_dvp, nd.nl_cred,
1963 NAMEI_DELETE);
dff430ab 1964 }
fad57d0e 1965 nlookup_done(&nd);
984263bc
MD
1966 return (error);
1967}
1968
984263bc 1969int
fad57d0e 1970kern_unlink(struct nlookupdata *nd)
984263bc 1971{
984263bc 1972 int error;
984263bc
MD
1973
1974 bwillwrite();
5312fa43 1975 nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
fad57d0e 1976 if ((error = nlookup(nd)) != 0)
984263bc 1977 return (error);
28623bf9 1978 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 1979 return (error);
5312fa43 1980 error = VOP_NREMOVE(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
984263bc
MD
1981 return (error);
1982}
1983
1984/*
9697c509 1985 * unlink_args(char *path)
41c20dac 1986 *
9697c509 1987 * Delete a name from the filesystem.
984263bc 1988 */
984263bc 1989int
753fd850 1990sys_unlink(struct unlink_args *uap)
9697c509 1991{
fad57d0e 1992 struct nlookupdata nd;
9697c509
DRJ
1993 int error;
1994
fad57d0e
MD
1995 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1996 if (error == 0)
1997 error = kern_unlink(&nd);
1998 nlookup_done(&nd);
9697c509
DRJ
1999 return (error);
2000}
2001
2002int
5a3fe67d 2003kern_lseek(int fd, off_t offset, int whence, off_t *res)
984263bc 2004{
dadab5e9
MD
2005 struct thread *td = curthread;
2006 struct proc *p = td->td_proc;
41c20dac 2007 struct file *fp;
984263bc
MD
2008 struct vattr vattr;
2009 int error;
2010
228b401d
MD
2011 fp = holdfp(p->p_fd, fd, -1);
2012 if (fp == NULL)
984263bc 2013 return (EBADF);
228b401d
MD
2014 if (fp->f_type != DTYPE_VNODE) {
2015 error = ESPIPE;
2016 goto done;
2017 }
2018
9697c509 2019 switch (whence) {
984263bc 2020 case L_INCR:
9697c509 2021 fp->f_offset += offset;
228b401d 2022 error = 0;
984263bc
MD
2023 break;
2024 case L_XTND:
228b401d
MD
2025 error = VOP_GETATTR((struct vnode *)fp->f_data, &vattr);
2026 if (error == 0)
2027 fp->f_offset = offset + vattr.va_size;
984263bc
MD
2028 break;
2029 case L_SET:
9697c509 2030 fp->f_offset = offset;
228b401d 2031 error = 0;
984263bc
MD
2032 break;
2033 default:
228b401d
MD
2034 error = EINVAL;
2035 break;
984263bc 2036 }
9697c509 2037 *res = fp->f_offset;
228b401d
MD
2038done:
2039 fdrop(fp);
2040 return (error);
984263bc
MD
2041}
2042
984263bc 2043/*
9697c509 2044 * lseek_args(int fd, int pad, off_t offset, int whence)
41c20dac 2045 *
9697c509 2046 * Reposition read/write file offset.
984263bc 2047 */
984263bc 2048int
753fd850 2049sys_lseek(struct lseek_args *uap)
984263bc 2050{
984263bc
MD
2051 int error;
2052
9697c509 2053 error = kern_lseek(uap->fd, uap->offset, uap->whence,
5a3fe67d 2054 &uap->sysmsg_offset);
9697c509 2055
984263bc
MD
2056 return (error);
2057}
984263bc 2058
984263bc 2059int
fad57d0e 2060kern_access(struct nlookupdata *nd, int aflags)
984263bc 2061{
41c20dac 2062 struct vnode *vp;
984263bc 2063 int error, flags;
984263bc 2064
fad57d0e
MD
2065 if ((error = nlookup(nd)) != 0)
2066 return (error);
806dcf9a 2067retry:
28623bf9 2068 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 2069 if (error)
fad57d0e 2070 return (error);
984263bc
MD
2071
2072 /* Flags == 0 means only check for existence. */
9697c509 2073 if (aflags) {
984263bc 2074 flags = 0;
9697c509 2075 if (aflags & R_OK)
984263bc 2076 flags |= VREAD;
9697c509 2077 if (aflags & W_OK)
984263bc 2078 flags |= VWRITE;
9697c509 2079 if (aflags & X_OK)
984263bc 2080 flags |= VEXEC;
468bb1f9 2081 if ((flags & VWRITE) == 0 ||
28623bf9 2082 (error = vn_writechk(vp, &nd->nl_nch)) == 0)
87de5057 2083 error = VOP_ACCESS(vp, flags, nd->nl_cred);
806dcf9a
MD
2084
2085 /*
2086 * If the file handle is stale we have to re-resolve the
2087 * entry. This is a hack at the moment.
2088 */
2089 if (error == ESTALE) {
5e1a76ea 2090 vput(vp);
28623bf9
MD
2091 cache_setunresolved(&nd->nl_nch);
2092 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
806dcf9a 2093 if (error == 0) {
806dcf9a
MD
2094 vp = NULL;
2095 goto retry;
2096 }
5e1a76ea 2097 return(error);
806dcf9a 2098 }
984263bc 2099 }
984263bc 2100 vput(vp);
984263bc
MD
2101 return (error);
2102}
2103
984263bc 2104/*
9697c509 2105 * access_args(char *path, int flags)
41c20dac 2106 *
9697c509 2107 * Check access permissions.
984263bc 2108 */
984263bc 2109int
753fd850 2110sys_access(struct access_args *uap)
984263bc 2111{
fad57d0e 2112 struct nlookupdata nd;
9697c509 2113 int error;
984263bc 2114
fad57d0e
MD
2115 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2116 if (error == 0)
2117 error = kern_access(&nd, uap->flags);
2118 nlookup_done(&nd);
984263bc
MD
2119 return (error);
2120}
2121
984263bc 2122int
21739618 2123kern_stat(struct nlookupdata *nd, struct stat *st)
984263bc 2124{
984263bc 2125 int error;
21739618
MD
2126 struct vnode *vp;
2127 thread_t td;
984263bc 2128
21739618 2129 if ((error = nlookup(nd)) != 0)
984263bc 2130 return (error);
806dcf9a 2131again:
28623bf9 2132 if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
21739618
MD
2133 return (ENOENT);
2134
2135 td = curthread;
87de5057 2136 if ((error = vget(vp, LK_SHARED)) != 0)
21739618 2137 return (error);
87de5057 2138 error = vn_stat(vp, st, nd->nl_cred);
806dcf9a
MD
2139
2140 /*
2141 * If the file handle is stale we have to re-resolve the entry. This
2142 * is a hack at the moment.
2143 */
2144 if (error == ESTALE) {
5dc91765 2145 vput(vp);
28623bf9
MD
2146 cache_setunresolved(&nd->nl_nch);
2147 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
5dc91765 2148 if (error == 0)
806dcf9a 2149 goto again;
5dc91765
MD
2150 } else {
2151 vput(vp);
806dcf9a 2152 }
984263bc
MD
2153 return (error);
2154}
2155
984263bc 2156/*
41c20dac
MD
2157 * stat_args(char *path, struct stat *ub)
2158 *
984263bc
MD
2159 * Get file status; this version follows links.
2160 */
984263bc 2161int
753fd850 2162sys_stat(struct stat_args *uap)
984263bc 2163{
21739618 2164 struct nlookupdata nd;
9697c509
DRJ
2165 struct stat st;
2166 int error;
984263bc 2167
21739618
MD
2168 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2169 if (error == 0) {
2170 error = kern_stat(&nd, &st);
2171 if (error == 0)
2172 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 2173 }
fad57d0e 2174 nlookup_done(&nd);
984263bc
MD
2175 return (error);
2176}
2177
2178/*
41c20dac
MD
2179 * lstat_args(char *path, struct stat *ub)
2180 *
984263bc
MD
2181 * Get file status; this version does not follow links.
2182 */
984263bc 2183int
753fd850 2184sys_lstat(struct lstat_args *uap)
984263bc 2185{
21739618 2186 struct nlookupdata nd;
9697c509
DRJ
2187 struct stat st;
2188 int error;
984263bc 2189
21739618
MD
2190 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2191 if (error == 0) {
2192 error = kern_stat(&nd, &st);
2193 if (error == 0)
2194 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 2195 }
fad57d0e 2196 nlookup_done(&nd);
984263bc
MD
2197 return (error);
2198}
2199
984263bc 2200/*
41c20dac
MD
2201 * pathconf_Args(char *path, int name)
2202 *
984263bc
MD
2203 * Get configurable pathname variables.
2204 */
984263bc
MD
2205/* ARGSUSED */
2206int
753fd850 2207sys_pathconf(struct pathconf_args *uap)
984263bc 2208{
fad57d0e
MD
2209 struct nlookupdata nd;
2210 struct vnode *vp;
984263bc 2211 int error;
984263bc 2212
fad57d0e 2213 vp = NULL;
ab2eb4eb 2214 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2215 if (error == 0)
2216 error = nlookup(&nd);
2217 if (error == 0)
28623bf9 2218 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
fad57d0e
MD
2219 nlookup_done(&nd);
2220 if (error == 0) {
ab2eb4eb 2221 error = VOP_PATHCONF(vp, uap->name, uap->sysmsg_fds);
fad57d0e
MD
2222 vput(vp);
2223 }
984263bc
MD
2224 return (error);
2225}
2226
bbab27f6
DRJ
2227/*
2228 * XXX: daver
2229 * kern_readlink isn't properly split yet. There is a copyin burried
2230 * in VOP_READLINK().
2231 */
984263bc 2232int
fad57d0e 2233kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
984263bc 2234{
dadab5e9
MD
2235 struct thread *td = curthread;
2236 struct proc *p = td->td_proc;
41c20dac 2237 struct vnode *vp;
984263bc
MD
2238 struct iovec aiov;
2239 struct uio auio;
2240 int error;
984263bc 2241
fad57d0e
MD
2242 if ((error = nlookup(nd)) != 0)
2243 return (error);
28623bf9 2244 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 2245 if (error)
984263bc 2246 return (error);
fad57d0e 2247 if (vp->v_type != VLNK) {
984263bc 2248 error = EINVAL;
fad57d0e 2249 } else {
9697c509
DRJ
2250 aiov.iov_base = buf;
2251 aiov.iov_len = count;
984263bc
MD
2252 auio.uio_iov = &aiov;
2253 auio.uio_iovcnt = 1;
2254 auio.uio_offset = 0;
2255 auio.uio_rw = UIO_READ;
2256 auio.uio_segflg = UIO_USERSPACE;
dadab5e9 2257 auio.uio_td = td;
9697c509 2258 auio.uio_resid = count;
984263bc
MD
2259 error = VOP_READLINK(vp, &auio, p->p_ucred);
2260 }
2261 vput(vp);
9697c509
DRJ
2262 *res = count - auio.uio_resid;
2263 return (error);
2264}
2265
2266/*
2267 * readlink_args(char *path, char *buf, int count)
2268 *
2269 * Return target name of a symbolic link.
2270 */
2271int
753fd850 2272sys_readlink(struct readlink_args *uap)
9697c509 2273{
fad57d0e 2274 struct nlookupdata nd;
9697c509
DRJ
2275 int error;
2276
fad57d0e
MD
2277 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2278 if (error == 0) {
2279 error = kern_readlink(&nd, uap->buf, uap->count,
2280 &uap->sysmsg_result);
2281 }
2282 nlookup_done(&nd);
984263bc
MD
2283 return (error);
2284}
2285
2286static int
41c20dac 2287setfflags(struct vnode *vp, int flags)
984263bc 2288{
dadab5e9
MD
2289 struct thread *td = curthread;
2290 struct proc *p = td->td_proc;
984263bc
MD
2291 int error;
2292 struct vattr vattr;
2293
2294 /*
2295 * Prevent non-root users from setting flags on devices. When
2296 * a device is reused, users can retain ownership of the device
2297 * if they are allowed to set flags and programs assume that
2298 * chown can't fail when done as root.
2299 */
2300 if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
dadab5e9 2301 ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
984263bc
MD
2302 return (error);
2303
186fccd6
MD
2304 /*
2305 * note: vget is required for any operation that might mod the vnode
2306 * so VINACTIVE is properly cleared.
2307 */
87de5057 2308 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2309 VATTR_NULL(&vattr);
2310 vattr.va_flags = flags;
87de5057 2311 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2312 vput(vp);
2313 }
984263bc
MD
2314 return (error);
2315}
2316
2317/*
41c20dac
MD
2318 * chflags(char *path, int flags)
2319 *
984263bc
MD
2320 * Change flags of a file given a path name.
2321 */
984263bc
MD
2322/* ARGSUSED */
2323int
753fd850 2324sys_chflags(struct chflags_args *uap)
984263bc 2325{
fad57d0e
MD
2326 struct nlookupdata nd;
2327 struct vnode *vp;
984263bc 2328 int error;
984263bc 2329
fad57d0e 2330 vp = NULL;
ab2eb4eb 2331 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e
MD
2332 /* XXX Add NLC flag indicating modifying operation? */
2333 if (error == 0)
2334 error = nlookup(&nd);
72310cfb 2335 if (error == 0)
28623bf9 2336 error = ncp_writechk(&nd.nl_nch);
fad57d0e 2337 if (error == 0)
28623bf9 2338 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
fad57d0e
MD
2339 nlookup_done(&nd);
2340 if (error == 0) {
ab2eb4eb 2341 error = setfflags(vp, uap->flags);
fad57d0e
MD
2342 vrele(vp);
2343 }
2344 return (error);
984263bc
MD
2345}
2346
2347/*
41c20dac
MD
2348 * fchflags_args(int fd, int flags)
2349 *
984263bc
MD
2350 * Change flags of a file given a file descriptor.
2351 */
984263bc
MD
2352/* ARGSUSED */
2353int
753fd850 2354sys_fchflags(struct fchflags_args *uap)
984263bc 2355{
dadab5e9
MD
2356 struct thread *td = curthread;
2357 struct proc *p = td->td_proc;
984263bc
MD
2358 struct file *fp;
2359 int error;
2360
5b287bba 2361 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2362 return (error);
28623bf9
MD
2363 if (fp->f_nchandle.ncp)
2364 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
2365 if (error == 0)
2366 error = setfflags((struct vnode *) fp->f_data, uap->flags);
5b287bba
MD
2367 fdrop(fp);
2368 return (error);
984263bc
MD
2369}
2370
2371static int
41c20dac 2372setfmode(struct vnode *vp, int mode)
984263bc 2373{
dadab5e9
MD
2374 struct thread *td = curthread;
2375 struct proc *p = td->td_proc;
984263bc
MD
2376 int error;
2377 struct vattr vattr;
2378
186fccd6
MD
2379 /*
2380 * note: vget is required for any operation that might mod the vnode
2381 * so VINACTIVE is properly cleared.
2382 */
87de5057 2383 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2384 VATTR_NULL(&vattr);
2385 vattr.va_mode = mode & ALLPERMS;
87de5057 2386 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2387 vput(vp);
2388 }
984263bc
MD
2389 return error;
2390}
2391
9697c509 2392int
fad57d0e 2393kern_chmod(struct nlookupdata *nd, int mode)
9697c509 2394{
fad57d0e 2395 struct vnode *vp;
9697c509
DRJ
2396 int error;
2397
fad57d0e
MD
2398 /* XXX Add NLC flag indicating modifying operation? */
2399 if ((error = nlookup(nd)) != 0)
9697c509 2400 return (error);
28623bf9 2401 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 2402 return (error);
28623bf9 2403 if ((error = ncp_writechk(&nd->nl_nch)) == 0)
72310cfb 2404 error = setfmode(vp, mode);
fad57d0e
MD
2405 vrele(vp);
2406 return (error);
9697c509
DRJ
2407}
2408
984263bc 2409/*
41c20dac
MD
2410 * chmod_args(char *path, int mode)
2411 *
984263bc
MD
2412 * Change mode of a file given path name.
2413 */
984263bc
MD
2414/* ARGSUSED */
2415int
753fd850 2416sys_chmod(struct chmod_args *uap)
984263bc 2417{
fad57d0e 2418 struct nlookupdata nd;
9697c509 2419 int error;
984263bc 2420
fad57d0e
MD
2421 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2422 if (error == 0)
2423 error = kern_chmod(&nd, uap->mode);
2424 nlookup_done(&nd);
9697c509 2425 return (error);
984263bc
MD
2426}
2427
2428/*
41c20dac
MD
2429 * lchmod_args(char *path, int mode)
2430 *
984263bc
MD
2431 * Change mode of a file given path name (don't follow links.)
2432 */
984263bc
MD
2433/* ARGSUSED */
2434int
753fd850 2435sys_lchmod(struct lchmod_args *uap)
984263bc 2436{
fad57d0e 2437 struct nlookupdata nd;
984263bc 2438 int error;
984263bc 2439
fad57d0e
MD
2440 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2441 if (error == 0)
2442 error = kern_chmod(&nd, uap->mode);
2443 nlookup_done(&nd);
2444 return (error);
984263bc
MD
2445}
2446
2447/*
41c20dac
MD
2448 * fchmod_args(int fd, int mode)
2449 *
984263bc
MD
2450 * Change mode of a file given a file descriptor.
2451 */
984263bc
MD
2452/* ARGSUSED */
2453int
753fd850 2454sys_fchmod(struct fchmod_args *uap)
984263bc 2455{
dadab5e9
MD
2456 struct thread *td = curthread;
2457 struct proc *p = td->td_proc;
984263bc
MD
2458 struct file *fp;
2459 int error;
2460
5b287bba 2461 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2462 return (error);
28623bf9
MD
2463 if (fp->f_nchandle.ncp)
2464 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
2465 if (error == 0)
2466 error = setfmode((struct vnode *)fp->f_data, uap->mode);
5b287bba
MD
2467 fdrop(fp);
2468 return (error);
984263bc
MD
2469}
2470
2471static int
41c20dac 2472setfown(struct vnode *vp, uid_t uid, gid_t gid)
984263bc 2473{
dadab5e9
MD
2474 struct thread *td = curthread;
2475 struct proc *p = td->td_proc;
984263bc
MD
2476 int error;
2477 struct vattr vattr;
2478
186fccd6
MD
2479 /*
2480 * note: vget is required for any operation that might mod the vnode
2481 * so VINACTIVE is properly cleared.
2482 */
87de5057 2483 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2484 VATTR_NULL(&vattr);
2485 vattr.va_uid = uid;
2486 vattr.va_gid = gid;
87de5057 2487 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2488 vput(vp);
2489 }
984263bc
MD
2490 return error;
2491}
2492
9697c509 2493int
fad57d0e 2494kern_chown(struct nlookupdata *nd, int uid, int gid)
9697c509 2495{
fad57d0e 2496 struct vnode *vp;
9697c509
DRJ
2497 int error;
2498
fad57d0e
MD
2499 /* XXX Add NLC flag indicating modifying operation? */
2500 if ((error = nlookup(nd)) != 0)
9697c509 2501 return (error);
28623bf9 2502 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 2503 return (error);
28623bf9 2504 if ((error = ncp_writechk(&nd->nl_nch)) == 0)
72310cfb 2505 error = setfown(vp, uid, gid);
fad57d0e 2506 vrele(vp);
9697c509
DRJ
2507 return (error);
2508}
2509
984263bc 2510/*
41c20dac
MD
2511 * chown(char *path, int uid, int gid)
2512 *
984263bc
MD
2513 * Set ownership given a path name.
2514 */
984263bc 2515int
753fd850 2516sys_chown(struct chown_args *uap)
984263bc 2517{
fad57d0e 2518 struct nlookupdata nd;
9697c509
DRJ
2519 int error;
2520
fad57d0e
MD
2521 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2522 if (error == 0)
2523 error = kern_chown(&nd, uap->uid, uap->gid);
2524 nlookup_done(&nd);
984263bc
MD
2525 return (error);
2526}
2527
2528/*
41c20dac
MD
2529 * lchown_args(char *path, int uid, int gid)
2530 *
984263bc
MD
2531 * Set ownership given a path name, do not cross symlinks.
2532 */
984263bc 2533int
753fd850 2534sys_lchown(struct lchown_args *uap)
984263bc 2535{
fad57d0e 2536 struct nlookupdata nd;
984263bc 2537 int error;
9697c509 2538
fad57d0e
MD
2539 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2540 if (error == 0)
2541 error = kern_chown(&nd, uap->uid, uap->gid);
2542 nlookup_done(&nd);
984263bc
MD
2543 return (error);
2544}
2545
2546/*
41c20dac
MD
2547 * fchown_args(int fd, int uid, int gid)
2548 *
984263bc
MD
2549 * Set ownership given a file descriptor.
2550 */
984263bc
MD
2551/* ARGSUSED */
2552int
753fd850 2553sys_fchown(struct fchown_args *uap)
984263bc 2554{
dadab5e9
MD
2555 struct thread *td = curthread;
2556 struct proc *p = td->td_proc;
984263bc
MD
2557 struct file *fp;
2558 int error;
2559
5b287bba 2560 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc 2561 return (error);
28623bf9
MD
2562 if (fp->f_nchandle.ncp)
2563 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
2564 if (error == 0)
2565 error = setfown((struct vnode *)fp->f_data, uap->uid, uap->gid);
5b287bba
MD
2566 fdrop(fp);
2567 return (error);
984263bc
MD
2568}
2569
2570static int
9697c509 2571getutimes(const struct timeval *tvp, struct timespec *tsp)
984263bc
MD
2572{
2573 struct timeval tv[2];
984263bc 2574
9697c509 2575 if (tvp == NULL) {
984263bc
MD
2576 microtime(&tv[0]);
2577 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2578 tsp[1] = tsp[0];
2579 } else {
9697c509
DRJ
2580 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2581 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
984263bc
MD
2582 }
2583 return 0;
2584}
2585
2586static int
41c20dac 2587setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
984263bc 2588{
dadab5e9
MD
2589 struct thread *td = curthread;
2590 struct proc *p = td->td_proc;
984263bc
MD
2591 int error;
2592 struct vattr vattr;
2593
186fccd6
MD
2594 /*
2595 * note: vget is required for any operation that might mod the vnode
2596 * so VINACTIVE is properly cleared.
2597 */
87de5057 2598 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
186fccd6
MD
2599 VATTR_NULL(&vattr);
2600 vattr.va_atime = ts[0];
2601 vattr.va_mtime = ts[1];
2602 if (nullflag)
2603 vattr.va_vaflags |= VA_UTIMES_NULL;
87de5057 2604 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
186fccd6
MD
2605 vput(vp);
2606 }
984263bc
MD
2607 return error;
2608}
2609
9697c509 2610int
fad57d0e 2611kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
9697c509
DRJ
2612{
2613 struct timespec ts[2];
fad57d0e 2614 struct vnode *vp;
9697c509
DRJ
2615 int error;
2616
fad57d0e 2617 if ((error = getutimes(tptr, ts)) != 0)
9697c509 2618 return (error);
fad57d0e
MD
2619 /* XXX Add NLC flag indicating modifying operation? */
2620 if ((error = nlookup(nd)) != 0)
9697c509 2621 return (error);
28623bf9 2622 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2623 return (error);
28623bf9 2624 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e
MD
2625 return (error);
2626 error = setutimes(vp, ts, tptr == NULL);
2627 vrele(vp);
9697c509
DRJ
2628 return (error);
2629}
2630
984263bc 2631/*
41c20dac
MD
2632 * utimes_args(char *path, struct timeval *tptr)
2633 *
984263bc
MD
2634 * Set the access and modification times of a file.
2635 */
984263bc 2636int
753fd850 2637sys_utimes(struct utimes_args *uap)
984263bc 2638{
9697c509 2639 struct timeval tv[2];
fad57d0e 2640 struct nlookupdata nd;
9697c509
DRJ
2641 int error;
2642
2643 if (uap->tptr) {
2644 error = copyin(uap->tptr, tv, sizeof(tv));
2645 if (error)
2646 return (error);
2647 }
fad57d0e
MD
2648 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2649 if (error == 0)
2650 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2651 nlookup_done(&nd);
984263bc
MD
2652 return (error);
2653}
2654
2655/*
41c20dac
MD
2656 * lutimes_args(char *path, struct timeval *tptr)
2657 *
984263bc
MD
2658 * Set the access and modification times of a file.
2659 */
984263bc 2660int
753fd850 2661sys_lutimes(struct lutimes_args *uap)
984263bc 2662{
9697c509 2663 struct timeval tv[2];
fad57d0e 2664 struct nlookupdata nd;
9697c509
DRJ
2665 int error;
2666
2667 if (uap->tptr) {
2668 error = copyin(uap->tptr, tv, sizeof(tv));
2669 if (error)
2670 return (error);
2671 }
fad57d0e
MD
2672 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2673 if (error == 0)
2674 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2675 nlookup_done(&nd);
984263bc
MD
2676 return (error);
2677}
2678
984263bc 2679int
9697c509 2680kern_futimes(int fd, struct timeval *tptr)
984263bc 2681{
dadab5e9
MD
2682 struct thread *td = curthread;
2683 struct proc *p = td->td_proc;
984263bc
MD
2684 struct timespec ts[2];
2685 struct file *fp;
984263bc
MD
2686 int error;
2687
9697c509
DRJ
2688 error = getutimes(tptr, ts);
2689 if (error)
984263bc 2690 return (error);
5b287bba 2691 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc 2692 return (error);
28623bf9
MD
2693 if (fp->f_nchandle.ncp)
2694 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
2695 if (error == 0)
2696 error = setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
5b287bba 2697 fdrop(fp);
9697c509 2698 return (error);
984263bc
MD
2699}
2700
2701/*
9697c509 2702 * futimes_args(int fd, struct timeval *tptr)
41c20dac 2703 *
9697c509 2704 * Set the access and modification times of a file.
984263bc 2705 */
984263bc 2706int
753fd850 2707sys_futimes(struct futimes_args *uap)
9697c509
DRJ
2708{
2709 struct timeval tv[2];
2710 int error;
2711
2712 if (uap->tptr) {
2713 error = copyin(uap->tptr, tv, sizeof(tv));
2714 if (error)
2715 return (error);
2716 }
2717
b651c485 2718 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
9697c509
DRJ
2719
2720 return (error);
2721}
2722
2723int
fad57d0e 2724kern_truncate(struct nlookupdata *nd, off_t length)
984263bc 2725{
41c20dac 2726 struct vnode *vp;
984263bc
MD
2727 struct vattr vattr;
2728 int error;
984263bc 2729
9697c509 2730 if (length < 0)
984263bc 2731 return(EINVAL);
fad57d0e
MD
2732 /* XXX Add NLC flag indicating modifying operation? */
2733 if ((error = nlookup(nd)) != 0)
984263bc 2734 return (error);
28623bf9 2735 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 2736 return (error);
28623bf9 2737 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
fad57d0e 2738 return (error);
ca466bae 2739 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) != 0) {
fad57d0e
MD
2740 vrele(vp);
2741 return (error);
2742 }
2743 if (vp->v_type == VDIR) {
984263bc 2744 error = EISDIR;
28623bf9 2745 } else if ((error = vn_writechk(vp, &nd->nl_nch)) == 0 &&
87de5057 2746 (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred)) == 0) {
984263bc 2747 VATTR_NULL(&vattr);
9697c509 2748 vattr.va_size = length;
87de5057 2749 error = VOP_SETATTR(vp, &vattr, nd->nl_cred);
984263bc
MD
2750 }
2751 vput(vp);
2752 return (error);
2753}
2754
9697c509
DRJ
2755/*
2756 * truncate(char *path, int pad, off_t length)
2757 *
2758 * Truncate a file given its path name.
2759 */
2760int
753fd850 2761sys_truncate(struct truncate_args *uap)
9697c509 2762{
fad57d0e 2763 struct nlookupdata nd;
9697c509
DRJ
2764 int error;
2765
fad57d0e
MD
2766 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2767 if (error == 0)
2768 error = kern_truncate(&nd, uap->length);
2769 nlookup_done(&nd);
9697c509
DRJ
2770 return error;
2771}
2772
984263bc 2773int
8f6f8622 2774kern_ftruncate(int fd, off_t length)
984263bc 2775{
dadab5e9
MD
2776 struct thread *td = curthread;
2777 struct proc *p = td->td_proc;
984263bc
MD
2778 struct vattr vattr;
2779 struct vnode *vp;
2780 struct file *fp;
2781 int error;
2782
8f6f8622 2783 if (length < 0)
984263bc 2784 return(EINVAL);
5b287bba 2785 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc 2786 return (error);
28623bf9
MD
2787 if (fp->f_nchandle.ncp) {
2788 error = ncp_writechk(&fp->f_nchandle);
72310cfb
MD
2789 if (error)
2790 goto done;
2791 }
5b287bba
MD
2792 if ((fp->f_flag & FWRITE) == 0) {
2793 error = EINVAL;
2794 goto done;
2795 }
984263bc 2796 vp = (struct vnode *)fp->f_data;
ca466bae 2797 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
5b287bba 2798 if (vp->v_type == VDIR) {
984263bc 2799 error = EISDIR;
468bb1f9 2800 } else if ((error = vn_writechk(vp, NULL)) == 0) {
984263bc 2801 VATTR_NULL(&vattr);
8f6f8622 2802 vattr.va_size = length;
87de5057 2803 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
984263bc 2804 }
a11aaa81 2805 vn_unlock(vp);
5b287bba
MD
2806done:
2807 fdrop(fp);
984263bc
MD
2808 return (error);
2809}
2810
984263bc 2811/*
8f6f8622 2812 * ftruncate_args(int fd, int pad, off_t length)
41c20dac 2813 *
8f6f8622 2814 * Truncate a file given a file descriptor.
984263bc 2815 */
984263bc 2816int
753fd850 2817sys_ftruncate(struct ftruncate_args *uap)
984263bc 2818{
8f6f8622 2819 int error;
984263bc 2820
8f6f8622
DRJ
2821 error = kern_ftruncate(uap->fd, uap->length);
2822
2823 return (error);
984263bc
MD
2824}
2825
984263bc 2826/*
41c20dac
MD
2827 * fsync(int fd)
2828 *
984263bc
MD
2829 * Sync an open file.
2830 */
984263bc
MD
2831/* ARGSUSED */
2832int
753fd850 2833sys_fsync(struct fsync_args *uap)
984263bc 2834{
dadab5e9
MD
2835 struct thread *td = curthread;
2836 struct proc *p = td->td_proc;
41c20dac 2837 struct vnode *vp;
984263bc
MD
2838 struct file *fp;
2839 vm_object_t obj;
2840 int error;
2841
5b287bba 2842 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
984263bc
MD
2843 return (error);
2844 vp = (struct vnode *)fp->f_data;
ca466bae 2845 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
7540ab49 2846 if ((obj = vp->v_object) != NULL)
984263bc 2847 vm_object_page_clean(obj, 0, 0, 0);
408357d8
MD
2848 if ((error = VOP_FSYNC(vp, MNT_WAIT)) == 0 && vp->v_mount)
2849 error = buf_fsync(vp);
a11aaa81 2850 vn_unlock(vp);
5b287bba 2851 fdrop(fp);
984263bc
MD
2852 return (error);
2853}
2854
984263bc 2855int
fad57d0e 2856kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
984263bc 2857{
28623bf9
MD
2858 struct nchandle fnchd;
2859 struct nchandle tnchd;
fad57d0e 2860 struct namecache *ncp;
dff430ab
MD
2861 struct vnode *fdvp;
2862 struct vnode *tdvp;
fad57d0e 2863 struct mount *mp;
984263bc
MD
2864 int error;
2865
2866 bwillwrite();
5312fa43 2867 fromnd->nl_flags |= NLC_REFDVP;
fad57d0e 2868 if ((error = nlookup(fromnd)) != 0)
984263bc 2869 return (error);
28623bf9 2870 if ((fnchd.ncp = fromnd->nl_nch.ncp->nc_parent) == NULL)
fad57d0e 2871 return (ENOENT);
28623bf9
MD
2872 fnchd.mount = fromnd->nl_nch.mount;
2873 cache_hold(&fnchd);
fad57d0e
MD
2874
2875 /*
28623bf9 2876 * unlock the source nch so we can lookup the target nch without
fad57d0e
MD
2877 * deadlocking. The target may or may not exist so we do not check
2878 * for a target vp like kern_mkdir() and other creation functions do.
2879 *
2880 * The source and target directories are ref'd and rechecked after
2881 * everything is relocked to determine if the source or target file
2882 * has been renamed.
2883 */
2884 KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
2885 fromnd->nl_flags &= ~NLC_NCPISLOCKED;
28623bf9 2886 cache_unlock(&fromnd->nl_nch);
fad57d0e 2887
5312fa43 2888 tond->nl_flags |= NLC_CREATE | NLC_REFDVP;
fad57d0e 2889 if ((error = nlookup(tond)) != 0) {
28623bf9 2890 cache_drop(&fnchd);
fad57d0e
MD
2891 return (error);
2892 }
28623bf9
MD
2893 if ((tnchd.ncp = tond->nl_nch.ncp->nc_parent) == NULL) {
2894 cache_drop(&fnchd);
fad57d0e
MD
2895 return (ENOENT);
2896 }
28623bf9
MD
2897 tnchd.mount = tond->nl_nch.mount;
2898 cache_hold(&tnchd);
fad57d0e
MD
2899
2900 /*
2901 * If the source and target are the same there is nothing to do
2902 */
28623bf9
MD
2903 if (fromnd->nl_nch.ncp == tond->nl_nch.ncp) {
2904 cache_drop(&fnchd);
2905 cache_drop(&tnchd);
fad57d0e
MD
2906 return (0);
2907 }
2908
1d505369
MD
2909 /*
2910 * Mount points cannot be renamed or overwritten
2911 */
28623bf9
MD
2912 if ((fromnd->nl_nch.ncp->nc_flag | tond->nl_nch.ncp->nc_flag) &
2913 NCF_ISMOUNTPT
1d505369 2914 ) {
28623bf9
MD
2915 cache_drop(&fnchd);
2916 cache_drop(&tnchd);
1d505369
MD
2917 return (EINVAL);
2918 }
2919
fad57d0e 2920 /*
aa40af33
MD
2921 * relock the source ncp. NOTE AFTER RELOCKING: the source ncp
2922 * may have become invalid while it was unlocked, nc_vp and nc_mount
2923 * could be NULL.
fad57d0e 2924 */
28623bf9
MD
2925 if (cache_lock_nonblock(&fromnd->nl_nch) == 0) {
2926 cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
2927 } else if (fromnd->nl_nch.ncp > tond->nl_nch.ncp) {
2928 cache_lock(&fromnd->nl_nch);
2929 cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
fad57d0e 2930 } else {
28623bf9
MD
2931 cache_unlock(&tond->nl_nch);
2932 cache_lock(&fromnd->nl_nch);
2933 cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
2934 cache_lock(&tond->nl_nch);
2935 cache_resolve(&tond->nl_nch, tond->nl_cred);
fad57d0e
MD
2936 }
2937 fromnd->nl_flags |= NLC_NCPISLOCKED;
2938
2939 /*
2940 * make sure the parent directories linkages are the same
2941 */
28623bf9
MD
2942 if (fnchd.ncp != fromnd->nl_nch.ncp->nc_parent ||
2943 tnchd.ncp != tond->nl_nch.ncp->nc_parent) {
2944 cache_drop(&fnchd);
2945 cache_drop(&tnchd);
fad57d0e
MD
2946 return (ENOENT);
2947 }
2948
2949 /*
2950 * Both the source and target must be within the same filesystem and
2951 * in the same filesystem as their parent directories within the
2952 * namecache topology.
aa40af33
MD
2953 *
2954 * NOTE: fromnd's nc_mount or nc_vp could be NULL.
fad57d0e 2955 */
28623bf9
MD
2956 mp = fnchd.mount;
2957 if (mp != tnchd.mount || mp != fromnd->nl_nch.mount ||
2958 mp != tond->nl_nch.mount) {
2959 cache_drop(&fnchd);
2960 cache_drop(&tnchd);
fad57d0e
MD
2961 return (EXDEV);
2962 }
2963
72310cfb
MD
2964 /*
2965 * Make sure the mount point is writable
2966 */
28623bf9
MD
2967 if ((error = ncp_writechk(&tond->nl_nch)) != 0) {
2968 cache_drop(&fnchd);
2969 cache_drop(&tnchd);
72310cfb
MD
2970 return (error);
2971 }
2972
fad57d0e
MD
2973 /*
2974 * If the target exists and either the source or target is a directory,
2975 * then both must be directories.
aa40af33 2976 *
28623bf9
MD
2977 * Due to relocking of the source, fromnd->nl_nch.ncp->nc_vp might h
2978 * have become NULL.
fad57d0e 2979 */
28623bf9
MD
2980 if (tond->nl_nch.ncp->nc_vp) {
2981 if (fromnd->nl_nch.ncp->nc_vp == NULL) {
aa40af33 2982 error = ENOENT;
28623bf9
MD
2983 } else if (fromnd->nl_nch.ncp->nc_vp->v_type == VDIR) {
2984 if (tond->nl_nch.ncp->nc_vp->v_type != VDIR)
fad57d0e 2985 error = ENOTDIR;
28623bf9 2986 } else if (tond->nl_nch.ncp->nc_vp->v_type == VDIR) {
984263bc 2987 error = EISDIR;
984263bc
MD
2988 }
2989 }
fad57d0e 2990
984263bc 2991 /*
fad57d0e
MD
2992 * You cannot rename a source into itself or a subdirectory of itself.
2993 * We check this by travsersing the target directory upwards looking
2994 * for a match against the source.
984263bc 2995 */
fad57d0e 2996 if (error == 0) {
28623bf9
MD
2997 for (ncp = tnchd.ncp; ncp; ncp = ncp->nc_parent) {
2998 if (fromnd->nl_nch.ncp == ncp) {
fad57d0e
MD
2999 error = EINVAL;
3000 break;
3001 }
984263bc 3002 }
fad57d0e
MD
3003 }
3004
28623bf9
MD
3005 cache_drop(&fnchd);
3006 cache_drop(&tnchd);
bc8fd9e0
MD
3007
3008 /*
3009 * Even though the namespaces are different, they may still represent
3010 * hardlinks to the same file. The filesystem might have a hard time
3011 * with this so we issue a NREMOVE of the source instead of a NRENAME
3012 * when we detect the situation.
3013 */
3014 if (error == 0) {
5312fa43
MD
3015 fdvp = fromnd->nl_dvp;
3016 tdvp = tond->nl_dvp;
dff430ab
MD
3017 if (fdvp == NULL || tdvp == NULL) {
3018 error = EPERM;
3019 } else if (fromnd->nl_nch.ncp->nc_vp == tond->nl_nch.ncp->nc_vp) {
dff430ab
MD
3020 error = VOP_NREMOVE(&fromnd->nl_nch, fdvp,
3021 fromnd->nl_cred);
bc8fd9e0 3022 } else {
28623bf9 3023 error = VOP_NRENAME(&fromnd->nl_nch, &tond->nl_nch,
dff430ab 3024 fdvp, tdvp, tond->nl_cred);
bc8fd9e0
MD
3025 }
3026 }
984263bc
MD
3027 return (error);
3028}
3029
3030/*
9697c509 3031 * rename_args(char *from, char *to)
41c20dac 3032 *
9697c509
DRJ
3033 * Rename files. Source and destination must either both be directories,
3034 * or both not be directories. If target is a directory, it must be empty.
984263bc 3035 */
984263bc 3036int
753fd850 3037sys_rename(struct rename_args *uap)
9697c509 3038{
fad57d0e 3039 struct nlookupdata fromnd, tond;
9697c509
DRJ
3040 int error;
3041
fad57d0e
MD
3042 error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
3043 if (error == 0) {
3044 error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
3045 if (error == 0)
3046 error = kern_rename(&fromnd, &tond);
3047 nlookup_done(&tond);
3048 }
3049 nlookup_done(&fromnd);
9697c509
DRJ
3050 return (error);
3051}
3052
3053int
fad57d0e 3054kern_mkdir(struct nlookupdata *nd, int mode)
984263bc 3055{
dadab5e9
MD
3056 struct thread *td = curthread;
3057 struct proc *p = td->td_proc;
41c20dac 3058 struct vnode *vp;
984263bc
MD
3059 struct vattr vattr;
3060 int error;
984263bc
MD
3061
3062 bwillwrite();
5312fa43 3063 nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE | NLC_REFDVP;
fad57d0e 3064 if ((error = nlookup(nd)) != 0)
984263bc 3065 return (error);
fad57d0e 3066
28623bf9 3067 if (nd->nl_nch.ncp->nc_vp)
984263bc 3068 return (EEXIST);
28623bf9 3069 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 3070 return (error);
984263bc
MD
3071 VATTR_NULL(&vattr);
3072 vattr.va_type = VDIR;
9697c509 3073 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
fad57d0e
MD
3074
3075 vp = NULL;
5312fa43 3076 error = VOP_NMKDIR(&nd->nl_nch, nd->nl_dvp, &vp, p->p_ucred, &vattr);
9697c509 3077 if (error == 0)
fad57d0e 3078 vput(vp);
984263bc
MD
3079 return (error);
3080}
3081
3082/*
9697c509 3083 * mkdir_args(char *path, int mode)
41c20dac 3084 *
9697c509 3085 * Make a directory file.
984263bc 3086 */
984263bc
MD
3087/* ARGSUSED */
3088int
753fd850 3089sys_mkdir(struct mkdir_args *uap)
9697c509 3090{
fad57d0e 3091 struct nlookupdata nd;
9697c509
DRJ
3092 int error;
3093
fad57d0e
MD
3094 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3095 if (error == 0)
3096 error = kern_mkdir(&nd, uap->mode);
3097 nlookup_done(&nd);
9697c509
DRJ
3098 return (error);
3099}
3100
3101int
fad57d0e 3102kern_rmdir(struct nlookupdata *nd)
984263bc 3103{
984263bc 3104 int error;
984263bc
MD
3105
3106 bwillwrite();
5312fa43 3107 nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
fad57d0e 3108 if ((error = nlookup(nd)) != 0)
984263bc 3109 return (error);
1d505369
MD
3110
3111 /*
3112 * Do not allow directories representing mount points to be
3113 * deleted, even if empty. Check write perms on mount point
3114 * in case the vnode is aliased (aka nullfs).
3115 */
28623bf9 3116 if (nd->nl_nch.ncp->nc_flag & (NCF_ISMOUNTPT))
1d505369 3117 return (EINVAL);
28623bf9 3118 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
72310cfb 3119 return (error);
5312fa43 3120 error = VOP_NRMDIR(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
984263bc
MD
3121 return (error);
3122}
3123
984263bc 3124/*
9697c509 3125 * rmdir_args(char *path)
41c20dac 3126 *
9697c509 3127 * Remove a directory file.
984263bc 3128 */
9697c509 3129/* ARGSUSED */
984263bc 3130int
753fd850 3131sys_rmdir(struct rmdir_args *uap)
984263bc 3132{
fad57d0e 3133 struct nlookupdata nd;
9697c509
DRJ
3134 int error;
3135
fad57d0e
MD
3136 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3137 if (error == 0)
3138 error = kern_rmdir(&nd);
3139 nlookup_done(&nd);
984263bc
MD
3140 return (error);
3141}
984263bc 3142
984263bc 3143int
53b02c48
JS
3144kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res,
3145 enum uio_seg direction)
984263bc 3146{
dadab5e9
MD
3147 struct thread *td = curthread;
3148 struct proc *p = td->td_proc;
984263bc
MD
3149 struct vnode *vp;
3150 struct file *fp;
3151 struct uio auio;
3152 struct iovec aiov;
7fd3e801 3153 off_t loff;
984263bc
MD
3154 int error, eofflag;
3155
5b287bba 3156 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
984263bc 3157 return (error);
5b287bba
MD