Changed comment to match code.
[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 $
806dcf9a 40 * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.48 2004/11/24 08:37:16 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>
50#include <sys/sysproto.h>
984263bc
MD
51#include <sys/filedesc.h>
52#include <sys/kernel.h>
53#include <sys/fcntl.h>
54#include <sys/file.h>
55#include <sys/linker.h>
56#include <sys/stat.h>
57#include <sys/unistd.h>
58#include <sys/vnode.h>
59#include <sys/proc.h>
dadab5e9 60#include <sys/namei.h>
21739618 61#include <sys/nlookup.h>
984263bc
MD
62#include <sys/dirent.h>
63#include <sys/extattr.h>
8f6f8622 64#include <sys/kern_syscall.h>
984263bc
MD
65
66#include <machine/limits.h>
1f2de5d4 67#include <vfs/union/union.h>
984263bc
MD
68#include <sys/sysctl.h>
69#include <vm/vm.h>
70#include <vm/vm_object.h>
71#include <vm/vm_zone.h>
72#include <vm/vm_page.h>
73
dadab5e9
MD
74#include <sys/file2.h>
75
e24b948e 76static int checkvp_chdir (struct vnode *vn, struct thread *td);
21739618 77static void checkdirs (struct vnode *olddp, struct namecache *ncp);
402ed7e1
RG
78static int chroot_refuse_vdir_fds (struct filedesc *fdp);
79static int getutimes (const struct timeval *, struct timespec *);
80static int setfown (struct vnode *, uid_t, gid_t);
81static int setfmode (struct vnode *, int);
82static int setfflags (struct vnode *, int);
83static int setutimes (struct vnode *, const struct timespec *, int);
984263bc
MD
84static int usermount = 0; /* if 1, non-root can mount fs. */
85
402ed7e1 86int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
984263bc
MD
87
88SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
89
90/*
91 * Virtual File System System Calls
92 */
93
94/*
95 * Mount a file system.
96 */
41c20dac
MD
97/*
98 * mount_args(char *type, char *path, int flags, caddr_t data)
99 */
984263bc
MD
100/* ARGSUSED */
101int
41c20dac 102mount(struct mount_args *uap)
984263bc 103{
dadab5e9
MD
104 struct thread *td = curthread;
105 struct proc *p = td->td_proc;
984263bc 106 struct vnode *vp;
21739618 107 struct namecache *ncp;
984263bc
MD
108 struct mount *mp;
109 struct vfsconf *vfsp;
110 int error, flag = 0, flag2 = 0;
111 struct vattr va;
21739618 112 struct nlookupdata nd;
984263bc 113 char fstypename[MFSNAMELEN];
41a01a4d 114 lwkt_tokref ilock;
21739618 115 struct nlcomponent nlc;
984263bc 116
cd0d3e17 117 KKASSERT(p);
84c3a710
HP
118 if (p->p_ucred->cr_prison != NULL)
119 return (EPERM);
dadab5e9 120 if (usermount == 0 && (error = suser(td)))
984263bc
MD
121 return (error);
122 /*
123 * Do not allow NFS export by non-root users.
124 */
125 if (SCARG(uap, flags) & MNT_EXPORTED) {
dadab5e9 126 error = suser(td);
984263bc
MD
127 if (error)
128 return (error);
129 }
130 /*
131 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
132 */
dadab5e9 133 if (suser(td))
984263bc 134 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
21739618 135
984263bc 136 /*
21739618 137 * Lookup the requested path and extract the ncp and vnode.
984263bc 138 */
21739618
MD
139 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
140 if (error == 0) {
141 if ((error = nlookup(&nd)) == 0) {
142 if (nd.nl_ncp->nc_vp == NULL)
143 error = ENOENT;
144 }
145 }
146 if (error) {
147 nlookup_done(&nd);
984263bc 148 return (error);
21739618
MD
149 }
150
151 /*
152 * Extract the locked+refd ncp and cleanup the nd structure
153 */
154 ncp = nd.nl_ncp;
155 nd.nl_ncp = NULL;
156 nlookup_done(&nd);
157
158 /*
159 * now we have the locked ref'd ncp and unreferenced vnode.
160 */
161 vp = ncp->nc_vp;
5fd012e0 162 if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0) {
21739618
MD
163 cache_put(ncp);
164 return (error);
165 }
166 cache_unlock(ncp);
167
168 /*
169 * Now we have an unlocked ref'd ncp and a locked ref'd vp
170 */
984263bc
MD
171 if (SCARG(uap, flags) & MNT_UPDATE) {
172 if ((vp->v_flag & VROOT) == 0) {
21739618 173 cache_drop(ncp);
984263bc
MD
174 vput(vp);
175 return (EINVAL);
176 }
177 mp = vp->v_mount;
178 flag = mp->mnt_flag;
179 flag2 = mp->mnt_kern_flag;
180 /*
181 * We only allow the filesystem to be reloaded if it
182 * is currently mounted read-only.
183 */
184 if ((SCARG(uap, flags) & MNT_RELOAD) &&
185 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
21739618 186 cache_drop(ncp);
984263bc
MD
187 vput(vp);
188 return (EOPNOTSUPP); /* Needs translation */
189 }
190 /*
191 * Only root, or the user that did the original mount is
192 * permitted to update it.
193 */
194 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
dadab5e9 195 (error = suser(td))) {
21739618 196 cache_drop(ncp);
984263bc
MD
197 vput(vp);
198 return (error);
199 }
41a01a4d 200 if (vfs_busy(mp, LK_NOWAIT, NULL, td)) {
21739618 201 cache_drop(ncp);
984263bc
MD
202 vput(vp);
203 return (EBUSY);
204 }
984263bc
MD
205 if ((vp->v_flag & VMOUNT) != 0 ||
206 vp->v_mountedhere != NULL) {
21739618 207 cache_drop(ncp);
dadab5e9 208 vfs_unbusy(mp, td);
984263bc
MD
209 vput(vp);
210 return (EBUSY);
211 }
212 vp->v_flag |= VMOUNT;
984263bc
MD
213 mp->mnt_flag |=
214 SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
5fd012e0 215 VOP_UNLOCK(vp, 0, td);
984263bc
MD
216 goto update;
217 }
218 /*
219 * If the user is not root, ensure that they own the directory
220 * onto which we are attempting to mount.
221 */
3b568787 222 if ((error = VOP_GETATTR(vp, &va, td)) ||
984263bc 223 (va.va_uid != p->p_ucred->cr_uid &&
dadab5e9 224 (error = suser(td)))) {
21739618 225 cache_drop(ncp);
984263bc
MD
226 vput(vp);
227 return (error);
228 }
3b568787 229 if ((error = vinvalbuf(vp, V_SAVE, td, 0, 0)) != 0) {
21739618 230 cache_drop(ncp);
984263bc
MD
231 vput(vp);
232 return (error);
233 }
234 if (vp->v_type != VDIR) {
21739618 235 cache_drop(ncp);
984263bc
MD
236 vput(vp);
237 return (ENOTDIR);
238 }
984263bc 239 if ((error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL)) != 0) {
21739618 240 cache_drop(ncp);
984263bc
MD
241 vput(vp);
242 return (error);
243 }
21739618 244 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
984263bc
MD
245 if (!strcmp(vfsp->vfc_name, fstypename))
246 break;
21739618 247 }
984263bc
MD
248 if (vfsp == NULL) {
249 linker_file_t lf;
250
251 /* Only load modules for root (very important!) */
dadab5e9 252 if ((error = suser(td)) != 0) {
21739618 253 cache_drop(ncp);
984263bc
MD
254 vput(vp);
255 return error;
256 }
257 error = linker_load_file(fstypename, &lf);
258 if (error || lf == NULL) {
21739618 259 cache_drop(ncp);
984263bc
MD
260 vput(vp);
261 if (lf == NULL)
262 error = ENODEV;
263 return error;
264 }
265 lf->userrefs++;
266 /* lookup again, see if the VFS was loaded */
21739618 267 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
984263bc
MD
268 if (!strcmp(vfsp->vfc_name, fstypename))
269 break;
21739618 270 }
984263bc
MD
271 if (vfsp == NULL) {
272 lf->userrefs--;
273 linker_file_unload(lf);
21739618 274 cache_drop(ncp);
984263bc
MD
275 vput(vp);
276 return (ENODEV);
277 }
278 }
984263bc
MD
279 if ((vp->v_flag & VMOUNT) != 0 ||
280 vp->v_mountedhere != NULL) {
21739618 281 cache_drop(ncp);
984263bc
MD
282 vput(vp);
283 return (EBUSY);
284 }
285 vp->v_flag |= VMOUNT;
984263bc
MD
286
287 /*
288 * Allocate and initialize the filesystem.
289 */
0961aa92 290 mp = malloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
984263bc
MD
291 TAILQ_INIT(&mp->mnt_nvnodelist);
292 TAILQ_INIT(&mp->mnt_reservedvnlist);
293 mp->mnt_nvnodelistsize = 0;
377d4740 294 lockinit(&mp->mnt_lock, 0, "vfslock", 0, LK_NOPAUSE);
41a01a4d 295 vfs_busy(mp, LK_NOWAIT, NULL, td);
984263bc
MD
296 mp->mnt_op = vfsp->vfc_vfsops;
297 mp->mnt_vfc = vfsp;
298 vfsp->vfc_refcount++;
299 mp->mnt_stat.f_type = vfsp->vfc_typenum;
300 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
301 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
302 mp->mnt_vnodecovered = vp;
303 mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
304 mp->mnt_iosize_max = DFLTPHYS;
5fd012e0 305 VOP_UNLOCK(vp, 0, td);
984263bc
MD
306update:
307 /*
308 * Set the mount level flags.
309 */
310 if (SCARG(uap, flags) & MNT_RDONLY)
311 mp->mnt_flag |= MNT_RDONLY;
312 else if (mp->mnt_flag & MNT_RDONLY)
313 mp->mnt_kern_flag |= MNTK_WANTRDWR;
314 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
315 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
316 MNT_NOSYMFOLLOW | MNT_IGNORE |
317 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
318 mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
319 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
320 MNT_NOSYMFOLLOW | MNT_IGNORE |
321 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
322 /*
323 * Mount the filesystem.
324 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
fad57d0e 325 * get.
984263bc 326 */
21739618 327 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), td);
984263bc
MD
328 if (mp->mnt_flag & MNT_UPDATE) {
329 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
330 mp->mnt_flag &= ~MNT_RDONLY;
331 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
332 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
333 if (error) {
334 mp->mnt_flag = flag;
335 mp->mnt_kern_flag = flag2;
336 }
dadab5e9 337 vfs_unbusy(mp, td);
984263bc 338 vp->v_flag &= ~VMOUNT;
984263bc 339 vrele(vp);
21739618 340 cache_drop(ncp);
984263bc
MD
341 return (error);
342 }
5fd012e0 343 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc 344 /*
21739618
MD
345 * Put the new filesystem on the mount list after root. The mount
346 * point gets its own mnt_ncp which is a special ncp linking the
347 * vnode-under to the root of the new mount. The lookup code
348 * detects the mount point going forward and detects the special
349 * mnt_ncp via NCP_MOUNTPT going backwards.
8e005a45
MD
350 *
351 * It is not necessary to invalidate or purge the vnode underneath
352 * because elements under the mount will be given their own glue
353 * namecache record.
984263bc 354 */
984263bc 355 if (!error) {
21739618
MD
356 nlc.nlc_nameptr = "";
357 nlc.nlc_namelen = 0;
358 mp->mnt_ncp = cache_nlookup(ncp, &nlc);
359 mp->mnt_ncp->nc_flag |= NCF_MOUNTPT;
646a1cda 360 mp->mnt_ncp->nc_mount = mp;
21739618
MD
361 cache_drop(ncp);
362 /* XXX get the root of the fs and cache_setvp(mnt_ncp...) */
984263bc
MD
363 vp->v_flag &= ~VMOUNT;
364 vp->v_mountedhere = mp;
41a01a4d 365 lwkt_gettoken(&ilock, &mountlist_token);
984263bc 366 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
41a01a4d 367 lwkt_reltoken(&ilock);
21739618
MD
368 checkdirs(vp, mp->mnt_ncp);
369 cache_unlock(mp->mnt_ncp); /* leave ref intact */
5fd012e0 370 VOP_UNLOCK(vp, 0, td);
41a01a4d 371 error = vfs_allocate_syncvnode(mp);
dadab5e9
MD
372 vfs_unbusy(mp, td);
373 if ((error = VFS_START(mp, 0, td)) != 0)
984263bc
MD
374 vrele(vp);
375 } else {
0961aa92
MD
376 vfs_rm_vnodeops(&mp->mnt_vn_ops);
377 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
378 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
984263bc 379 vp->v_flag &= ~VMOUNT;
984263bc 380 mp->mnt_vfc->vfc_refcount--;
dadab5e9 381 vfs_unbusy(mp, td);
e38462a2 382 free(mp, M_MOUNT);
21739618 383 cache_drop(ncp);
984263bc
MD
384 vput(vp);
385 }
386 return (error);
387}
388
389/*
390 * Scan all active processes to see if any of them have a current
391 * or root directory onto which the new filesystem has just been
392 * mounted. If so, replace them with the new mount point.
21739618
MD
393 *
394 * The passed ncp is ref'd and locked (from the mount code) and
395 * must be associated with the vnode representing the root of the
396 * mount point.
984263bc
MD
397 */
398static void
21739618 399checkdirs(struct vnode *olddp, struct namecache *ncp)
984263bc
MD
400{
401 struct filedesc *fdp;
402 struct vnode *newdp;
690a3127 403 struct mount *mp;
984263bc
MD
404 struct proc *p;
405
406 if (olddp->v_usecount == 1)
407 return;
690a3127
MD
408 mp = olddp->v_mountedhere;
409 if (VFS_ROOT(mp, &newdp))
984263bc 410 panic("mount: lost mount");
21739618
MD
411 cache_setvp(ncp, newdp);
412
413 if (rootvnode == olddp) {
414 vref(newdp);
415 vfs_cache_setroot(newdp, cache_hold(ncp));
416 }
417
f62004ad 418 FOREACH_PROC_IN_SYSTEM(p) {
984263bc
MD
419 fdp = p->p_fd;
420 if (fdp->fd_cdir == olddp) {
421 vrele(fdp->fd_cdir);
597aea93 422 vref(newdp);
984263bc 423 fdp->fd_cdir = newdp;
690a3127 424 cache_drop(fdp->fd_ncdir);
21739618 425 fdp->fd_ncdir = cache_hold(ncp);
984263bc
MD
426 }
427 if (fdp->fd_rdir == olddp) {
428 vrele(fdp->fd_rdir);
597aea93 429 vref(newdp);
984263bc 430 fdp->fd_rdir = newdp;
690a3127 431 cache_drop(fdp->fd_nrdir);
21739618 432 fdp->fd_nrdir = cache_hold(ncp);
984263bc
MD
433 }
434 }
984263bc
MD
435 vput(newdp);
436}
437
438/*
439 * Unmount a file system.
440 *
441 * Note: unmount takes a path to the vnode mounted on as argument,
442 * not special file (as before).
443 */
41c20dac
MD
444/*
445 * umount_args(char *path, int flags)
446 */
984263bc
MD
447/* ARGSUSED */
448int
41c20dac 449unmount(struct unmount_args *uap)
984263bc 450{
dadab5e9
MD
451 struct thread *td = curthread;
452 struct proc *p = td->td_proc;
41c20dac 453 struct vnode *vp;
984263bc
MD
454 struct mount *mp;
455 int error;
fad57d0e 456 struct nlookupdata nd;
984263bc 457
dadab5e9 458 KKASSERT(p);
84c3a710
HP
459 if (p->p_ucred->cr_prison != NULL)
460 return (EPERM);
461 if (usermount == 0 && (error = suser(td)))
462 return (error);
463
fad57d0e
MD
464 vp = NULL;
465 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
466 if (error == 0)
467 error = nlookup(&nd);
468 if (error == 0)
469 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
470 nlookup_done(&nd);
471 if (error)
984263bc 472 return (error);
fad57d0e 473
984263bc
MD
474 mp = vp->v_mount;
475
476 /*
477 * Only root, or the user that did the original mount is
478 * permitted to unmount this filesystem.
479 */
480 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
dadab5e9 481 (error = suser(td))) {
984263bc
MD
482 vput(vp);
483 return (error);
484 }
485
486 /*
487 * Don't allow unmounting the root file system.
488 */
489 if (mp->mnt_flag & MNT_ROOTFS) {
490 vput(vp);
491 return (EINVAL);
492 }
493
494 /*
495 * Must be the root of the filesystem
496 */
497 if ((vp->v_flag & VROOT) == 0) {
498 vput(vp);
499 return (EINVAL);
500 }
501 vput(vp);
dadab5e9 502 return (dounmount(mp, SCARG(uap, flags), td));
984263bc
MD
503}
504
505/*
506 * Do the actual file system unmount.
507 */
508int
dadab5e9 509dounmount(struct mount *mp, int flags, struct thread *td)
984263bc
MD
510{
511 struct vnode *coveredvp;
512 int error;
513 int async_flag;
41a01a4d 514 lwkt_tokref ilock;
984263bc 515
41a01a4d 516 lwkt_gettoken(&ilock, &mountlist_token);
984263bc 517 if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
41a01a4d 518 lwkt_reltoken(&ilock);
984263bc
MD
519 return (EBUSY);
520 }
521 mp->mnt_kern_flag |= MNTK_UNMOUNT;
522 /* Allow filesystems to detect that a forced unmount is in progress. */
523 if (flags & MNT_FORCE)
524 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
525 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK |
41a01a4d 526 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &ilock, td);
984263bc
MD
527 if (error) {
528 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
529 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 530 wakeup(mp);
984263bc
MD
531 return (error);
532 }
533
534 if (mp->mnt_flag & MNT_EXPUBLIC)
535 vfs_setpublicfs(NULL, NULL, NULL);
536
537 vfs_msync(mp, MNT_WAIT);
538 async_flag = mp->mnt_flag & MNT_ASYNC;
539 mp->mnt_flag &=~ MNT_ASYNC;
540 cache_purgevfs(mp); /* remove cache entries for this file sys */
541 if (mp->mnt_syncer != NULL)
542 vrele(mp->mnt_syncer);
543 if (((mp->mnt_flag & MNT_RDONLY) ||
3b568787 544 (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) ||
984263bc 545 (flags & MNT_FORCE))
dadab5e9 546 error = VFS_UNMOUNT(mp, flags, td);
41a01a4d 547 lwkt_gettokref(&ilock);
984263bc 548 if (error) {
41a01a4d
MD
549 if (mp->mnt_syncer == NULL)
550 vfs_allocate_syncvnode(mp);
984263bc
MD
551 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
552 mp->mnt_flag |= async_flag;
553 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
41a01a4d 554 &ilock, td);
984263bc 555 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2 556 wakeup(mp);
984263bc
MD
557 return (error);
558 }
559 TAILQ_REMOVE(&mountlist, mp, mnt_list);
0961aa92
MD
560
561 /*
562 * Remove any installed vnode ops here so the individual VFSs don't
563 * have to.
564 */
565 vfs_rm_vnodeops(&mp->mnt_vn_ops);
566 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
567 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
568
984263bc 569 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
7ea21ed1 570 coveredvp->v_mountedhere = NULL;
984263bc 571 vrele(coveredvp);
21739618
MD
572 cache_drop(mp->mnt_ncp);
573 mp->mnt_ncp = NULL;
984263bc
MD
574 }
575 mp->mnt_vfc->vfc_refcount--;
576 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
577 panic("unmount: dangling vnode");
41a01a4d 578 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &ilock, td);
984263bc 579 if (mp->mnt_kern_flag & MNTK_MWAIT)
e38462a2
DR
580 wakeup(mp);
581 free(mp, M_MOUNT);
984263bc
MD
582 return (0);
583}
584
585/*
586 * Sync each mounted filesystem.
587 */
984263bc
MD
588
589#ifdef DEBUG
590static int syncprt = 0;
591SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
6bdbb368 592#endif /* DEBUG */
984263bc
MD
593
594/* ARGSUSED */
595int
41c20dac 596sync(struct sync_args *uap)
984263bc 597{
dadab5e9 598 struct thread *td = curthread;
41c20dac 599 struct mount *mp, *nmp;
41a01a4d 600 lwkt_tokref ilock;
984263bc
MD
601 int asyncflag;
602
41a01a4d 603 lwkt_gettoken(&ilock, &mountlist_token);
984263bc 604 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
41a01a4d 605 if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) {
984263bc
MD
606 nmp = TAILQ_NEXT(mp, mnt_list);
607 continue;
608 }
609 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
610 asyncflag = mp->mnt_flag & MNT_ASYNC;
611 mp->mnt_flag &= ~MNT_ASYNC;
612 vfs_msync(mp, MNT_NOWAIT);
3b568787 613 VFS_SYNC(mp, MNT_NOWAIT, td);
984263bc
MD
614 mp->mnt_flag |= asyncflag;
615 }
41a01a4d 616 lwkt_gettokref(&ilock);
984263bc 617 nmp = TAILQ_NEXT(mp, mnt_list);
dadab5e9 618 vfs_unbusy(mp, td);
984263bc 619 }
41a01a4d 620 lwkt_reltoken(&ilock);
984263bc 621/*
6bdbb368 622 * print out buffer pool stat information on each sync() call.
984263bc 623 */
6bdbb368 624#ifdef DEBUG
984263bc
MD
625 if (syncprt)
626 vfs_bufstats();
6bdbb368 627#endif /* DEBUG */
984263bc
MD
628 return (0);
629}
630
631/* XXX PRISON: could be per prison flag */
632static int prison_quotas;
633#if 0
634SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
635#endif
636
637/*
41c20dac
MD
638 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
639 *
984263bc
MD
640 * Change filesystem quotas.
641 */
984263bc
MD
642/* ARGSUSED */
643int
41c20dac 644quotactl(struct quotactl_args *uap)
984263bc 645{
fad57d0e
MD
646 struct nlookupdata nd;
647 struct thread *td;
648 struct proc *p;
41c20dac 649 struct mount *mp;
984263bc 650 int error;
984263bc 651
fad57d0e
MD
652 td = curthread;
653 p = td->td_proc;
41c20dac 654 if (p->p_ucred->cr_prison && !prison_quotas)
984263bc 655 return (EPERM);
fad57d0e
MD
656
657 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
658 if (error == 0)
659 error = nlookup(&nd);
660 if (error == 0) {
661 mp = nd.nl_ncp->nc_mount;
662 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
663 SCARG(uap, arg), nd.nl_td);
664 }
665 nlookup_done(&nd);
666 return (error);
984263bc
MD
667}
668
984263bc 669int
fad57d0e 670kern_statfs(struct nlookupdata *nd, struct statfs *buf)
984263bc 671{
dadab5e9 672 struct thread *td = curthread;
41c20dac
MD
673 struct mount *mp;
674 struct statfs *sp;
984263bc 675 int error;
984263bc 676
fad57d0e 677 if ((error = nlookup(nd)) != 0)
984263bc 678 return (error);
fad57d0e 679 mp = nd->nl_ncp->nc_mount;
984263bc 680 sp = &mp->mnt_stat;
fad57d0e 681 if ((error = VFS_STATFS(mp, sp, td)) != 0)
984263bc
MD
682 return (error);
683 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509
DRJ
684 bcopy(sp, buf, sizeof(*buf));
685 /* Only root should have access to the fsid's. */
686 if (suser(td))
687 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
688 return (0);
984263bc
MD
689}
690
691/*
9697c509 692 * statfs_args(char *path, struct statfs *buf)
41c20dac 693 *
984263bc
MD
694 * Get filesystem statistics.
695 */
984263bc 696int
9697c509
DRJ
697statfs(struct statfs_args *uap)
698{
fad57d0e 699 struct nlookupdata nd;
9697c509
DRJ
700 struct statfs buf;
701 int error;
702
fad57d0e
MD
703 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
704 if (error == 0)
705 error = kern_statfs(&nd, &buf);
706 nlookup_done(&nd);
9697c509
DRJ
707 if (error == 0)
708 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
709 return (error);
710}
711
712int
713kern_fstatfs(int fd, struct statfs *buf)
984263bc 714{
dadab5e9
MD
715 struct thread *td = curthread;
716 struct proc *p = td->td_proc;
984263bc
MD
717 struct file *fp;
718 struct mount *mp;
1fd87d54 719 struct statfs *sp;
984263bc 720 int error;
984263bc 721
dadab5e9 722 KKASSERT(p);
9697c509
DRJ
723 error = getvnode(p->p_fd, fd, &fp);
724 if (error)
984263bc
MD
725 return (error);
726 mp = ((struct vnode *)fp->f_data)->v_mount;
727 if (mp == NULL)
728 return (EBADF);
729 sp = &mp->mnt_stat;
dadab5e9 730 error = VFS_STATFS(mp, sp, td);
984263bc
MD
731 if (error)
732 return (error);
733 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
9697c509
DRJ
734 bcopy(sp, buf, sizeof(*buf));
735 /* Only root should have access to the fsid's. */
736 if (suser(td))
737 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
738 return (0);
739}
740
741/*
742 * fstatfs_args(int fd, struct statfs *buf)
743 *
744 * Get filesystem statistics.
745 */
746int
747fstatfs(struct fstatfs_args *uap)
748{
749 struct statfs buf;
750 int error;
751
752 error = kern_fstatfs(uap->fd, &buf);
753
754 if (error == 0)
755 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
756 return (error);
984263bc
MD
757}
758
759/*
41c20dac
MD
760 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
761 *
984263bc
MD
762 * Get statistics on all filesystems.
763 */
41c20dac 764/* ARGSUSED */
984263bc 765int
41c20dac 766getfsstat(struct getfsstat_args *uap)
984263bc 767{
dadab5e9 768 struct thread *td = curthread;
41c20dac
MD
769 struct mount *mp, *nmp;
770 struct statfs *sp;
984263bc 771 caddr_t sfsp;
41a01a4d 772 lwkt_tokref ilock;
984263bc
MD
773 long count, maxcount, error;
774
775 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
776 sfsp = (caddr_t)SCARG(uap, buf);
777 count = 0;
41a01a4d 778 lwkt_gettoken(&ilock, &mountlist_token);
984263bc 779 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
41a01a4d 780 if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) {
984263bc
MD
781 nmp = TAILQ_NEXT(mp, mnt_list);
782 continue;
783 }
784 if (sfsp && count < maxcount) {
785 sp = &mp->mnt_stat;
786 /*
787 * If MNT_NOWAIT or MNT_LAZY is specified, do not
788 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
789 * overrides MNT_WAIT.
790 */
791 if (((SCARG(uap, flags) & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
792 (SCARG(uap, flags) & MNT_WAIT)) &&
dadab5e9 793 (error = VFS_STATFS(mp, sp, td))) {
41a01a4d 794 lwkt_gettokref(&ilock);
984263bc 795 nmp = TAILQ_NEXT(mp, mnt_list);
dadab5e9 796 vfs_unbusy(mp, td);
984263bc
MD
797 continue;
798 }
799 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
e38462a2 800 error = copyout(sp, sfsp, sizeof(*sp));
984263bc 801 if (error) {
dadab5e9 802 vfs_unbusy(mp, td);
984263bc
MD
803 return (error);
804 }
805 sfsp += sizeof(*sp);
806 }
807 count++;
41a01a4d 808 lwkt_gettokref(&ilock);
984263bc 809 nmp = TAILQ_NEXT(mp, mnt_list);
dadab5e9 810 vfs_unbusy(mp, td);
984263bc 811 }
41a01a4d 812 lwkt_reltoken(&ilock);
984263bc 813 if (sfsp && count > maxcount)
c7114eea 814 uap->sysmsg_result = maxcount;
984263bc 815 else
c7114eea 816 uap->sysmsg_result = count;
984263bc
MD
817 return (0);
818}
819
820/*
41c20dac
MD
821 * fchdir_args(int fd)
822 *
984263bc
MD
823 * Change current working directory to a given file descriptor.
824 */
984263bc
MD
825/* ARGSUSED */
826int
41c20dac 827fchdir(struct fchdir_args *uap)
984263bc 828{
dadab5e9
MD
829 struct thread *td = curthread;
830 struct proc *p = td->td_proc;
41c20dac 831 struct filedesc *fdp = p->p_fd;
21739618 832 struct vnode *vp, *ovp;
984263bc
MD
833 struct mount *mp;
834 struct file *fp;
21739618
MD
835 struct namecache *ncp, *oncp;
836 struct namecache *nct;
984263bc
MD
837 int error;
838
839 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
840 return (error);
841 vp = (struct vnode *)fp->f_data;
597aea93 842 vref(vp);
5fd012e0 843 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
21739618 844 if (vp->v_type != VDIR || fp->f_ncp == NULL)
984263bc
MD
845 error = ENOTDIR;
846 else
dadab5e9 847 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td);
984263bc
MD
848 if (error) {
849 vput(vp);
850 return (error);
851 }
21739618
MD
852 ncp = cache_hold(fp->f_ncp);
853 while (!error && (mp = vp->v_mountedhere) != NULL) {
854 error = nlookup_mp(mp, &nct);
855 if (error == 0) {
856 cache_unlock(nct); /* leave ref intact */
857 vput(vp);
858 vp = nct->nc_vp;
5fd012e0 859 error = vget(vp, LK_SHARED, td);
21739618
MD
860 KKASSERT(error == 0);
861 cache_drop(ncp);
862 ncp = nct;
863 }
864 }
865 if (error == 0) {
866 ovp = fdp->fd_cdir;
867 oncp = fdp->fd_ncdir;
5fd012e0 868 VOP_UNLOCK(vp, 0, td); /* leave ref intact */
21739618
MD
869 fdp->fd_cdir = vp;
870 fdp->fd_ncdir = ncp;
871 cache_drop(oncp);
872 vrele(ovp);
873 } else {
874 cache_drop(ncp);
875 vput(vp);
876 }
877 return (error);
984263bc
MD
878}
879
9697c509 880int
21739618 881kern_chdir(struct nlookupdata *nd)
9697c509
DRJ
882{
883 struct thread *td = curthread;
884 struct proc *p = td->td_proc;
885 struct filedesc *fdp = p->p_fd;
21739618
MD
886 struct vnode *vp, *ovp;
887 struct namecache *oncp;
9697c509
DRJ
888 int error;
889
21739618 890 if ((error = nlookup(nd)) != 0)
9697c509 891 return (error);
21739618
MD
892 if ((vp = nd->nl_ncp->nc_vp) == NULL)
893 return (ENOENT);
5fd012e0 894 if ((error = vget(vp, LK_SHARED, td)) != 0)
21739618
MD
895 return (error);
896
897 error = checkvp_chdir(vp, td);
5fd012e0 898 VOP_UNLOCK(vp, 0, td);
21739618
MD
899 if (error == 0) {
900 ovp = fdp->fd_cdir;
901 oncp = fdp->fd_ncdir;
902 cache_unlock(nd->nl_ncp); /* leave reference intact */
903 fdp->fd_ncdir = nd->nl_ncp;
904 fdp->fd_cdir = vp;
905 cache_drop(oncp);
906 vrele(ovp);
907 nd->nl_ncp = NULL;
908 } else {
909 vrele(vp);
e24b948e 910 }
e24b948e 911 return (error);
9697c509
DRJ
912}
913
984263bc 914/*
41c20dac
MD
915 * chdir_args(char *path)
916 *
984263bc
MD
917 * Change current working directory (``.'').
918 */
984263bc 919int
41c20dac 920chdir(struct chdir_args *uap)
984263bc 921{
21739618 922 struct nlookupdata nd;
9697c509 923 int error;
984263bc 924
21739618 925 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e 926 if (error == 0)
21739618 927 error = kern_chdir(&nd);
fad57d0e 928 nlookup_done(&nd);
9697c509 929 return (error);
984263bc
MD
930}
931
932/*
933 * Helper function for raised chroot(2) security function: Refuse if
934 * any filedescriptors are open directories.
935 */
936static int
937chroot_refuse_vdir_fds(fdp)
938 struct filedesc *fdp;
939{
940 struct vnode *vp;
941 struct file *fp;
942 int error;
943 int fd;
944
945 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
946 error = getvnode(fdp, fd, &fp);
947 if (error)
948 continue;
949 vp = (struct vnode *)fp->f_data;
950 if (vp->v_type != VDIR)
951 continue;
952 return(EPERM);
953 }
954 return (0);
955}
956
957/*
958 * This sysctl determines if we will allow a process to chroot(2) if it
959 * has a directory open:
960 * 0: disallowed for all processes.
961 * 1: allowed for processes that were not already chroot(2)'ed.
962 * 2: allowed for all processes.
963 */
964
965static int chroot_allow_open_directories = 1;
966
967SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
968 &chroot_allow_open_directories, 0, "");
969
e24b948e 970/*
21739618
MD
971 * chroot to the specified namecache entry. We obtain the vp from the
972 * namecache data. The passed ncp must be locked and referenced and will
973 * remain locked and referenced on return.
e24b948e 974 */
21739618 975static int
524c845c 976kern_chroot(struct nlookupdata *nd)
e24b948e
MD
977{
978 struct thread *td = curthread;
979 struct proc *p = td->td_proc;
980 struct filedesc *fdp = p->p_fd;
524c845c 981 struct namecache *ncp;
21739618 982 struct vnode *vp;
e24b948e
MD
983 int error;
984
985 /*
986 * Only root can chroot
987 */
988 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
989 return (error);
990
524c845c
MD
991 if ((error = nlookup(nd)) != 0)
992 return (error);
993 ncp = nd->nl_ncp;
994
e24b948e
MD
995 /*
996 * Disallow open directory descriptors (fchdir() breakouts).
997 */
998 if (chroot_allow_open_directories == 0 ||
999 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1000 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1001 return (error);
1002 }
21739618
MD
1003 if ((vp = ncp->nc_vp) == NULL)
1004 return (ENOENT);
1005
5fd012e0 1006 if ((error = vget(vp, LK_SHARED, td)) != 0)
21739618 1007 return (error);
e24b948e
MD
1008
1009 /*
1010 * Check the validity of vp as a directory to change to and
1011 * associate it with rdir/jdir.
1012 */
21739618 1013 error = checkvp_chdir(vp, td);
5fd012e0 1014 VOP_UNLOCK(vp, 0, td); /* leave reference intact */
21739618 1015 if (error == 0) {
e24b948e 1016 vrele(fdp->fd_rdir);
21739618 1017 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */
690a3127 1018 cache_drop(fdp->fd_nrdir);
21739618 1019 fdp->fd_nrdir = cache_hold(ncp);
e24b948e
MD
1020 if (fdp->fd_jdir == NULL) {
1021 fdp->fd_jdir = vp;
597aea93 1022 vref(fdp->fd_jdir);
21739618 1023 fdp->fd_njdir = cache_hold(ncp);
e24b948e 1024 }
21739618
MD
1025 } else {
1026 vrele(vp);
e24b948e
MD
1027 }
1028 return (error);
1029}
1030
984263bc 1031/*
41c20dac
MD
1032 * chroot_args(char *path)
1033 *
984263bc
MD
1034 * Change notion of root (``/'') directory.
1035 */
984263bc
MD
1036/* ARGSUSED */
1037int
41c20dac 1038chroot(struct chroot_args *uap)
984263bc 1039{
dadab5e9 1040 struct thread *td = curthread;
21739618 1041 struct nlookupdata nd;
e24b948e 1042 int error;
984263bc 1043
e24b948e 1044 KKASSERT(td->td_proc);
21739618 1045 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
fad57d0e 1046 if (error == 0)
524c845c 1047 error = kern_chroot(&nd);
fad57d0e 1048 nlookup_done(&nd);
e24b948e 1049 return (error);
984263bc
MD
1050}
1051
1052/*
e24b948e
MD
1053 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1054 * determine whether it is legal to chdir to the vnode. The vnode's state
1055 * is not changed by this call.
984263bc 1056 */
e24b948e
MD
1057int
1058checkvp_chdir(struct vnode *vp, struct thread *td)
984263bc 1059{
984263bc
MD
1060 int error;
1061
984263bc
MD
1062 if (vp->v_type != VDIR)
1063 error = ENOTDIR;
1064 else
e24b948e 1065 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred, td);
984263bc
MD
1066 return (error);
1067}
1068
984263bc 1069int
fad57d0e 1070kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
984263bc 1071{
dadab5e9
MD
1072 struct thread *td = curthread;
1073 struct proc *p = td->td_proc;
41c20dac 1074 struct filedesc *fdp = p->p_fd;
9697c509 1075 int cmode, flags;
984263bc 1076 struct file *nfp;
fad57d0e
MD
1077 struct file *fp;
1078 struct vnode *vp;
984263bc
MD
1079 int type, indx, error;
1080 struct flock lf;
984263bc 1081
984263bc
MD
1082 if ((oflags & O_ACCMODE) == O_ACCMODE)
1083 return (EINVAL);
1084 flags = FFLAGS(oflags);
fad57d0e 1085 error = falloc(p, &nfp, NULL);
984263bc
MD
1086 if (error)
1087 return (error);
1088 fp = nfp;
9697c509 1089 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
fad57d0e 1090
984263bc 1091 /*
fad57d0e
MD
1092 * XXX p_dupfd is a real mess. It allows a device to return a
1093 * file descriptor to be duplicated rather then doing the open
1094 * itself.
984263bc 1095 */
fad57d0e 1096 p->p_dupfd = -1;
21739618
MD
1097
1098 /*
fad57d0e
MD
1099 * Call vn_open() to do the lookup and assign the vnode to the
1100 * file pointer. vn_open() does not change the ref count on fp
1101 * and the vnode, on success, will be inherited by the file pointer
1102 * and unlocked.
21739618 1103 */
fad57d0e
MD
1104 nd->nl_flags |= NLC_LOCKVP;
1105 error = vn_open(nd, fp, flags, cmode);
1106 nlookup_done(nd);
984263bc 1107 if (error) {
984263bc
MD
1108 /*
1109 * handle special fdopen() case. bleh. dupfdopen() is
1110 * responsible for dropping the old contents of ofiles[indx]
1111 * if it succeeds.
fad57d0e
MD
1112 *
1113 * Note that if fsetfd() succeeds it will add a ref to fp
1114 * which represents the fd_ofiles[] assignment. We must still
1115 * drop our reference.
984263bc 1116 */
fad57d0e
MD
1117 if ((error == ENODEV || error == ENXIO) && p->p_dupfd >= 0) {
1118 if (fsetfd(p, fp, &indx) == 0) {
1119 error = dupfdopen(fdp, indx, p->p_dupfd, flags, error);
1120 if (error == 0) {
1121 *res = indx;
1122 fdrop(fp, td); /* our ref */
1123 return (0);
1124 }
1125 if (fdp->fd_ofiles[indx] == fp) {
1126 fdp->fd_ofiles[indx] = NULL;
1127 fdrop(fp, td); /* fd_ofiles[] ref */
1128 }
1129 }
984263bc 1130 }
fad57d0e 1131 fdrop(fp, td); /* our ref */
984263bc
MD
1132 if (error == ERESTART)
1133 error = EINTR;
1134 return (error);
1135 }
fad57d0e
MD
1136
1137 /*
1138 * ref the vnode for ourselves so it can't be ripped out from under
1139 * is. XXX need an ND flag to request that the vnode be returned
1140 * anyway.
1141 */
1142 vp = (struct vnode *)fp->f_data;
1143 vref(vp);
1144 if ((error = fsetfd(p, fp, &indx)) != 0) {
1145 fdrop(fp, td);
1146 vrele(vp);
1147 return (error);
1148 }
1149
1150 /*
1151 * If no error occurs the vp will have been assigned to the file
1152 * pointer.
1153 */
984263bc 1154 p->p_dupfd = 0;
984263bc
MD
1155
1156 /*
1157 * There should be 2 references on the file, one from the descriptor
1158 * table, and one for us.
1159 *
1160 * Handle the case where someone closed the file (via its file
1161 * descriptor) while we were blocked. The end result should look
1162 * like opening the file succeeded but it was immediately closed.
1163 */
1164 if (fp->f_count == 1) {
1165 KASSERT(fdp->fd_ofiles[indx] != fp,
1166 ("Open file descriptor lost all refs"));
fad57d0e
MD
1167 vrele(vp);
1168 fo_close(fp, td);
dadab5e9 1169 fdrop(fp, td);
9697c509 1170 *res = indx;
984263bc
MD
1171 return 0;
1172 }
1173
984263bc
MD
1174 if (flags & (O_EXLOCK | O_SHLOCK)) {
1175 lf.l_whence = SEEK_SET;
1176 lf.l_start = 0;
1177 lf.l_len = 0;
1178 if (flags & O_EXLOCK)
1179 lf.l_type = F_WRLCK;
1180 else
1181 lf.l_type = F_RDLCK;
1182 type = F_FLOCK;
1183 if ((flags & FNONBLOCK) == 0)
1184 type |= F_WAIT;
fad57d0e 1185
984263bc
MD
1186 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1187 /*
1188 * lock request failed. Normally close the descriptor
1189 * but handle the case where someone might have dup()d
1190 * it when we weren't looking. One reference is
1191 * owned by the descriptor array, the other by us.
1192 */
fad57d0e 1193 vrele(vp);
984263bc
MD
1194 if (fdp->fd_ofiles[indx] == fp) {
1195 fdp->fd_ofiles[indx] = NULL;
dadab5e9 1196 fdrop(fp, td);
984263bc 1197 }
dadab5e9 1198 fdrop(fp, td);
984263bc
MD
1199 return (error);
1200 }
984263bc
MD
1201 fp->f_flag |= FHASLOCK;
1202 }
1203 /* assert that vn_open created a backing object if one is needed */
1204 KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0,
1205 ("open: vmio vnode has no backing object after vn_open"));
984263bc 1206
fad57d0e 1207 vrele(vp);
21739618 1208
984263bc
MD
1209 /*
1210 * release our private reference, leaving the one associated with the
1211 * descriptor table intact.
1212 */
dadab5e9 1213 fdrop(fp, td);
9697c509 1214 *res = indx;
984263bc
MD
1215 return (0);
1216}
1217
984263bc 1218/*
9697c509 1219 * open_args(char *path, int flags, int mode)
41c20dac 1220 *
9697c509
DRJ
1221 * Check permissions, allocate an open file structure,
1222 * and call the device open routine if any.
984263bc 1223 */
984263bc 1224int
9697c509 1225open(struct open_args *uap)
984263bc 1226{
fad57d0e 1227 struct nlookupdata nd;
9697c509
DRJ
1228 int error;
1229
fad57d0e
MD
1230 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1231 if (error == 0) {
1232 error = kern_open(&nd, uap->flags,
1233 uap->mode, &uap->sysmsg_result);
1234 }
1235 nlookup_done(&nd);
9697c509 1236 return (error);
984263bc 1237}
984263bc 1238
984263bc 1239int
fad57d0e 1240kern_mknod(struct nlookupdata *nd, int mode, int dev)
984263bc 1241{
fad57d0e 1242 struct namecache *ncp;
dadab5e9
MD
1243 struct thread *td = curthread;
1244 struct proc *p = td->td_proc;
41c20dac 1245 struct vnode *vp;
984263bc
MD
1246 struct vattr vattr;
1247 int error;
1248 int whiteout = 0;
984263bc 1249
dadab5e9
MD
1250 KKASSERT(p);
1251
9697c509 1252 switch (mode & S_IFMT) {
984263bc
MD
1253 case S_IFCHR:
1254 case S_IFBLK:
dadab5e9 1255 error = suser(td);
984263bc
MD
1256 break;
1257 default:
dadab5e9 1258 error = suser_cred(p->p_ucred, PRISON_ROOT);
984263bc
MD
1259 break;
1260 }
1261 if (error)
1262 return (error);
fad57d0e 1263
984263bc 1264 bwillwrite();
fad57d0e
MD
1265 nd->nl_flags |= NLC_CREATE;
1266 if ((error = nlookup(nd)) != 0)
984263bc 1267 return (error);
fad57d0e
MD
1268 ncp = nd->nl_ncp;
1269 if (ncp->nc_vp)
1270 return (EEXIST);
984263bc 1271
fad57d0e
MD
1272 VATTR_NULL(&vattr);
1273 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1274 vattr.va_rdev = dev;
1275 whiteout = 0;
1276
1277 switch (mode & S_IFMT) {
1278 case S_IFMT: /* used by badsect to flag bad sectors */
1279 vattr.va_type = VBAD;
1280 break;
1281 case S_IFCHR:
1282 vattr.va_type = VCHR;
1283 break;
1284 case S_IFBLK:
1285 vattr.va_type = VBLK;
1286 break;
1287 case S_IFWHT:
1288 whiteout = 1;
1289 break;
1290 default:
1291 error = EINVAL;
1292 break;
984263bc 1293 }
9697c509 1294 if (error == 0) {
fad57d0e
MD
1295 if (whiteout) {
1296 error = VOP_NWHITEOUT(ncp, nd->nl_cred, NAMEI_CREATE);
1297 } else {
1298 vp = NULL;
1299 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
984263bc 1300 if (error == 0)
fad57d0e 1301 vput(vp);
984263bc 1302 }
984263bc 1303 }
9697c509
DRJ
1304 return (error);
1305}
1306
1307/*
1308 * mknod_args(char *path, int mode, int dev)
1309 *
1310 * Create a special file.
1311 */
1312int
1313mknod(struct mknod_args *uap)
1314{
fad57d0e 1315 struct nlookupdata nd;
9697c509
DRJ
1316 int error;
1317
fad57d0e
MD
1318 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1319 if (error == 0)
1320 error = kern_mknod(&nd, uap->mode, uap->dev);
1321 nlookup_done(&nd);
984263bc
MD
1322 return (error);
1323}
1324
984263bc 1325int
fad57d0e 1326kern_mkfifo(struct nlookupdata *nd, int mode)
984263bc 1327{
fad57d0e 1328 struct namecache *ncp;
dadab5e9
MD
1329 struct thread *td = curthread;
1330 struct proc *p = td->td_proc;
984263bc 1331 struct vattr vattr;
fad57d0e 1332 struct vnode *vp;
984263bc 1333 int error;
984263bc
MD
1334
1335 bwillwrite();
fad57d0e
MD
1336
1337 nd->nl_flags |= NLC_CREATE;
1338 if ((error = nlookup(nd)) != 0)
984263bc 1339 return (error);
fad57d0e
MD
1340 ncp = nd->nl_ncp;
1341 if (ncp->nc_vp)
984263bc 1342 return (EEXIST);
fad57d0e 1343
984263bc
MD
1344 VATTR_NULL(&vattr);
1345 vattr.va_type = VFIFO;
136178b3 1346 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
fad57d0e
MD
1347 vp = NULL;
1348 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
984263bc 1349 if (error == 0)
fad57d0e 1350 vput(vp);
136178b3
DRJ
1351 return (error);
1352}
1353
1354/*
1355 * mkfifo_args(char *path, int mode)
1356 *
1357 * Create a named pipe.
1358 */
1359int
1360mkfifo(struct mkfifo_args *uap)
1361{
fad57d0e 1362 struct nlookupdata nd;
136178b3
DRJ
1363 int error;
1364
fad57d0e
MD
1365 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1366 if (error == 0)
1367 error = kern_mkfifo(&nd, uap->mode);
1368 nlookup_done(&nd);
984263bc
MD
1369 return (error);
1370}
1371
984263bc 1372int
fad57d0e 1373kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
984263bc 1374{
dadab5e9 1375 struct thread *td = curthread;
41c20dac 1376 struct vnode *vp;
984263bc
MD
1377 int error;
1378
fad57d0e
MD
1379 /*
1380 * Lookup the source and obtained a locked vnode.
1381 *
1382 * XXX relookup on vget failure / race ?
1383 */
984263bc 1384 bwillwrite();
fad57d0e 1385 if ((error = nlookup(nd)) != 0)
984263bc 1386 return (error);
fad57d0e
MD
1387 vp = nd->nl_ncp->nc_vp;
1388 KKASSERT(vp != NULL);
984263bc 1389 if (vp->v_type == VDIR)
fad57d0e
MD
1390 return (EPERM); /* POSIX */
1391 if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0)
1392 return (error);
1393
1394 /*
1395 * Unlock the source so we can lookup the target without deadlocking
1396 * (XXX vp is locked already, possible other deadlock?). The target
1397 * must not exist.
1398 */
1399 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
1400 nd->nl_flags &= ~NLC_NCPISLOCKED;
1401 cache_unlock(nd->nl_ncp);
1402
1403 linknd->nl_flags |= NLC_CREATE;
1404 if ((error = nlookup(linknd)) != 0) {
1405 vput(vp);
1406 return (error);
984263bc 1407 }
fad57d0e
MD
1408 if (linknd->nl_ncp->nc_vp) {
1409 vput(vp);
1410 return (EEXIST);
1411 }
1412
1413 /*
1414 * Finally run the new API VOP.
1415 */
1416 error = VOP_NLINK(linknd->nl_ncp, vp, linknd->nl_cred);
1417 vput(vp);
984263bc
MD
1418 return (error);
1419}
1420
1421/*
9697c509 1422 * link_args(char *path, char *link)
41c20dac 1423 *
9697c509 1424 * Make a hard file link.
984263bc 1425 */
984263bc 1426int
9697c509
DRJ
1427link(struct link_args *uap)
1428{
fad57d0e 1429 struct nlookupdata nd, linknd;
9697c509
DRJ
1430 int error;
1431
fad57d0e
MD
1432 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1433 if (error == 0) {
1434 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
1435 if (error == 0)
1436 error = kern_link(&nd, &linknd);
1437 nlookup_done(&linknd);
1438 }
1439 nlookup_done(&nd);
9697c509
DRJ
1440 return (error);
1441}
1442
1443int
fad57d0e 1444kern_symlink(struct nlookupdata *nd, char *path, int mode)
984263bc 1445{
fad57d0e 1446 struct namecache *ncp;
984263bc 1447 struct vattr vattr;
fad57d0e 1448 struct vnode *vp;
984263bc 1449 int error;
984263bc 1450
984263bc 1451 bwillwrite();
fad57d0e
MD
1452 nd->nl_flags |= NLC_CREATE;
1453 if ((error = nlookup(nd)) != 0)
9697c509 1454 return (error);
fad57d0e
MD
1455 ncp = nd->nl_ncp;
1456 if (ncp->nc_vp)
9697c509 1457 return (EEXIST);
fad57d0e 1458
984263bc 1459 VATTR_NULL(&vattr);
fad57d0e
MD
1460 vattr.va_mode = mode;
1461 error = VOP_NSYMLINK(ncp, &vp, nd->nl_cred, &vattr, path);
984263bc 1462 if (error == 0)
fad57d0e 1463 vput(vp);
9697c509
DRJ
1464 return (error);
1465}
1466
1467/*
1468 * symlink(char *path, char *link)
1469 *
1470 * Make a symbolic link.
1471 */
1472int
1473symlink(struct symlink_args *uap)
1474{
1475 struct thread *td = curthread;
fad57d0e 1476 struct nlookupdata nd;
9697c509
DRJ
1477 char *path;
1478 int error;
fad57d0e 1479 int mode;
9697c509
DRJ
1480
1481 path = zalloc(namei_zone);
1482 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
e24b948e 1483 if (error == 0) {
fad57d0e
MD
1484 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
1485 if (error == 0) {
1486 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
1487 error = kern_symlink(&nd, path, mode);
1488 }
1489 nlookup_done(&nd);
e24b948e 1490 }
984263bc
MD
1491 zfree(namei_zone, path);
1492 return (error);
1493}
1494
1495/*
41c20dac
MD
1496 * undelete_args(char *path)
1497 *
984263bc
MD
1498 * Delete a whiteout from the filesystem.
1499 */
1500/* ARGSUSED */
1501int
41c20dac 1502undelete(struct undelete_args *uap)
984263bc 1503{
fad57d0e 1504 struct nlookupdata nd;
984263bc 1505 int error;
984263bc 1506
fad57d0e 1507 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, 0);
984263bc 1508 bwillwrite();
fad57d0e
MD
1509 nd.nl_flags |= NLC_DELETE;
1510 if (error == 0)
1511 error = nlookup(&nd);
b8997912 1512 if (error == 0)
fad57d0e
MD
1513 error = VOP_NWHITEOUT(nd.nl_ncp, nd.nl_cred, NAMEI_DELETE);
1514 nlookup_done(&nd);
984263bc
MD
1515 return (error);
1516}
1517
984263bc 1518int
fad57d0e 1519kern_unlink(struct nlookupdata *nd)
984263bc 1520{
fad57d0e 1521 struct namecache *ncp;
984263bc 1522 int error;
984263bc
MD
1523
1524 bwillwrite();
fad57d0e
MD
1525 nd->nl_flags |= NLC_DELETE;
1526 if ((error = nlookup(nd)) != 0)
984263bc 1527 return (error);
fad57d0e
MD
1528 ncp = nd->nl_ncp;
1529 error = VOP_NREMOVE(ncp, nd->nl_cred);
984263bc
MD
1530 return (error);
1531}
1532
1533/*
9697c509 1534 * unlink_args(char *path)
41c20dac 1535 *
9697c509 1536 * Delete a name from the filesystem.
984263bc 1537 */
984263bc 1538int
9697c509
DRJ
1539unlink(struct unlink_args *uap)
1540{
fad57d0e 1541 struct nlookupdata nd;
9697c509
DRJ
1542 int error;
1543
fad57d0e
MD
1544 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1545 if (error == 0)
1546 error = kern_unlink(&nd);
1547 nlookup_done(&nd);
9697c509
DRJ
1548 return (error);
1549}
1550
1551int
5a3fe67d 1552kern_lseek(int fd, off_t offset, int whence, off_t *res)
984263bc 1553{
dadab5e9
MD
1554 struct thread *td = curthread;
1555 struct proc *p = td->td_proc;
41c20dac
MD
1556 struct filedesc *fdp = p->p_fd;
1557 struct file *fp;
984263bc
MD
1558 struct vattr vattr;
1559 int error;
1560
9697c509
DRJ
1561 if (fd >= fdp->fd_nfiles ||
1562 (fp = fdp->fd_ofiles[fd]) == NULL)
984263bc
MD
1563 return (EBADF);
1564 if (fp->f_type != DTYPE_VNODE)
1565 return (ESPIPE);
9697c509 1566 switch (whence) {
984263bc 1567 case L_INCR:
9697c509 1568 fp->f_offset += offset;
984263bc
MD
1569 break;
1570 case L_XTND:
3b568787 1571 error=VOP_GETATTR((struct vnode *)fp->f_data, &vattr, td);
984263bc
MD
1572 if (error)
1573 return (error);
9697c509 1574 fp->f_offset = offset + vattr.va_size;
984263bc
MD
1575 break;
1576 case L_SET:
9697c509 1577 fp->f_offset = offset;
984263bc
MD
1578 break;
1579 default:
1580 return (EINVAL);
1581 }
9697c509 1582 *res = fp->f_offset;
984263bc
MD
1583 return (0);
1584}
1585
984263bc 1586/*
9697c509 1587 * lseek_args(int fd, int pad, off_t offset, int whence)
41c20dac 1588 *
9697c509 1589 * Reposition read/write file offset.
984263bc 1590 */
984263bc 1591int
9697c509 1592lseek(struct lseek_args *uap)
984263bc 1593{
984263bc
MD
1594 int error;
1595
9697c509 1596 error = kern_lseek(uap->fd, uap->offset, uap->whence,
5a3fe67d 1597 &uap->sysmsg_offset);
9697c509 1598
984263bc
MD
1599 return (error);
1600}
984263bc 1601
984263bc 1602int
fad57d0e 1603kern_access(struct nlookupdata *nd, int aflags)
984263bc 1604{
dadab5e9 1605 struct thread *td = curthread;
41c20dac 1606 struct vnode *vp;
984263bc 1607 int error, flags;
984263bc 1608
fad57d0e
MD
1609 if ((error = nlookup(nd)) != 0)
1610 return (error);
806dcf9a 1611retry:
fad57d0e 1612 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 1613 if (error)
fad57d0e 1614 return (error);
984263bc
MD
1615
1616 /* Flags == 0 means only check for existence. */
9697c509 1617 if (aflags) {
984263bc 1618 flags = 0;
9697c509 1619 if (aflags & R_OK)
984263bc 1620 flags |= VREAD;
9697c509 1621 if (aflags & W_OK)
984263bc 1622 flags |= VWRITE;
9697c509 1623 if (aflags & X_OK)
984263bc
MD
1624 flags |= VEXEC;
1625 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
fad57d0e 1626 error = VOP_ACCESS(vp, flags, nd->nl_cred, td);
806dcf9a
MD
1627
1628 /*
1629 * If the file handle is stale we have to re-resolve the
1630 * entry. This is a hack at the moment.
1631 */
1632 if (error == ESTALE) {
1633 cache_setunresolved(nd->nl_ncp);
1634 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1635 if (error == 0) {
1636 vput(vp);
1637 vp = NULL;
1638 goto retry;
1639 }
1640 }
984263bc 1641 }
984263bc 1642 vput(vp);
984263bc
MD
1643 return (error);
1644}
1645
984263bc 1646/*
9697c509 1647 * access_args(char *path, int flags)
41c20dac 1648 *
9697c509 1649 * Check access permissions.
984263bc 1650 */
984263bc 1651int
9697c509 1652access(struct access_args *uap)
984263bc 1653{
fad57d0e 1654 struct nlookupdata nd;
9697c509 1655 int error;
984263bc 1656
fad57d0e
MD
1657 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1658 if (error == 0)
1659 error = kern_access(&nd, uap->flags);
1660 nlookup_done(&nd);
984263bc
MD
1661 return (error);
1662}
1663
984263bc 1664int
21739618 1665kern_stat(struct nlookupdata *nd, struct stat *st)
984263bc 1666{
984263bc 1667 int error;
21739618
MD
1668 struct vnode *vp;
1669 thread_t td;
984263bc 1670
21739618 1671 if ((error = nlookup(nd)) != 0)
984263bc 1672 return (error);
806dcf9a 1673again:
21739618
MD
1674 if ((vp = nd->nl_ncp->nc_vp) == NULL)
1675 return (ENOENT);
1676
1677 td = curthread;
5fd012e0 1678 if ((error = vget(vp, LK_SHARED, td)) != 0)
21739618
MD
1679 return (error);
1680 error = vn_stat(vp, st, td);
806dcf9a
MD
1681
1682 /*
1683 * If the file handle is stale we have to re-resolve the entry. This
1684 * is a hack at the moment.
1685 */
1686 if (error == ESTALE) {
1687 cache_setunresolved(nd->nl_ncp);
1688 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1689 if (error == 0) {
1690 vput(vp);
1691 goto again;
1692 }
1693 }
21739618 1694 vput(vp);
984263bc
MD
1695 return (error);
1696}
1697
984263bc 1698/*
41c20dac
MD
1699 * stat_args(char *path, struct stat *ub)
1700 *
984263bc
MD
1701 * Get file status; this version follows links.
1702 */
984263bc 1703int
41c20dac 1704stat(struct stat_args *uap)
984263bc 1705{
21739618 1706 struct nlookupdata nd;
9697c509
DRJ
1707 struct stat st;
1708 int error;
984263bc 1709
21739618
MD
1710 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1711 if (error == 0) {
1712 error = kern_stat(&nd, &st);
1713 if (error == 0)
1714 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 1715 }
fad57d0e 1716 nlookup_done(&nd);
984263bc
MD
1717 return (error);
1718}
1719
1720/*
41c20dac
MD
1721 * lstat_args(char *path, struct stat *ub)
1722 *
984263bc
MD
1723 * Get file status; this version does not follow links.
1724 */
984263bc 1725int
41c20dac 1726lstat(struct lstat_args *uap)
984263bc 1727{
21739618 1728 struct nlookupdata nd;
9697c509
DRJ
1729 struct stat st;
1730 int error;
984263bc 1731
21739618
MD
1732 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1733 if (error == 0) {
1734 error = kern_stat(&nd, &st);
1735 if (error == 0)
1736 error = copyout(&st, uap->ub, sizeof(*uap->ub));
21739618 1737 }
fad57d0e 1738 nlookup_done(&nd);
984263bc
MD
1739 return (error);
1740}
1741
1742void
1743cvtnstat(sb, nsb)
1744 struct stat *sb;
1745 struct nstat *nsb;
1746{
1747 nsb->st_dev = sb->st_dev;
1748 nsb->st_ino = sb->st_ino;
1749 nsb->st_mode = sb->st_mode;
1750 nsb->st_nlink = sb->st_nlink;
1751 nsb->st_uid = sb->st_uid;
1752 nsb->st_gid = sb->st_gid;
1753 nsb->st_rdev = sb->st_rdev;
1754 nsb->st_atimespec = sb->st_atimespec;
1755 nsb->st_mtimespec = sb->st_mtimespec;
1756 nsb->st_ctimespec = sb->st_ctimespec;
1757 nsb->st_size = sb->st_size;
1758 nsb->st_blocks = sb->st_blocks;
1759 nsb->st_blksize = sb->st_blksize;
1760 nsb->st_flags = sb->st_flags;
1761 nsb->st_gen = sb->st_gen;
1762 nsb->st_qspare[0] = sb->st_qspare[0];
1763 nsb->st_qspare[1] = sb->st_qspare[1];
1764}
1765
41c20dac
MD
1766/*
1767 * nstat_args(char *path, struct nstat *ub)
1768 */
984263bc
MD
1769/* ARGSUSED */
1770int
41c20dac 1771nstat(struct nstat_args *uap)
984263bc 1772{
dadab5e9 1773 struct thread *td = curthread;
fad57d0e 1774 struct vnode *vp;
984263bc
MD
1775 struct stat sb;
1776 struct nstat nsb;
fad57d0e 1777 struct nlookupdata nd;
984263bc 1778 int error;
984263bc 1779
fad57d0e
MD
1780 vp = NULL;
1781 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
1782 if (error == 0)
1783 error = nlookup(&nd);
1784 if (error == 0)
1785 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
1786 nlookup_done(&nd);
1787 if (error == 0) {
1788 error = vn_stat(vp, &sb, td);
1789 vput(vp);
1790 if (error == 0) {
1791 cvtnstat(&sb, &nsb);
1792 error = copyout(&nsb, SCARG(uap, ub), sizeof(nsb));
1793 }
1794 }
1795 return (error);
984263bc
MD
1796}
1797
1798/*
41c20dac
MD
1799 * lstat_args(char *path, struct stat *ub)
1800 *
984263bc
MD
1801 * Get file status; this version does not follow links.
1802 */
984263bc
MD
1803/* ARGSUSED */
1804int
41c20dac 1805nlstat(struct nlstat_args *uap)
984263bc 1806{
dadab5e9 1807 struct thread *td = curthread;
984263bc
MD
1808 struct vnode *vp;
1809 struct stat sb;
1810 struct nstat nsb;
fad57d0e
MD
1811 struct nlookupdata nd;
1812 int error;
984263bc 1813
fad57d0e
MD
1814 vp = NULL;
1815 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, 0);
1816 if (error == 0)
1817 error = nlookup(&nd);
1818 if (error == 0)
1819 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
1820 nlookup_done(&nd);
1821 if (error == 0) {
1822 error = vn_stat(vp, &sb, td);
1823 vput(vp);
1824 if (error == 0) {
1825 cvtnstat(&sb, &nsb);
1826 error = copyout(&nsb, SCARG(uap, ub), sizeof(nsb));
1827 }
1828 }
984263bc
MD
1829 return (error);
1830}
1831
1832/*
41c20dac
MD
1833 * pathconf_Args(char *path, int name)
1834 *
984263bc
MD
1835 * Get configurable pathname variables.
1836 */
984263bc
MD
1837/* ARGSUSED */
1838int
41c20dac 1839pathconf(struct pathconf_args *uap)
984263bc 1840{
fad57d0e
MD
1841 struct nlookupdata nd;
1842 struct vnode *vp;
984263bc 1843 int error;
984263bc 1844
fad57d0e
MD
1845 vp = NULL;
1846 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
1847 if (error == 0)
1848 error = nlookup(&nd);
1849 if (error == 0)
1850 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
1851 nlookup_done(&nd);
1852 if (error == 0) {
1853 error = VOP_PATHCONF(vp, SCARG(uap, name), uap->sysmsg_fds);
1854 vput(vp);
1855 }
984263bc
MD
1856 return (error);
1857}
1858
bbab27f6
DRJ
1859/*
1860 * XXX: daver
1861 * kern_readlink isn't properly split yet. There is a copyin burried
1862 * in VOP_READLINK().
1863 */
984263bc 1864int
fad57d0e 1865kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
984263bc 1866{
dadab5e9
MD
1867 struct thread *td = curthread;
1868 struct proc *p = td->td_proc;
41c20dac 1869 struct vnode *vp;
984263bc
MD
1870 struct iovec aiov;
1871 struct uio auio;
1872 int error;
984263bc 1873
fad57d0e
MD
1874 if ((error = nlookup(nd)) != 0)
1875 return (error);
1876 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
9697c509 1877 if (error)
984263bc 1878 return (error);
fad57d0e 1879 if (vp->v_type != VLNK) {
984263bc 1880 error = EINVAL;
fad57d0e 1881 } else {
9697c509
DRJ
1882 aiov.iov_base = buf;
1883 aiov.iov_len = count;
984263bc
MD
1884 auio.uio_iov = &aiov;
1885 auio.uio_iovcnt = 1;
1886 auio.uio_offset = 0;
1887 auio.uio_rw = UIO_READ;
1888 auio.uio_segflg = UIO_USERSPACE;
dadab5e9 1889 auio.uio_td = td;
9697c509 1890 auio.uio_resid = count;
984263bc
MD
1891 error = VOP_READLINK(vp, &auio, p->p_ucred);
1892 }
1893 vput(vp);
9697c509
DRJ
1894 *res = count - auio.uio_resid;
1895 return (error);
1896}
1897
1898/*
1899 * readlink_args(char *path, char *buf, int count)
1900 *
1901 * Return target name of a symbolic link.
1902 */
1903int
1904readlink(struct readlink_args *uap)
1905{
fad57d0e 1906 struct nlookupdata nd;
9697c509
DRJ
1907 int error;
1908
fad57d0e
MD
1909 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1910 if (error == 0) {
1911 error = kern_readlink(&nd, uap->buf, uap->count,
1912 &uap->sysmsg_result);
1913 }
1914 nlookup_done(&nd);
984263bc
MD
1915 return (error);
1916}
1917
1918static int
41c20dac 1919setfflags(struct vnode *vp, int flags)
984263bc 1920{
dadab5e9
MD
1921 struct thread *td = curthread;
1922 struct proc *p = td->td_proc;
984263bc
MD
1923 int error;
1924 struct vattr vattr;
1925
1926 /*
1927 * Prevent non-root users from setting flags on devices. When
1928 * a device is reused, users can retain ownership of the device
1929 * if they are allowed to set flags and programs assume that
1930 * chown can't fail when done as root.
1931 */
1932 if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
dadab5e9 1933 ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
984263bc
MD
1934 return (error);
1935
186fccd6
MD
1936 /*
1937 * note: vget is required for any operation that might mod the vnode
1938 * so VINACTIVE is properly cleared.
1939 */
dadab5e9 1940 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
186fccd6
MD
1941 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
1942 VATTR_NULL(&vattr);
1943 vattr.va_flags = flags;
1944 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
1945 vput(vp);
1946 }
984263bc
MD
1947 return (error);
1948}
1949
1950/*
41c20dac
MD
1951 * chflags(char *path, int flags)
1952 *
984263bc
MD
1953 * Change flags of a file given a path name.
1954 */
984263bc
MD
1955/* ARGSUSED */
1956int
41c20dac 1957chflags(struct chflags_args *uap)
984263bc 1958{
fad57d0e
MD
1959 struct nlookupdata nd;
1960 struct vnode *vp;
984263bc 1961 int error;
984263bc 1962
fad57d0e
MD
1963 vp = NULL;
1964 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
1965 /* XXX Add NLC flag indicating modifying operation? */
1966 if (error == 0)
1967 error = nlookup(&nd);
1968 if (error == 0)
1969 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
1970 nlookup_done(&nd);
1971 if (error == 0) {
1972 error = setfflags(vp, SCARG(uap, flags));
1973 vrele(vp);
1974 }
1975 return (error);
984263bc
MD
1976}
1977
1978/*
41c20dac
MD
1979 * fchflags_args(int fd, int flags)
1980 *
984263bc
MD
1981 * Change flags of a file given a file descriptor.
1982 */
984263bc
MD
1983/* ARGSUSED */
1984int
41c20dac 1985fchflags(struct fchflags_args *uap)
984263bc 1986{
dadab5e9
MD
1987 struct thread *td = curthread;
1988 struct proc *p = td->td_proc;
984263bc
MD
1989 struct file *fp;
1990 int error;
1991
1992 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1993 return (error);
41c20dac 1994 return setfflags((struct vnode *) fp->f_data, SCARG(uap, flags));
984263bc
MD
1995}
1996
1997static int
41c20dac 1998setfmode(struct vnode *vp, int mode)
984263bc 1999{
dadab5e9
MD
2000 struct thread *td = curthread;
2001 struct proc *p = td->td_proc;
984263bc
MD
2002 int error;
2003 struct vattr vattr;
2004
186fccd6
MD
2005 /*
2006 * note: vget is required for any operation that might mod the vnode
2007 * so VINACTIVE is properly cleared.
2008 */
dadab5e9 2009 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
186fccd6
MD
2010 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2011 VATTR_NULL(&vattr);
2012 vattr.va_mode = mode & ALLPERMS;
2013 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2014 vput(vp);
2015 }
984263bc
MD
2016 return error;
2017}
2018
9697c509 2019int
fad57d0e 2020kern_chmod(struct nlookupdata *nd, int mode)
9697c509 2021{
fad57d0e 2022 struct vnode *vp;
9697c509
DRJ
2023 int error;
2024
fad57d0e
MD
2025 /* XXX Add NLC flag indicating modifying operation? */
2026 if ((error = nlookup(nd)) != 0)
9697c509 2027 return (error);
fad57d0e
MD
2028 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2029 return (error);
2030 error = setfmode(vp, mode);
2031 vrele(vp);
2032 return (error);
9697c509
DRJ
2033}
2034
984263bc 2035/*
41c20dac
MD
2036 * chmod_args(char *path, int mode)
2037 *
984263bc
MD
2038 * Change mode of a file given path name.
2039 */
984263bc
MD
2040/* ARGSUSED */
2041int
41c20dac 2042chmod(struct chmod_args *uap)
984263bc 2043{
fad57d0e 2044 struct nlookupdata nd;
9697c509 2045 int error;
984263bc 2046
fad57d0e
MD
2047 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2048 if (error == 0)
2049 error = kern_chmod(&nd, uap->mode);
2050 nlookup_done(&nd);
9697c509 2051 return (error);
984263bc
MD
2052}
2053
2054/*
41c20dac
MD
2055 * lchmod_args(char *path, int mode)
2056 *
984263bc
MD
2057 * Change mode of a file given path name (don't follow links.)
2058 */
984263bc
MD
2059/* ARGSUSED */
2060int
41c20dac 2061lchmod(struct lchmod_args *uap)
984263bc 2062{
fad57d0e 2063 struct nlookupdata nd;
984263bc 2064 int error;
984263bc 2065
fad57d0e
MD
2066 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2067 if (error == 0)
2068 error = kern_chmod(&nd, uap->mode);
2069 nlookup_done(&nd);
2070 return (error);
984263bc
MD
2071}
2072
2073/*
41c20dac
MD
2074 * fchmod_args(int fd, int mode)
2075 *
984263bc
MD
2076 * Change mode of a file given a file descriptor.
2077 */
984263bc
MD
2078/* ARGSUSED */
2079int
41c20dac 2080fchmod(struct fchmod_args *uap)
984263bc 2081{
dadab5e9
MD
2082 struct thread *td = curthread;
2083 struct proc *p = td->td_proc;
984263bc
MD
2084 struct file *fp;
2085 int error;
2086
2087 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2088 return (error);
41c20dac 2089 return setfmode((struct vnode *)fp->f_data, SCARG(uap, mode));
984263bc
MD
2090}
2091
2092static int
41c20dac 2093setfown(struct vnode *vp, uid_t uid, gid_t gid)
984263bc 2094{
dadab5e9
MD
2095 struct thread *td = curthread;
2096 struct proc *p = td->td_proc;
984263bc
MD
2097 int error;
2098 struct vattr vattr;
2099
186fccd6
MD
2100 /*
2101 * note: vget is required for any operation that might mod the vnode
2102 * so VINACTIVE is properly cleared.
2103 */
dadab5e9 2104 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
186fccd6
MD
2105 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2106 VATTR_NULL(&vattr);
2107 vattr.va_uid = uid;
2108 vattr.va_gid = gid;
2109 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2110 vput(vp);
2111 }
984263bc
MD
2112 return error;
2113}
2114
9697c509 2115int
fad57d0e 2116kern_chown(struct nlookupdata *nd, int uid, int gid)
9697c509 2117{
fad57d0e 2118 struct vnode *vp;
9697c509
DRJ
2119 int error;
2120
fad57d0e
MD
2121 /* XXX Add NLC flag indicating modifying operation? */
2122 if ((error = nlookup(nd)) != 0)
9697c509 2123 return (error);
fad57d0e
MD
2124 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2125 return (error);
2126 error = setfown(vp, uid, gid);
2127 vrele(vp);
9697c509
DRJ
2128 return (error);
2129}
2130
984263bc 2131/*
41c20dac
MD
2132 * chown(char *path, int uid, int gid)
2133 *
984263bc
MD
2134 * Set ownership given a path name.
2135 */
984263bc 2136int
41c20dac 2137chown(struct chown_args *uap)
984263bc 2138{
fad57d0e 2139 struct nlookupdata nd;
9697c509
DRJ
2140 int error;
2141
fad57d0e
MD
2142 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2143 if (error == 0)
2144 error = kern_chown(&nd, uap->uid, uap->gid);
2145 nlookup_done(&nd);
984263bc
MD
2146 return (error);
2147}
2148
2149/*
41c20dac
MD
2150 * lchown_args(char *path, int uid, int gid)
2151 *
984263bc
MD
2152 * Set ownership given a path name, do not cross symlinks.
2153 */
984263bc 2154int
41c20dac 2155lchown(struct lchown_args *uap)
984263bc 2156{
fad57d0e 2157 struct nlookupdata nd;
984263bc 2158 int error;
9697c509 2159
fad57d0e
MD
2160 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2161 if (error == 0)
2162 error = kern_chown(&nd, uap->uid, uap->gid);
2163 nlookup_done(&nd);
984263bc
MD
2164 return (error);
2165}
2166
2167/*
41c20dac
MD
2168 * fchown_args(int fd, int uid, int gid)
2169 *
984263bc
MD
2170 * Set ownership given a file descriptor.
2171 */
984263bc
MD
2172/* ARGSUSED */
2173int
41c20dac 2174fchown(struct fchown_args *uap)
984263bc 2175{
dadab5e9
MD
2176 struct thread *td = curthread;
2177 struct proc *p = td->td_proc;
984263bc
MD
2178 struct file *fp;
2179 int error;
2180
2181 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2182 return (error);
41c20dac 2183 return setfown((struct vnode *)fp->f_data,
984263bc
MD
2184 SCARG(uap, uid), SCARG(uap, gid));
2185}
2186
2187static int
9697c509 2188getutimes(const struct timeval *tvp, struct timespec *tsp)
984263bc
MD
2189{
2190 struct timeval tv[2];
984263bc 2191
9697c509 2192 if (tvp == NULL) {
984263bc
MD
2193 microtime(&tv[0]);
2194 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2195 tsp[1] = tsp[0];
2196 } else {
9697c509
DRJ
2197 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2198 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
984263bc
MD
2199 }
2200 return 0;
2201}
2202
2203static int
41c20dac 2204setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
984263bc 2205{
dadab5e9
MD
2206 struct thread *td = curthread;
2207 struct proc *p = td->td_proc;
984263bc
MD
2208 int error;
2209 struct vattr vattr;
2210
186fccd6
MD
2211 /*
2212 * note: vget is required for any operation that might mod the vnode
2213 * so VINACTIVE is properly cleared.
2214 */
dadab5e9 2215 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
186fccd6
MD
2216 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2217 VATTR_NULL(&vattr);
2218 vattr.va_atime = ts[0];
2219 vattr.va_mtime = ts[1];
2220 if (nullflag)
2221 vattr.va_vaflags |= VA_UTIMES_NULL;
2222 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2223 vput(vp);
2224 }
984263bc
MD
2225 return error;
2226}
2227
9697c509 2228int
fad57d0e 2229kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
9697c509
DRJ
2230{
2231 struct timespec ts[2];
fad57d0e 2232 struct vnode *vp;
9697c509
DRJ
2233 int error;
2234
fad57d0e 2235 if ((error = getutimes(tptr, ts)) != 0)
9697c509 2236 return (error);
fad57d0e
MD
2237 /* XXX Add NLC flag indicating modifying operation? */
2238 if ((error = nlookup(nd)) != 0)
9697c509 2239 return (error);
fad57d0e
MD
2240 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2241 return (error);
2242 error = setutimes(vp, ts, tptr == NULL);
2243 vrele(vp);
9697c509
DRJ
2244 return (error);
2245}
2246
984263bc 2247/*
41c20dac
MD
2248 * utimes_args(char *path, struct timeval *tptr)
2249 *
984263bc
MD
2250 * Set the access and modification times of a file.
2251 */
984263bc 2252int
41c20dac 2253utimes(struct utimes_args *uap)
984263bc 2254{
9697c509 2255 struct timeval tv[2];
fad57d0e 2256 struct nlookupdata nd;
9697c509
DRJ
2257 int error;
2258
2259 if (uap->tptr) {
2260 error = copyin(uap->tptr, tv, sizeof(tv));
2261 if (error)
2262 return (error);
2263 }
fad57d0e
MD
2264 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2265 if (error == 0)
2266 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2267 nlookup_done(&nd);
984263bc
MD
2268 return (error);
2269}
2270
2271/*
41c20dac
MD
2272 * lutimes_args(char *path, struct timeval *tptr)
2273 *
984263bc
MD
2274 * Set the access and modification times of a file.
2275 */
984263bc 2276int
41c20dac 2277lutimes(struct lutimes_args *uap)
984263bc 2278{
9697c509 2279 struct timeval tv[2];
fad57d0e 2280 struct nlookupdata nd;
9697c509
DRJ
2281 int error;
2282
2283 if (uap->tptr) {
2284 error = copyin(uap->tptr, tv, sizeof(tv));
2285 if (error)
2286 return (error);
2287 }
fad57d0e
MD
2288 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2289 if (error == 0)
2290 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2291 nlookup_done(&nd);
984263bc
MD
2292 return (error);
2293}
2294
984263bc 2295int
9697c509 2296kern_futimes(int fd, struct timeval *tptr)
984263bc 2297{
dadab5e9
MD
2298 struct thread *td = curthread;
2299 struct proc *p = td->td_proc;
984263bc
MD
2300 struct timespec ts[2];
2301 struct file *fp;
984263bc
MD
2302 int error;
2303
9697c509
DRJ
2304 error = getutimes(tptr, ts);
2305 if (error)
984263bc 2306 return (error);
9697c509
DRJ
2307 error = getvnode(p->p_fd, fd, &fp);
2308 if (error)
984263bc 2309 return (error);
9697c509
DRJ
2310 error = setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
2311 return (error);
984263bc
MD
2312}
2313
2314/*
9697c509 2315 * futimes_args(int fd, struct timeval *tptr)
41c20dac 2316 *
9697c509 2317 * Set the access and modification times of a file.
984263bc 2318 */
984263bc 2319int
9697c509
DRJ
2320futimes(struct futimes_args *uap)
2321{
2322 struct timeval tv[2];
2323 int error;
2324
2325 if (uap->tptr) {
2326 error = copyin(uap->tptr, tv, sizeof(tv));
2327 if (error)
2328 return (error);
2329 }
2330
b651c485 2331 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
9697c509
DRJ
2332
2333 return (error);
2334}
2335
2336int
fad57d0e 2337kern_truncate(struct nlookupdata *nd, off_t length)
984263bc 2338{
41c20dac 2339 struct vnode *vp;
984263bc
MD
2340 struct vattr vattr;
2341 int error;
984263bc 2342
9697c509 2343 if (length < 0)
984263bc 2344 return(EINVAL);
fad57d0e
MD
2345 /* XXX Add NLC flag indicating modifying operation? */
2346 if ((error = nlookup(nd)) != 0)
984263bc 2347 return (error);
fad57d0e
MD
2348 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2349 return (error);
2350 VOP_LEASE(vp, nd->nl_td, nd->nl_cred, LEASE_WRITE);
2351 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, nd->nl_td)) != 0) {
2352 vrele(vp);
2353 return (error);
2354 }
2355 if (vp->v_type == VDIR) {
984263bc 2356 error = EISDIR;
fad57d0e
MD
2357 } else if ((error = vn_writechk(vp)) == 0 &&
2358 (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred, nd->nl_td)) == 0) {
984263bc 2359 VATTR_NULL(&vattr);
9697c509 2360 vattr.va_size = length;
fad57d0e 2361 error = VOP_SETATTR(vp, &vattr, nd->nl_cred, nd->nl_td);
984263bc
MD
2362 }
2363 vput(vp);
2364 return (error);
2365}
2366
9697c509
DRJ
2367/*
2368 * truncate(char *path, int pad, off_t length)
2369 *
2370 * Truncate a file given its path name.
2371 */
2372int
2373truncate(struct truncate_args *uap)
2374{
fad57d0e 2375 struct nlookupdata nd;
9697c509
DRJ
2376 int error;
2377
fad57d0e
MD
2378 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2379 if (error == 0)
2380 error = kern_truncate(&nd, uap->length);
2381 nlookup_done(&nd);
9697c509
DRJ
2382 return error;
2383}
2384
984263bc 2385int
8f6f8622 2386kern_ftruncate(int fd, off_t length)
984263bc 2387{
dadab5e9
MD
2388 struct thread *td = curthread;
2389 struct proc *p = td->td_proc;
984263bc
MD
2390 struct vattr vattr;
2391 struct vnode *vp;
2392 struct file *fp;
2393 int error;
2394
8f6f8622 2395 if (length < 0)
984263bc 2396 return(EINVAL);
8f6f8622 2397 if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
984263bc
MD
2398 return (error);
2399 if ((fp->f_flag & FWRITE) == 0)
2400 return (EINVAL);
2401 vp = (struct vnode *)fp->f_data;
dadab5e9 2402 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
5fd012e0 2403 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
2404 if (vp->v_type == VDIR)
2405 error = EISDIR;
2406 else if ((error = vn_writechk(vp)) == 0) {
2407 VATTR_NULL(&vattr);
8f6f8622 2408 vattr.va_size = length;
dadab5e9 2409 error = VOP_SETATTR(vp, &vattr, fp->f_cred, td);
984263bc 2410 }
5fd012e0 2411 VOP_UNLOCK(vp, 0, td);
984263bc
MD
2412 return (error);
2413}
2414
984263bc 2415/*
8f6f8622 2416 * ftruncate_args(int fd, int pad, off_t length)
41c20dac 2417 *
8f6f8622 2418 * Truncate a file given a file descriptor.
984263bc 2419 */
984263bc 2420int
8f6f8622 2421ftruncate(struct ftruncate_args *uap)
984263bc 2422{
8f6f8622 2423 int error;
984263bc 2424
8f6f8622
DRJ
2425 error = kern_ftruncate(uap->fd, uap->length);
2426
2427 return (error);
984263bc
MD
2428}
2429
984263bc 2430/*
41c20dac
MD
2431 * fsync(int fd)
2432 *
984263bc
MD
2433 * Sync an open file.
2434 */
984263bc
MD
2435/* ARGSUSED */
2436int
41c20dac 2437fsync(struct fsync_args *uap)
984263bc 2438{
dadab5e9
MD
2439 struct thread *td = curthread;
2440 struct proc *p = td->td_proc;
41c20dac 2441 struct vnode *vp;
984263bc
MD
2442 struct file *fp;
2443 vm_object_t obj;
2444 int error;
2445
2446 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2447 return (error);
2448 vp = (struct vnode *)fp->f_data;
5fd012e0 2449 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
2450 if (VOP_GETVOBJECT(vp, &obj) == 0)
2451 vm_object_page_clean(obj, 0, 0, 0);
3b568787 2452 if ((error = VOP_FSYNC(vp, MNT_WAIT, td)) == 0 &&
984263bc
MD
2453 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) &&
2454 bioops.io_fsync)
2455 error = (*bioops.io_fsync)(vp);
5fd012e0 2456 VOP_UNLOCK(vp, 0, td);
984263bc
MD
2457 return (error);
2458}
2459
984263bc 2460int
fad57d0e 2461kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
984263bc 2462{
fad57d0e
MD
2463 struct namecache *fncpd;
2464 struct namecache *tncpd;
2465 struct namecache *ncp;
2466 struct mount *mp;
984263bc
MD
2467 int error;
2468
2469 bwillwrite();
fad57d0e 2470 if ((error = nlookup(fromnd)) != 0)
984263bc 2471 return (error);
fad57d0e
MD
2472 if ((fncpd = fromnd->nl_ncp->nc_parent) == NULL)
2473 return (ENOENT);
2474 cache_hold(fncpd);
2475
2476 /*
2477 * unlock the source ncp so we can lookup the target ncp without
2478 * deadlocking. The target may or may not exist so we do not check
2479 * for a target vp like kern_mkdir() and other creation functions do.
2480 *
2481 * The source and target directories are ref'd and rechecked after
2482 * everything is relocked to determine if the source or target file
2483 * has been renamed.
2484 */
2485 KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
2486 fromnd->nl_flags &= ~NLC_NCPISLOCKED;
2487 cache_unlock(fromnd->nl_ncp);
2488
2489 tond->nl_flags |= NLC_CREATE;
2490 if ((error = nlookup(tond)) != 0) {
2491 cache_drop(fncpd);
2492 return (error);
2493 }
2494 if ((tncpd = tond->nl_ncp->nc_parent) == NULL) {
2495 cache_drop(fncpd);
2496 return (ENOENT);
2497 }
2498 cache_hold(tncpd);
2499
2500 /*
2501 * If the source and target are the same there is nothing to do
2502 */
2503 if (fromnd->nl_ncp == tond->nl_ncp) {
2504 cache_drop(fncpd);
2505 cache_drop(tncpd);
2506 return (0);
2507 }
2508
2509 /*
2510 * relock the source ncp
2511 */
2512 if (cache_lock_nonblock(fromnd->nl_ncp) == 0) {
2513 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2514 } else if (fromnd->nl_ncp > tond->nl_ncp) {
2515 cache_lock(fromnd->nl_ncp);
2516 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2517 } else {
2518 cache_unlock(tond->nl_ncp);
2519 cache_lock(fromnd->nl_ncp);
2520 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2521 cache_lock(tond->nl_ncp);
2522 cache_resolve(tond->nl_ncp, tond->nl_cred);
2523 }
2524 fromnd->nl_flags |= NLC_NCPISLOCKED;
2525
2526 /*
2527 * make sure the parent directories linkages are the same
2528 */
2529 if (fncpd != fromnd->nl_ncp->nc_parent ||
2530 tncpd != tond->nl_ncp->nc_parent) {
2531 cache_drop(fncpd);
2532 cache_drop(tncpd);
2533 return (ENOENT);
2534 }
2535
2536 /*
2537 * Both the source and target must be within the same filesystem and
2538 * in the same filesystem as their parent directories within the
2539 * namecache topology.
2540 */
2541 mp = fncpd->nc_mount;
2542 if (mp != tncpd->nc_mount || mp != fromnd->nl_ncp->nc_mount ||
2543 mp != tond->nl_ncp->nc_mount) {
2544 cache_drop(fncpd);
2545 cache_drop(tncpd);
2546 return (EXDEV);
2547 }
2548
2549 /*
2550 * If the target exists and either the source or target is a directory,
2551 * then both must be directories.
2552 */
2553 if (tond->nl_ncp->nc_vp) {
2554 if (fromnd->nl_ncp->nc_vp->v_type == VDIR) {
2555 if (tond->nl_ncp->nc_vp->v_type != VDIR)
2556 error = ENOTDIR;
2557 } else if (tond->nl_ncp->nc_vp->v_type == VDIR) {
984263bc 2558 error = EISDIR;
984263bc
MD
2559 }
2560 }
fad57d0e 2561
984263bc 2562 /*
fad57d0e
MD
2563 * You cannot rename a source into itself or a subdirectory of itself.
2564 * We check this by travsersing the target directory upwards looking
2565 * for a match against the source.
984263bc 2566 */
fad57d0e
MD
2567 if (error == 0) {
2568 for (ncp = tncpd; ncp; ncp = ncp->nc_parent) {
2569 if (fromnd->nl_ncp == ncp) {
2570 error = EINVAL;
2571 break;
2572 }
984263bc 2573 }
fad57d0e
MD
2574 }
2575
2576 cache_drop(fncpd);
2577 cache_drop(tncpd);
2578 if (error)
2579 return (error);
2580 error = VOP_NRENAME(fromnd->nl_ncp, tond->nl_ncp, tond->nl_cred);
984263bc
MD
2581 return (error);
2582}
2583
2584/*
9697c509 2585 * rename_args(char *from, char *to)
41c20dac 2586 *
9697c509
DRJ
2587 * Rename files. Source and destination must either both be directories,
2588 * or both not be directories. If target is a directory, it must be empty.
984263bc 2589 */
984263bc 2590int
9697c509
DRJ
2591rename(struct rename_args *uap)
2592{
fad57d0e 2593 struct nlookupdata fromnd, tond;
9697c509
DRJ
2594 int error;
2595
fad57d0e
MD
2596 error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
2597 if (error == 0) {
2598 error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
2599 if (error == 0)
2600 error = kern_rename(&fromnd, &tond);
2601 nlookup_done(&tond);
2602 }
2603 nlookup_done(&fromnd);
9697c509
DRJ
2604 return (error);
2605}
2606
2607int
fad57d0e 2608kern_mkdir(struct nlookupdata *nd, int mode)
984263bc 2609{
dadab5e9
MD
2610 struct thread *td = curthread;
2611 struct proc *p = td->td_proc;
fad57d0e 2612 struct namecache *ncp;
41c20dac 2613 struct vnode *vp;
984263bc
MD
2614 struct vattr vattr;
2615 int error;
984263bc
MD
2616
2617 bwillwrite();
fad57d0e
MD
2618 nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE;
2619 if ((error = nlookup(nd)) != 0)
984263bc 2620 return (error);
fad57d0e
MD
2621
2622 ncp = nd->nl_ncp;
2623 if (ncp->nc_vp)
984263bc 2624 return (EEXIST);
fad57d0e 2625
984263bc
MD
2626 VATTR_NULL(&vattr);
2627 vattr.va_type = VDIR;
9697c509 2628 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
fad57d0e
MD
2629
2630 vp = NULL;
2631 error = VOP_NMKDIR(ncp, &vp, p->p_ucred, &vattr);
9697c509 2632 if (error == 0)
fad57d0e 2633 vput(vp);
984263bc
MD
2634 return (error);
2635}
2636
2637/*
9697c509 2638 * mkdir_args(char *path, int mode)
41c20dac 2639 *
9697c509 2640 * Make a directory file.
984263bc 2641 */
984263bc
MD
2642/* ARGSUSED */
2643int
9697c509
DRJ
2644mkdir(struct mkdir_args *uap)
2645{
fad57d0e 2646 struct nlookupdata nd;
9697c509
DRJ
2647 int error;
2648
fad57d0e
MD
2649 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2650 if (error == 0)
2651 error = kern_mkdir(&nd, uap->mode);
2652 nlookup_done(&nd);
9697c509
DRJ
2653 return (error);
2654}
2655
2656int
fad57d0e 2657kern_rmdir(struct nlookupdata *nd)
984263bc 2658{
fad57d0e 2659 struct namecache *ncp;
984263bc 2660 int error;
984263bc
MD
2661
2662 bwillwrite();
fad57d0e
MD
2663 nd->nl_flags |= NLC_DELETE;
2664 if ((error = nlookup(nd)) != 0)
984263bc 2665 return (error);
fad57d0e
MD
2666
2667 ncp = nd->nl_ncp;
2668 error = VOP_NRMDIR(ncp, nd->nl_cred);
984263bc
MD
2669 return (error);
2670}
2671
984263bc 2672/*
9697c509 2673 * rmdir_args(char *path)
41c20dac 2674 *
9697c509 2675 * Remove a directory file.
984263bc 2676 */
9697c509 2677/* ARGSUSED */
984263bc 2678int
9697c509 2679rmdir(struct rmdir_args *uap)
984263bc 2680{
fad57d0e 2681 struct nlookupdata nd;
9697c509
DRJ
2682 int error;
2683
fad57d0e
MD
2684 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2685 if (error == 0)
2686 error = kern_rmdir(&nd);
2687 nlookup_done(&nd);
984263bc
MD
2688 return (error);
2689}
984263bc 2690
984263bc 2691int
9697c509 2692kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res)
984263bc 2693{
dadab5e9
MD
2694 struct thread *td = curthread;
2695 struct proc *p = td->td_proc;
984263bc
MD
2696 struct vnode *vp;
2697 struct file *fp;
2698 struct uio auio;
2699 struct iovec aiov;
2700 long loff;
2701 int error, eofflag;
2702
9697c509 2703 if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
984263bc
MD
2704 return (error);
2705 if ((fp->f_flag & FREAD) == 0)
2706 return (EBADF);
2707 vp = (struct vnode *)fp->f_data;
2708unionread:
2709 if (vp->v_type != VDIR)
2710 return (EINVAL);
9697c509
DRJ
2711 aiov.iov_base = buf;
2712 aiov.iov_len = count;
984263bc
MD
2713 auio.uio_iov = &aiov;
2714 auio.uio_iovcnt = 1;
2715 auio.uio_rw = UIO_READ;
2716 auio.uio_segflg = UIO_USERSPACE;
dadab5e9 2717 auio.uio_td = td;
9697c509 2718 auio.uio_resid = count;
5fd012e0
MD
2719 /* vn_lock(vp, LK_SHARED | LK_RETRY, td); */
2720 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
2721 loff = auio.uio_offset = fp->f_offset;
2722 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
2723 fp->f_offset = auio.uio_offset;
5fd012e0 2724 VOP_UNLOCK(vp, 0, td);
984263bc
MD
2725 if (error)
2726 return (error);
9697c509 2727 if (count == auio.uio_resid) {
984263bc 2728 if (union_dircheckp) {
dadab5e9 2729 error = union_dircheckp(td, &vp, fp);
984263bc
MD
2730 if (error == -1)
2731 goto unionread;
2732 if (error)
2733 return (error);
2734 }
2735 if ((vp->v_flag & VROOT) &&
2736 (vp->v_mount->mnt_flag & MNT_UNION)) {
2737 struct vnode *tvp = vp;
2738 vp = vp->v_mount->mnt_vnodecovered;
597aea93 2739 vref(vp);
e38462a2 2740 fp->f_data = (caddr_t)vp;
984263bc
MD
2741 fp->f_offset = 0;
2742 vrele(tvp);
2743 goto unionread;
2744 }
2745 }
9697c509
DRJ
2746 if (basep) {
2747 *basep = loff;
984263bc 2748 }
9697c509
DRJ
2749 *res = count - auio.uio_resid;
2750 return (error);
2751}
2752
2753/*
2754 * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
2755 *
2756 * Read a block of directory entries in a file system independent format.
2757 */
2758int
2759getdirentries(struct getdirentries_args *uap)
2760{
2761 long base;
2762 int error;
2763
2764 error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base,
2765 &uap->sysmsg_result);
2766
2767 if (error == 0)
2768 error = copyout(&base, uap->basep, sizeof(*uap->basep));
984263bc
MD
2769 return (error);
2770}
41c20dac
MD
2771
2772/*
2773 * getdents_args(int fd, char *buf, size_t count)
2774 */
984263bc 2775int
41c20dac 2776getdents(struct getdents_args *uap)
984263bc 2777{
9697c509
DRJ
2778 int error;
2779
2780 error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL,
2781 &uap->sysmsg_result);
41c20dac 2782
9697c509 2783 return (error);
984263bc
MD
2784}
2785
2786/*
41c20dac
MD
2787 * umask(int newmask)
2788 *
984263bc
MD
2789 * Set the mode mask for creation of filesystem nodes.
2790 *
2791 * MP SAFE
2792 */
984263bc 2793int
41c20dac 2794umask(struct umask_args *uap)
984263bc 2795{
dadab5e9
MD
2796 struct thread *td = curthread;
2797 struct proc *p = td->td_proc;
41c20dac 2798 struct filedesc *fdp;
984263bc
MD
2799
2800 fdp = p->p_fd;
c7114eea 2801 uap->sysmsg_result = fdp->fd_cmask;
984263bc
MD
2802 fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS;
2803 return (0);
2804}
2805
2806/*
41c20dac
MD
2807 * revoke(char *path)
2808 *
984263bc
MD
2809 * Void all references to file by ripping underlying filesystem
2810 * away from vnode.
2811 */
984263bc
MD
2812/* ARGSUSED */
2813int
41c20dac 2814revoke(struct revoke_args *uap)
984263bc 2815{
dadab5e9 2816 struct thread *td = curthread;
fad57d0e 2817 struct nlookupdata nd;
984263bc 2818 struct vattr vattr;
fad57d0e
MD
2819 struct vnode *vp;
2820 struct ucred *cred;
984263bc 2821 int error;
984263bc 2822
fad57d0e
MD
2823 vp = NULL;
2824 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
2825 if (error == 0)
2826 error = nlookup(&nd);
2827 if (error == 0)
2828 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
2829 cred = crhold(nd.nl_cred);
2830 nlookup_done(&nd);
2831 if (error == 0) {
2832 if (vp->v_type != VCHR && vp->v_type != VBLK)
2833 error = EINVAL;
2834 if (error == 0)
2835 error = VOP_GETATTR(vp, &vattr, td);
2836 if (error == 0 && cred->cr_uid != vattr.va_uid)
2837 error = suser_cred(cred, PRISON_ROOT);
2838 if (error == 0 && count_udev(vp->v_udev) > 0) {
2839 if ((error = vx_lock(vp)) == 0) {
2840 VOP_REVOKE(vp, REVOKEALL);
2841 vx_unlock(vp);
2842 }
5fd012e0 2843 }
fad57d0e 2844 vrele(vp);
5fd012e0 2845 }
fad57d0e 2846 crfree(cred);
984263bc
MD
2847 return (error);
2848}
2849
2850/*
2851 * Convert a user file descriptor to a kernel file entry.
2852 */
2853int
41c20dac 2854getvnode(struct filedesc *fdp, int fd, struct file **fpp)
984263bc
MD
2855{
2856 struct file *fp;
2857
2858 if ((u_int)fd >= fdp->fd_nfiles ||
2859 (fp = fdp->fd_ofiles[fd]) == NULL)
2860 return (EBADF);
2861 if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO)
2862 return (EINVAL);
2863 *fpp = fp;
2864 return (0);
2865}
2866/*
41c20dac
MD
2867 * getfh_args(char *fname, fhandle_t *fhp)
2868 *
984263bc
MD
2869 * Get (NFS) file handle
2870 */
984263bc 2871int
41c20dac 2872getfh(struct getfh_args *uap)
984263bc 2873{
dadab5e9 2874 struct thread *td = curthread;
fad57d0e 2875 struct nlookupdata nd;
984263bc 2876 fhandle_t fh;
41c20dac 2877 struct vnode *vp;
984263bc
MD
2878 int error;
2879
2880 /*
2881 * Must be super user
2882 */
fad57d0e 2883 if ((error = suser(td)) != 0)
984263bc 2884 return (error);
fad57d0e
MD
2885
2886 vp = NULL;
2887 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
2888 if (error == 0)
2889 error = nlookup(&nd);
2890 if (error == 0)
2891 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
2892 nlookup_done(&nd);
2893 if (error == 0) {
2894 bzero(&fh, sizeof(fh));
2895 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
2896 error = VFS_VPTOFH(vp, &fh.fh_fid);
2897 vput(vp);
2898 if (error == 0)
2899 error = copyout(&fh, uap->fhp, sizeof(fh));
2900 }
984263bc
MD
2901 return (error);
2902}
2903
2904/*
41c20dac
MD
2905 * fhopen_args(const struct fhandle *u_fhp, int flags)
2906 *
984263bc
MD
2907 * syscall for the rpc.lockd to use to translate a NFS file handle into
2908 * an open descriptor.
2909 *
2910 * warning: do not remove the suser() call or this becomes one giant
2911 * security hole.
2912 */
984263bc 2913int
41c20dac 2914fhopen(struct fhopen_args *uap)
984263bc 2915{
dadab5e9
MD
2916 struct thread *td = curthread;
2917 struct proc *p = td->td_proc;
984263bc
MD
2918 struct mount *mp;
2919 struct vnode *vp;
2920 struct fhandle fhp;
2921 struct vattr vat;
2922 struct vattr *vap = &vat;
2923 struct flock lf;
41c20dac 2924 struct filedesc *fdp = p->p_fd;
984263bc
MD
2925 int fmode, mode, error, type;
2926 struct file *nfp;
fad57d0e 2927 struct file *fp;
984263bc
MD
2928 int indx;
2929
2930 /*
2931 * Must be super user
2932 */
dadab5e9 2933 error = suser(td);
984263bc
MD
2934 if (error)
2935 return (error);
2936
2937 fmode = FFLAGS(SCARG(uap, flags));
2938 /* why not allow a non-read/write open for our lockd? */
2939 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
2940 return (EINVAL);
2941 error = copyin(SCARG(uap,u_fhp), &fhp, sizeof(fhp));
2942 if (error)
2943 return(error);
2944 /* find the mount point */
2945 mp = vfs_getvfs(&fhp.fh_fsid);
2946 if (mp == NULL)
2947 return (ESTALE);
2948 /* now give me my vnode, it gets returned to me locked */
2949 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
2950 if (error)
2951 return (error);
2952 /*
2953 * from now on we have to make sure not
2954 * to forget about the vnode
2955 * any error that causes an abort must vput(vp)
2956 * just set error = err and 'goto bad;'.
2957 */
2958
2959 /*
2960 * from vn_open
2961 */
2962 if (vp->v_type == VLNK) {
2963 error = EMLINK;
2964 goto bad;
2965 }
2966 if (vp->v_type == VSOCK) {
2967 error = EOPNOTSUPP;
2968 goto bad;
2969 }
2970 mode = 0;
2971 if (fmode & (FWRITE | O_TRUNC)) {
2972 if (vp->v_type == VDIR) {
2973 error = EISDIR;
2974 goto bad;
2975 }
2976 error = vn_writechk(vp);
2977 if (error)
2978 goto bad;
2979 mode |= VWRITE;
2980 }
2981 if (fmode & FREAD)
2982 mode |= VREAD;
2983 if (mode) {
dadab5e9 2984 error = VOP_ACCESS(vp, mode, p->p_ucred, td);
984263bc
MD
2985 if (error)
2986 goto bad;
2987 }
2988 if (fmode & O_TRUNC) {
5fd012e0 2989 VOP_UNLOCK(vp, 0, td); /* XXX */
dadab5e9 2990 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
5fd012e0 2991 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); /* XXX */
984263bc
MD
2992 VATTR_NULL(vap);
2993 vap->va_size = 0;
dadab5e9 2994 error = VOP_SETATTR(vp, vap, p->p_ucred, td);
984263bc
MD
2995 if (error)
2996 goto bad;
2997 }
fad57d0e 2998
984263bc 2999 /*
fad57d0e
MD
3000 * VOP_OPEN needs the file pointer so it can potentially override
3001 * it.
3002 *
3003 * WARNING! no f_ncp will be associated when fhopen()ing a directory.
3004 * XXX
984263bc 3005 */
fad57d0e
MD
3006 if ((error = falloc(p, &nfp, NULL)) != 0)
3007 goto bad;
3008 fp = nfp;
3009
3010 fp->f_data = (caddr_t)vp;
3011 fp->f_flag = fmode & FMASK;
3012 fp->f_ops = &vnode_fileops;
3013 fp->f_type = DTYPE_VNODE;
3014
3015 error = VOP_OPEN(vp, fmode, p->p_ucred, fp, td);
3016 if (error) {
3017 /*
3018 * setting f_ops this way prevents VOP_CLOSE from being
3019 * called or fdrop() releasing the vp from v_data. Since
3020 * the VOP_OPEN failed we don't want to VOP_CLOSE.
3021 */
3022 fp->f_ops = &badfileops;
3023 fp->f_data = NULL;
3024 fdrop(fp, td);
3025 goto bad;
984263bc
MD
3026 }
3027 if (fmode & FWRITE)
3028 vp->v_writecount++;
3029
3030 /*
fad57d0e
MD
3031 * The fp now owns a reference on the vnode. We still have our own
3032 * ref+lock.
3033 */
3034 vref(vp);
3035
3036 /*
3037 * Make sure that a VM object is created for VMIO support. If this
3038 * fails just fdrop() normally to clean up.
984263bc 3039 */
fad57d0e
MD
3040 if (vn_canvmio(vp) == TRUE) {
3041 if ((error = vfs_object_create(vp, td)) != 0) {
3042 fdrop(fp, td);
3043 goto bad;
3044 }
3045 }
984263bc 3046
fad57d0e
MD
3047 /*
3048 * The open was successful, associate it with a file descriptor.
3049 */
3050 if ((error = fsetfd(p, fp, &indx)) != 0) {
984263bc
MD
3051 if (fmode & FWRITE)
3052 vp->v_writecount--;
fad57d0e 3053 fdrop(fp, td);
984263bc
MD
3054 goto bad;
3055 }
984263bc 3056
984263bc
MD
3057 if (fmode & (O_EXLOCK | O_SHLOCK)) {
3058 lf.l_whence = SEEK_SET;
3059 lf.l_start = 0;
3060 lf.l_len = 0;
3061 if (fmode & O_EXLOCK)
3062 lf.l_type = F_WRLCK;
3063 else
3064 lf.l_type = F_RDLCK;
3065 type = F_FLOCK;
3066 if ((fmode & FNONBLOCK) == 0)
3067 type |= F_WAIT;
5fd012e0 3068 VOP_UNLOCK(vp, 0, td);
984263bc
MD
3069 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
3070 /*
3071 * lock request failed. Normally close the descriptor
3072 * but handle the case where someone might have dup()d
3073 * or close()d it when we weren't looking.
3074 */
3075 if (fdp->fd_ofiles[indx] == fp) {
3076 fdp->fd_ofiles[indx] = NULL;
dadab5e9 3077 fdrop(fp, td);
984263bc
MD
3078 }
3079
3080 /*
3081 * release our private reference.
3082 */
dadab5e9 3083 fdrop(fp, td);
fad57d0e 3084 vrele(vp);
984263bc
MD
3085 return (error);
3086 }
5fd012e0 3087 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
3088 fp->f_flag |= FHASLOCK;
3089 }
3090 if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0))
3b568787 3091 vfs_object_create(vp, td);
984263bc 3092
fad57d0e 3093 vput(vp);
dadab5e9 3094 fdrop(fp, td);
c7114eea 3095 uap->sysmsg_result = indx;
984263bc
MD
3096 return (0);
3097
3098bad:
3099 vput(vp);
3100 return (error);
3101}
3102
41c20dac
MD
3103/*
3104 * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
3105 */
984263bc 3106int
41c20dac 3107fhstat(struct fhstat_args *uap)
984263bc 3108{
dadab5e9 3109 struct thread *td = curthread;
984263bc
MD
3110 struct stat sb;
3111 fhandle_t fh;
3112 struct mount *mp;
3113 struct vnode *vp;
3114 int error;
3115
3116 /*
3117 * Must be super user
3118 */
dadab5e9 3119 error = suser(td);
984263bc
MD
3120 if (error)
3121 return (error);
3122
3123 error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t));
3124 if (error)
3125 return (error);
3126
3127 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3128 return (ESTALE);
3129 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3130 return (error);
dadab5e9 3131 error = vn_stat(vp, &sb, td);
984263bc
MD
3132 vput(vp);
3133 if (error)
3134 return (error);
3135 error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
3136 return (error);
3137}
3138
41c20dac
MD
3139/*
3140 * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
3141 */
984263bc 3142int
41c20dac 3143fhstatfs(struct fhstatfs_args *uap)
984263bc 3144{
dadab5e9 3145 struct thread *td = curthread;
984263bc
MD
3146 struct statfs *sp;
3147 struct mount *mp;
3148 struct vnode *vp;
3149 struct statfs sb;
3150 fhandle_t fh;
3151 int error;
3152
3153 /*
3154 * Must be super user
3155 */
dadab5e9 3156 if ((error = suser(td)))
984263bc
MD
3157 return (error);
3158
3159 if ((error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t))) != 0)
3160 return (error);
3161
3162 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3163 return (ESTALE);
3164 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3165 return (error);
3166 mp = vp->v_mount;
3167 sp = &mp->mnt_stat;
3168 vput(vp);
dadab5e9 3169 if ((error = VFS_STATFS(mp, sp, td)) != 0)
984263bc
MD
3170 return (error);
3171 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
dadab5e9 3172 if (suser(td)) {
e38462a2 3173 bcopy(sp, &sb, sizeof(sb));
984263bc
MD
3174 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
3175 sp = &sb;
3176 }
3177 return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
3178}
3179
3180/*
3181 * Syscall to push extended attribute configuration information into the
3182 * VFS. Accepts a path, which it converts to a mountpoint, as well as
3183 * a command (int cmd), and attribute name and misc data. For now, the
3184 * attribute name is left in userspace for consumption by the VFS_op.
3185 * It will probably be changed to be copied into sysspace by the
3186 * syscall in the future, once issues with various consumers of the
3187 * attribute code have raised their hands.
3188 *
3189 * Currently this is used only by UFS Extended Attributes.
3190 */
3191int
41c20dac 3192extattrctl(struct extattrctl_args *uap)
984263bc 3193{
fad57d0e 3194 struct nlookupdata nd;
984263bc 3195 struct mount *mp;
fad57d0e 3196 struct vnode *vp;
984263bc
MD
3197 int error;
3198
fad57d0e
MD
3199 vp = NULL;
3200 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3201 if (error == 0)
3202 error = nlookup(&nd);
3203 if (error == 0) {
3204 mp = nd.nl_ncp->nc_mount;
3205 error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd),
3206 SCARG(uap, attrname), SCARG(uap, arg),
3207 nd.nl_td);
3208 }
3209 nlookup_done(&nd);
3210 return (error);
984263bc
MD
3211}
3212
3213/*
3214 * Syscall to set a named extended attribute on a file or directory.
3215 * Accepts attribute name, and a uio structure pointing to the data to set.
3216 * The uio is consumed in the style of writev(). The real work happens
3217 * in VOP_SETEXTATTR().
3218 */
3219int
41c20dac 3220extattr_set_file(struct extattr_set_file_args *uap)
984263bc 3221{
984263bc 3222 char attrname[EXTATTR_MAXNAMELEN];
fad57d0e
MD
3223 struct iovec aiov[UIO_SMALLIOV];
3224 struct iovec *needfree;
3225 struct nlookupdata nd;
3226 struct iovec *iov;
3227 struct vnode *vp;
3228 struct uio auio;
3229 u_int iovlen;
3230 u_int cnt;
3231 int error;
3232 int i;
984263bc
MD
3233
3234 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3235 if (error)
3236 return (error);
fad57d0e
MD
3237
3238 vp = NULL;
3239 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3240 if (error == 0)
3241 error = nlookup(&nd);
3242 if (error == 0)
3243 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3244 if (error) {
3245 nlookup_done(&nd);
3246 return (error);
3247 }
3248
3249 needfree = NULL;
984263bc
MD
3250 iovlen = uap->iovcnt * sizeof(struct iovec);
3251 if (uap->iovcnt > UIO_SMALLIOV) {
3252 if (uap->iovcnt > UIO_MAXIOV) {
3253 error = EINVAL;
3254 goto done;
3255 }
3256 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3257 needfree = iov;
fad57d0e 3258 } else {
984263bc 3259 iov = aiov;
fad57d0e 3260 }
984263bc
MD
3261 auio.uio_iov = iov;
3262 auio.uio_iovcnt = uap->iovcnt;
3263 auio.uio_rw = UIO_WRITE;
3264 auio.uio_segflg = UIO_USERSPACE;
fad57d0e 3265 auio.uio_td = nd.nl_td;
984263bc 3266 auio.uio_offset = 0;
e38462a2 3267 if ((error = copyin(uap->iovp, iov, iovlen)))
984263bc
MD
3268 goto done;
3269 auio.uio_resid = 0;
3270 for (i = 0; i < uap->iovcnt; i++) {
3271 if (iov->iov_len > INT_MAX - auio.uio_resid) {
3272 error = EINVAL;
3273 goto done;
3274 }
3275 auio.uio_resid += iov->iov_len;
3276 iov++;
3277 }
3278 cnt = auio.uio_resid;
fad57d0e 3279 error = VOP_SETEXTATTR(vp, attrname, &auio, nd.nl_cred, nd.nl_td);
984263bc 3280 cnt -= auio.uio_resid;
c7114eea 3281 uap->sysmsg_result = cnt;
984263bc 3282done:
fad57d0e
MD
3283 vput(vp);
3284 nlookup_done(&nd);
984263bc
MD
3285 if (needfree)
3286 FREE(needfree, M_IOV);
984263bc
MD
3287 return (error);
3288}
3289
3290/*
3291 * Syscall to get a named extended attribute on a file or directory.
3292 * Accepts attribute name, and a uio structure pointing to a buffer for the
3293 * data. The uio is consumed in the style of readv(). The real work
3294 * happens in VOP_GETEXTATTR();
3295 */
3296int