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 1.51 2004/12/28 04:39:59 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/sysproto.h>
51 #include <sys/filedesc.h>
52 #include <sys/kernel.h>
53 #include <sys/fcntl.h>
55 #include <sys/linker.h>
57 #include <sys/unistd.h>
58 #include <sys/vnode.h>
60 #include <sys/namei.h>
61 #include <sys/nlookup.h>
62 #include <sys/dirent.h>
63 #include <sys/extattr.h>
64 #include <sys/kern_syscall.h>
66 #include <machine/limits.h>
67 #include <vfs/union/union.h>
68 #include <sys/sysctl.h>
70 #include <vm/vm_object.h>
71 #include <vm/vm_zone.h>
72 #include <vm/vm_page.h>
74 #include <sys/file2.h>
76 static int checkvp_chdir (struct vnode *vn, struct thread *td);
77 static void checkdirs (struct vnode *olddp, struct namecache *ncp);
78 static int chroot_refuse_vdir_fds (struct filedesc *fdp);
79 static int getutimes (const struct timeval *, struct timespec *);
80 static int setfown (struct vnode *, uid_t, gid_t);
81 static int setfmode (struct vnode *, int);
82 static int setfflags (struct vnode *, int);
83 static int setutimes (struct vnode *, const struct timespec *, int);
84 static int usermount = 0; /* if 1, non-root can mount fs. */
86 int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
88 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
91 * Virtual File System System Calls
95 * Mount a file system.
98 * mount_args(char *type, char *path, int flags, caddr_t data)
102 mount(struct mount_args *uap)
104 struct thread *td = curthread;
105 struct proc *p = td->td_proc;
107 struct namecache *ncp;
109 struct vfsconf *vfsp;
110 int error, flag = 0, flag2 = 0;
112 struct nlookupdata nd;
113 char fstypename[MFSNAMELEN];
115 struct nlcomponent nlc;
118 if (p->p_ucred->cr_prison != NULL)
120 if (usermount == 0 && (error = suser(td)))
123 * Do not allow NFS export by non-root users.
125 if (SCARG(uap, flags) & MNT_EXPORTED) {
131 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
134 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
137 * Lookup the requested path and extract the ncp and vnode.
139 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
141 if ((error = nlookup(&nd)) == 0) {
142 if (nd.nl_ncp->nc_vp == NULL)
152 * Extract the locked+refd ncp and cleanup the nd structure
159 * now we have the locked ref'd ncp and unreferenced vnode.
162 if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0) {
169 * Now we have an unlocked ref'd ncp and a locked ref'd vp
171 if (SCARG(uap, flags) & MNT_UPDATE) {
172 if ((vp->v_flag & VROOT) == 0) {
179 flag2 = mp->mnt_kern_flag;
181 * We only allow the filesystem to be reloaded if it
182 * is currently mounted read-only.
184 if ((SCARG(uap, flags) & MNT_RELOAD) &&
185 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
188 return (EOPNOTSUPP); /* Needs translation */
191 * Only root, or the user that did the original mount is
192 * permitted to update it.
194 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
195 (error = suser(td))) {
200 if (vfs_busy(mp, LK_NOWAIT, NULL, td)) {
205 if ((vp->v_flag & VMOUNT) != 0 ||
206 vp->v_mountedhere != NULL) {
212 vp->v_flag |= VMOUNT;
214 SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
215 VOP_UNLOCK(vp, 0, td);
219 * If the user is not root, ensure that they own the directory
220 * onto which we are attempting to mount.
222 if ((error = VOP_GETATTR(vp, &va, td)) ||
223 (va.va_uid != p->p_ucred->cr_uid &&
224 (error = suser(td)))) {
229 if ((error = vinvalbuf(vp, V_SAVE, td, 0, 0)) != 0) {
234 if (vp->v_type != VDIR) {
239 if ((error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL)) != 0) {
244 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
245 if (!strcmp(vfsp->vfc_name, fstypename))
251 /* Only load modules for root (very important!) */
252 if ((error = suser(td)) != 0) {
257 error = linker_load_file(fstypename, &lf);
258 if (error || lf == NULL) {
266 /* lookup again, see if the VFS was loaded */
267 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
268 if (!strcmp(vfsp->vfc_name, fstypename))
273 linker_file_unload(lf);
279 if ((vp->v_flag & VMOUNT) != 0 ||
280 vp->v_mountedhere != NULL) {
285 vp->v_flag |= VMOUNT;
288 * Allocate and initialize the filesystem.
290 mp = malloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
291 TAILQ_INIT(&mp->mnt_nvnodelist);
292 TAILQ_INIT(&mp->mnt_reservedvnlist);
293 mp->mnt_nvnodelistsize = 0;
294 lockinit(&mp->mnt_lock, 0, "vfslock", 0, LK_NOPAUSE);
295 vfs_busy(mp, LK_NOWAIT, NULL, td);
296 mp->mnt_op = vfsp->vfc_vfsops;
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;
305 VOP_UNLOCK(vp, 0, td);
308 * Set the mount level flags.
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);
323 * Mount the filesystem.
324 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
327 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), td);
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;
335 mp->mnt_kern_flag = flag2;
338 vp->v_flag &= ~VMOUNT;
343 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
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.
351 * It is not necessary to invalidate or purge the vnode underneath
352 * because elements under the mount will be given their own glue
356 nlc.nlc_nameptr = "";
358 mp->mnt_ncp = cache_nlookup(ncp, &nlc);
359 mp->mnt_ncp->nc_flag |= NCF_MOUNTPT;
360 mp->mnt_ncp->nc_mount = mp;
362 /* XXX get the root of the fs and cache_setvp(mnt_ncp...) */
363 vp->v_flag &= ~VMOUNT;
364 vp->v_mountedhere = mp;
365 lwkt_gettoken(&ilock, &mountlist_token);
366 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
367 lwkt_reltoken(&ilock);
368 checkdirs(vp, mp->mnt_ncp);
369 cache_unlock(mp->mnt_ncp); /* leave ref intact */
370 VOP_UNLOCK(vp, 0, td);
371 error = vfs_allocate_syncvnode(mp);
373 if ((error = VFS_START(mp, 0, td)) != 0)
376 vfs_rm_vnodeops(&mp->mnt_vn_coherency_ops);
377 vfs_rm_vnodeops(&mp->mnt_vn_journal_ops);
378 vfs_rm_vnodeops(&mp->mnt_vn_norm_ops);
379 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
380 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
381 vp->v_flag &= ~VMOUNT;
382 mp->mnt_vfc->vfc_refcount--;
392 * Scan all active processes to see if any of them have a current
393 * or root directory onto which the new filesystem has just been
394 * mounted. If so, replace them with the new mount point.
396 * The passed ncp is ref'd and locked (from the mount code) and
397 * must be associated with the vnode representing the root of the
401 checkdirs(struct vnode *olddp, struct namecache *ncp)
403 struct filedesc *fdp;
408 if (olddp->v_usecount == 1)
410 mp = olddp->v_mountedhere;
411 if (VFS_ROOT(mp, &newdp))
412 panic("mount: lost mount");
413 cache_setvp(ncp, newdp);
415 if (rootvnode == olddp) {
417 vfs_cache_setroot(newdp, cache_hold(ncp));
420 FOREACH_PROC_IN_SYSTEM(p) {
422 if (fdp->fd_cdir == olddp) {
425 fdp->fd_cdir = newdp;
426 cache_drop(fdp->fd_ncdir);
427 fdp->fd_ncdir = cache_hold(ncp);
429 if (fdp->fd_rdir == olddp) {
432 fdp->fd_rdir = newdp;
433 cache_drop(fdp->fd_nrdir);
434 fdp->fd_nrdir = cache_hold(ncp);
441 * Unmount a file system.
443 * Note: unmount takes a path to the vnode mounted on as argument,
444 * not special file (as before).
447 * umount_args(char *path, int flags)
451 unmount(struct unmount_args *uap)
453 struct thread *td = curthread;
454 struct proc *p = td->td_proc;
458 struct nlookupdata nd;
461 if (p->p_ucred->cr_prison != NULL)
463 if (usermount == 0 && (error = suser(td)))
467 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
469 error = nlookup(&nd);
471 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
479 * Only root, or the user that did the original mount is
480 * permitted to unmount this filesystem.
482 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
483 (error = suser(td))) {
489 * Don't allow unmounting the root file system.
491 if (mp->mnt_flag & MNT_ROOTFS) {
497 * Must be the root of the filesystem
499 if ((vp->v_flag & VROOT) == 0) {
504 return (dounmount(mp, SCARG(uap, flags), td));
508 * Do the actual file system unmount.
511 dounmount(struct mount *mp, int flags, struct thread *td)
513 struct vnode *coveredvp;
518 lwkt_gettoken(&ilock, &mountlist_token);
519 if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
520 lwkt_reltoken(&ilock);
523 mp->mnt_kern_flag |= MNTK_UNMOUNT;
524 /* Allow filesystems to detect that a forced unmount is in progress. */
525 if (flags & MNT_FORCE)
526 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
527 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK |
528 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &ilock, td);
530 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
531 if (mp->mnt_kern_flag & MNTK_MWAIT)
536 if (mp->mnt_flag & MNT_EXPUBLIC)
537 vfs_setpublicfs(NULL, NULL, NULL);
539 vfs_msync(mp, MNT_WAIT);
540 async_flag = mp->mnt_flag & MNT_ASYNC;
541 mp->mnt_flag &=~ MNT_ASYNC;
542 cache_purgevfs(mp); /* remove cache entries for this file sys */
543 if (mp->mnt_syncer != NULL)
544 vrele(mp->mnt_syncer);
545 if (((mp->mnt_flag & MNT_RDONLY) ||
546 (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) ||
548 error = VFS_UNMOUNT(mp, flags, td);
549 lwkt_gettokref(&ilock);
551 if (mp->mnt_syncer == NULL)
552 vfs_allocate_syncvnode(mp);
553 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
554 mp->mnt_flag |= async_flag;
555 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE,
557 if (mp->mnt_kern_flag & MNTK_MWAIT)
561 TAILQ_REMOVE(&mountlist, mp, mnt_list);
564 * Remove any installed vnode ops here so the individual VFSs don't
567 vfs_rm_vnodeops(&mp->mnt_vn_coherency_ops);
568 vfs_rm_vnodeops(&mp->mnt_vn_journal_ops);
569 vfs_rm_vnodeops(&mp->mnt_vn_norm_ops);
570 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops);
571 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops);
573 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
574 coveredvp->v_mountedhere = NULL;
576 cache_drop(mp->mnt_ncp);
579 mp->mnt_vfc->vfc_refcount--;
580 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
581 panic("unmount: dangling vnode");
582 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &ilock, td);
583 if (mp->mnt_kern_flag & MNTK_MWAIT)
590 * Sync each mounted filesystem.
594 static int syncprt = 0;
595 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
600 sync(struct sync_args *uap)
602 struct thread *td = curthread;
603 struct mount *mp, *nmp;
607 lwkt_gettoken(&ilock, &mountlist_token);
608 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
609 if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) {
610 nmp = TAILQ_NEXT(mp, mnt_list);
613 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
614 asyncflag = mp->mnt_flag & MNT_ASYNC;
615 mp->mnt_flag &= ~MNT_ASYNC;
616 vfs_msync(mp, MNT_NOWAIT);
617 VFS_SYNC(mp, MNT_NOWAIT, td);
618 mp->mnt_flag |= asyncflag;
620 lwkt_gettokref(&ilock);
621 nmp = TAILQ_NEXT(mp, mnt_list);
624 lwkt_reltoken(&ilock);
626 * print out buffer pool stat information on each sync() call.
635 /* XXX PRISON: could be per prison flag */
636 static int prison_quotas;
638 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
642 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
644 * Change filesystem quotas.
648 quotactl(struct quotactl_args *uap)
650 struct nlookupdata nd;
658 if (p->p_ucred->cr_prison && !prison_quotas)
661 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
663 error = nlookup(&nd);
665 mp = nd.nl_ncp->nc_mount;
666 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
667 SCARG(uap, arg), nd.nl_td);
674 * mountctl(char *path, int op, const void *ctl, int ctllen,
675 * void *buf, int buflen)
677 * This function operates on a mount point and executes the specified
678 * operation using the specified control data, and possibly returns data.
680 * The actual number of bytes stored in the result buffer is returned, 0
681 * if none, otherwise an error is returned.
685 mountctl(struct mountctl_args *uap)
687 struct thread *td = curthread;
688 struct proc *p = td->td_proc;
695 * Sanity and permissions checks. We must be root.
698 if (p->p_ucred->cr_prison != NULL)
700 if ((error = suser(td)) != 0)
704 * Argument length checks
706 if (uap->ctllen < 0 || uap->ctllen > MAXPATHLEN)
708 if (uap->buflen < 0 || uap->buflen > MAXPATHLEN)
710 if (uap->path == NULL)
714 * Allocate the necessary buffers and copyin data
716 path = zalloc(namei_zone);
717 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
722 ctl = malloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
723 error = copyin(uap->ctl, ctl, uap->ctllen);
728 buf = malloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
731 * Execute the internal kernel function and clean up.
733 error = kern_mountctl(path, uap->op, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
734 if (error == 0 && uap->sysmsg_result > 0)
735 error = copyout(buf, uap->buf, uap->sysmsg_result);
738 zfree(namei_zone, path);
747 * Execute a mount control operation by resolving the path to a mount point
748 * and calling vop_mountctl().
751 kern_mountctl(const char *path, int op, const void *ctl, int ctllen,
752 void *buf, int buflen, int *res)
754 struct thread *td = curthread;
757 struct nlookupdata nd;
762 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
764 error = nlookup(&nd);
766 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
774 * Must be the root of the filesystem
776 if ((vp->v_flag & VROOT) == 0) {
780 error = vop_mountctl(mp->mnt_vn_use_ops, op, ctl, ctllen,
787 kern_statfs(struct nlookupdata *nd, struct statfs *buf)
789 struct thread *td = curthread;
794 if ((error = nlookup(nd)) != 0)
796 mp = nd->nl_ncp->nc_mount;
798 if ((error = VFS_STATFS(mp, sp, td)) != 0)
800 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
801 bcopy(sp, buf, sizeof(*buf));
802 /* Only root should have access to the fsid's. */
804 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
809 * statfs_args(char *path, struct statfs *buf)
811 * Get filesystem statistics.
814 statfs(struct statfs_args *uap)
816 struct nlookupdata nd;
820 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
822 error = kern_statfs(&nd, &buf);
825 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
830 kern_fstatfs(int fd, struct statfs *buf)
832 struct thread *td = curthread;
833 struct proc *p = td->td_proc;
840 error = getvnode(p->p_fd, fd, &fp);
843 mp = ((struct vnode *)fp->f_data)->v_mount;
847 error = VFS_STATFS(mp, sp, td);
850 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
851 bcopy(sp, buf, sizeof(*buf));
852 /* Only root should have access to the fsid's. */
854 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
859 * fstatfs_args(int fd, struct statfs *buf)
861 * Get filesystem statistics.
864 fstatfs(struct fstatfs_args *uap)
869 error = kern_fstatfs(uap->fd, &buf);
872 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
877 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
879 * Get statistics on all filesystems.
883 getfsstat(struct getfsstat_args *uap)
885 struct thread *td = curthread;
886 struct mount *mp, *nmp;
890 long count, maxcount, error;
892 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
893 sfsp = (caddr_t)SCARG(uap, buf);
895 lwkt_gettoken(&ilock, &mountlist_token);
896 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
897 if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) {
898 nmp = TAILQ_NEXT(mp, mnt_list);
901 if (sfsp && count < maxcount) {
904 * If MNT_NOWAIT or MNT_LAZY is specified, do not
905 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
906 * overrides MNT_WAIT.
908 if (((SCARG(uap, flags) & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
909 (SCARG(uap, flags) & MNT_WAIT)) &&
910 (error = VFS_STATFS(mp, sp, td))) {
911 lwkt_gettokref(&ilock);
912 nmp = TAILQ_NEXT(mp, mnt_list);
916 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
917 error = copyout(sp, sfsp, sizeof(*sp));
925 lwkt_gettokref(&ilock);
926 nmp = TAILQ_NEXT(mp, mnt_list);
929 lwkt_reltoken(&ilock);
930 if (sfsp && count > maxcount)
931 uap->sysmsg_result = maxcount;
933 uap->sysmsg_result = count;
938 * fchdir_args(int fd)
940 * Change current working directory to a given file descriptor.
944 fchdir(struct fchdir_args *uap)
946 struct thread *td = curthread;
947 struct proc *p = td->td_proc;
948 struct filedesc *fdp = p->p_fd;
949 struct vnode *vp, *ovp;
952 struct namecache *ncp, *oncp;
953 struct namecache *nct;
956 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
958 vp = (struct vnode *)fp->f_data;
960 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
961 if (vp->v_type != VDIR || fp->f_ncp == NULL)
964 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td);
969 ncp = cache_hold(fp->f_ncp);
970 while (!error && (mp = vp->v_mountedhere) != NULL) {
971 error = nlookup_mp(mp, &nct);
973 cache_unlock(nct); /* leave ref intact */
976 error = vget(vp, LK_SHARED, td);
977 KKASSERT(error == 0);
984 oncp = fdp->fd_ncdir;
985 VOP_UNLOCK(vp, 0, td); /* leave ref intact */
998 kern_chdir(struct nlookupdata *nd)
1000 struct thread *td = curthread;
1001 struct proc *p = td->td_proc;
1002 struct filedesc *fdp = p->p_fd;
1003 struct vnode *vp, *ovp;
1004 struct namecache *oncp;
1007 if ((error = nlookup(nd)) != 0)
1009 if ((vp = nd->nl_ncp->nc_vp) == NULL)
1011 if ((error = vget(vp, LK_SHARED, td)) != 0)
1014 error = checkvp_chdir(vp, td);
1015 VOP_UNLOCK(vp, 0, td);
1018 oncp = fdp->fd_ncdir;
1019 cache_unlock(nd->nl_ncp); /* leave reference intact */
1020 fdp->fd_ncdir = nd->nl_ncp;
1032 * chdir_args(char *path)
1034 * Change current working directory (``.'').
1037 chdir(struct chdir_args *uap)
1039 struct nlookupdata nd;
1042 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1044 error = kern_chdir(&nd);
1050 * Helper function for raised chroot(2) security function: Refuse if
1051 * any filedescriptors are open directories.
1054 chroot_refuse_vdir_fds(fdp)
1055 struct filedesc *fdp;
1062 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
1063 error = getvnode(fdp, fd, &fp);
1066 vp = (struct vnode *)fp->f_data;
1067 if (vp->v_type != VDIR)
1075 * This sysctl determines if we will allow a process to chroot(2) if it
1076 * has a directory open:
1077 * 0: disallowed for all processes.
1078 * 1: allowed for processes that were not already chroot(2)'ed.
1079 * 2: allowed for all processes.
1082 static int chroot_allow_open_directories = 1;
1084 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1085 &chroot_allow_open_directories, 0, "");
1088 * chroot to the specified namecache entry. We obtain the vp from the
1089 * namecache data. The passed ncp must be locked and referenced and will
1090 * remain locked and referenced on return.
1093 kern_chroot(struct nlookupdata *nd)
1095 struct thread *td = curthread;
1096 struct proc *p = td->td_proc;
1097 struct filedesc *fdp = p->p_fd;
1098 struct namecache *ncp;
1103 * Only root can chroot
1105 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
1108 if ((error = nlookup(nd)) != 0)
1113 * Disallow open directory descriptors (fchdir() breakouts).
1115 if (chroot_allow_open_directories == 0 ||
1116 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1117 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1120 if ((vp = ncp->nc_vp) == NULL)
1123 if ((error = vget(vp, LK_SHARED, td)) != 0)
1127 * Check the validity of vp as a directory to change to and
1128 * associate it with rdir/jdir.
1130 error = checkvp_chdir(vp, td);
1131 VOP_UNLOCK(vp, 0, td); /* leave reference intact */
1133 vrele(fdp->fd_rdir);
1134 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */
1135 cache_drop(fdp->fd_nrdir);
1136 fdp->fd_nrdir = cache_hold(ncp);
1137 if (fdp->fd_jdir == NULL) {
1140 fdp->fd_njdir = cache_hold(ncp);
1149 * chroot_args(char *path)
1151 * Change notion of root (``/'') directory.
1155 chroot(struct chroot_args *uap)
1157 struct thread *td = curthread;
1158 struct nlookupdata nd;
1161 KKASSERT(td->td_proc);
1162 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1164 error = kern_chroot(&nd);
1170 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1171 * determine whether it is legal to chdir to the vnode. The vnode's state
1172 * is not changed by this call.
1175 checkvp_chdir(struct vnode *vp, struct thread *td)
1179 if (vp->v_type != VDIR)
1182 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred, td);
1187 kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
1189 struct thread *td = curthread;
1190 struct proc *p = td->td_proc;
1191 struct filedesc *fdp = p->p_fd;
1196 int type, indx, error;
1199 if ((oflags & O_ACCMODE) == O_ACCMODE)
1201 flags = FFLAGS(oflags);
1202 error = falloc(p, &nfp, NULL);
1206 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1209 * XXX p_dupfd is a real mess. It allows a device to return a
1210 * file descriptor to be duplicated rather then doing the open
1216 * Call vn_open() to do the lookup and assign the vnode to the
1217 * file pointer. vn_open() does not change the ref count on fp
1218 * and the vnode, on success, will be inherited by the file pointer
1221 nd->nl_flags |= NLC_LOCKVP;
1222 error = vn_open(nd, fp, flags, cmode);
1226 * handle special fdopen() case. bleh. dupfdopen() is
1227 * responsible for dropping the old contents of ofiles[indx]
1230 * Note that if fsetfd() succeeds it will add a ref to fp
1231 * which represents the fd_ofiles[] assignment. We must still
1232 * drop our reference.
1234 if ((error == ENODEV || error == ENXIO) && p->p_dupfd >= 0) {
1235 if (fsetfd(p, fp, &indx) == 0) {
1236 error = dupfdopen(fdp, indx, p->p_dupfd, flags, error);
1239 fdrop(fp, td); /* our ref */
1242 if (fdp->fd_ofiles[indx] == fp) {
1243 fdp->fd_ofiles[indx] = NULL;
1244 fdrop(fp, td); /* fd_ofiles[] ref */
1248 fdrop(fp, td); /* our ref */
1249 if (error == ERESTART)
1255 * ref the vnode for ourselves so it can't be ripped out from under
1256 * is. XXX need an ND flag to request that the vnode be returned
1259 vp = (struct vnode *)fp->f_data;
1261 if ((error = fsetfd(p, fp, &indx)) != 0) {
1268 * If no error occurs the vp will have been assigned to the file
1274 * There should be 2 references on the file, one from the descriptor
1275 * table, and one for us.
1277 * Handle the case where someone closed the file (via its file
1278 * descriptor) while we were blocked. The end result should look
1279 * like opening the file succeeded but it was immediately closed.
1281 if (fp->f_count == 1) {
1282 KASSERT(fdp->fd_ofiles[indx] != fp,
1283 ("Open file descriptor lost all refs"));
1291 if (flags & (O_EXLOCK | O_SHLOCK)) {
1292 lf.l_whence = SEEK_SET;
1295 if (flags & O_EXLOCK)
1296 lf.l_type = F_WRLCK;
1298 lf.l_type = F_RDLCK;
1300 if ((flags & FNONBLOCK) == 0)
1303 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1305 * lock request failed. Normally close the descriptor
1306 * but handle the case where someone might have dup()d
1307 * it when we weren't looking. One reference is
1308 * owned by the descriptor array, the other by us.
1311 if (fdp->fd_ofiles[indx] == fp) {
1312 fdp->fd_ofiles[indx] = NULL;
1318 fp->f_flag |= FHASLOCK;
1320 /* assert that vn_open created a backing object if one is needed */
1321 KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0,
1322 ("open: vmio vnode has no backing object after vn_open"));
1327 * release our private reference, leaving the one associated with the
1328 * descriptor table intact.
1336 * open_args(char *path, int flags, int mode)
1338 * Check permissions, allocate an open file structure,
1339 * and call the device open routine if any.
1342 open(struct open_args *uap)
1344 struct nlookupdata nd;
1347 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1349 error = kern_open(&nd, uap->flags,
1350 uap->mode, &uap->sysmsg_result);
1357 kern_mknod(struct nlookupdata *nd, int mode, int dev)
1359 struct namecache *ncp;
1360 struct thread *td = curthread;
1361 struct proc *p = td->td_proc;
1369 switch (mode & S_IFMT) {
1375 error = suser_cred(p->p_ucred, PRISON_ROOT);
1382 nd->nl_flags |= NLC_CREATE;
1383 if ((error = nlookup(nd)) != 0)
1390 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1391 vattr.va_rdev = dev;
1394 switch (mode & S_IFMT) {
1395 case S_IFMT: /* used by badsect to flag bad sectors */
1396 vattr.va_type = VBAD;
1399 vattr.va_type = VCHR;
1402 vattr.va_type = VBLK;
1413 error = VOP_NWHITEOUT(ncp, nd->nl_cred, NAMEI_CREATE);
1416 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
1425 * mknod_args(char *path, int mode, int dev)
1427 * Create a special file.
1430 mknod(struct mknod_args *uap)
1432 struct nlookupdata nd;
1435 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1437 error = kern_mknod(&nd, uap->mode, uap->dev);
1443 kern_mkfifo(struct nlookupdata *nd, int mode)
1445 struct namecache *ncp;
1446 struct thread *td = curthread;
1447 struct proc *p = td->td_proc;
1454 nd->nl_flags |= NLC_CREATE;
1455 if ((error = nlookup(nd)) != 0)
1462 vattr.va_type = VFIFO;
1463 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1465 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr);
1472 * mkfifo_args(char *path, int mode)
1474 * Create a named pipe.
1477 mkfifo(struct mkfifo_args *uap)
1479 struct nlookupdata nd;
1482 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1484 error = kern_mkfifo(&nd, uap->mode);
1490 kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
1492 struct thread *td = curthread;
1497 * Lookup the source and obtained a locked vnode.
1499 * XXX relookup on vget failure / race ?
1502 if ((error = nlookup(nd)) != 0)
1504 vp = nd->nl_ncp->nc_vp;
1505 KKASSERT(vp != NULL);
1506 if (vp->v_type == VDIR)
1507 return (EPERM); /* POSIX */
1508 if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0)
1512 * Unlock the source so we can lookup the target without deadlocking
1513 * (XXX vp is locked already, possible other deadlock?). The target
1516 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
1517 nd->nl_flags &= ~NLC_NCPISLOCKED;
1518 cache_unlock(nd->nl_ncp);
1520 linknd->nl_flags |= NLC_CREATE;
1521 if ((error = nlookup(linknd)) != 0) {
1525 if (linknd->nl_ncp->nc_vp) {
1531 * Finally run the new API VOP.
1533 error = VOP_NLINK(linknd->nl_ncp, vp, linknd->nl_cred);
1539 * link_args(char *path, char *link)
1541 * Make a hard file link.
1544 link(struct link_args *uap)
1546 struct nlookupdata nd, linknd;
1549 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1551 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
1553 error = kern_link(&nd, &linknd);
1554 nlookup_done(&linknd);
1561 kern_symlink(struct nlookupdata *nd, char *path, int mode)
1563 struct namecache *ncp;
1569 nd->nl_flags |= NLC_CREATE;
1570 if ((error = nlookup(nd)) != 0)
1577 vattr.va_mode = mode;
1578 error = VOP_NSYMLINK(ncp, &vp, nd->nl_cred, &vattr, path);
1585 * symlink(char *path, char *link)
1587 * Make a symbolic link.
1590 symlink(struct symlink_args *uap)
1592 struct thread *td = curthread;
1593 struct nlookupdata nd;
1598 path = zalloc(namei_zone);
1599 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
1601 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
1603 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
1604 error = kern_symlink(&nd, path, mode);
1608 zfree(namei_zone, path);
1613 * undelete_args(char *path)
1615 * Delete a whiteout from the filesystem.
1619 undelete(struct undelete_args *uap)
1621 struct nlookupdata nd;
1624 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, 0);
1626 nd.nl_flags |= NLC_DELETE;
1628 error = nlookup(&nd);
1630 error = VOP_NWHITEOUT(nd.nl_ncp, nd.nl_cred, NAMEI_DELETE);
1636 kern_unlink(struct nlookupdata *nd)
1638 struct namecache *ncp;
1642 nd->nl_flags |= NLC_DELETE;
1643 if ((error = nlookup(nd)) != 0)
1646 error = VOP_NREMOVE(ncp, nd->nl_cred);
1651 * unlink_args(char *path)
1653 * Delete a name from the filesystem.
1656 unlink(struct unlink_args *uap)
1658 struct nlookupdata nd;
1661 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1663 error = kern_unlink(&nd);
1669 kern_lseek(int fd, off_t offset, int whence, off_t *res)
1671 struct thread *td = curthread;
1672 struct proc *p = td->td_proc;
1673 struct filedesc *fdp = p->p_fd;
1678 if ((u_int)fd >= fdp->fd_nfiles ||
1679 (fp = fdp->fd_ofiles[fd]) == NULL)
1681 if (fp->f_type != DTYPE_VNODE)
1685 fp->f_offset += offset;
1688 error=VOP_GETATTR((struct vnode *)fp->f_data, &vattr, td);
1691 fp->f_offset = offset + vattr.va_size;
1694 fp->f_offset = offset;
1699 *res = fp->f_offset;
1704 * lseek_args(int fd, int pad, off_t offset, int whence)
1706 * Reposition read/write file offset.
1709 lseek(struct lseek_args *uap)
1713 error = kern_lseek(uap->fd, uap->offset, uap->whence,
1714 &uap->sysmsg_offset);
1720 kern_access(struct nlookupdata *nd, int aflags)
1722 struct thread *td = curthread;
1726 if ((error = nlookup(nd)) != 0)
1729 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
1733 /* Flags == 0 means only check for existence. */
1742 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1743 error = VOP_ACCESS(vp, flags, nd->nl_cred, td);
1746 * If the file handle is stale we have to re-resolve the
1747 * entry. This is a hack at the moment.
1749 if (error == ESTALE) {
1750 cache_setunresolved(nd->nl_ncp);
1751 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1764 * access_args(char *path, int flags)
1766 * Check access permissions.
1769 access(struct access_args *uap)
1771 struct nlookupdata nd;
1774 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1776 error = kern_access(&nd, uap->flags);
1782 kern_stat(struct nlookupdata *nd, struct stat *st)
1788 if ((error = nlookup(nd)) != 0)
1791 if ((vp = nd->nl_ncp->nc_vp) == NULL)
1795 if ((error = vget(vp, LK_SHARED, td)) != 0)
1797 error = vn_stat(vp, st, td);
1800 * If the file handle is stale we have to re-resolve the entry. This
1801 * is a hack at the moment.
1803 if (error == ESTALE) {
1804 cache_setunresolved(nd->nl_ncp);
1805 error = cache_resolve(nd->nl_ncp, nd->nl_cred);
1816 * stat_args(char *path, struct stat *ub)
1818 * Get file status; this version follows links.
1821 stat(struct stat_args *uap)
1823 struct nlookupdata nd;
1827 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1829 error = kern_stat(&nd, &st);
1831 error = copyout(&st, uap->ub, sizeof(*uap->ub));
1838 * lstat_args(char *path, struct stat *ub)
1840 * Get file status; this version does not follow links.
1843 lstat(struct lstat_args *uap)
1845 struct nlookupdata nd;
1849 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1851 error = kern_stat(&nd, &st);
1853 error = copyout(&st, uap->ub, sizeof(*uap->ub));
1864 nsb->st_dev = sb->st_dev;
1865 nsb->st_ino = sb->st_ino;
1866 nsb->st_mode = sb->st_mode;
1867 nsb->st_nlink = sb->st_nlink;
1868 nsb->st_uid = sb->st_uid;
1869 nsb->st_gid = sb->st_gid;
1870 nsb->st_rdev = sb->st_rdev;
1871 nsb->st_atimespec = sb->st_atimespec;
1872 nsb->st_mtimespec = sb->st_mtimespec;
1873 nsb->st_ctimespec = sb->st_ctimespec;
1874 nsb->st_size = sb->st_size;
1875 nsb->st_blocks = sb->st_blocks;
1876 nsb->st_blksize = sb->st_blksize;
1877 nsb->st_flags = sb->st_flags;
1878 nsb->st_gen = sb->st_gen;
1879 nsb->st_qspare[0] = sb->st_qspare[0];
1880 nsb->st_qspare[1] = sb->st_qspare[1];
1884 * nstat_args(char *path, struct nstat *ub)
1888 nstat(struct nstat_args *uap)
1890 struct thread *td = curthread;
1894 struct nlookupdata nd;
1898 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
1900 error = nlookup(&nd);
1902 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
1905 error = vn_stat(vp, &sb, td);
1908 cvtnstat(&sb, &nsb);
1909 error = copyout(&nsb, SCARG(uap, ub), sizeof(nsb));
1916 * lstat_args(char *path, struct stat *ub)
1918 * Get file status; this version does not follow links.
1922 nlstat(struct nlstat_args *uap)
1924 struct thread *td = curthread;
1928 struct nlookupdata nd;
1932 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, 0);
1934 error = nlookup(&nd);
1936 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
1939 error = vn_stat(vp, &sb, td);
1942 cvtnstat(&sb, &nsb);
1943 error = copyout(&nsb, SCARG(uap, ub), sizeof(nsb));
1950 * pathconf_Args(char *path, int name)
1952 * Get configurable pathname variables.
1956 pathconf(struct pathconf_args *uap)
1958 struct nlookupdata nd;
1963 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
1965 error = nlookup(&nd);
1967 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
1970 error = VOP_PATHCONF(vp, SCARG(uap, name), uap->sysmsg_fds);
1978 * kern_readlink isn't properly split yet. There is a copyin burried
1979 * in VOP_READLINK().
1982 kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
1984 struct thread *td = curthread;
1985 struct proc *p = td->td_proc;
1991 if ((error = nlookup(nd)) != 0)
1993 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp);
1996 if (vp->v_type != VLNK) {
1999 aiov.iov_base = buf;
2000 aiov.iov_len = count;
2001 auio.uio_iov = &aiov;
2002 auio.uio_iovcnt = 1;
2003 auio.uio_offset = 0;
2004 auio.uio_rw = UIO_READ;
2005 auio.uio_segflg = UIO_USERSPACE;
2007 auio.uio_resid = count;
2008 error = VOP_READLINK(vp, &auio, p->p_ucred);
2011 *res = count - auio.uio_resid;
2016 * readlink_args(char *path, char *buf, int count)
2018 * Return target name of a symbolic link.
2021 readlink(struct readlink_args *uap)
2023 struct nlookupdata nd;
2026 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2028 error = kern_readlink(&nd, uap->buf, uap->count,
2029 &uap->sysmsg_result);
2036 setfflags(struct vnode *vp, int flags)
2038 struct thread *td = curthread;
2039 struct proc *p = td->td_proc;
2044 * Prevent non-root users from setting flags on devices. When
2045 * a device is reused, users can retain ownership of the device
2046 * if they are allowed to set flags and programs assume that
2047 * chown can't fail when done as root.
2049 if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
2050 ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
2054 * note: vget is required for any operation that might mod the vnode
2055 * so VINACTIVE is properly cleared.
2057 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2058 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2060 vattr.va_flags = flags;
2061 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2068 * chflags(char *path, int flags)
2070 * Change flags of a file given a path name.
2074 chflags(struct chflags_args *uap)
2076 struct nlookupdata nd;
2081 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
2082 /* XXX Add NLC flag indicating modifying operation? */
2084 error = nlookup(&nd);
2086 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
2089 error = setfflags(vp, SCARG(uap, flags));
2096 * fchflags_args(int fd, int flags)
2098 * Change flags of a file given a file descriptor.
2102 fchflags(struct fchflags_args *uap)
2104 struct thread *td = curthread;
2105 struct proc *p = td->td_proc;
2109 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2111 return setfflags((struct vnode *) fp->f_data, SCARG(uap, flags));
2115 setfmode(struct vnode *vp, int mode)
2117 struct thread *td = curthread;
2118 struct proc *p = td->td_proc;
2123 * note: vget is required for any operation that might mod the vnode
2124 * so VINACTIVE is properly cleared.
2126 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2127 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2129 vattr.va_mode = mode & ALLPERMS;
2130 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2137 kern_chmod(struct nlookupdata *nd, int mode)
2142 /* XXX Add NLC flag indicating modifying operation? */
2143 if ((error = nlookup(nd)) != 0)
2145 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2147 error = setfmode(vp, mode);
2153 * chmod_args(char *path, int mode)
2155 * Change mode of a file given path name.
2159 chmod(struct chmod_args *uap)
2161 struct nlookupdata nd;
2164 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2166 error = kern_chmod(&nd, uap->mode);
2172 * lchmod_args(char *path, int mode)
2174 * Change mode of a file given path name (don't follow links.)
2178 lchmod(struct lchmod_args *uap)
2180 struct nlookupdata nd;
2183 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2185 error = kern_chmod(&nd, uap->mode);
2191 * fchmod_args(int fd, int mode)
2193 * Change mode of a file given a file descriptor.
2197 fchmod(struct fchmod_args *uap)
2199 struct thread *td = curthread;
2200 struct proc *p = td->td_proc;
2204 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2206 return setfmode((struct vnode *)fp->f_data, SCARG(uap, mode));
2210 setfown(struct vnode *vp, uid_t uid, gid_t gid)
2212 struct thread *td = curthread;
2213 struct proc *p = td->td_proc;
2218 * note: vget is required for any operation that might mod the vnode
2219 * so VINACTIVE is properly cleared.
2221 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2222 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2226 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2233 kern_chown(struct nlookupdata *nd, int uid, int gid)
2238 /* XXX Add NLC flag indicating modifying operation? */
2239 if ((error = nlookup(nd)) != 0)
2241 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2243 error = setfown(vp, uid, gid);
2249 * chown(char *path, int uid, int gid)
2251 * Set ownership given a path name.
2254 chown(struct chown_args *uap)
2256 struct nlookupdata nd;
2259 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2261 error = kern_chown(&nd, uap->uid, uap->gid);
2267 * lchown_args(char *path, int uid, int gid)
2269 * Set ownership given a path name, do not cross symlinks.
2272 lchown(struct lchown_args *uap)
2274 struct nlookupdata nd;
2277 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2279 error = kern_chown(&nd, uap->uid, uap->gid);
2285 * fchown_args(int fd, int uid, int gid)
2287 * Set ownership given a file descriptor.
2291 fchown(struct fchown_args *uap)
2293 struct thread *td = curthread;
2294 struct proc *p = td->td_proc;
2298 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2300 return setfown((struct vnode *)fp->f_data,
2301 SCARG(uap, uid), SCARG(uap, gid));
2305 getutimes(const struct timeval *tvp, struct timespec *tsp)
2307 struct timeval tv[2];
2311 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2314 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2315 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2321 setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
2323 struct thread *td = curthread;
2324 struct proc *p = td->td_proc;
2329 * note: vget is required for any operation that might mod the vnode
2330 * so VINACTIVE is properly cleared.
2332 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2333 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) {
2335 vattr.va_atime = ts[0];
2336 vattr.va_mtime = ts[1];
2338 vattr.va_vaflags |= VA_UTIMES_NULL;
2339 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td);
2346 kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
2348 struct timespec ts[2];
2352 if ((error = getutimes(tptr, ts)) != 0)
2354 /* XXX Add NLC flag indicating modifying operation? */
2355 if ((error = nlookup(nd)) != 0)
2357 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2359 error = setutimes(vp, ts, tptr == NULL);
2365 * utimes_args(char *path, struct timeval *tptr)
2367 * Set the access and modification times of a file.
2370 utimes(struct utimes_args *uap)
2372 struct timeval tv[2];
2373 struct nlookupdata nd;
2377 error = copyin(uap->tptr, tv, sizeof(tv));
2381 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2383 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2389 * lutimes_args(char *path, struct timeval *tptr)
2391 * Set the access and modification times of a file.
2394 lutimes(struct lutimes_args *uap)
2396 struct timeval tv[2];
2397 struct nlookupdata nd;
2401 error = copyin(uap->tptr, tv, sizeof(tv));
2405 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2407 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2413 kern_futimes(int fd, struct timeval *tptr)
2415 struct thread *td = curthread;
2416 struct proc *p = td->td_proc;
2417 struct timespec ts[2];
2421 error = getutimes(tptr, ts);
2424 error = getvnode(p->p_fd, fd, &fp);
2427 error = setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
2432 * futimes_args(int fd, struct timeval *tptr)
2434 * Set the access and modification times of a file.
2437 futimes(struct futimes_args *uap)
2439 struct timeval tv[2];
2443 error = copyin(uap->tptr, tv, sizeof(tv));
2448 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
2454 kern_truncate(struct nlookupdata *nd, off_t length)
2462 /* XXX Add NLC flag indicating modifying operation? */
2463 if ((error = nlookup(nd)) != 0)
2465 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0)
2467 VOP_LEASE(vp, nd->nl_td, nd->nl_cred, LEASE_WRITE);
2468 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, nd->nl_td)) != 0) {
2472 if (vp->v_type == VDIR) {
2474 } else if ((error = vn_writechk(vp)) == 0 &&
2475 (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred, nd->nl_td)) == 0) {
2477 vattr.va_size = length;
2478 error = VOP_SETATTR(vp, &vattr, nd->nl_cred, nd->nl_td);
2485 * truncate(char *path, int pad, off_t length)
2487 * Truncate a file given its path name.
2490 truncate(struct truncate_args *uap)
2492 struct nlookupdata nd;
2495 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2497 error = kern_truncate(&nd, uap->length);
2503 kern_ftruncate(int fd, off_t length)
2505 struct thread *td = curthread;
2506 struct proc *p = td->td_proc;
2514 if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
2516 if ((fp->f_flag & FWRITE) == 0)
2518 vp = (struct vnode *)fp->f_data;
2519 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
2520 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2521 if (vp->v_type == VDIR)
2523 else if ((error = vn_writechk(vp)) == 0) {
2525 vattr.va_size = length;
2526 error = VOP_SETATTR(vp, &vattr, fp->f_cred, td);
2528 VOP_UNLOCK(vp, 0, td);
2533 * ftruncate_args(int fd, int pad, off_t length)
2535 * Truncate a file given a file descriptor.
2538 ftruncate(struct ftruncate_args *uap)
2542 error = kern_ftruncate(uap->fd, uap->length);
2550 * Sync an open file.
2554 fsync(struct fsync_args *uap)
2556 struct thread *td = curthread;
2557 struct proc *p = td->td_proc;
2563 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2565 vp = (struct vnode *)fp->f_data;
2566 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2567 if (VOP_GETVOBJECT(vp, &obj) == 0)
2568 vm_object_page_clean(obj, 0, 0, 0);
2569 if ((error = VOP_FSYNC(vp, MNT_WAIT, td)) == 0 &&
2570 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) &&
2572 error = (*bioops.io_fsync)(vp);
2573 VOP_UNLOCK(vp, 0, td);
2578 kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
2580 struct namecache *fncpd;
2581 struct namecache *tncpd;
2582 struct namecache *ncp;
2587 if ((error = nlookup(fromnd)) != 0)
2589 if ((fncpd = fromnd->nl_ncp->nc_parent) == NULL)
2594 * unlock the source ncp so we can lookup the target ncp without
2595 * deadlocking. The target may or may not exist so we do not check
2596 * for a target vp like kern_mkdir() and other creation functions do.
2598 * The source and target directories are ref'd and rechecked after
2599 * everything is relocked to determine if the source or target file
2602 KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
2603 fromnd->nl_flags &= ~NLC_NCPISLOCKED;
2604 cache_unlock(fromnd->nl_ncp);
2606 tond->nl_flags |= NLC_CREATE;
2607 if ((error = nlookup(tond)) != 0) {
2611 if ((tncpd = tond->nl_ncp->nc_parent) == NULL) {
2618 * If the source and target are the same there is nothing to do
2620 if (fromnd->nl_ncp == tond->nl_ncp) {
2627 * relock the source ncp
2629 if (cache_lock_nonblock(fromnd->nl_ncp) == 0) {
2630 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2631 } else if (fromnd->nl_ncp > tond->nl_ncp) {
2632 cache_lock(fromnd->nl_ncp);
2633 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2635 cache_unlock(tond->nl_ncp);
2636 cache_lock(fromnd->nl_ncp);
2637 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred);
2638 cache_lock(tond->nl_ncp);
2639 cache_resolve(tond->nl_ncp, tond->nl_cred);
2641 fromnd->nl_flags |= NLC_NCPISLOCKED;
2644 * make sure the parent directories linkages are the same
2646 if (fncpd != fromnd->nl_ncp->nc_parent ||
2647 tncpd != tond->nl_ncp->nc_parent) {
2654 * Both the source and target must be within the same filesystem and
2655 * in the same filesystem as their parent directories within the
2656 * namecache topology.
2658 mp = fncpd->nc_mount;
2659 if (mp != tncpd->nc_mount || mp != fromnd->nl_ncp->nc_mount ||
2660 mp != tond->nl_ncp->nc_mount) {
2667 * If the target exists and either the source or target is a directory,
2668 * then both must be directories.
2670 if (tond->nl_ncp->nc_vp) {
2671 if (fromnd->nl_ncp->nc_vp->v_type == VDIR) {
2672 if (tond->nl_ncp->nc_vp->v_type != VDIR)
2674 } else if (tond->nl_ncp->nc_vp->v_type == VDIR) {
2680 * You cannot rename a source into itself or a subdirectory of itself.
2681 * We check this by travsersing the target directory upwards looking
2682 * for a match against the source.
2685 for (ncp = tncpd; ncp; ncp = ncp->nc_parent) {
2686 if (fromnd->nl_ncp == ncp) {
2697 error = VOP_NRENAME(fromnd->nl_ncp, tond->nl_ncp, tond->nl_cred);
2702 * rename_args(char *from, char *to)
2704 * Rename files. Source and destination must either both be directories,
2705 * or both not be directories. If target is a directory, it must be empty.
2708 rename(struct rename_args *uap)
2710 struct nlookupdata fromnd, tond;
2713 error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
2715 error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
2717 error = kern_rename(&fromnd, &tond);
2718 nlookup_done(&tond);
2720 nlookup_done(&fromnd);
2725 kern_mkdir(struct nlookupdata *nd, int mode)
2727 struct thread *td = curthread;
2728 struct proc *p = td->td_proc;
2729 struct namecache *ncp;
2735 nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE;
2736 if ((error = nlookup(nd)) != 0)
2744 vattr.va_type = VDIR;
2745 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
2748 error = VOP_NMKDIR(ncp, &vp, p->p_ucred, &vattr);
2755 * mkdir_args(char *path, int mode)
2757 * Make a directory file.
2761 mkdir(struct mkdir_args *uap)
2763 struct nlookupdata nd;
2766 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2768 error = kern_mkdir(&nd, uap->mode);
2774 kern_rmdir(struct nlookupdata *nd)
2776 struct namecache *ncp;
2780 nd->nl_flags |= NLC_DELETE;
2781 if ((error = nlookup(nd)) != 0)
2785 error = VOP_NRMDIR(ncp, nd->nl_cred);
2790 * rmdir_args(char *path)
2792 * Remove a directory file.
2796 rmdir(struct rmdir_args *uap)
2798 struct nlookupdata nd;
2801 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2803 error = kern_rmdir(&nd);
2809 kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res)
2811 struct thread *td = curthread;
2812 struct proc *p = td->td_proc;
2820 if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
2822 if ((fp->f_flag & FREAD) == 0)
2824 vp = (struct vnode *)fp->f_data;
2826 if (vp->v_type != VDIR)
2828 aiov.iov_base = buf;
2829 aiov.iov_len = count;
2830 auio.uio_iov = &aiov;
2831 auio.uio_iovcnt = 1;
2832 auio.uio_rw = UIO_READ;
2833 auio.uio_segflg = UIO_USERSPACE;
2835 auio.uio_resid = count;
2836 /* vn_lock(vp, LK_SHARED | LK_RETRY, td); */
2837 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
2838 loff = auio.uio_offset = fp->f_offset;
2839 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
2840 fp->f_offset = auio.uio_offset;
2841 VOP_UNLOCK(vp, 0, td);
2844 if (count == auio.uio_resid) {
2845 if (union_dircheckp) {
2846 error = union_dircheckp(td, &vp, fp);
2852 if ((vp->v_flag & VROOT) &&
2853 (vp->v_mount->mnt_flag & MNT_UNION)) {
2854 struct vnode *tvp = vp;
2855 vp = vp->v_mount->mnt_vnodecovered;
2857 fp->f_data = (caddr_t)vp;
2866 *res = count - auio.uio_resid;
2871 * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
2873 * Read a block of directory entries in a file system independent format.
2876 getdirentries(struct getdirentries_args *uap)
2881 error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base,
2882 &uap->sysmsg_result);
2885 error = copyout(&base, uap->basep, sizeof(*uap->basep));
2890 * getdents_args(int fd, char *buf, size_t count)
2893 getdents(struct getdents_args *uap)
2897 error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL,
2898 &uap->sysmsg_result);
2904 * umask(int newmask)
2906 * Set the mode mask for creation of filesystem nodes.
2911 umask(struct umask_args *uap)
2913 struct thread *td = curthread;
2914 struct proc *p = td->td_proc;
2915 struct filedesc *fdp;
2918 uap->sysmsg_result = fdp->fd_cmask;
2919 fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS;
2924 * revoke(char *path)
2926 * Void all references to file by ripping underlying filesystem
2931 revoke(struct revoke_args *uap)
2933 struct thread *td = curthread;
2934 struct nlookupdata nd;
2941 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW);
2943 error = nlookup(&nd);
2945 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
2946 cred = crhold(nd.nl_cred);
2949 if (vp->v_type != VCHR && vp->v_type != VBLK)
2952 error = VOP_GETATTR(vp, &vattr, td);
2953 if (error == 0 && cred->cr_uid != vattr.va_uid)
2954 error = suser_cred(cred, PRISON_ROOT);
2955 if (error == 0 && count_udev(vp->v_udev) > 0) {
2956 if ((error = vx_lock(vp)) == 0) {
2957 VOP_REVOKE(vp, REVOKEALL);
2968 * Convert a user file descriptor to a kernel file entry.
2971 getvnode(struct filedesc *fdp, int fd, struct file **fpp)
2975 if ((u_int)fd >= fdp->fd_nfiles ||
2976 (fp = fdp->fd_ofiles[fd]) == NULL)
2978 if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO)
2984 * getfh_args(char *fname, fhandle_t *fhp)
2986 * Get (NFS) file handle
2989 getfh(struct getfh_args *uap)
2991 struct thread *td = curthread;
2992 struct nlookupdata nd;
2998 * Must be super user
3000 if ((error = suser(td)) != 0)
3004 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
3006 error = nlookup(&nd);
3008 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3011 bzero(&fh, sizeof(fh));
3012 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3013 error = VFS_VPTOFH(vp, &fh.fh_fid);
3016 error = copyout(&fh, uap->fhp, sizeof(fh));
3022 * fhopen_args(const struct fhandle *u_fhp, int flags)
3024 * syscall for the rpc.lockd to use to translate a NFS file handle into
3025 * an open descriptor.
3027 * warning: do not remove the suser() call or this becomes one giant
3031 fhopen(struct fhopen_args *uap)
3033 struct thread *td = curthread;
3034 struct proc *p = td->td_proc;
3039 struct vattr *vap = &vat;
3041 struct filedesc *fdp = p->p_fd;
3042 int fmode, mode, error, type;
3048 * Must be super user
3054 fmode = FFLAGS(SCARG(uap, flags));
3055 /* why not allow a non-read/write open for our lockd? */
3056 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3058 error = copyin(SCARG(uap,u_fhp), &fhp, sizeof(fhp));
3061 /* find the mount point */
3062 mp = vfs_getvfs(&fhp.fh_fsid);
3065 /* now give me my vnode, it gets returned to me locked */
3066 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
3070 * from now on we have to make sure not
3071 * to forget about the vnode
3072 * any error that causes an abort must vput(vp)
3073 * just set error = err and 'goto bad;'.
3079 if (vp->v_type == VLNK) {
3083 if (vp->v_type == VSOCK) {
3088 if (fmode & (FWRITE | O_TRUNC)) {
3089 if (vp->v_type == VDIR) {
3093 error = vn_writechk(vp);
3101 error = VOP_ACCESS(vp, mode, p->p_ucred, td);
3105 if (fmode & O_TRUNC) {
3106 VOP_UNLOCK(vp, 0, td); /* XXX */
3107 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
3108 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); /* XXX */
3111 error = VOP_SETATTR(vp, vap, p->p_ucred, td);
3117 * VOP_OPEN needs the file pointer so it can potentially override
3120 * WARNING! no f_ncp will be associated when fhopen()ing a directory.
3123 if ((error = falloc(p, &nfp, NULL)) != 0)
3127 fp->f_data = (caddr_t)vp;
3128 fp->f_flag = fmode & FMASK;
3129 fp->f_ops = &vnode_fileops;
3130 fp->f_type = DTYPE_VNODE;
3132 error = VOP_OPEN(vp, fmode, p->p_ucred, fp, td);
3135 * setting f_ops this way prevents VOP_CLOSE from being
3136 * called or fdrop() releasing the vp from v_data. Since
3137 * the VOP_OPEN failed we don't want to VOP_CLOSE.
3139 fp->f_ops = &badfileops;
3148 * The fp now owns a reference on the vnode. We still have our own
3154 * Make sure that a VM object is created for VMIO support. If this
3155 * fails just fdrop() normally to clean up.
3157 if (vn_canvmio(vp) == TRUE) {
3158 if ((error = vfs_object_create(vp, td)) != 0) {
3165 * The open was successful, associate it with a file descriptor.
3167 if ((error = fsetfd(p, fp, &indx)) != 0) {
3174 if (fmode & (O_EXLOCK | O_SHLOCK)) {
3175 lf.l_whence = SEEK_SET;
3178 if (fmode & O_EXLOCK)
3179 lf.l_type = F_WRLCK;
3181 lf.l_type = F_RDLCK;
3183 if ((fmode & FNONBLOCK) == 0)
3185 VOP_UNLOCK(vp, 0, td);
3186 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
3188 * lock request failed. Normally close the descriptor
3189 * but handle the case where someone might have dup()d
3190 * or close()d it when we weren't looking.
3192 if (fdp->fd_ofiles[indx] == fp) {
3193 fdp->fd_ofiles[indx] = NULL;
3198 * release our private reference.
3204 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
3205 fp->f_flag |= FHASLOCK;
3207 if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0))
3208 vfs_object_create(vp, td);
3212 uap->sysmsg_result = indx;
3221 * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
3224 fhstat(struct fhstat_args *uap)
3226 struct thread *td = curthread;
3234 * Must be super user
3240 error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t));
3244 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3246 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3248 error = vn_stat(vp, &sb, td);
3252 error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
3257 * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
3260 fhstatfs(struct fhstatfs_args *uap)
3262 struct thread *td = curthread;
3271 * Must be super user
3273 if ((error = suser(td)))
3276 if ((error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t))) != 0)
3279 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3281 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3286 if ((error = VFS_STATFS(mp, sp, td)) != 0)
3288 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
3290 bcopy(sp, &sb, sizeof(sb));
3291 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
3294 return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
3298 * Syscall to push extended attribute configuration information into the
3299 * VFS. Accepts a path, which it converts to a mountpoint, as well as
3300 * a command (int cmd), and attribute name and misc data. For now, the
3301 * attribute name is left in userspace for consumption by the VFS_op.
3302 * It will probably be changed to be copied into sysspace by the
3303 * syscall in the future, once issues with various consumers of the
3304 * attribute code have raised their hands.
3306 * Currently this is used only by UFS Extended Attributes.
3309 extattrctl(struct extattrctl_args *uap)
3311 struct nlookupdata nd;
3317 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3319 error = nlookup(&nd);
3321 mp = nd.nl_ncp->nc_mount;
3322 error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd),
3323 SCARG(uap, attrname), SCARG(uap, arg),
3331 * Syscall to set a named extended attribute on a file or directory.
3332 * Accepts attribute name, and a uio structure pointing to the data to set.
3333 * The uio is consumed in the style of writev(). The real work happens
3334 * in VOP_SETEXTATTR().
3337 extattr_set_file(struct extattr_set_file_args *uap)
3339 char attrname[EXTATTR_MAXNAMELEN];
3340 struct iovec aiov[UIO_SMALLIOV];
3341 struct iovec *needfree;
3342 struct nlookupdata nd;
3351 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3356 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3358 error = nlookup(&nd);
3360 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3367 iovlen = uap->iovcnt * sizeof(struct iovec);
3368 if (uap->iovcnt > UIO_SMALLIOV) {
3369 if (uap->iovcnt > UIO_MAXIOV) {
3373 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3379 auio.uio_iovcnt = uap->iovcnt;
3380 auio.uio_rw = UIO_WRITE;
3381 auio.uio_segflg = UIO_USERSPACE;
3382 auio.uio_td = nd.nl_td;
3383 auio.uio_offset = 0;
3384 if ((error = copyin(uap->iovp, iov, iovlen)))
3387 for (i = 0; i < uap->iovcnt; i++) {
3388 if (iov->iov_len > INT_MAX - auio.uio_resid) {
3392 auio.uio_resid += iov->iov_len;
3395 cnt = auio.uio_resid;
3396 error = VOP_SETEXTATTR(vp, attrname, &auio, nd.nl_cred, nd.nl_td);
3397 cnt -= auio.uio_resid;
3398 uap->sysmsg_result = cnt;
3403 FREE(needfree, M_IOV);
3408 * Syscall to get a named extended attribute on a file or directory.
3409 * Accepts attribute name, and a uio structure pointing to a buffer for the
3410 * data. The uio is consumed in the style of readv(). The real work
3411 * happens in VOP_GETEXTATTR();
3414 extattr_get_file(struct extattr_get_file_args *uap)
3416 char attrname[EXTATTR_MAXNAMELEN];
3417 struct iovec aiov[UIO_SMALLIOV];
3418 struct iovec *needfree;
3419 struct nlookupdata nd;
3428 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3433 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3435 error = nlookup(&nd);
3437 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3443 iovlen = uap->iovcnt * sizeof (struct iovec);
3445 if (uap->iovcnt > UIO_SMALLIOV) {
3446 if (uap->iovcnt > UIO_MAXIOV) {
3450 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3456 auio.uio_iovcnt = uap->iovcnt;
3457 auio.uio_rw = UIO_READ;
3458 auio.uio_segflg = UIO_USERSPACE;
3459 auio.uio_td = nd.nl_td;
3460 auio.uio_offset = 0;
3461 if ((error = copyin(uap->iovp, iov, iovlen)))
3464 for (i = 0; i < uap->iovcnt; i++) {
3465 if (iov->iov_len > INT_MAX - auio.uio_resid) {
3469 auio.uio_resid += iov->iov_len;
3472 cnt = auio.uio_resid;
3473 error = VOP_GETEXTATTR(vp, attrname, &auio, nd.nl_cred, nd.nl_td);
3474 cnt -= auio.uio_resid;
3475 uap->sysmsg_result = cnt;
3480 FREE(needfree, M_IOV);
3485 * Syscall to delete a named extended attribute from a file or directory.
3486 * Accepts attribute name. The real work happens in VOP_SETEXTATTR().
3489 extattr_delete_file(struct extattr_delete_file_args *uap)
3491 char attrname[EXTATTR_MAXNAMELEN];
3492 struct nlookupdata nd;
3496 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN);
3501 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3503 error = nlookup(&nd);
3505 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp);
3511 error = VOP_SETEXTATTR(vp, attrname, NULL, nd.nl_cred, nd.nl_td);
3518 * print out statistics from the current status of the buffer pool
3519 * this can be toggeled by the system control option debug.syncprt
3528 int counts[(MAXBSIZE / PAGE_SIZE) + 1];
3529 static char *bname[3] = { "LOCKED", "LRU", "AGE" };
3531 for (dp = bufqueues, i = 0; dp < &bufqueues[3]; dp++, i++) {
3533 for (j = 0; j <= MAXBSIZE/PAGE_SIZE; j++)
3536 TAILQ_FOREACH(bp, dp, b_freelist) {
3537 counts[bp->b_bufsize/PAGE_SIZE]++;
3541 printf("%s: total-%d", bname[i], count);
3542 for (j = 0; j <= MAXBSIZE/PAGE_SIZE; j++)
3544 printf(", %d-%d", j * PAGE_SIZE, counts[j]);