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.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
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.
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
38 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
39 * $FreeBSD: src/sys/kern/vfs_syscalls.c,v 1.151.2.18 2003/04/04 20:35:58 tegge Exp $
40 * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.101 2006/08/19 17:27:23 dillon Exp $
43 #include <sys/param.h>
44 #include <sys/systm.h>
47 #include <sys/sysent.h>
48 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/mountctl.h>
51 #include <sys/sysproto.h>
52 #include <sys/filedesc.h>
53 #include <sys/kernel.h>
54 #include <sys/fcntl.h>
56 #include <sys/linker.h>
58 #include <sys/unistd.h>
59 #include <sys/vnode.h>
61 #include <sys/namei.h>
62 #include <sys/nlookup.h>
63 #include <sys/dirent.h>
64 #include <sys/extattr.h>
65 #include <sys/spinlock.h>
66 #include <sys/kern_syscall.h>
67 #include <sys/objcache.h>
69 #include <machine/limits.h>
70 #include <vfs/union/union.h>
71 #include <sys/sysctl.h>
73 #include <vm/vm_object.h>
74 #include <vm/vm_page.h>
76 #include <sys/file2.h>
77 #include <sys/spinlock2.h>
79 static int checkvp_chdir (struct vnode *vn, struct thread *td);
80 static void checkdirs (struct vnode *olddp, struct namecache *ncp);
81 static int chroot_refuse_vdir_fds (struct filedesc *fdp);
82 static int chroot_visible_mnt(struct mount *mp, struct proc *p);
83 static int getutimes (const struct timeval *, struct timespec *);
84 static int setfown (struct vnode *, uid_t, gid_t);
85 static int setfmode (struct vnode *, int);
86 static int setfflags (struct vnode *, int);
87 static int setutimes (struct vnode *, const struct timespec *, int);
88 static int usermount = 0; /* if 1, non-root can mount fs. */
90 int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
92 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
95 * Virtual File System System Calls
99 * Mount a file system.
102 * mount_args(char *type, char *path, int flags, caddr_t data)
106 sys_mount(struct mount_args *uap)
108 struct thread *td = curthread;
109 struct proc *p = td->td_proc;
111 struct namecache *ncp;
113 struct vfsconf *vfsp;
114 int error, flag = 0, flag2 = 0;
116 struct nlookupdata nd;
117 char fstypename[MFSNAMELEN];
118 struct nlcomponent nlc;
119 struct ucred *cred = p->p_ucred;
122 if (cred->cr_prison != NULL)
124 if (usermount == 0 && (error = suser(td)))
127 * Do not allow NFS export by non-root users.
129 if (uap->flags & MNT_EXPORTED) {
135 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
138 uap->flags |= MNT_NOSUID | MNT_NODEV;
141 * Lookup the requested path and extract the ncp and vnode.
143 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
145 if ((error = nlookup(&nd)) == 0) {
146 if (nd.nl_ncp->nc_vp == NULL)
156 * Extract the locked+refd ncp and cleanup the nd structure
163 * now we have the locked ref'd ncp and unreferenced vnode.
166 if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
173 * Now we have an unlocked ref'd ncp and a locked ref'd vp
175 if (uap->flags & MNT_UPDATE) {
176 if ((vp->v_flag & VROOT) == 0) {
183 flag2 = mp->mnt_kern_flag;
185 * We only allow the filesystem to be reloaded if it
186 * is currently mounted read-only.
188 if ((uap->flags & MNT_RELOAD) &&
189 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
192 return (EOPNOTSUPP); /* Needs translation */
195 * Only root, or the user that did the original mount is
196 * permitted to update it.
198 if (mp->mnt_stat.f_owner != cred->cr_uid &&
199 (error = suser(td))) {
204 if (vfs_busy(mp, LK_NOWAIT)) {
209 if ((vp->v_flag & VMOUNT) != 0 ||
210 vp->v_mountedhere != NULL) {
216 vp->v_flag |= VMOUNT;
218 uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
223 * If the user is not root, ensure that they own the directory
224 * onto which we are attempting to mount.
226 if ((error = VOP_GETATTR(vp, &va)) ||
227 (va.va_uid != cred->cr_uid && (error = suser(td)))) {
232 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
237 if (vp->v_type != VDIR) {
242 if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
247 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
248 if (!strcmp(vfsp->vfc_name, fstypename))
254 /* Only load modules for root (very important!) */
255 if ((error = suser(td)) != 0) {
260 error = linker_load_file(fstypename, &lf);
261 if (error || lf == NULL) {
269 /* lookup again, see if the VFS was loaded */
270 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
271 if (!strcmp(vfsp->vfc_name, fstypename))
276 linker_file_unload(lf);
282 if ((vp->v_flag & VMOUNT) != 0 ||
283 vp->v_mountedhere != NULL) {
288 vp->v_flag |= VMOUNT;
291 * Allocate and initialize the filesystem.
293 mp = malloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
294 TAILQ_INIT(&mp->mnt_nvnodelist);
295 TAILQ_INIT(&mp->mnt_reservedvnlist);
296 TAILQ_INIT(&mp->mnt_jlist);
297 mp->mnt_nvnodelistsize = 0;
298 lockinit(&mp->mnt_lock, "vfslock", 0, 0);
299 vfs_busy(mp, LK_NOWAIT);
300 mp->mnt_op = vfsp->vfc_vfsops;
302 vfsp->vfc_refcount++;
303 mp->mnt_stat.f_type = vfsp->vfc_typenum;
304 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
305 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
306 mp->mnt_vnodecovered = vp;
307 mp->mnt_stat.f_owner = cred->cr_uid;
308 mp->mnt_iosize_max = DFLTPHYS;
312 * Set the mount level flags.
314 if (uap->flags & MNT_RDONLY)
315 mp->mnt_flag |= MNT_RDONLY;
316 else if (mp->mnt_flag & MNT_RDONLY)
317 mp->mnt_kern_flag |= MNTK_WANTRDWR;
318 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
319 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
320 MNT_NOSYMFOLLOW | MNT_IGNORE |
321 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
322 mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
323 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
324 MNT_NOSYMFOLLOW | MNT_IGNORE |
325 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
327 * Mount the filesystem.
328 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
331 error = VFS_MOUNT(mp, uap->path, uap->data, cred);
332 if (mp->mnt_flag & MNT_UPDATE) {
333 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
334 mp->mnt_flag &= ~MNT_RDONLY;
335 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
336 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
339 mp->mnt_kern_flag = flag2;
342 vp->v_flag &= ~VMOUNT;
347 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
349 * Put the new filesystem on the mount list after root. The mount
350 * point gets its own mnt_ncp which is a special ncp linking the
351 * vnode-under to the root of the new mount. The lookup code
352 * detects the mount point going forward and detects the special
353 * mnt_ncp via NCP_MOUNTPT going backwards.
355 * It is not necessary to invalidate or purge the vnode underneath
356 * because elements under the mount will be given their own glue
360 nlc.nlc_nameptr = "";
362 mp->mnt_ncp = cache_nlookup(ncp, &nlc);
363 cache_setunresolved(mp->mnt_ncp);
364 mp->mnt_ncp->nc_flag |= NCF_MOUNTPT;
365 mp->mnt_ncp->nc_mount = mp;
367 /* XXX get the root of the fs and cache_setvp(mnt_ncp...) */
368 vp->v_flag &= ~VMOUNT;
369 vp->v_mountedhere = mp;
370 mountlist_insert(mp, MNTINS_LAST);
371 checkdirs(vp, mp->mnt_ncp);
372 cache_unlock(mp->mnt_ncp); /* leave ref intact */
374 error = vfs_allocate_syncvnode(mp);
376 if ((error = VFS_START(mp, 0)) != 0)
379 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
380 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
381 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
382 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
383 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
384 vp->v_flag &= ~VMOUNT;
385 mp->mnt_vfc->vfc_refcount--;
395 * Scan all active processes to see if any of them have a current
396 * or root directory onto which the new filesystem has just been
397 * mounted. If so, replace them with the new mount point.
399 * The passed ncp is ref'd and locked (from the mount code) and
400 * must be associated with the vnode representing the root of the
403 struct checkdirs_info {
406 struct namecache *ncp;
409 static int checkdirs_callback(struct proc *p, void *data);
412 checkdirs(struct vnode *olddp, struct namecache *ncp)
414 struct checkdirs_info info;
418 if (olddp->v_usecount == 1)
420 mp = olddp->v_mountedhere;
421 if (VFS_ROOT(mp, &newdp))
422 panic("mount: lost mount");
423 cache_setvp(ncp, newdp);
425 if (rootvnode == olddp) {
427 vfs_cache_setroot(newdp, cache_hold(ncp));
433 allproc_scan(checkdirs_callback, &info);
438 * NOTE: callback is not MP safe because the scanned process's filedesc
439 * structure can be ripped out from under us, amoung other things.
442 checkdirs_callback(struct proc *p, void *data)
444 struct checkdirs_info *info = data;
445 struct filedesc *fdp;
446 struct namecache *ncdrop1;
447 struct namecache *ncdrop2;
448 struct vnode *vprele1;
449 struct vnode *vprele2;
451 if ((fdp = p->p_fd) != NULL) {
458 * MPUNSAFE - XXX fdp can be pulled out from under a
461 * A shared filedesc is ok, we don't have to copy it
462 * because we are making this change globally.
464 spin_lock_wr(&fdp->fd_spin);
465 if (fdp->fd_cdir == info->olddp) {
466 vprele1 = fdp->fd_cdir;
468 fdp->fd_cdir = info->newdp;
469 ncdrop1 = fdp->fd_ncdir;
470 fdp->fd_ncdir = cache_hold(info->ncp);
472 if (fdp->fd_rdir == info->olddp) {
473 vprele2 = fdp->fd_rdir;
475 fdp->fd_rdir = info->newdp;
476 ncdrop2 = fdp->fd_nrdir;
477 fdp->fd_nrdir = cache_hold(info->ncp);
479 spin_unlock_wr(&fdp->fd_spin);
493 * Unmount a file system.
495 * Note: unmount takes a path to the vnode mounted on as argument,
496 * not special file (as before).
499 * umount_args(char *path, int flags)
503 sys_unmount(struct unmount_args *uap)
505 struct thread *td = curthread;
506 struct proc *p = td->td_proc;
507 struct mount *mp = NULL;
509 struct nlookupdata nd;
512 if (p->p_ucred->cr_prison != NULL)
514 if (usermount == 0 && (error = suser(td)))
517 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
519 error = nlookup(&nd);
523 mp = nd.nl_ncp->nc_mount;
526 * Only root, or the user that did the original mount is
527 * permitted to unmount this filesystem.
529 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
534 * Don't allow unmounting the root file system.
536 if (mp->mnt_flag & MNT_ROOTFS) {
542 * Must be the root of the filesystem
544 if (! (nd.nl_ncp->nc_flag & NCF_MOUNTPT)) {
553 return (dounmount(mp, uap->flags));
557 * Do the actual file system unmount.
560 dounmount_interlock(struct mount *mp)
562 if (mp->mnt_kern_flag & MNTK_UNMOUNT)
564 mp->mnt_kern_flag |= MNTK_UNMOUNT;
569 dounmount(struct mount *mp, int flags)
571 struct vnode *coveredvp;
577 * Exclusive access for unmounting purposes
579 if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
583 * Allow filesystems to detect that a forced unmount is in progress.
585 if (flags & MNT_FORCE)
586 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
587 lflags = LK_EXCLUSIVE | ((flags & MNT_FORCE) ? 0 : LK_NOWAIT);
588 error = lockmgr(&mp->mnt_lock, lflags);
590 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
591 if (mp->mnt_kern_flag & MNTK_MWAIT)
596 if (mp->mnt_flag & MNT_EXPUBLIC)
597 vfs_setpublicfs(NULL, NULL, NULL);
599 vfs_msync(mp, MNT_WAIT);
600 async_flag = mp->mnt_flag & MNT_ASYNC;
601 mp->mnt_flag &=~ MNT_ASYNC;
602 cache_purgevfs(mp); /* remove cache entries for this file sys */
603 if (mp->mnt_syncer != NULL)
604 vrele(mp->mnt_syncer);
605 if (((mp->mnt_flag & MNT_RDONLY) ||
606 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) ||
608 error = VFS_UNMOUNT(mp, flags);
610 if (mp->mnt_syncer == NULL)
611 vfs_allocate_syncvnode(mp);
612 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
613 mp->mnt_flag |= async_flag;
614 lockmgr(&mp->mnt_lock, LK_RELEASE);
615 if (mp->mnt_kern_flag & MNTK_MWAIT)
620 * Clean up any journals still associated with the mount after
621 * filesystem activity has ceased.
623 journal_remove_all_journals(mp,
624 ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));
626 mountlist_remove(mp);
629 * Remove any installed vnode ops here so the individual VFSs don't
632 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
633 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
634 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
635 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
636 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
638 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
639 coveredvp->v_mountedhere = NULL;
641 cache_drop(mp->mnt_ncp);
644 mp->mnt_vfc->vfc_refcount--;
645 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
646 panic("unmount: dangling vnode");
647 lockmgr(&mp->mnt_lock, LK_RELEASE);
648 if (mp->mnt_kern_flag & MNTK_MWAIT)
655 * Sync each mounted filesystem.
659 static int syncprt = 0;
660 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
663 static int sync_callback(struct mount *mp, void *data);
667 sys_sync(struct sync_args *uap)
669 mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
672 * print out buffer pool stat information on each sync() call.
682 sync_callback(struct mount *mp, void *data __unused)
686 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
687 asyncflag = mp->mnt_flag & MNT_ASYNC;
688 mp->mnt_flag &= ~MNT_ASYNC;
689 vfs_msync(mp, MNT_NOWAIT);
690 VFS_SYNC(mp, MNT_NOWAIT);
691 mp->mnt_flag |= asyncflag;
696 /* XXX PRISON: could be per prison flag */
697 static int prison_quotas;
699 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
703 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
705 * Change filesystem quotas.
709 sys_quotactl(struct quotactl_args *uap)
711 struct nlookupdata nd;
719 if (p->p_ucred->cr_prison && !prison_quotas)
722 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
724 error = nlookup(&nd);
726 mp = nd.nl_ncp->nc_mount;
727 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
728 uap->arg, nd.nl_cred);
735 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
736 * void *buf, int buflen)
738 * This function operates on a mount point and executes the specified
739 * operation using the specified control data, and possibly returns data.
741 * The actual number of bytes stored in the result buffer is returned, 0
742 * if none, otherwise an error is returned.
746 sys_mountctl(struct mountctl_args *uap)
748 struct thread *td = curthread;
749 struct proc *p = td->td_proc;
757 * Sanity and permissions checks. We must be root.
760 if (p->p_ucred->cr_prison != NULL)
762 if ((error = suser(td)) != 0)
766 * Argument length checks
768 if (uap->ctllen < 0 || uap->ctllen > 1024)
770 if (uap->buflen < 0 || uap->buflen > 16 * 1024)
772 if (uap->path == NULL)
776 * Allocate the necessary buffers and copyin data
778 path = objcache_get(namei_oc, M_WAITOK);
779 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
784 ctl = malloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
785 error = copyin(uap->ctl, ctl, uap->ctllen);
790 buf = malloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
793 * Validate the descriptor
795 fp = holdfp(p->p_fd, uap->fd, -1);
802 * Execute the internal kernel function and clean up.
804 error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
807 if (error == 0 && uap->sysmsg_result > 0)
808 error = copyout(buf, uap->buf, uap->sysmsg_result);
811 objcache_put(namei_oc, path);
820 * Execute a mount control operation by resolving the path to a mount point
821 * and calling vop_mountctl().
824 kern_mountctl(const char *path, int op, struct file *fp,
825 const void *ctl, int ctllen,
826 void *buf, int buflen, int *res)
830 struct nlookupdata nd;
835 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
837 error = nlookup(&nd);
839 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
847 * Must be the root of the filesystem
849 if ((vp->v_flag & VROOT) == 0) {
853 error = vop_mountctl(mp->mnt_vn_use_ops, op, fp, ctl, ctllen,
860 kern_statfs(struct nlookupdata *nd, struct statfs *buf)
862 struct thread *td = curthread;
863 struct proc *p = td->td_proc;
866 char *fullpath, *freepath;
869 if ((error = nlookup(nd)) != 0)
871 mp = nd->nl_ncp->nc_mount;
873 if ((error = VFS_STATFS(mp, sp, nd->nl_cred)) != 0)
876 error = cache_fullpath(p, mp->mnt_ncp, &fullpath, &freepath);
879 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
880 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
881 free(freepath, M_TEMP);
883 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
884 bcopy(sp, buf, sizeof(*buf));
885 /* Only root should have access to the fsid's. */
887 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
892 * statfs_args(char *path, struct statfs *buf)
894 * Get filesystem statistics.
897 sys_statfs(struct statfs_args *uap)
899 struct nlookupdata nd;
903 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
905 error = kern_statfs(&nd, &buf);
908 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
913 kern_fstatfs(int fd, struct statfs *buf)
915 struct thread *td = curthread;
916 struct proc *p = td->td_proc;
920 char *fullpath, *freepath;
924 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
926 mp = ((struct vnode *)fp->f_data)->v_mount;
931 if (fp->f_cred == NULL) {
936 if ((error = VFS_STATFS(mp, sp, fp->f_cred)) != 0)
939 if ((error = cache_fullpath(p, mp->mnt_ncp, &fullpath, &freepath)) != 0)
941 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
942 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
943 free(freepath, M_TEMP);
945 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
946 bcopy(sp, buf, sizeof(*buf));
948 /* Only root should have access to the fsid's. */
950 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
958 * fstatfs_args(int fd, struct statfs *buf)
960 * Get filesystem statistics.
963 sys_fstatfs(struct fstatfs_args *uap)
968 error = kern_fstatfs(uap->fd, &buf);
971 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
976 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
978 * Get statistics on all filesystems.
981 struct getfsstat_info {
991 static int getfsstat_callback(struct mount *, void *);
995 sys_getfsstat(struct getfsstat_args *uap)
997 struct thread *td = curthread;
998 struct proc *p = td->td_proc;
999 struct getfsstat_info info;
1001 bzero(&info, sizeof(info));
1002 if (p != NULL && (p->p_fd->fd_nrdir->nc_flag & NCF_ROOT) == 0)
1003 info.is_chrooted = 1;
1005 info.is_chrooted = 0;
1007 info.maxcount = uap->bufsize / sizeof(struct statfs);
1008 info.sfsp = uap->buf;
1010 info.flags = uap->flags;
1013 mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
1014 if (info.sfsp && info.count > info.maxcount)
1015 uap->sysmsg_result = info.maxcount;
1017 uap->sysmsg_result = info.count;
1018 return (info.error);
1022 getfsstat_callback(struct mount *mp, void *data)
1024 struct getfsstat_info *info = data;
1030 if (info->sfsp && info->count < info->maxcount) {
1031 if (info->is_chrooted && !chroot_visible_mnt(mp, info->p))
1036 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1037 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1038 * overrides MNT_WAIT.
1040 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1041 (info->flags & MNT_WAIT)) &&
1042 (error = VFS_STATFS(mp, sp, info->p->p_ucred))) {
1045 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1047 error = cache_fullpath(info->p, mp->mnt_ncp,
1048 &fullpath, &freepath);
1050 info->error = error;
1053 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1054 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1055 free(freepath, M_TEMP);
1057 error = copyout(sp, info->sfsp, sizeof(*sp));
1059 info->error = error;
1069 * fchdir_args(int fd)
1071 * Change current working directory to a given file descriptor.
1075 sys_fchdir(struct fchdir_args *uap)
1077 struct thread *td = curthread;
1078 struct proc *p = td->td_proc;
1079 struct filedesc *fdp = p->p_fd;
1080 struct vnode *vp, *ovp;
1083 struct namecache *ncp, *oncp;
1084 struct namecache *nct;
1087 if ((error = holdvnode(fdp, uap->fd, &fp)) != 0)
1089 vp = (struct vnode *)fp->f_data;
1091 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1092 if (vp->v_type != VDIR || fp->f_ncp == NULL)
1095 error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
1101 ncp = cache_hold(fp->f_ncp);
1102 while (!error && (mp = vp->v_mountedhere) != NULL) {
1103 error = nlookup_mp(mp, &nct);
1105 cache_unlock(nct); /* leave ref intact */
1108 error = vget(vp, LK_SHARED);
1109 KKASSERT(error == 0);
1116 oncp = fdp->fd_ncdir;
1117 vn_unlock(vp); /* leave ref intact */
1119 fdp->fd_ncdir = ncp;
1131 kern_chdir(struct nlookupdata *nd)
1133 struct thread *td = curthread;
1134 struct proc *p = td->td_proc;
1135 struct filedesc *fdp = p->p_fd;
1136 struct vnode *vp, *ovp;
1137 struct namecache *oncp;
1140 if ((error = nlookup(nd)) != 0)
1142 if ((vp = nd->nl_ncp->nc_vp) == NULL)
1144 if ((error = vget(vp, LK_SHARED)) != 0)
1147 error = checkvp_chdir(vp, td);
1151 oncp = fdp->fd_ncdir;
1152 cache_unlock(nd->nl_ncp); /* leave reference intact */
1153 fdp->fd_ncdir = nd->nl_ncp;
1165 * chdir_args(char *path)
1167 * Change current working directory (``.'').
1170 sys_chdir(struct chdir_args *uap)
1172 struct nlookupdata nd;
1175 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1177 error = kern_chdir(&nd);
1183 * Helper function for raised chroot(2) security function: Refuse if
1184 * any filedescriptors are open directories.
1187 chroot_refuse_vdir_fds(fdp)
1188 struct filedesc *fdp;
1195 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
1196 if ((error = holdvnode(fdp, fd, &fp)) != 0)
1198 vp = (struct vnode *)fp->f_data;
1199 if (vp->v_type != VDIR) {
1210 * This sysctl determines if we will allow a process to chroot(2) if it
1211 * has a directory open:
1212 * 0: disallowed for all processes.
1213 * 1: allowed for processes that were not already chroot(2)'ed.
1214 * 2: allowed for all processes.
1217 static int chroot_allow_open_directories = 1;
1219 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1220 &chroot_allow_open_directories, 0, "");
1223 * chroot to the specified namecache entry. We obtain the vp from the
1224 * namecache data. The passed ncp must be locked and referenced and will
1225 * remain locked and referenced on return.
1228 kern_chroot(struct namecache *ncp)
1230 struct thread *td = curthread;
1231 struct proc *p = td->td_proc;
1232 struct filedesc *fdp = p->p_fd;
1237 * Only root can chroot
1239 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
1243 * Disallow open directory descriptors (fchdir() breakouts).
1245 if (chroot_allow_open_directories == 0 ||
1246 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1247 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1250 if ((vp = ncp->nc_vp) == NULL)
1253 if ((error = vget(vp, LK_SHARED)) != 0)
1257 * Check the validity of vp as a directory to change to and
1258 * associate it with rdir/jdir.
1260 error = checkvp_chdir(vp, td);
1261 vn_unlock(vp); /* leave reference intact */
1263 vrele(fdp->fd_rdir);
1264 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */
1265 cache_drop(fdp->fd_nrdir);
1266 fdp->fd_nrdir = cache_hold(ncp);
1267 if (fdp->fd_jdir == NULL) {
1270 fdp->fd_njdir = cache_hold(ncp);
1279 * chroot_args(char *path)
1281 * Change notion of root (``/'') directory.
1285 sys_chroot(struct chroot_args *uap)
1287 struct thread *td = curthread;
1288 struct nlookupdata nd;
1291 KKASSERT(td->td_proc);
1292 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1297 error = nlookup(&nd);
1299 error = kern_chroot(nd.nl_ncp);
1305 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1306 * determine whether it is legal to chdir to the vnode. The vnode's state
1307 * is not changed by this call.
1310 checkvp_chdir(struct vnode *vp, struct thread *td)
1314 if (vp->v_type != VDIR)
1317 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred);
1322 kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
1324 struct thread *td = curthread;
1325 struct proc *p = td->td_proc;
1326 struct lwp *lp = td->td_lwp;
1327 struct filedesc *fdp = p->p_fd;
1332 int type, indx, error;
1335 if ((oflags & O_ACCMODE) == O_ACCMODE)
1337 flags = FFLAGS(oflags);
1338 error = falloc(p, &nfp, NULL);
1342 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1345 * XXX p_dupfd is a real mess. It allows a device to return a
1346 * file descriptor to be duplicated rather then doing the open
1352 * Call vn_open() to do the lookup and assign the vnode to the
1353 * file pointer. vn_open() does not change the ref count on fp
1354 * and the vnode, on success, will be inherited by the file pointer
1357 nd->nl_flags |= NLC_LOCKVP;
1358 error = vn_open(nd, fp, flags, cmode);
1362 * handle special fdopen() case. bleh. dupfdopen() is
1363 * responsible for dropping the old contents of ofiles[indx]
1366 * Note that fsetfd() will add a ref to fp which represents
1367 * the fd_files[] assignment. We must still drop our
1370 if ((error == ENODEV || error == ENXIO) && lp->lwp_dupfd >= 0) {
1371 if (fdalloc(p, 0, &indx) == 0) {
1372 error = dupfdopen(p, indx, lp->lwp_dupfd, flags, error);
1375 fdrop(fp); /* our ref */
1378 fsetfd(p, NULL, indx);
1381 fdrop(fp); /* our ref */
1382 if (error == ERESTART)
1388 * ref the vnode for ourselves so it can't be ripped out from under
1389 * is. XXX need an ND flag to request that the vnode be returned
1392 * Reserve a file descriptor but do not assign it until the open
1395 vp = (struct vnode *)fp->f_data;
1397 if ((error = fdalloc(p, 0, &indx)) != 0) {
1404 * If no error occurs the vp will have been assigned to the file
1409 if (flags & (O_EXLOCK | O_SHLOCK)) {
1410 lf.l_whence = SEEK_SET;
1413 if (flags & O_EXLOCK)
1414 lf.l_type = F_WRLCK;
1416 lf.l_type = F_RDLCK;
1417 if (flags & FNONBLOCK)
1422 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1424 * lock request failed. Clean up the reserved
1428 fsetfd(p, NULL, indx);
1432 fp->f_flag |= FHASLOCK;
1436 * Assert that all regular file vnodes were created with a object.
1438 KASSERT(vp->v_type != VREG || vp->v_object != NULL,
1439 ("open: regular file has no backing object after vn_open"));
1445 * release our private reference, leaving the one associated with the
1446 * descriptor table intact.
1448 fsetfd(p, fp, indx);
1455 * open_args(char *path, int flags, int mode)
1457 * Check permissions, allocate an open file structure,
1458 * and call the device open routine if any.
1461 sys_open(struct open_args *uap)
1463 struct nlookupdata nd;
1466 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1468 error = kern_open(&nd, uap->flags,
1469 uap->mode, &uap->sysmsg_result);
1476 kern_mknod(struct nlookupdata *nd, int mode, int dev)
1478 struct namecache *ncp;
1479 struct thread *td = curthread;
1480 struct proc *p = td->td_proc;
1488 switch (mode & S_IFMT) {
1494 error = suser_cred(p->p_ucred, PRISON_ROOT);
1501 nd->nl_flags |= NLC_CREATE;
1502 if ((error = nlookup(nd)) != 0)
1509 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1510 vattr.va_rdev = dev;
1513 switch (mode & S_IFMT) {
1514 case S_IFMT: /* used by badsect to flag bad sectors */
1515 vattr.va_type = VBAD;
1518 vattr.va_type = VCHR;
1521 vattr.va_type = VBLK;
1532 error = VOP_NWHITEOUT(ncp, nd->nl_cred, NAMEI_CREATE);
1535 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
1544 * mknod_args(char *path, int mode, int dev)
1546 * Create a special file.
1549 sys_mknod(struct mknod_args *uap)
1551 struct nlookupdata nd;
1554 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1556 error = kern_mknod(&nd, uap->mode, uap->dev);
1562 kern_mkfifo(struct nlookupdata *nd, int mode)
1564 struct namecache *ncp;
1565 struct thread *td = curthread;
1566 struct proc *p = td->td_proc;
1573 nd->nl_flags |= NLC_CREATE;
1574 if ((error = nlookup(nd)) != 0)
1581 vattr.va_type = VFIFO;
1582 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1584 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
1591 * mkfifo_args(char *path, int mode)
1593 * Create a named pipe.
1596 sys_mkfifo(struct mkfifo_args *uap)
1598 struct nlookupdata nd;
1601 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1603 error = kern_mkfifo(&nd, uap->mode);
1608 static int hardlink_check_uid = 0;
1609 SYSCTL_INT(_kern, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1610 &hardlink_check_uid, 0,
1611 "Unprivileged processes cannot create hard links to files owned by other "
1613 static int hardlink_check_gid = 0;
1614 SYSCTL_INT(_kern, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1615 &hardlink_check_gid, 0,
1616 "Unprivileged processes cannot create hard links to files owned by other "
1620 can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
1626 * Shortcut if disabled
1628 if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
1632 * root cred can always hardlink
1634 if (suser_cred(cred, PRISON_ROOT) == 0)
1638 * Otherwise only if the originating file is owned by the
1639 * same user or group. Note that any group is allowed if
1640 * the file is owned by the caller.
1642 error = VOP_GETATTR(vp, &va);
1646 if (hardlink_check_uid) {
1647 if (cred->cr_uid != va.va_uid)
1651 if (hardlink_check_gid) {
1652 if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
1660 kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
1662 struct thread *td = curthread;
1667 * Lookup the source and obtained a locked vnode.
1669 * XXX relookup on vget failure / race ?
1672 if ((error = nlookup(nd)) != 0)
1674 vp = nd->nl_ncp->nc_vp;
1675 KKASSERT(vp != NULL);
1676 if (vp->v_type == VDIR)
1677 return (EPERM); /* POSIX */
1678 if ((error = vget(vp, LK_EXCLUSIVE)) != 0)
1682 * Unlock the source so we can lookup the target without deadlocking
1683 * (XXX vp is locked already, possible other deadlock?). The target
1686 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
1687 nd->nl_flags &= ~NLC_NCPISLOCKED;
1688 cache_unlock(nd->nl_ncp);
1690 linknd->nl_flags |= NLC_CREATE;
1691 if ((error = nlookup(linknd)) != 0) {
1695 if (linknd->nl_ncp->nc_vp) {
1701 * Finally run the new API VOP.
1703 error = can_hardlink(vp, td, td->td_proc->p_ucred);
1705 error = VOP_NLINK(linknd->nl_ncp, vp, linknd->nl_cred);
1711 * link_args(char *path, char *link)
1713 * Make a hard file link.
1716 sys_link(struct link_args *uap)
1718 struct nlookupdata nd, linknd;
1721 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1723 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
1725 error = kern_link(&nd, &linknd);
1726 nlookup_done(&linknd);
1733 kern_symlink(struct nlookupdata *nd, char *path, int mode)
1735 struct namecache *ncp;
1741 nd->nl_flags |= NLC_CREATE;
1742 if ((error = nlookup(nd)) != 0)
1749 vattr.va_mode = mode;
1750 error = VOP_NSYMLINK(ncp, &vp, nd->nl_cred, &vattr, path);
1757 * symlink(char *path, char *link)
1759 * Make a symbolic link.
1762 sys_symlink(struct symlink_args *uap)
1764 struct thread *td = curthread;
1765 struct nlookupdata nd;
1770 path = objcache_get(namei_oc, M_WAITOK);
1771 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
1773 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
1775 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
1776 error = kern_symlink(&nd, path, mode);
1780 objcache_put(namei_oc, path);
1785 * undelete_args(char *path)
1787 * Delete a whiteout from the filesystem.
1791 sys_undelete(struct undelete_args *uap)
1793 struct nlookupdata nd;
1796 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1798 nd.nl_flags |= NLC_DELETE;
1800 error = nlookup(&nd);
1802 error = VOP_NWHITEOUT(nd.nl_ncp, nd.nl_cred, NAMEI_DELETE);
1808 kern_unlink(struct nlookupdata *nd)
1810 struct namecache *ncp;
1814 nd->nl_flags |= NLC_DELETE;
1815 if ((error = nlookup(nd)) != 0)
1818 error = VOP_NREMOVE(ncp, nd->nl_cred);
1823 * unlink_args(char *path)
1825 * Delete a name from the filesystem.
1828 sys_unlink(struct unlink_args *uap)
1830 struct nlookupdata nd;
1833 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1835 error = kern_unlink(&nd);
1841 kern_lseek(int fd, off_t offset, int whence, off_t *res)
1843 struct thread *td = curthread;
1844 struct proc *p = td->td_proc;
1849 fp = holdfp(p->p_fd, fd, -1);
1852 if (fp->f_type != DTYPE_VNODE) {
1859 fp->f_offset += offset;
1863 error = VOP_GETATTR((struct vnode *)fp->f_data, &vattr);
1865 fp->f_offset = offset + vattr.va_size;
1868 fp->f_offset = offset;
1875 *res = fp->f_offset;
1882 * lseek_args(int fd, int pad, off_t offset, int whence)
1884 * Reposition read/write file offset.
1887 sys_lseek(struct lseek_args *uap)
1891 error = kern_lseek(uap->fd, uap->offset, uap->whence,
1892 &uap->sysmsg_offset);
1898 kern_access(struct nlookupdata *nd, int aflags)
1903 if ((error = nlookup(nd)) != 0)
1906 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
1910 /* Flags == 0 means only check for existence. */
1919 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1920 error = VOP_ACCESS(vp, flags, nd->nl_cred);
1923 * If the file handle is stale we have to re-resolve the
1924 * entry. This is a hack at the moment.
1926 if (error == ESTALE) {
1927 cache_setunresolved(nd->nl_ncp);
1928 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1941 * access_args(char *path, int flags)
1943 * Check access permissions.
1946 sys_access(struct access_args *uap)
1948 struct nlookupdata nd;
1951 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1953 error = kern_access(&nd, uap->flags);
1959 kern_stat(struct nlookupdata *nd, struct stat *st)
1965 if ((error = nlookup(nd)) != 0)
1968 if ((vp = nd->nl_ncp->nc_vp) == NULL)
1972 if ((error = vget(vp, LK_SHARED)) != 0)
1974 error = vn_stat(vp, st, nd->nl_cred);
1977 * If the file handle is stale we have to re-resolve the entry. This
1978 * is a hack at the moment.
1980 if (error == ESTALE) {
1981 cache_setunresolved(nd->nl_ncp);
1982 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1993 * stat_args(char *path, struct stat *ub)
1995 * Get file status; this version follows links.
1998 sys_stat(struct stat_args *uap)
2000 struct nlookupdata nd;
2004 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2006 error = kern_stat(&nd, &st);
2008 error = copyout(&st, uap->ub, sizeof(*uap->ub));
2015 * lstat_args(char *path, struct stat *ub)
2017 * Get file status; this version does not follow links.
2020 sys_lstat(struct lstat_args *uap)
2022 struct nlookupdata nd;
2026 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2028 error = kern_stat(&nd, &st);
2030 error = copyout(&st, uap->ub, sizeof(*uap->ub));
2037 * pathconf_Args(char *path, int name)
2039 * Get configurable pathname variables.
2043 sys_pathconf(struct pathconf_args *uap)
2045 struct nlookupdata nd;
2050 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2052 error = nlookup(&nd);
2054 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
2057 error = VOP_PATHCONF(vp, uap->name, uap->sysmsg_fds);
2065 * kern_readlink isn't properly split yet. There is a copyin burried
2066 * in VOP_READLINK().
2069 kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
2071 struct thread *td = curthread;
2072 struct proc *p = td->td_proc;
2078 if ((error = nlookup(nd)) != 0)
2080 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
2083 if (vp->v_type != VLNK) {
2086 aiov.iov_base = buf;
2087 aiov.iov_len = count;
2088 auio.uio_iov = &aiov;
2089 auio.uio_iovcnt = 1;
2090 auio.uio_offset = 0;
2091 auio.uio_rw = UIO_READ;
2092 auio.uio_segflg = UIO_USERSPACE;
2094 auio.uio_resid = count;
2095 error = VOP_READLINK(vp, &auio, p->p_ucred);
2098 *res = count - auio.uio_resid;
2103 * readlink_args(char *path, char *buf, int count)
2105 * Return target name of a symbolic link.
2108 sys_readlink(struct readlink_args *uap)
2110 struct nlookupdata nd;
2113 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2115 error = kern_readlink(&nd, uap->buf, uap->count,
2116 &uap->sysmsg_result);
2123 setfflags(struct vnode *vp, int flags)
2125 struct thread *td = curthread;
2126 struct proc *p = td->td_proc;
2131 * Prevent non-root users from setting flags on devices. When
2132 * a device is reused, users can retain ownership of the device
2133 * if they are allowed to set flags and programs assume that
2134 * chown can't fail when done as root.
2136 if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
2137 ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
2141 * note: vget is required for any operation that might mod the vnode
2142 * so VINACTIVE is properly cleared.
2144 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2146 vattr.va_flags = flags;
2147 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2154 * chflags(char *path, int flags)
2156 * Change flags of a file given a path name.
2160 sys_chflags(struct chflags_args *uap)
2162 struct nlookupdata nd;
2167 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2168 /* XXX Add NLC flag indicating modifying operation? */
2170 error = nlookup(&nd);
2172 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
2175 error = setfflags(vp, uap->flags);
2182 * fchflags_args(int fd, int flags)
2184 * Change flags of a file given a file descriptor.
2188 sys_fchflags(struct fchflags_args *uap)
2190 struct thread *td = curthread;
2191 struct proc *p = td->td_proc;
2195 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2197 error = setfflags((struct vnode *) fp->f_data, uap->flags);
2203 setfmode(struct vnode *vp, int mode)
2205 struct thread *td = curthread;
2206 struct proc *p = td->td_proc;
2211 * note: vget is required for any operation that might mod the vnode
2212 * so VINACTIVE is properly cleared.
2214 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2216 vattr.va_mode = mode & ALLPERMS;
2217 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2224 kern_chmod(struct nlookupdata *nd, int mode)
2229 /* XXX Add NLC flag indicating modifying operation? */
2230 if ((error = nlookup(nd)) != 0)
2232 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2234 error = setfmode(vp, mode);
2240 * chmod_args(char *path, int mode)
2242 * Change mode of a file given path name.
2246 sys_chmod(struct chmod_args *uap)
2248 struct nlookupdata nd;
2251 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2253 error = kern_chmod(&nd, uap->mode);
2259 * lchmod_args(char *path, int mode)
2261 * Change mode of a file given path name (don't follow links.)
2265 sys_lchmod(struct lchmod_args *uap)
2267 struct nlookupdata nd;
2270 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2272 error = kern_chmod(&nd, uap->mode);
2278 * fchmod_args(int fd, int mode)
2280 * Change mode of a file given a file descriptor.
2284 sys_fchmod(struct fchmod_args *uap)
2286 struct thread *td = curthread;
2287 struct proc *p = td->td_proc;
2291 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2293 error = setfmode((struct vnode *)fp->f_data, uap->mode);
2299 setfown(struct vnode *vp, uid_t uid, gid_t gid)
2301 struct thread *td = curthread;
2302 struct proc *p = td->td_proc;
2307 * note: vget is required for any operation that might mod the vnode
2308 * so VINACTIVE is properly cleared.
2310 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2314 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2321 kern_chown(struct nlookupdata *nd, int uid, int gid)
2326 /* XXX Add NLC flag indicating modifying operation? */
2327 if ((error = nlookup(nd)) != 0)
2329 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2331 error = setfown(vp, uid, gid);
2337 * chown(char *path, int uid, int gid)
2339 * Set ownership given a path name.
2342 sys_chown(struct chown_args *uap)
2344 struct nlookupdata nd;
2347 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2349 error = kern_chown(&nd, uap->uid, uap->gid);
2355 * lchown_args(char *path, int uid, int gid)
2357 * Set ownership given a path name, do not cross symlinks.
2360 sys_lchown(struct lchown_args *uap)
2362 struct nlookupdata nd;
2365 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2367 error = kern_chown(&nd, uap->uid, uap->gid);
2373 * fchown_args(int fd, int uid, int gid)
2375 * Set ownership given a file descriptor.
2379 sys_fchown(struct fchown_args *uap)
2381 struct thread *td = curthread;
2382 struct proc *p = td->td_proc;
2386 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2388 error = setfown((struct vnode *)fp->f_data, uap->uid, uap->gid);
2394 getutimes(const struct timeval *tvp, struct timespec *tsp)
2396 struct timeval tv[2];
2400 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2403 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2404 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2410 setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
2412 struct thread *td = curthread;
2413 struct proc *p = td->td_proc;
2418 * note: vget is required for any operation that might mod the vnode
2419 * so VINACTIVE is properly cleared.
2421 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2423 vattr.va_atime = ts[0];
2424 vattr.va_mtime = ts[1];
2426 vattr.va_vaflags |= VA_UTIMES_NULL;
2427 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2434 kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
2436 struct timespec ts[2];
2440 if ((error = getutimes(tptr, ts)) != 0)
2442 /* XXX Add NLC flag indicating modifying operation? */
2443 if ((error = nlookup(nd)) != 0)
2445 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2447 error = setutimes(vp, ts, tptr == NULL);
2453 * utimes_args(char *path, struct timeval *tptr)
2455 * Set the access and modification times of a file.
2458 sys_utimes(struct utimes_args *uap)
2460 struct timeval tv[2];
2461 struct nlookupdata nd;
2465 error = copyin(uap->tptr, tv, sizeof(tv));
2469 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2471 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2477 * lutimes_args(char *path, struct timeval *tptr)
2479 * Set the access and modification times of a file.
2482 sys_lutimes(struct lutimes_args *uap)
2484 struct timeval tv[2];
2485 struct nlookupdata nd;
2489 error = copyin(uap->tptr, tv, sizeof(tv));
2493 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2495 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2501 kern_futimes(int fd, struct timeval *tptr)
2503 struct thread *td = curthread;
2504 struct proc *p = td->td_proc;
2505 struct timespec ts[2];
2509 error = getutimes(tptr, ts);
2512 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
2514 error = setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
2520 * futimes_args(int fd, struct timeval *tptr)
2522 * Set the access and modification times of a file.
2525 sys_futimes(struct futimes_args *uap)
2527 struct timeval tv[2];
2531 error = copyin(uap->tptr, tv, sizeof(tv));
2536 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
2542 kern_truncate(struct nlookupdata *nd, off_t length)
2550 /* XXX Add NLC flag indicating modifying operation? */
2551 if ((error = nlookup(nd)) != 0)
2553 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2555 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) != 0) {
2559 if (vp->v_type == VDIR) {
2561 } else if ((error = vn_writechk(vp)) == 0 &&
2562 (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred)) == 0) {
2564 vattr.va_size = length;
2565 error = VOP_SETATTR(vp, &vattr, nd->nl_cred);
2572 * truncate(char *path, int pad, off_t length)
2574 * Truncate a file given its path name.
2577 sys_truncate(struct truncate_args *uap)
2579 struct nlookupdata nd;
2582 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2584 error = kern_truncate(&nd, uap->length);
2590 kern_ftruncate(int fd, off_t length)
2592 struct thread *td = curthread;
2593 struct proc *p = td->td_proc;
2601 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
2603 if ((fp->f_flag & FWRITE) == 0) {
2607 vp = (struct vnode *)fp->f_data;
2608 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2609 if (vp->v_type == VDIR) {
2611 } else if ((error = vn_writechk(vp)) == 0) {
2613 vattr.va_size = length;
2614 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
2623 * ftruncate_args(int fd, int pad, off_t length)
2625 * Truncate a file given a file descriptor.
2628 sys_ftruncate(struct ftruncate_args *uap)
2632 error = kern_ftruncate(uap->fd, uap->length);
2640 * Sync an open file.
2644 sys_fsync(struct fsync_args *uap)
2646 struct thread *td = curthread;
2647 struct proc *p = td->td_proc;
2653 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2655 vp = (struct vnode *)fp->f_data;
2656 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2657 if ((obj = vp->v_object) != NULL)
2658 vm_object_page_clean(obj, 0, 0, 0);
2659 if ((error = VOP_FSYNC(vp, MNT_WAIT)) == 0 &&
2660 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) &&
2662 error = (*bioops.io_fsync)(vp);
2670 kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
2672 struct namecache *fncpd;
2673 struct namecache *tncpd;
2674 struct namecache *ncp;
2679 if ((error = nlookup(fromnd)) != 0)
2681 if ((fncpd = fromnd->nl_ncp->nc_parent) == NULL)
2686 * unlock the source ncp so we can lookup the target ncp without
2687 * deadlocking. The target may or may not exist so we do not check
2688 * for a target vp like kern_mkdir() and other creation functions do.
2690 * The source and target directories are ref'd and rechecked after
2691 * everything is relocked to determine if the source or target file
2694 KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
2695 fromnd->nl_flags &= ~NLC_NCPISLOCKED;
2696 cache_unlock(fromnd->nl_ncp);
2698 tond->nl_flags |= NLC_CREATE;
2699 if ((error = nlookup(tond)) != 0) {
2703 if ((tncpd = tond->nl_ncp->nc_parent) == NULL) {
2710 * If the source and target are the same there is nothing to do
2712 if (fromnd->nl_ncp == tond->nl_ncp) {
2719 * relock the source ncp. NOTE AFTER RELOCKING: the source ncp
2720 * may have become invalid while it was unlocked, nc_vp and nc_mount
2723 if (cache_lock_nonblock(fromnd->nl_ncp) == 0) {
2724 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2725 } else if (fromnd->nl_ncp > tond->nl_ncp) {
2726 cache_lock(fromnd->nl_ncp);
2727 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2729 cache_unlock(tond->nl_ncp);
2730 cache_lock(fromnd->nl_ncp);
2731 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2732 cache_lock(tond->nl_ncp);
2733 cache_resolve(tond->nl_ncp, tond->nl_cred);
2735 fromnd->nl_flags |= NLC_NCPISLOCKED;
2738 * make sure the parent directories linkages are the same
2740 if (fncpd != fromnd->nl_ncp->nc_parent ||
2741 tncpd != tond->nl_ncp->nc_parent) {
2748 * Both the source and target must be within the same filesystem and
2749 * in the same filesystem as their parent directories within the
2750 * namecache topology.
2752 * NOTE: fromnd's nc_mount or nc_vp could be NULL.
2754 mp = fncpd->nc_mount;
2755 if (mp != tncpd->nc_mount || mp != fromnd->nl_ncp->nc_mount ||
2756 mp != tond->nl_ncp->nc_mount) {
2763 * If the target exists and either the source or target is a directory,
2764 * then both must be directories.
2766 * Due to relocking of the source, fromnd->nl_ncp->nc_vp might have
2769 if (tond->nl_ncp->nc_vp) {
2770 if (fromnd->nl_ncp->nc_vp == NULL) {
2772 } else if (fromnd->nl_ncp->nc_vp->v_type == VDIR) {
2773 if (tond->nl_ncp->nc_vp->v_type != VDIR)
2775 } else if (tond->nl_ncp->nc_vp->v_type == VDIR) {
2781 * You cannot rename a source into itself or a subdirectory of itself.
2782 * We check this by travsersing the target directory upwards looking
2783 * for a match against the source.
2786 for (ncp = tncpd; ncp; ncp = ncp->nc_parent) {
2787 if (fromnd->nl_ncp == ncp) {
2798 * Even though the namespaces are different, they may still represent
2799 * hardlinks to the same file. The filesystem might have a hard time
2800 * with this so we issue a NREMOVE of the source instead of a NRENAME
2801 * when we detect the situation.
2804 if (fromnd->nl_ncp->nc_vp == tond->nl_ncp->nc_vp) {
2805 error = VOP_NREMOVE(fromnd->nl_ncp, fromnd->nl_cred);
2807 error = VOP_NRENAME(fromnd->nl_ncp, tond->nl_ncp,
2815 * rename_args(char *from, char *to)
2817 * Rename files. Source and destination must either both be directories,
2818 * or both not be directories. If target is a directory, it must be empty.
2821 sys_rename(struct rename_args *uap)
2823 struct nlookupdata fromnd, tond;
2826 error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
2828 error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
2830 error = kern_rename(&fromnd, &tond);
2831 nlookup_done(&tond);
2833 nlookup_done(&fromnd);
2838 kern_mkdir(struct nlookupdata *nd, int mode)
2840 struct thread *td = curthread;
2841 struct proc *p = td->td_proc;
2842 struct namecache *ncp;
2848 nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE;
2849 if ((error = nlookup(nd)) != 0)
2857 vattr.va_type = VDIR;
2858 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
2861 error = VOP_NMKDIR(ncp, &vp, p->p_ucred, &vattr);
2868 * mkdir_args(char *path, int mode)
2870 * Make a directory file.
2874 sys_mkdir(struct mkdir_args *uap)
2876 struct nlookupdata nd;
2879 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2881 error = kern_mkdir(&nd, uap->mode);
2887 kern_rmdir(struct nlookupdata *nd)
2889 struct namecache *ncp;
2893 nd->nl_flags |= NLC_DELETE;
2894 if ((error = nlookup(nd)) != 0)
2898 error = VOP_NRMDIR(ncp, nd->nl_cred);
2903 * rmdir_args(char *path)
2905 * Remove a directory file.
2909 sys_rmdir(struct rmdir_args *uap)
2911 struct nlookupdata nd;
2914 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2916 error = kern_rmdir(&nd);
2922 kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res,
2923 enum uio_seg direction)
2925 struct thread *td = curthread;
2926 struct proc *p = td->td_proc;
2934 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
2936 if ((fp->f_flag & FREAD) == 0) {
2940 vp = (struct vnode *)fp->f_data;
2942 if (vp->v_type != VDIR) {
2946 aiov.iov_base = buf;
2947 aiov.iov_len = count;
2948 auio.uio_iov = &aiov;
2949 auio.uio_iovcnt = 1;
2950 auio.uio_rw = UIO_READ;
2951 auio.uio_segflg = direction;
2953 auio.uio_resid = count;
2954 loff = auio.uio_offset = fp->f_offset;
2955 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
2956 fp->f_offset = auio.uio_offset;
2959 if (count == auio.uio_resid) {
2960 if (union_dircheckp) {
2961 error = union_dircheckp(td, &vp, fp);
2967 if ((vp->v_flag & VROOT) &&
2968 (vp->v_mount->mnt_flag & MNT_UNION)) {
2969 struct vnode *tvp = vp;
2970 vp = vp->v_mount->mnt_vnodecovered;
2981 *res = count - auio.uio_resid;
2988 * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
2990 * Read a block of directory entries in a file system independent format.
2993 sys_getdirentries(struct getdirentries_args *uap)
2998 error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base,
2999 &uap->sysmsg_result, UIO_USERSPACE);
3002 error = copyout(&base, uap->basep, sizeof(*uap->basep));
3007 * getdents_args(int fd, char *buf, size_t count)
3010 sys_getdents(struct getdents_args *uap)
3014 error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL,
3015 &uap->sysmsg_result, UIO_USERSPACE);
3021 * umask(int newmask)
3023 * Set the mode mask for creation of filesystem nodes.
3028 sys_umask(struct umask_args *uap)
3030 struct thread *td = curthread;
3031 struct proc *p = td->td_proc;
3032 struct filedesc *fdp;
3035 uap->sysmsg_result = fdp->fd_cmask;
3036 fdp->fd_cmask = uap->newmask & ALLPERMS;
3041 * revoke(char *path)
3043 * Void all references to file by ripping underlying filesystem
3048 sys_revoke(struct revoke_args *uap)
3050 struct nlookupdata nd;
3057 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3059 error = nlookup(&nd);
3061 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
3062 cred = crhold(nd.nl_cred);
3065 if (vp->v_type != VCHR && vp->v_type != VBLK)
3068 error = VOP_GETATTR(vp, &vattr);
3069 if (error == 0 && cred->cr_uid != vattr.va_uid)
3070 error = suser_cred(cred, PRISON_ROOT);
3071 if (error == 0 && count_udev(vp->v_udev) > 0) {
3074 VOP_REVOKE(vp, REVOKEALL);
3085 * getfh_args(char *fname, fhandle_t *fhp)
3087 * Get (NFS) file handle
3090 sys_getfh(struct getfh_args *uap)
3092 struct thread *td = curthread;
3093 struct nlookupdata nd;
3099 * Must be super user
3101 if ((error = suser(td)) != 0)
3105 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
3107 error = nlookup(&nd);
3109 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3112 bzero(&fh, sizeof(fh));
3113 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3114 error = VFS_VPTOFH(vp, &fh.fh_fid);
3117 error = copyout(&fh, uap->fhp, sizeof(fh));
3123 * fhopen_args(const struct fhandle *u_fhp, int flags)
3125 * syscall for the rpc.lockd to use to translate a NFS file handle into
3126 * an open descriptor.
3128 * warning: do not remove the suser() call or this becomes one giant
3132 sys_fhopen(struct fhopen_args *uap)
3134 struct thread *td = curthread;
3135 struct proc *p = td->td_proc;
3140 struct vattr *vap = &vat;
3142 int fmode, mode, error, type;
3148 * Must be super user
3154 fmode = FFLAGS(uap->flags);
3155 /* why not allow a non-read/write open for our lockd? */
3156 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3158 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
3161 /* find the mount point */
3162 mp = vfs_getvfs(&fhp.fh_fsid);
3165 /* now give me my vnode, it gets returned to me locked */
3166 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
3170 * from now on we have to make sure not
3171 * to forget about the vnode
3172 * any error that causes an abort must vput(vp)
3173 * just set error = err and 'goto bad;'.
3179 if (vp->v_type == VLNK) {
3183 if (vp->v_type == VSOCK) {
3188 if (fmode & (FWRITE | O_TRUNC)) {
3189 if (vp->v_type == VDIR) {
3193 error = vn_writechk(vp);
3201 error = VOP_ACCESS(vp, mode, p->p_ucred);
3205 if (fmode & O_TRUNC) {
3206 vn_unlock(vp); /* XXX */
3207 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
3210 error = VOP_SETATTR(vp, vap, p->p_ucred);
3216 * VOP_OPEN needs the file pointer so it can potentially override
3219 * WARNING! no f_ncp will be associated when fhopen()ing a directory.
3222 if ((error = falloc(p, &nfp, &indx)) != 0)
3226 error = VOP_OPEN(vp, fmode, p->p_ucred, fp);
3229 * setting f_ops this way prevents VOP_CLOSE from being
3230 * called or fdrop() releasing the vp from v_data. Since
3231 * the VOP_OPEN failed we don't want to VOP_CLOSE.
3233 fp->f_ops = &badfileops;
3239 * The fp is given its own reference, we still have our ref and lock.
3241 * Assert that all regular files must be created with a VM object.
3243 if (vp->v_type == VREG && vp->v_object == NULL) {
3244 printf("fhopen: regular file did not have VM object: %p\n", vp);
3249 * The open was successful. Handle any locking requirements.
3251 if (fmode & (O_EXLOCK | O_SHLOCK)) {
3252 lf.l_whence = SEEK_SET;
3255 if (fmode & O_EXLOCK)
3256 lf.l_type = F_WRLCK;
3258 lf.l_type = F_RDLCK;
3259 if (fmode & FNONBLOCK)
3264 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
3266 * release our private reference.
3268 fsetfd(p, NULL, indx);
3273 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3274 fp->f_flag |= FHASLOCK;
3278 * Clean up. Associate the file pointer with the previously
3279 * reserved descriptor and return it.
3282 fsetfd(p, fp, indx);
3284 uap->sysmsg_result = indx;
3288 fsetfd(p, NULL, indx);
3296 * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
3299 sys_fhstat(struct fhstat_args *uap)
3301 struct thread *td = curthread;
3309 * Must be super user
3315 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
3319 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3321 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3323 error = vn_stat(vp, &sb, td->td_proc->p_ucred);
3327 error = copyout(&sb, uap->sb, sizeof(sb));
3332 * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
3335 sys_fhstatfs(struct fhstatfs_args *uap)
3337 struct thread *td = curthread;
3338 struct proc *p = td->td_proc;
3343 char *fullpath, *freepath;
3348 * Must be super user
3350 if ((error = suser(td)))
3353 if ((error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t))) != 0)
3356 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3359 if (p != NULL && (p->p_fd->fd_nrdir->nc_flag & NCF_ROOT) == 0 &&
3360 !chroot_visible_mnt(mp, p))
3363 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3368 if ((error = VFS_STATFS(mp, sp, p->p_ucred)) != 0)
3371 error = cache_fullpath(p, mp->mnt_ncp, &fullpath, &freepath);
3374 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3375 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
3376 free(freepath, M_TEMP);
3378 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
3380 bcopy(sp, &sb, sizeof(sb));
3381 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
3384 return (copyout(sp, uap->buf, sizeof(*sp)));
3388 * Syscall to push extended attribute configuration information into the
3389 * VFS. Accepts a path, which it converts to a mountpoint, as well as
3390 * a command (int cmd), and attribute name and misc data. For now, the
3391 * attribute name is left in userspace for consumption by the VFS_op.
3392 * It will probably be changed to be copied into sysspace by the
3393 * syscall in the future, once issues with various consumers of the
3394 * attribute code have raised their hands.
3396 * Currently this is used only by UFS Extended Attributes.
3399 sys_extattrctl(struct extattrctl_args *uap)
3401 struct nlookupdata nd;
3407 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3409 error = nlookup(&nd);
3411 mp = nd.nl_ncp->nc_mount;
3412 error = VFS_EXTATTRCTL(mp, uap->cmd,
3413 uap->attrname, uap->arg,
3421 * Syscall to set a named extended attribute on a file or directory.
3422 * Accepts attribute name, and a uio structure pointing to the data to set.
3423 * The uio is consumed in the style of writev(). The real work happens
3424 * in VOP_SETEXTATTR().
3427 sys_extattr_set_file(struct extattr_set_file_args *uap)
3429 char attrname[EXTATTR_MAXNAMELEN];
3430 struct iovec aiov[UIO_SMALLIOV];
3431 struct iovec *needfree;
3432 struct nlookupdata nd;
3441 error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
3446 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3448 error = nlookup(&nd);
3450 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3457 iovlen = uap->iovcnt * sizeof(struct iovec);
3458 if (uap->iovcnt > UIO_SMALLIOV) {
3459 if (uap->iovcnt > UIO_MAXIOV) {
3463 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3469 auio.uio_iovcnt = uap->iovcnt;
3470 auio.uio_rw = UIO_WRITE;
3471 auio.uio_segflg = UIO_USERSPACE;
3472 auio.uio_td = nd.nl_td;
3473 auio.uio_offset = 0;
3474 if ((error = copyin(uap->iovp, iov, iovlen)))
3477 for (i = 0; i < uap->iovcnt; i++) {
3478 if (iov->iov_len > INT_MAX - auio.uio_resid) {
3482 auio.uio_resid += iov->iov_len;
3485 cnt = auio.uio_resid;
3486 error = VOP_SETEXTATTR(vp, attrname, &auio, nd.nl_cred);
3487 cnt -= auio.uio_resid;
3488 uap->sysmsg_result = cnt;
3493 FREE(needfree, M_IOV);
3498 * Syscall to get a named extended attribute on a file or directory.
3499 * Accepts attribute name, and a uio structure pointing to a buffer for the
3500 * data. The uio is consumed in the style of readv(). The real work
3501 * happens in VOP_GETEXTATTR();
3504 sys_extattr_get_file(struct extattr_get_file_args *uap)
3506 char attrname[EXTATTR_MAXNAMELEN];
3507 struct iovec aiov[UIO_SMALLIOV];
3508 struct iovec *needfree;
3509 struct nlookupdata nd;
3518 error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
3523 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3525 error = nlookup(&nd);
3527 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3533 iovlen = uap->iovcnt * sizeof (struct iovec);
3535 if (uap->iovcnt > UIO_SMALLIOV) {
3536 if (uap->iovcnt > UIO_MAXIOV) {
3540 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3546 auio.uio_iovcnt = uap->iovcnt;
3547 auio.uio_rw = UIO_READ;
3548 auio.uio_segflg = UIO_USERSPACE;
3549 auio.uio_td = nd.nl_td;
3550 auio.uio_offset = 0;
3551 if ((error = copyin(uap->iovp, iov, iovlen)))
3554 for (i = 0; i < uap->iovcnt; i++) {
3555 if (iov->iov_len > INT_MAX - auio.uio_resid) {
3559 auio.uio_resid += iov->iov_len;
3562 cnt = auio.uio_resid;
3563 error = VOP_GETEXTATTR(vp, attrname, &auio, nd.nl_cred);
3564 cnt -= auio.uio_resid;
3565 uap->sysmsg_result = cnt;
3570 FREE(needfree, M_IOV);
3575 * Syscall to delete a named extended attribute from a file or directory.
3576 * Accepts attribute name. The real work happens in VOP_SETEXTATTR().
3579 sys_extattr_delete_file(struct extattr_delete_file_args *uap)
3581 char attrname[EXTATTR_MAXNAMELEN];
3582 struct nlookupdata nd;
3586 error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
3591 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3593 error = nlookup(&nd);
3595 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3601 error = VOP_SETEXTATTR(vp, attrname, NULL, nd.nl_cred);
3608 chroot_visible_mnt(struct mount *mp, struct proc *p)
3610 struct namecache *ncp;
3612 * First check if this file system is below
3616 while (ncp != NULL && ncp != p->p_fd->fd_nrdir)
3617 ncp = ncp->nc_parent;
3620 * This is not below the chroot path.
3622 * Check if the chroot path is on the same filesystem,
3623 * by determing if we have to cross a mount point
3624 * before reaching mp->mnt_ncp.
3626 ncp = p->p_fd->fd_nrdir;
3627 while (ncp != NULL && ncp != mp->mnt_ncp) {
3628 if (ncp->nc_flag & NCF_MOUNTPT) {
3632 ncp = ncp->nc_parent;
3635 return(ncp != NULL);