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 1.126 2008/05/09 17:52:17 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>
68 #include <sys/sysctl.h>
71 #include <sys/file2.h>
72 #include <sys/spinlock2.h>
75 #include <vm/vm_object.h>
76 #include <vm/vm_page.h>
78 #include <machine/limits.h>
79 #include <machine/stdarg.h>
81 #include <vfs/union/union.h>
83 static void mount_warning(struct mount *mp, const char *ctl, ...);
84 static int mount_path(struct proc *p, struct mount *mp, char **rb, char **fb);
85 static int checkvp_chdir (struct vnode *vn, struct thread *td);
86 static void checkdirs (struct nchandle *old_nch, struct nchandle *new_nch);
87 static int chroot_refuse_vdir_fds (struct filedesc *fdp);
88 static int chroot_visible_mnt(struct mount *mp, struct proc *p);
89 static int getutimes (const struct timeval *, struct timespec *);
90 static int setfown (struct vnode *, uid_t, gid_t);
91 static int setfmode (struct vnode *, int);
92 static int setfflags (struct vnode *, int);
93 static int setutimes (struct vnode *, const struct timespec *, int);
94 static int usermount = 0; /* if 1, non-root can mount fs. */
96 int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);
98 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
101 * Virtual File System System Calls
105 * Mount a file system.
108 * mount_args(char *type, char *path, int flags, caddr_t data)
112 sys_mount(struct mount_args *uap)
114 struct thread *td = curthread;
115 struct proc *p = td->td_proc;
119 struct vfsconf *vfsp;
120 int error, flag = 0, flag2 = 0;
123 struct nlookupdata nd;
124 char fstypename[MFSNAMELEN];
125 struct ucred *cred = p->p_ucred;
128 if (cred->cr_prison != NULL)
130 if (usermount == 0 && (error = suser(td)))
133 * Do not allow NFS export by non-root users.
135 if (uap->flags & MNT_EXPORTED) {
141 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
144 uap->flags |= MNT_NOSUID | MNT_NODEV;
147 * Lookup the requested path and extract the nch and vnode.
149 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
151 if ((error = nlookup(&nd)) == 0) {
152 if (nd.nl_nch.ncp->nc_vp == NULL)
162 * Extract the locked+refd ncp and cleanup the nd structure
165 cache_zero(&nd.nl_nch);
168 if ((nch.ncp->nc_flag & NCF_ISMOUNTPT) && cache_findmount(&nch))
175 * now we have the locked ref'd nch and unreferenced vnode.
178 if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
185 * Now we have an unlocked ref'd nch and a locked ref'd vp
187 if (uap->flags & MNT_UPDATE) {
188 if ((vp->v_flag & VROOT) == 0) {
195 flag2 = mp->mnt_kern_flag;
197 * We only allow the filesystem to be reloaded if it
198 * is currently mounted read-only.
200 if ((uap->flags & MNT_RELOAD) &&
201 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
204 return (EOPNOTSUPP); /* Needs translation */
207 * Only root, or the user that did the original mount is
208 * permitted to update it.
210 if (mp->mnt_stat.f_owner != cred->cr_uid &&
211 (error = suser(td))) {
216 if (vfs_busy(mp, LK_NOWAIT)) {
221 if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
227 vp->v_flag |= VMOUNT;
229 uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
234 * If the user is not root, ensure that they own the directory
235 * onto which we are attempting to mount.
237 if ((error = VOP_GETATTR(vp, &va)) ||
238 (va.va_uid != cred->cr_uid && (error = suser(td)))) {
243 if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
248 if (vp->v_type != VDIR) {
253 if (vp->v_mount->mnt_kern_flag & MNTK_NOSTKMNT) {
258 if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
263 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
264 if (!strcmp(vfsp->vfc_name, fstypename))
270 /* Only load modules for root (very important!) */
271 if ((error = suser(td)) != 0) {
276 error = linker_load_file(fstypename, &lf);
277 if (error || lf == NULL) {
285 /* lookup again, see if the VFS was loaded */
286 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
287 if (!strcmp(vfsp->vfc_name, fstypename))
292 linker_file_unload(lf);
298 if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
303 vp->v_flag |= VMOUNT;
306 * Allocate and initialize the filesystem.
308 mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
309 TAILQ_INIT(&mp->mnt_nvnodelist);
310 TAILQ_INIT(&mp->mnt_reservedvnlist);
311 TAILQ_INIT(&mp->mnt_jlist);
312 mp->mnt_nvnodelistsize = 0;
313 lockinit(&mp->mnt_lock, "vfslock", 0, 0);
314 vfs_busy(mp, LK_NOWAIT);
315 mp->mnt_op = vfsp->vfc_vfsops;
317 vfsp->vfc_refcount++;
318 mp->mnt_stat.f_type = vfsp->vfc_typenum;
319 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
320 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
321 mp->mnt_stat.f_owner = cred->cr_uid;
322 mp->mnt_iosize_max = DFLTPHYS;
326 * Set the mount level flags.
328 if (uap->flags & MNT_RDONLY)
329 mp->mnt_flag |= MNT_RDONLY;
330 else if (mp->mnt_flag & MNT_RDONLY)
331 mp->mnt_kern_flag |= MNTK_WANTRDWR;
332 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
333 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
334 MNT_NOSYMFOLLOW | MNT_IGNORE |
335 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
336 mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
337 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
338 MNT_NOSYMFOLLOW | MNT_IGNORE |
339 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
341 * Mount the filesystem.
342 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
345 error = VFS_MOUNT(mp, uap->path, uap->data, cred);
346 if (mp->mnt_flag & MNT_UPDATE) {
347 if (mp->mnt_kern_flag & MNTK_WANTRDWR)
348 mp->mnt_flag &= ~MNT_RDONLY;
349 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
350 mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
353 mp->mnt_kern_flag = flag2;
356 vp->v_flag &= ~VMOUNT;
361 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
363 * Put the new filesystem on the mount list after root. The mount
364 * point gets its own mnt_ncmountpt (unless the VFS already set one
365 * up) which represents the root of the mount. The lookup code
366 * detects the mount point going forward and checks the root of
367 * the mount going backwards.
369 * It is not necessary to invalidate or purge the vnode underneath
370 * because elements under the mount will be given their own glue
374 if (mp->mnt_ncmountpt.ncp == NULL) {
376 * allocate, then unlock, but leave the ref intact
378 cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
379 cache_unlock(&mp->mnt_ncmountpt);
381 mp->mnt_ncmounton = nch; /* inherits ref */
382 nch.ncp->nc_flag |= NCF_ISMOUNTPT;
384 /* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
385 vp->v_flag &= ~VMOUNT;
386 mountlist_insert(mp, MNTINS_LAST);
387 checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
389 error = vfs_allocate_syncvnode(mp);
391 error = VFS_START(mp, 0);
394 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
395 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
396 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
397 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
398 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
399 vp->v_flag &= ~VMOUNT;
400 mp->mnt_vfc->vfc_refcount--;
410 * Scan all active processes to see if any of them have a current
411 * or root directory onto which the new filesystem has just been
412 * mounted. If so, replace them with the new mount point.
414 * The passed ncp is ref'd and locked (from the mount code) and
415 * must be associated with the vnode representing the root of the
418 struct checkdirs_info {
419 struct nchandle old_nch;
420 struct nchandle new_nch;
421 struct vnode *old_vp;
422 struct vnode *new_vp;
425 static int checkdirs_callback(struct proc *p, void *data);
428 checkdirs(struct nchandle *old_nch, struct nchandle *new_nch)
430 struct checkdirs_info info;
436 * If the old mount point's vnode has a usecount of 1, it is not
437 * being held as a descriptor anywhere.
439 olddp = old_nch->ncp->nc_vp;
440 if (olddp == NULL || olddp->v_sysref.refcnt == 1)
444 * Force the root vnode of the new mount point to be resolved
445 * so we can update any matching processes.
448 if (VFS_ROOT(mp, &newdp))
449 panic("mount: lost mount");
450 cache_setunresolved(new_nch);
451 cache_setvp(new_nch, newdp);
454 * Special handling of the root node
456 if (rootvnode == olddp) {
458 vfs_cache_setroot(newdp, cache_hold(new_nch));
462 * Pass newdp separately so the callback does not have to access
463 * it via new_nch->ncp->nc_vp.
465 info.old_nch = *old_nch;
466 info.new_nch = *new_nch;
468 allproc_scan(checkdirs_callback, &info);
473 * NOTE: callback is not MP safe because the scanned process's filedesc
474 * structure can be ripped out from under us, amoung other things.
477 checkdirs_callback(struct proc *p, void *data)
479 struct checkdirs_info *info = data;
480 struct filedesc *fdp;
481 struct nchandle ncdrop1;
482 struct nchandle ncdrop2;
483 struct vnode *vprele1;
484 struct vnode *vprele2;
486 if ((fdp = p->p_fd) != NULL) {
487 cache_zero(&ncdrop1);
488 cache_zero(&ncdrop2);
493 * MPUNSAFE - XXX fdp can be pulled out from under a
496 * A shared filedesc is ok, we don't have to copy it
497 * because we are making this change globally.
499 spin_lock_wr(&fdp->fd_spin);
500 if (fdp->fd_ncdir.mount == info->old_nch.mount &&
501 fdp->fd_ncdir.ncp == info->old_nch.ncp) {
502 vprele1 = fdp->fd_cdir;
504 fdp->fd_cdir = info->new_vp;
505 ncdrop1 = fdp->fd_ncdir;
506 cache_copy(&info->new_nch, &fdp->fd_ncdir);
508 if (fdp->fd_nrdir.mount == info->old_nch.mount &&
509 fdp->fd_nrdir.ncp == info->old_nch.ncp) {
510 vprele2 = fdp->fd_rdir;
512 fdp->fd_rdir = info->new_vp;
513 ncdrop2 = fdp->fd_nrdir;
514 cache_copy(&info->new_nch, &fdp->fd_nrdir);
516 spin_unlock_wr(&fdp->fd_spin);
518 cache_drop(&ncdrop1);
520 cache_drop(&ncdrop2);
530 * Unmount a file system.
532 * Note: unmount takes a path to the vnode mounted on as argument,
533 * not special file (as before).
536 * umount_args(char *path, int flags)
540 sys_unmount(struct unmount_args *uap)
542 struct thread *td = curthread;
543 struct proc *p = td->td_proc;
544 struct mount *mp = NULL;
546 struct nlookupdata nd;
549 if (p->p_ucred->cr_prison != NULL)
551 if (usermount == 0 && (error = suser(td)))
554 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
556 error = nlookup(&nd);
560 mp = nd.nl_nch.mount;
563 * Only root, or the user that did the original mount is
564 * permitted to unmount this filesystem.
566 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
571 * Don't allow unmounting the root file system.
573 if (mp->mnt_flag & MNT_ROOTFS) {
579 * Must be the root of the filesystem
581 if (nd.nl_nch.ncp != mp->mnt_ncmountpt.ncp) {
590 return (dounmount(mp, uap->flags));
594 * Do the actual file system unmount.
597 dounmount_interlock(struct mount *mp)
599 if (mp->mnt_kern_flag & MNTK_UNMOUNT)
601 mp->mnt_kern_flag |= MNTK_UNMOUNT;
606 dounmount(struct mount *mp, int flags)
608 struct namecache *ncp;
616 * Exclusive access for unmounting purposes
618 if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
622 * Allow filesystems to detect that a forced unmount is in progress.
624 if (flags & MNT_FORCE)
625 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
626 lflags = LK_EXCLUSIVE | ((flags & MNT_FORCE) ? 0 : LK_NOWAIT);
627 error = lockmgr(&mp->mnt_lock, lflags);
629 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
630 if (mp->mnt_kern_flag & MNTK_MWAIT)
635 if (mp->mnt_flag & MNT_EXPUBLIC)
636 vfs_setpublicfs(NULL, NULL, NULL);
638 vfs_msync(mp, MNT_WAIT);
639 async_flag = mp->mnt_flag & MNT_ASYNC;
640 mp->mnt_flag &=~ MNT_ASYNC;
643 * If this filesystem isn't aliasing other filesystems,
644 * try to invalidate any remaining namecache entries and
645 * check the count afterwords.
647 if ((mp->mnt_kern_flag & MNTK_NCALIASED) == 0) {
648 cache_lock(&mp->mnt_ncmountpt);
649 cache_inval(&mp->mnt_ncmountpt, CINV_DESTROY|CINV_CHILDREN);
650 cache_unlock(&mp->mnt_ncmountpt);
652 if ((ncp = mp->mnt_ncmountpt.ncp) != NULL &&
653 (ncp->nc_refs != 1 || TAILQ_FIRST(&ncp->nc_list))) {
655 if ((flags & MNT_FORCE) == 0) {
657 mount_warning(mp, "Cannot unmount: "
663 mount_warning(mp, "Forced unmount: "
674 * nchandle records ref the mount structure. Expect a count of 1
675 * (our mount->mnt_ncmountpt).
677 if (mp->mnt_refs != 1) {
678 if ((flags & MNT_FORCE) == 0) {
679 mount_warning(mp, "Cannot unmount: "
680 "%d process references still "
681 "present", mp->mnt_refs);
684 mount_warning(mp, "Forced unmount: "
685 "%d process references still "
686 "present", mp->mnt_refs);
692 if (mp->mnt_syncer != NULL)
693 vrele(mp->mnt_syncer);
694 if (((mp->mnt_flag & MNT_RDONLY) ||
695 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) ||
696 (flags & MNT_FORCE)) {
697 error = VFS_UNMOUNT(mp, flags);
701 if (mp->mnt_syncer == NULL)
702 vfs_allocate_syncvnode(mp);
703 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
704 mp->mnt_flag |= async_flag;
705 lockmgr(&mp->mnt_lock, LK_RELEASE);
706 if (mp->mnt_kern_flag & MNTK_MWAIT)
711 * Clean up any journals still associated with the mount after
712 * filesystem activity has ceased.
714 journal_remove_all_journals(mp,
715 ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));
717 mountlist_remove(mp);
720 * Remove any installed vnode ops here so the individual VFSs don't
723 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
724 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
725 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
726 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
727 vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
729 if (mp->mnt_ncmountpt.ncp != NULL) {
730 nch = mp->mnt_ncmountpt;
731 cache_zero(&mp->mnt_ncmountpt);
732 cache_clrmountpt(&nch);
735 if (mp->mnt_ncmounton.ncp != NULL) {
736 nch = mp->mnt_ncmounton;
737 cache_zero(&mp->mnt_ncmounton);
738 cache_clrmountpt(&nch);
742 mp->mnt_vfc->vfc_refcount--;
743 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
744 panic("unmount: dangling vnode");
745 lockmgr(&mp->mnt_lock, LK_RELEASE);
746 if (mp->mnt_kern_flag & MNTK_MWAIT)
755 mount_warning(struct mount *mp, const char *ctl, ...)
762 if (cache_fullpath(NULL, &mp->mnt_ncmounton, &ptr, &buf) == 0) {
763 kprintf("unmount(%s): ", ptr);
768 kprintf("unmount(%p", mp);
769 if (mp->mnt_ncmounton.ncp && mp->mnt_ncmounton.ncp->nc_name)
770 kprintf(",%s", mp->mnt_ncmounton.ncp->nc_name);
779 * Shim cache_fullpath() to handle the case where a process is chrooted into
780 * a subdirectory of a mount. In this case if the root mount matches the
781 * process root directory's mount we have to specify the process's root
782 * directory instead of the mount point, because the mount point might
783 * be above the root directory.
787 mount_path(struct proc *p, struct mount *mp, char **rb, char **fb)
789 struct nchandle *nch;
791 if (p && p->p_fd->fd_nrdir.mount == mp)
792 nch = &p->p_fd->fd_nrdir;
794 nch = &mp->mnt_ncmountpt;
795 return(cache_fullpath(p, nch, rb, fb));
799 * Sync each mounted filesystem.
803 static int syncprt = 0;
804 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
807 static int sync_callback(struct mount *mp, void *data);
811 sys_sync(struct sync_args *uap)
813 mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
816 * print out buffer pool stat information on each sync() call.
826 sync_callback(struct mount *mp, void *data __unused)
830 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
831 asyncflag = mp->mnt_flag & MNT_ASYNC;
832 mp->mnt_flag &= ~MNT_ASYNC;
833 vfs_msync(mp, MNT_NOWAIT);
834 VFS_SYNC(mp, MNT_NOWAIT);
835 mp->mnt_flag |= asyncflag;
840 /* XXX PRISON: could be per prison flag */
841 static int prison_quotas;
843 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
847 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
849 * Change filesystem quotas.
853 sys_quotactl(struct quotactl_args *uap)
855 struct nlookupdata nd;
863 if (p->p_ucred->cr_prison && !prison_quotas)
866 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
868 error = nlookup(&nd);
870 mp = nd.nl_nch.mount;
871 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
872 uap->arg, nd.nl_cred);
879 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
880 * void *buf, int buflen)
882 * This function operates on a mount point and executes the specified
883 * operation using the specified control data, and possibly returns data.
885 * The actual number of bytes stored in the result buffer is returned, 0
886 * if none, otherwise an error is returned.
890 sys_mountctl(struct mountctl_args *uap)
892 struct thread *td = curthread;
893 struct proc *p = td->td_proc;
901 * Sanity and permissions checks. We must be root.
904 if (p->p_ucred->cr_prison != NULL)
906 if ((error = suser(td)) != 0)
910 * Argument length checks
912 if (uap->ctllen < 0 || uap->ctllen > 1024)
914 if (uap->buflen < 0 || uap->buflen > 16 * 1024)
916 if (uap->path == NULL)
920 * Allocate the necessary buffers and copyin data
922 path = objcache_get(namei_oc, M_WAITOK);
923 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
928 ctl = kmalloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
929 error = copyin(uap->ctl, ctl, uap->ctllen);
934 buf = kmalloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);
937 * Validate the descriptor
940 fp = holdfp(p->p_fd, uap->fd, -1);
950 * Execute the internal kernel function and clean up.
952 error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
955 if (error == 0 && uap->sysmsg_result > 0)
956 error = copyout(buf, uap->buf, uap->sysmsg_result);
959 objcache_put(namei_oc, path);
968 * Execute a mount control operation by resolving the path to a mount point
969 * and calling vop_mountctl().
972 kern_mountctl(const char *path, int op, struct file *fp,
973 const void *ctl, int ctllen,
974 void *buf, int buflen, int *res)
978 struct nlookupdata nd;
983 error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
985 error = nlookup(&nd);
987 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
995 * Must be the root of the filesystem
997 if ((vp->v_flag & VROOT) == 0) {
1001 error = vop_mountctl(mp->mnt_vn_use_ops, op, fp, ctl, ctllen,
1008 kern_statfs(struct nlookupdata *nd, struct statfs *buf)
1010 struct thread *td = curthread;
1011 struct proc *p = td->td_proc;
1014 char *fullpath, *freepath;
1017 if ((error = nlookup(nd)) != 0)
1019 mp = nd->nl_nch.mount;
1021 if ((error = VFS_STATFS(mp, sp, nd->nl_cred)) != 0)
1024 error = mount_path(p, mp, &fullpath, &freepath);
1027 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1028 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1029 kfree(freepath, M_TEMP);
1031 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1032 bcopy(sp, buf, sizeof(*buf));
1033 /* Only root should have access to the fsid's. */
1035 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
1040 * statfs_args(char *path, struct statfs *buf)
1042 * Get filesystem statistics.
1045 sys_statfs(struct statfs_args *uap)
1047 struct nlookupdata nd;
1051 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1053 error = kern_statfs(&nd, &buf);
1056 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1061 kern_fstatfs(int fd, struct statfs *buf)
1063 struct thread *td = curthread;
1064 struct proc *p = td->td_proc;
1068 char *fullpath, *freepath;
1072 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
1074 mp = ((struct vnode *)fp->f_data)->v_mount;
1079 if (fp->f_cred == NULL) {
1084 if ((error = VFS_STATFS(mp, sp, fp->f_cred)) != 0)
1087 if ((error = mount_path(p, mp, &fullpath, &freepath)) != 0)
1089 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1090 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1091 kfree(freepath, M_TEMP);
1093 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1094 bcopy(sp, buf, sizeof(*buf));
1096 /* Only root should have access to the fsid's. */
1098 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
1106 * fstatfs_args(int fd, struct statfs *buf)
1108 * Get filesystem statistics.
1111 sys_fstatfs(struct fstatfs_args *uap)
1116 error = kern_fstatfs(uap->fd, &buf);
1119 error = copyout(&buf, uap->buf, sizeof(*uap->buf));
1124 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
1126 * Get statistics on all filesystems.
1129 struct getfsstat_info {
1130 struct statfs *sfsp;
1138 static int getfsstat_callback(struct mount *, void *);
1142 sys_getfsstat(struct getfsstat_args *uap)
1144 struct thread *td = curthread;
1145 struct proc *p = td->td_proc;
1146 struct getfsstat_info info;
1148 bzero(&info, sizeof(info));
1150 info.maxcount = uap->bufsize / sizeof(struct statfs);
1151 info.sfsp = uap->buf;
1153 info.flags = uap->flags;
1156 mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
1157 if (info.sfsp && info.count > info.maxcount)
1158 uap->sysmsg_result = info.maxcount;
1160 uap->sysmsg_result = info.count;
1161 return (info.error);
1165 getfsstat_callback(struct mount *mp, void *data)
1167 struct getfsstat_info *info = data;
1173 if (info->sfsp && info->count < info->maxcount) {
1174 if (info->p && !chroot_visible_mnt(mp, info->p))
1179 * If MNT_NOWAIT or MNT_LAZY is specified, do not
1180 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
1181 * overrides MNT_WAIT.
1183 if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
1184 (info->flags & MNT_WAIT)) &&
1185 (error = VFS_STATFS(mp, sp, info->p->p_ucred))) {
1188 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1190 error = mount_path(info->p, mp, &fullpath, &freepath);
1192 info->error = error;
1195 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1196 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
1197 kfree(freepath, M_TEMP);
1199 error = copyout(sp, info->sfsp, sizeof(*sp));
1201 info->error = error;
1211 * fchdir_args(int fd)
1213 * Change current working directory to a given file descriptor.
1217 sys_fchdir(struct fchdir_args *uap)
1219 struct thread *td = curthread;
1220 struct proc *p = td->td_proc;
1221 struct filedesc *fdp = p->p_fd;
1222 struct vnode *vp, *ovp;
1225 struct nchandle nch, onch, tnch;
1228 if ((error = holdvnode(fdp, uap->fd, &fp)) != 0)
1230 vp = (struct vnode *)fp->f_data;
1232 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1233 if (vp->v_type != VDIR || fp->f_nchandle.ncp == NULL)
1236 error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
1242 cache_copy(&fp->f_nchandle, &nch);
1245 * If the ncp has become a mount point, traverse through
1249 while (!error && (nch.ncp->nc_flag & NCF_ISMOUNTPT) &&
1250 (mp = cache_findmount(&nch)) != NULL
1252 error = nlookup_mp(mp, &tnch);
1254 cache_unlock(&tnch); /* leave ref intact */
1256 vp = tnch.ncp->nc_vp;
1257 error = vget(vp, LK_SHARED);
1258 KKASSERT(error == 0);
1265 onch = fdp->fd_ncdir;
1266 vn_unlock(vp); /* leave ref intact */
1268 fdp->fd_ncdir = nch;
1280 kern_chdir(struct nlookupdata *nd)
1282 struct thread *td = curthread;
1283 struct proc *p = td->td_proc;
1284 struct filedesc *fdp = p->p_fd;
1285 struct vnode *vp, *ovp;
1286 struct nchandle onch;
1289 if ((error = nlookup(nd)) != 0)
1291 if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
1293 if ((error = vget(vp, LK_SHARED)) != 0)
1296 error = checkvp_chdir(vp, td);
1300 onch = fdp->fd_ncdir;
1301 cache_unlock(&nd->nl_nch); /* leave reference intact */
1302 fdp->fd_ncdir = nd->nl_nch;
1306 cache_zero(&nd->nl_nch);
1314 * chdir_args(char *path)
1316 * Change current working directory (``.'').
1319 sys_chdir(struct chdir_args *uap)
1321 struct nlookupdata nd;
1324 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1326 error = kern_chdir(&nd);
1332 * Helper function for raised chroot(2) security function: Refuse if
1333 * any filedescriptors are open directories.
1336 chroot_refuse_vdir_fds(struct filedesc *fdp)
1343 for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
1344 if ((error = holdvnode(fdp, fd, &fp)) != 0)
1346 vp = (struct vnode *)fp->f_data;
1347 if (vp->v_type != VDIR) {
1358 * This sysctl determines if we will allow a process to chroot(2) if it
1359 * has a directory open:
1360 * 0: disallowed for all processes.
1361 * 1: allowed for processes that were not already chroot(2)'ed.
1362 * 2: allowed for all processes.
1365 static int chroot_allow_open_directories = 1;
1367 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
1368 &chroot_allow_open_directories, 0, "");
1371 * chroot to the specified namecache entry. We obtain the vp from the
1372 * namecache data. The passed ncp must be locked and referenced and will
1373 * remain locked and referenced on return.
1376 kern_chroot(struct nchandle *nch)
1378 struct thread *td = curthread;
1379 struct proc *p = td->td_proc;
1380 struct filedesc *fdp = p->p_fd;
1385 * Only root can chroot
1387 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
1391 * Disallow open directory descriptors (fchdir() breakouts).
1393 if (chroot_allow_open_directories == 0 ||
1394 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
1395 if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
1398 if ((vp = nch->ncp->nc_vp) == NULL)
1401 if ((error = vget(vp, LK_SHARED)) != 0)
1405 * Check the validity of vp as a directory to change to and
1406 * associate it with rdir/jdir.
1408 error = checkvp_chdir(vp, td);
1409 vn_unlock(vp); /* leave reference intact */
1411 vrele(fdp->fd_rdir);
1412 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */
1413 cache_drop(&fdp->fd_nrdir);
1414 cache_copy(nch, &fdp->fd_nrdir);
1415 if (fdp->fd_jdir == NULL) {
1418 cache_copy(nch, &fdp->fd_njdir);
1427 * chroot_args(char *path)
1429 * Change notion of root (``/'') directory.
1433 sys_chroot(struct chroot_args *uap)
1435 struct thread *td = curthread;
1436 struct nlookupdata nd;
1439 KKASSERT(td->td_proc);
1440 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1445 error = nlookup(&nd);
1447 error = kern_chroot(&nd.nl_nch);
1453 * Common routine for chroot and chdir. Given a locked, referenced vnode,
1454 * determine whether it is legal to chdir to the vnode. The vnode's state
1455 * is not changed by this call.
1458 checkvp_chdir(struct vnode *vp, struct thread *td)
1462 if (vp->v_type != VDIR)
1465 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred);
1470 kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
1472 struct thread *td = curthread;
1473 struct proc *p = td->td_proc;
1474 struct lwp *lp = td->td_lwp;
1475 struct filedesc *fdp = p->p_fd;
1480 int type, indx, error;
1483 if ((oflags & O_ACCMODE) == O_ACCMODE)
1485 flags = FFLAGS(oflags);
1486 error = falloc(p, &nfp, NULL);
1490 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1493 * XXX p_dupfd is a real mess. It allows a device to return a
1494 * file descriptor to be duplicated rather then doing the open
1500 * Call vn_open() to do the lookup and assign the vnode to the
1501 * file pointer. vn_open() does not change the ref count on fp
1502 * and the vnode, on success, will be inherited by the file pointer
1505 nd->nl_flags |= NLC_LOCKVP;
1506 error = vn_open(nd, fp, flags, cmode);
1510 * handle special fdopen() case. bleh. dupfdopen() is
1511 * responsible for dropping the old contents of ofiles[indx]
1514 * Note that fsetfd() will add a ref to fp which represents
1515 * the fd_files[] assignment. We must still drop our
1518 if ((error == ENODEV || error == ENXIO) && lp->lwp_dupfd >= 0) {
1519 if (fdalloc(p, 0, &indx) == 0) {
1520 error = dupfdopen(p, indx, lp->lwp_dupfd, flags, error);
1523 fdrop(fp); /* our ref */
1526 fsetfd(p, NULL, indx);
1529 fdrop(fp); /* our ref */
1530 if (error == ERESTART)
1536 * ref the vnode for ourselves so it can't be ripped out from under
1537 * is. XXX need an ND flag to request that the vnode be returned
1540 * Reserve a file descriptor but do not assign it until the open
1543 vp = (struct vnode *)fp->f_data;
1545 if ((error = fdalloc(p, 0, &indx)) != 0) {
1552 * If no error occurs the vp will have been assigned to the file
1557 if (flags & (O_EXLOCK | O_SHLOCK)) {
1558 lf.l_whence = SEEK_SET;
1561 if (flags & O_EXLOCK)
1562 lf.l_type = F_WRLCK;
1564 lf.l_type = F_RDLCK;
1565 if (flags & FNONBLOCK)
1570 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
1572 * lock request failed. Clean up the reserved
1576 fsetfd(p, NULL, indx);
1580 fp->f_flag |= FHASLOCK;
1584 * Assert that all regular file vnodes were created with a object.
1586 KASSERT(vp->v_type != VREG || vp->v_object != NULL,
1587 ("open: regular file has no backing object after vn_open"));
1593 * release our private reference, leaving the one associated with the
1594 * descriptor table intact.
1596 fsetfd(p, fp, indx);
1603 * open_args(char *path, int flags, int mode)
1605 * Check permissions, allocate an open file structure,
1606 * and call the device open routine if any.
1609 sys_open(struct open_args *uap)
1611 struct nlookupdata nd;
1614 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1616 error = kern_open(&nd, uap->flags,
1617 uap->mode, &uap->sysmsg_result);
1624 kern_mknod(struct nlookupdata *nd, int mode, int rmajor, int rminor)
1626 struct thread *td = curthread;
1627 struct proc *p = td->td_proc;
1635 switch (mode & S_IFMT) {
1641 error = suser_cred(p->p_ucred, PRISON_ROOT);
1648 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
1649 if ((error = nlookup(nd)) != 0)
1651 if (nd->nl_nch.ncp->nc_vp)
1653 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
1657 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1658 vattr.va_rmajor = rmajor;
1659 vattr.va_rminor = rminor;
1662 switch (mode & S_IFMT) {
1663 case S_IFMT: /* used by badsect to flag bad sectors */
1664 vattr.va_type = VBAD;
1667 vattr.va_type = VCHR;
1670 vattr.va_type = VBLK;
1681 error = VOP_NWHITEOUT(&nd->nl_nch, nd->nl_dvp,
1682 nd->nl_cred, NAMEI_CREATE);
1685 error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp,
1686 &vp, nd->nl_cred, &vattr);
1695 * mknod_args(char *path, int mode, int dev)
1697 * Create a special file.
1700 sys_mknod(struct mknod_args *uap)
1702 struct nlookupdata nd;
1705 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1707 error = kern_mknod(&nd, uap->mode,
1708 umajor(uap->dev), uminor(uap->dev));
1715 kern_mkfifo(struct nlookupdata *nd, int mode)
1717 struct thread *td = curthread;
1718 struct proc *p = td->td_proc;
1725 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
1726 if ((error = nlookup(nd)) != 0)
1728 if (nd->nl_nch.ncp->nc_vp)
1730 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
1734 vattr.va_type = VFIFO;
1735 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1737 error = VOP_NMKNOD(&nd->nl_nch, nd->nl_dvp, &vp, nd->nl_cred, &vattr);
1744 * mkfifo_args(char *path, int mode)
1746 * Create a named pipe.
1749 sys_mkfifo(struct mkfifo_args *uap)
1751 struct nlookupdata nd;
1754 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1756 error = kern_mkfifo(&nd, uap->mode);
1761 static int hardlink_check_uid = 0;
1762 SYSCTL_INT(_security, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1763 &hardlink_check_uid, 0,
1764 "Unprivileged processes cannot create hard links to files owned by other "
1766 static int hardlink_check_gid = 0;
1767 SYSCTL_INT(_security, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1768 &hardlink_check_gid, 0,
1769 "Unprivileged processes cannot create hard links to files owned by other "
1773 can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
1779 * Shortcut if disabled
1781 if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
1785 * root cred can always hardlink
1787 if (suser_cred(cred, PRISON_ROOT) == 0)
1791 * Otherwise only if the originating file is owned by the
1792 * same user or group. Note that any group is allowed if
1793 * the file is owned by the caller.
1795 error = VOP_GETATTR(vp, &va);
1799 if (hardlink_check_uid) {
1800 if (cred->cr_uid != va.va_uid)
1804 if (hardlink_check_gid) {
1805 if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
1813 kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
1815 struct thread *td = curthread;
1820 * Lookup the source and obtained a locked vnode.
1822 * XXX relookup on vget failure / race ?
1825 if ((error = nlookup(nd)) != 0)
1827 vp = nd->nl_nch.ncp->nc_vp;
1828 KKASSERT(vp != NULL);
1829 if (vp->v_type == VDIR)
1830 return (EPERM); /* POSIX */
1831 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
1833 if ((error = vget(vp, LK_EXCLUSIVE)) != 0)
1837 * Unlock the source so we can lookup the target without deadlocking
1838 * (XXX vp is locked already, possible other deadlock?). The target
1841 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
1842 nd->nl_flags &= ~NLC_NCPISLOCKED;
1843 cache_unlock(&nd->nl_nch);
1845 linknd->nl_flags |= NLC_CREATE | NLC_REFDVP;
1846 if ((error = nlookup(linknd)) != 0) {
1850 if (linknd->nl_nch.ncp->nc_vp) {
1856 * Finally run the new API VOP.
1858 error = can_hardlink(vp, td, td->td_proc->p_ucred);
1860 error = VOP_NLINK(&linknd->nl_nch, linknd->nl_dvp,
1861 vp, linknd->nl_cred);
1868 * link_args(char *path, char *link)
1870 * Make a hard file link.
1873 sys_link(struct link_args *uap)
1875 struct nlookupdata nd, linknd;
1878 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
1880 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
1882 error = kern_link(&nd, &linknd);
1883 nlookup_done(&linknd);
1890 kern_symlink(struct nlookupdata *nd, char *path, int mode)
1898 nd->nl_flags |= NLC_CREATE | NLC_REFDVP;
1899 if ((error = nlookup(nd)) != 0)
1901 if (nd->nl_nch.ncp->nc_vp)
1903 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
1907 vattr.va_mode = mode;
1908 error = VOP_NSYMLINK(&nd->nl_nch, dvp, &vp, nd->nl_cred, &vattr, path);
1915 * symlink(char *path, char *link)
1917 * Make a symbolic link.
1920 sys_symlink(struct symlink_args *uap)
1922 struct thread *td = curthread;
1923 struct nlookupdata nd;
1928 path = objcache_get(namei_oc, M_WAITOK);
1929 error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
1931 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
1933 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
1934 error = kern_symlink(&nd, path, mode);
1938 objcache_put(namei_oc, path);
1943 * undelete_args(char *path)
1945 * Delete a whiteout from the filesystem.
1949 sys_undelete(struct undelete_args *uap)
1951 struct nlookupdata nd;
1954 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1956 nd.nl_flags |= NLC_DELETE | NLC_REFDVP;
1958 error = nlookup(&nd);
1960 error = ncp_writechk(&nd.nl_nch);
1962 error = VOP_NWHITEOUT(&nd.nl_nch, nd.nl_dvp, nd.nl_cred,
1970 kern_unlink(struct nlookupdata *nd)
1975 nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
1976 if ((error = nlookup(nd)) != 0)
1978 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
1980 error = VOP_NREMOVE(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
1985 * unlink_args(char *path)
1987 * Delete a name from the filesystem.
1990 sys_unlink(struct unlink_args *uap)
1992 struct nlookupdata nd;
1995 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
1997 error = kern_unlink(&nd);
2003 kern_lseek(int fd, off_t offset, int whence, off_t *res)
2005 struct thread *td = curthread;
2006 struct proc *p = td->td_proc;
2011 fp = holdfp(p->p_fd, fd, -1);
2014 if (fp->f_type != DTYPE_VNODE) {
2021 fp->f_offset += offset;
2025 error = VOP_GETATTR((struct vnode *)fp->f_data, &vattr);
2027 fp->f_offset = offset + vattr.va_size;
2030 fp->f_offset = offset;
2037 *res = fp->f_offset;
2044 * lseek_args(int fd, int pad, off_t offset, int whence)
2046 * Reposition read/write file offset.
2049 sys_lseek(struct lseek_args *uap)
2053 error = kern_lseek(uap->fd, uap->offset, uap->whence,
2054 &uap->sysmsg_offset);
2060 kern_access(struct nlookupdata *nd, int aflags)
2065 if ((error = nlookup(nd)) != 0)
2068 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
2072 /* Flags == 0 means only check for existence. */
2081 if ((flags & VWRITE) == 0 ||
2082 (error = vn_writechk(vp, &nd->nl_nch)) == 0)
2083 error = VOP_ACCESS(vp, flags, nd->nl_cred);
2086 * If the file handle is stale we have to re-resolve the
2087 * entry. This is a hack at the moment.
2089 if (error == ESTALE) {
2091 cache_setunresolved(&nd->nl_nch);
2092 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
2105 * access_args(char *path, int flags)
2107 * Check access permissions.
2110 sys_access(struct access_args *uap)
2112 struct nlookupdata nd;
2115 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2117 error = kern_access(&nd, uap->flags);
2123 kern_stat(struct nlookupdata *nd, struct stat *st)
2129 if ((error = nlookup(nd)) != 0)
2132 if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
2136 if ((error = vget(vp, LK_SHARED)) != 0)
2138 error = vn_stat(vp, st, nd->nl_cred);
2141 * If the file handle is stale we have to re-resolve the entry. This
2142 * is a hack at the moment.
2144 if (error == ESTALE) {
2146 cache_setunresolved(&nd->nl_nch);
2147 error = cache_resolve(&nd->nl_nch, nd->nl_cred);
2157 * stat_args(char *path, struct stat *ub)
2159 * Get file status; this version follows links.
2162 sys_stat(struct stat_args *uap)
2164 struct nlookupdata nd;
2168 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2170 error = kern_stat(&nd, &st);
2172 error = copyout(&st, uap->ub, sizeof(*uap->ub));
2179 * lstat_args(char *path, struct stat *ub)
2181 * Get file status; this version does not follow links.
2184 sys_lstat(struct lstat_args *uap)
2186 struct nlookupdata nd;
2190 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2192 error = kern_stat(&nd, &st);
2194 error = copyout(&st, uap->ub, sizeof(*uap->ub));
2201 * pathconf_Args(char *path, int name)
2203 * Get configurable pathname variables.
2207 sys_pathconf(struct pathconf_args *uap)
2209 struct nlookupdata nd;
2214 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2216 error = nlookup(&nd);
2218 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
2221 error = VOP_PATHCONF(vp, uap->name, uap->sysmsg_fds);
2229 * kern_readlink isn't properly split yet. There is a copyin burried
2230 * in VOP_READLINK().
2233 kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
2235 struct thread *td = curthread;
2236 struct proc *p = td->td_proc;
2242 if ((error = nlookup(nd)) != 0)
2244 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
2247 if (vp->v_type != VLNK) {
2250 aiov.iov_base = buf;
2251 aiov.iov_len = count;
2252 auio.uio_iov = &aiov;
2253 auio.uio_iovcnt = 1;
2254 auio.uio_offset = 0;
2255 auio.uio_rw = UIO_READ;
2256 auio.uio_segflg = UIO_USERSPACE;
2258 auio.uio_resid = count;
2259 error = VOP_READLINK(vp, &auio, p->p_ucred);
2262 *res = count - auio.uio_resid;
2267 * readlink_args(char *path, char *buf, int count)
2269 * Return target name of a symbolic link.
2272 sys_readlink(struct readlink_args *uap)
2274 struct nlookupdata nd;
2277 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2279 error = kern_readlink(&nd, uap->buf, uap->count,
2280 &uap->sysmsg_result);
2287 setfflags(struct vnode *vp, int flags)
2289 struct thread *td = curthread;
2290 struct proc *p = td->td_proc;
2295 * Prevent non-root users from setting flags on devices. When
2296 * a device is reused, users can retain ownership of the device
2297 * if they are allowed to set flags and programs assume that
2298 * chown can't fail when done as root.
2300 if ((vp->v_type == VCHR || vp->v_type == VBLK) &&
2301 ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
2305 * note: vget is required for any operation that might mod the vnode
2306 * so VINACTIVE is properly cleared.
2308 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2310 vattr.va_flags = flags;
2311 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2318 * chflags(char *path, int flags)
2320 * Change flags of a file given a path name.
2324 sys_chflags(struct chflags_args *uap)
2326 struct nlookupdata nd;
2331 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2332 /* XXX Add NLC flag indicating modifying operation? */
2334 error = nlookup(&nd);
2336 error = ncp_writechk(&nd.nl_nch);
2338 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
2341 error = setfflags(vp, uap->flags);
2348 * fchflags_args(int fd, int flags)
2350 * Change flags of a file given a file descriptor.
2354 sys_fchflags(struct fchflags_args *uap)
2356 struct thread *td = curthread;
2357 struct proc *p = td->td_proc;
2361 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2363 if (fp->f_nchandle.ncp)
2364 error = ncp_writechk(&fp->f_nchandle);
2366 error = setfflags((struct vnode *) fp->f_data, uap->flags);
2372 setfmode(struct vnode *vp, int mode)
2374 struct thread *td = curthread;
2375 struct proc *p = td->td_proc;
2380 * note: vget is required for any operation that might mod the vnode
2381 * so VINACTIVE is properly cleared.
2383 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2385 vattr.va_mode = mode & ALLPERMS;
2386 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2393 kern_chmod(struct nlookupdata *nd, int mode)
2398 /* XXX Add NLC flag indicating modifying operation? */
2399 if ((error = nlookup(nd)) != 0)
2401 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
2403 if ((error = ncp_writechk(&nd->nl_nch)) == 0)
2404 error = setfmode(vp, mode);
2410 * chmod_args(char *path, int mode)
2412 * Change mode of a file given path name.
2416 sys_chmod(struct chmod_args *uap)
2418 struct nlookupdata nd;
2421 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2423 error = kern_chmod(&nd, uap->mode);
2429 * lchmod_args(char *path, int mode)
2431 * Change mode of a file given path name (don't follow links.)
2435 sys_lchmod(struct lchmod_args *uap)
2437 struct nlookupdata nd;
2440 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2442 error = kern_chmod(&nd, uap->mode);
2448 * fchmod_args(int fd, int mode)
2450 * Change mode of a file given a file descriptor.
2454 sys_fchmod(struct fchmod_args *uap)
2456 struct thread *td = curthread;
2457 struct proc *p = td->td_proc;
2461 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2463 if (fp->f_nchandle.ncp)
2464 error = ncp_writechk(&fp->f_nchandle);
2466 error = setfmode((struct vnode *)fp->f_data, uap->mode);
2472 setfown(struct vnode *vp, uid_t uid, gid_t gid)
2474 struct thread *td = curthread;
2475 struct proc *p = td->td_proc;
2480 * note: vget is required for any operation that might mod the vnode
2481 * so VINACTIVE is properly cleared.
2483 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2487 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2494 kern_chown(struct nlookupdata *nd, int uid, int gid)
2499 /* XXX Add NLC flag indicating modifying operation? */
2500 if ((error = nlookup(nd)) != 0)
2502 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
2504 if ((error = ncp_writechk(&nd->nl_nch)) == 0)
2505 error = setfown(vp, uid, gid);
2511 * chown(char *path, int uid, int gid)
2513 * Set ownership given a path name.
2516 sys_chown(struct chown_args *uap)
2518 struct nlookupdata nd;
2521 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2523 error = kern_chown(&nd, uap->uid, uap->gid);
2529 * lchown_args(char *path, int uid, int gid)
2531 * Set ownership given a path name, do not cross symlinks.
2534 sys_lchown(struct lchown_args *uap)
2536 struct nlookupdata nd;
2539 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2541 error = kern_chown(&nd, uap->uid, uap->gid);
2547 * fchown_args(int fd, int uid, int gid)
2549 * Set ownership given a file descriptor.
2553 sys_fchown(struct fchown_args *uap)
2555 struct thread *td = curthread;
2556 struct proc *p = td->td_proc;
2560 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2562 if (fp->f_nchandle.ncp)
2563 error = ncp_writechk(&fp->f_nchandle);
2565 error = setfown((struct vnode *)fp->f_data, uap->uid, uap->gid);
2571 getutimes(const struct timeval *tvp, struct timespec *tsp)
2573 struct timeval tv[2];
2577 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2580 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2581 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2587 setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
2589 struct thread *td = curthread;
2590 struct proc *p = td->td_proc;
2595 * note: vget is required for any operation that might mod the vnode
2596 * so VINACTIVE is properly cleared.
2598 if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
2600 vattr.va_atime = ts[0];
2601 vattr.va_mtime = ts[1];
2603 vattr.va_vaflags |= VA_UTIMES_NULL;
2604 error = VOP_SETATTR(vp, &vattr, p->p_ucred);
2611 kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
2613 struct timespec ts[2];
2617 if ((error = getutimes(tptr, ts)) != 0)
2619 /* XXX Add NLC flag indicating modifying operation? */
2620 if ((error = nlookup(nd)) != 0)
2622 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2624 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
2626 error = setutimes(vp, ts, tptr == NULL);
2632 * utimes_args(char *path, struct timeval *tptr)
2634 * Set the access and modification times of a file.
2637 sys_utimes(struct utimes_args *uap)
2639 struct timeval tv[2];
2640 struct nlookupdata nd;
2644 error = copyin(uap->tptr, tv, sizeof(tv));
2648 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2650 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2656 * lutimes_args(char *path, struct timeval *tptr)
2658 * Set the access and modification times of a file.
2661 sys_lutimes(struct lutimes_args *uap)
2663 struct timeval tv[2];
2664 struct nlookupdata nd;
2668 error = copyin(uap->tptr, tv, sizeof(tv));
2672 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
2674 error = kern_utimes(&nd, uap->tptr ? tv : NULL);
2680 kern_futimes(int fd, struct timeval *tptr)
2682 struct thread *td = curthread;
2683 struct proc *p = td->td_proc;
2684 struct timespec ts[2];
2688 error = getutimes(tptr, ts);
2691 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
2693 if (fp->f_nchandle.ncp)
2694 error = ncp_writechk(&fp->f_nchandle);
2696 error = setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
2702 * futimes_args(int fd, struct timeval *tptr)
2704 * Set the access and modification times of a file.
2707 sys_futimes(struct futimes_args *uap)
2709 struct timeval tv[2];
2713 error = copyin(uap->tptr, tv, sizeof(tv));
2718 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);
2724 kern_truncate(struct nlookupdata *nd, off_t length)
2732 /* XXX Add NLC flag indicating modifying operation? */
2733 if ((error = nlookup(nd)) != 0)
2735 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
2737 if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
2739 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) != 0) {
2743 if (vp->v_type == VDIR) {
2745 } else if ((error = vn_writechk(vp, &nd->nl_nch)) == 0 &&
2746 (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred)) == 0) {
2748 vattr.va_size = length;
2749 error = VOP_SETATTR(vp, &vattr, nd->nl_cred);
2756 * truncate(char *path, int pad, off_t length)
2758 * Truncate a file given its path name.
2761 sys_truncate(struct truncate_args *uap)
2763 struct nlookupdata nd;
2766 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
2768 error = kern_truncate(&nd, uap->length);
2774 kern_ftruncate(int fd, off_t length)
2776 struct thread *td = curthread;
2777 struct proc *p = td->td_proc;
2785 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
2787 if (fp->f_nchandle.ncp) {
2788 error = ncp_writechk(&fp->f_nchandle);
2792 if ((fp->f_flag & FWRITE) == 0) {
2796 vp = (struct vnode *)fp->f_data;
2797 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2798 if (vp->v_type == VDIR) {
2800 } else if ((error = vn_writechk(vp, NULL)) == 0) {
2802 vattr.va_size = length;
2803 error = VOP_SETATTR(vp, &vattr, fp->f_cred);
2812 * ftruncate_args(int fd, int pad, off_t length)
2814 * Truncate a file given a file descriptor.
2817 sys_ftruncate(struct ftruncate_args *uap)
2821 error = kern_ftruncate(uap->fd, uap->length);
2829 * Sync an open file.
2833 sys_fsync(struct fsync_args *uap)
2835 struct thread *td = curthread;
2836 struct proc *p = td->td_proc;
2842 if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
2844 vp = (struct vnode *)fp->f_data;
2845 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2846 if ((obj = vp->v_object) != NULL)
2847 vm_object_page_clean(obj, 0, 0, 0);
2848 if ((error = VOP_FSYNC(vp, MNT_WAIT)) == 0 && vp->v_mount)
2849 error = buf_fsync(vp);
2856 kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
2858 struct nchandle fnchd;
2859 struct nchandle tnchd;
2860 struct namecache *ncp;
2867 fromnd->nl_flags |= NLC_REFDVP;
2868 if ((error = nlookup(fromnd)) != 0)
2870 if ((fnchd.ncp = fromnd->nl_nch.ncp->nc_parent) == NULL)
2872 fnchd.mount = fromnd->nl_nch.mount;
2876 * unlock the source nch so we can lookup the target nch without
2877 * deadlocking. The target may or may not exist so we do not check
2878 * for a target vp like kern_mkdir() and other creation functions do.
2880 * The source and target directories are ref'd and rechecked after
2881 * everything is relocked to determine if the source or target file
2884 KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
2885 fromnd->nl_flags &= ~NLC_NCPISLOCKED;
2886 cache_unlock(&fromnd->nl_nch);
2888 tond->nl_flags |= NLC_CREATE | NLC_REFDVP;
2889 if ((error = nlookup(tond)) != 0) {
2893 if ((tnchd.ncp = tond->nl_nch.ncp->nc_parent) == NULL) {
2897 tnchd.mount = tond->nl_nch.mount;
2901 * If the source and target are the same there is nothing to do
2903 if (fromnd->nl_nch.ncp == tond->nl_nch.ncp) {
2910 * Mount points cannot be renamed or overwritten
2912 if ((fromnd->nl_nch.ncp->nc_flag | tond->nl_nch.ncp->nc_flag) &
2921 * relock the source ncp. NOTE AFTER RELOCKING: the source ncp
2922 * may have become invalid while it was unlocked, nc_vp and nc_mount
2925 if (cache_lock_nonblock(&fromnd->nl_nch) == 0) {
2926 cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
2927 } else if (fromnd->nl_nch.ncp > tond->nl_nch.ncp) {
2928 cache_lock(&fromnd->nl_nch);
2929 cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
2931 cache_unlock(&tond->nl_nch);
2932 cache_lock(&fromnd->nl_nch);
2933 cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
2934 cache_lock(&tond->nl_nch);
2935 cache_resolve(&tond->nl_nch, tond->nl_cred);
2937 fromnd->nl_flags |= NLC_NCPISLOCKED;
2940 * make sure the parent directories linkages are the same
2942 if (fnchd.ncp != fromnd->nl_nch.ncp->nc_parent ||
2943 tnchd.ncp != tond->nl_nch.ncp->nc_parent) {
2950 * Both the source and target must be within the same filesystem and
2951 * in the same filesystem as their parent directories within the
2952 * namecache topology.
2954 * NOTE: fromnd's nc_mount or nc_vp could be NULL.
2957 if (mp != tnchd.mount || mp != fromnd->nl_nch.mount ||
2958 mp != tond->nl_nch.mount) {
2965 * Make sure the mount point is writable
2967 if ((error = ncp_writechk(&tond->nl_nch)) != 0) {
2974 * If the target exists and either the source or target is a directory,
2975 * then both must be directories.
2977 * Due to relocking of the source, fromnd->nl_nch.ncp->nc_vp might h
2980 if (tond->nl_nch.ncp->nc_vp) {
2981 if (fromnd->nl_nch.ncp->nc_vp == NULL) {
2983 } else if (fromnd->nl_nch.ncp->nc_vp->v_type == VDIR) {
2984 if (tond->nl_nch.ncp->nc_vp->v_type != VDIR)
2986 } else if (tond->nl_nch.ncp->nc_vp->v_type == VDIR) {
2992 * You cannot rename a source into itself or a subdirectory of itself.
2993 * We check this by travsersing the target directory upwards looking
2994 * for a match against the source.
2997 for (ncp = tnchd.ncp; ncp; ncp = ncp->nc_parent) {
2998 if (fromnd->nl_nch.ncp == ncp) {
3009 * Even though the namespaces are different, they may still represent
3010 * hardlinks to the same file. The filesystem might have a hard time
3011 * with this so we issue a NREMOVE of the source instead of a NRENAME
3012 * when we detect the situation.
3015 fdvp = fromnd->nl_dvp;
3016 tdvp = tond->nl_dvp;
3017 if (fdvp == NULL || tdvp == NULL) {
3019 } else if (fromnd->nl_nch.ncp->nc_vp == tond->nl_nch.ncp->nc_vp) {
3020 error = VOP_NREMOVE(&fromnd->nl_nch, fdvp,
3023 error = VOP_NRENAME(&fromnd->nl_nch, &tond->nl_nch,
3024 fdvp, tdvp, tond->nl_cred);
3031 * rename_args(char *from, char *to)
3033 * Rename files. Source and destination must either both be directories,
3034 * or both not be directories. If target is a directory, it must be empty.
3037 sys_rename(struct rename_args *uap)
3039 struct nlookupdata fromnd, tond;
3042 error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
3044 error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
3046 error = kern_rename(&fromnd, &tond);
3047 nlookup_done(&tond);
3049 nlookup_done(&fromnd);
3054 kern_mkdir(struct nlookupdata *nd, int mode)
3056 struct thread *td = curthread;
3057 struct proc *p = td->td_proc;
3063 nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE | NLC_REFDVP;
3064 if ((error = nlookup(nd)) != 0)
3067 if (nd->nl_nch.ncp->nc_vp)
3069 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
3072 vattr.va_type = VDIR;
3073 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
3076 error = VOP_NMKDIR(&nd->nl_nch, nd->nl_dvp, &vp, p->p_ucred, &vattr);
3083 * mkdir_args(char *path, int mode)
3085 * Make a directory file.
3089 sys_mkdir(struct mkdir_args *uap)
3091 struct nlookupdata nd;
3094 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3096 error = kern_mkdir(&nd, uap->mode);
3102 kern_rmdir(struct nlookupdata *nd)
3107 nd->nl_flags |= NLC_DELETE | NLC_REFDVP;
3108 if ((error = nlookup(nd)) != 0)
3112 * Do not allow directories representing mount points to be
3113 * deleted, even if empty. Check write perms on mount point
3114 * in case the vnode is aliased (aka nullfs).
3116 if (nd->nl_nch.ncp->nc_flag & (NCF_ISMOUNTPT))
3118 if ((error = ncp_writechk(&nd->nl_nch)) != 0)
3120 error = VOP_NRMDIR(&nd->nl_nch, nd->nl_dvp, nd->nl_cred);
3125 * rmdir_args(char *path)
3127 * Remove a directory file.
3131 sys_rmdir(struct rmdir_args *uap)
3133 struct nlookupdata nd;
3136 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
3138 error = kern_rmdir(&nd);
3144 kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res,
3145 enum uio_seg direction)
3147 struct thread *td = curthread;
3148 struct proc *p = td->td_proc;
3156 if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
3158 if ((fp->f_flag & FREAD) == 0) {
3162 vp = (struct vnode *)fp->f_data;
3164 if (vp->v_type != VDIR) {
3168 aiov.iov_base = buf;
3169 aiov.iov_len = count;
3170 auio.uio_iov = &aiov;
3171 auio.uio_iovcnt = 1;
3172 auio.uio_rw = UIO_READ;
3173 auio.uio_segflg = direction;
3175 auio.uio_resid = count;
3176 loff = auio.uio_offset = fp->f_offset;
3177 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
3178 fp->f_offset = auio.uio_offset;
3181 if (count == auio.uio_resid) {
3182 if (union_dircheckp) {
3183 error = union_dircheckp(td, &vp, fp);
3190 if ((vp->v_flag & VROOT) &&
3191 (vp->v_mount->mnt_flag & MNT_UNION)) {
3192 struct vnode *tvp = vp;
3193 vp = vp->v_mount->mnt_vnodecovered;
3204 * WARNING! *basep may not be wide enough to accomodate the
3205 * seek offset. XXX should we hack this to return the upper 32 bits
3206 * for offsets greater then 4G?
3209 *basep = (long)loff;
3211 *res = count - auio.uio_resid;
3218 * getdirentries_args(int fd, char *buf, u_int conut, long *basep)
3220 * Read a block of directory entries in a file system independent format.
3223 sys_getdirentries(struct getdirentries_args *uap)
3228 error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base,
3229 &uap->sysmsg_result, UIO_USERSPACE);
3231 if (error == 0 && uap->basep)
3232 error = copyout(&base, uap->basep, sizeof(*uap->basep));
3237 * getdents_args(int fd, char *buf, size_t count)
3240 sys_getdents(struct getdents_args *uap)
3244 error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL,
3245 &uap->sysmsg_result, UIO_USERSPACE);
3251 * umask(int newmask)
3253 * Set the mode mask for creation of filesystem nodes.
3258 sys_umask(struct umask_args *uap)
3260 struct thread *td = curthread;
3261 struct proc *p = td->td_proc;
3262 struct filedesc *fdp;
3265 uap->sysmsg_result = fdp->fd_cmask;
3266 fdp->fd_cmask = uap->newmask & ALLPERMS;
3271 * revoke(char *path)
3273 * Void all references to file by ripping underlying filesystem
3278 sys_revoke(struct revoke_args *uap)
3280 struct nlookupdata nd;
3287 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3289 error = nlookup(&nd);
3291 error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
3292 cred = crhold(nd.nl_cred);
3295 if (vp->v_type != VCHR && vp->v_type != VBLK)
3298 error = VOP_GETATTR(vp, &vattr);
3299 if (error == 0 && cred->cr_uid != vattr.va_uid)
3300 error = suser_cred(cred, PRISON_ROOT);
3301 if (error == 0 && count_udev(vp->v_umajor, vp->v_uminor) > 0) {
3304 VOP_REVOKE(vp, REVOKEALL);
3315 * getfh_args(char *fname, fhandle_t *fhp)
3317 * Get (NFS) file handle
3320 sys_getfh(struct getfh_args *uap)
3322 struct thread *td = curthread;
3323 struct nlookupdata nd;
3329 * Must be super user
3331 if ((error = suser(td)) != 0)
3335 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
3337 error = nlookup(&nd);
3339 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
3342 bzero(&fh, sizeof(fh));
3343 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3344 error = VFS_VPTOFH(vp, &fh.fh_fid);
3347 error = copyout(&fh, uap->fhp, sizeof(fh));
3353 * fhopen_args(const struct fhandle *u_fhp, int flags)
3355 * syscall for the rpc.lockd to use to translate a NFS file handle into
3356 * an open descriptor.
3358 * warning: do not remove the suser() call or this becomes one giant
3362 sys_fhopen(struct fhopen_args *uap)
3364 struct thread *td = curthread;
3365 struct proc *p = td->td_proc;
3370 struct vattr *vap = &vat;
3372 int fmode, mode, error, type;
3378 * Must be super user
3384 fmode = FFLAGS(uap->flags);
3385 /* why not allow a non-read/write open for our lockd? */
3386 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
3388 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
3391 /* find the mount point */
3392 mp = vfs_getvfs(&fhp.fh_fsid);
3395 /* now give me my vnode, it gets returned to me locked */
3396 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
3400 * from now on we have to make sure not
3401 * to forget about the vnode
3402 * any error that causes an abort must vput(vp)
3403 * just set error = err and 'goto bad;'.
3409 if (vp->v_type == VLNK) {
3413 if (vp->v_type == VSOCK) {
3418 if (fmode & (FWRITE | O_TRUNC)) {
3419 if (vp->v_type == VDIR) {
3423 error = vn_writechk(vp, NULL);
3431 error = VOP_ACCESS(vp, mode, p->p_ucred);
3435 if (fmode & O_TRUNC) {
3436 vn_unlock(vp); /* XXX */
3437 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
3440 error = VOP_SETATTR(vp, vap, p->p_ucred);
3446 * VOP_OPEN needs the file pointer so it can potentially override
3449 * WARNING! no f_nchandle will be associated when fhopen()ing a
3452 if ((error = falloc(p, &nfp, &indx)) != 0)
3456 error = VOP_OPEN(vp, fmode, p->p_ucred, fp);
3459 * setting f_ops this way prevents VOP_CLOSE from being
3460 * called or fdrop() releasing the vp from v_data. Since
3461 * the VOP_OPEN failed we don't want to VOP_CLOSE.
3463 fp->f_ops = &badfileops;
3469 * The fp is given its own reference, we still have our ref and lock.
3471 * Assert that all regular files must be created with a VM object.
3473 if (vp->v_type == VREG && vp->v_object == NULL) {
3474 kprintf("fhopen: regular file did not have VM object: %p\n", vp);
3479 * The open was successful. Handle any locking requirements.
3481 if (fmode & (O_EXLOCK | O_SHLOCK)) {
3482 lf.l_whence = SEEK_SET;
3485 if (fmode & O_EXLOCK)
3486 lf.l_type = F_WRLCK;
3488 lf.l_type = F_RDLCK;
3489 if (fmode & FNONBLOCK)
3494 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
3496 * release our private reference.
3498 fsetfd(p, NULL, indx);
3503 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3504 fp->f_flag |= FHASLOCK;
3508 * Clean up. Associate the file pointer with the previously
3509 * reserved descriptor and return it.
3512 fsetfd(p, fp, indx);
3514 uap->sysmsg_result = indx;
3518 fsetfd(p, NULL, indx);
3526 * fhstat_args(struct fhandle *u_fhp, struct stat *sb)
3529 sys_fhstat(struct fhstat_args *uap)
3531 struct thread *td = curthread;
3539 * Must be super user
3545 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
3549 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3551 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3553 error = vn_stat(vp, &sb, td->td_proc->p_ucred);
3557 error = copyout(&sb, uap->sb, sizeof(sb));
3562 * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf)
3565 sys_fhstatfs(struct fhstatfs_args *uap)
3567 struct thread *td = curthread;
3568 struct proc *p = td->td_proc;
3573 char *fullpath, *freepath;
3578 * Must be super user
3580 if ((error = suser(td)))
3583 if ((error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t))) != 0)
3586 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
3589 if (p != NULL && !chroot_visible_mnt(mp, p))
3592 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
3597 if ((error = VFS_STATFS(mp, sp, p->p_ucred)) != 0)
3600 error = mount_path(p, mp, &fullpath, &freepath);
3603 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3604 strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
3605 kfree(freepath, M_TEMP);
3607 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
3609 bcopy(sp, &sb, sizeof(sb));
3610 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
3613 return (copyout(sp, uap->buf, sizeof(*sp)));
3617 * Syscall to push extended attribute configuration information into the
3618 * VFS. Accepts a path, which it converts to a mountpoint, as well as
3619 * a command (int cmd), and attribute name and misc data. For now, the
3620 * attribute name is left in userspace for consumption by the VFS_op.
3621 * It will probably be changed to be copied into sysspace by the
3622 * syscall in the future, once issues with various consumers of the
3623 * attribute code have raised their hands.
3625 * Currently this is used only by UFS Extended Attributes.
3628 sys_extattrctl(struct extattrctl_args *uap)
3630 struct nlookupdata nd;
3636 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3638 error = nlookup(&nd);
3640 mp = nd.nl_nch.mount;
3641 error = VFS_EXTATTRCTL(mp, uap->cmd,
3642 uap->attrname, uap->arg,
3650 * Syscall to set a named extended attribute on a file or directory.
3651 * Accepts attribute name, and a uio structure pointing to the data to set.
3652 * The uio is consumed in the style of writev(). The real work happens
3653 * in VOP_SETEXTATTR().
3656 sys_extattr_set_file(struct extattr_set_file_args *uap)
3658 char attrname[EXTATTR_MAXNAMELEN];
3659 struct iovec aiov[UIO_SMALLIOV];
3660 struct iovec *needfree;
3661 struct nlookupdata nd;
3670 error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
3675 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3677 error = nlookup(&nd);
3679 error = ncp_writechk(&nd.nl_nch);
3681 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
3688 iovlen = uap->iovcnt * sizeof(struct iovec);
3689 if (uap->iovcnt > UIO_SMALLIOV) {
3690 if (uap->iovcnt > UIO_MAXIOV) {
3694 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3700 auio.uio_iovcnt = uap->iovcnt;
3701 auio.uio_rw = UIO_WRITE;
3702 auio.uio_segflg = UIO_USERSPACE;
3703 auio.uio_td = nd.nl_td;
3704 auio.uio_offset = 0;
3705 if ((error = copyin(uap->iovp, iov, iovlen)))
3708 for (i = 0; i < uap->iovcnt; i++) {
3709 if (iov->iov_len > INT_MAX - auio.uio_resid) {
3713 auio.uio_resid += iov->iov_len;
3716 cnt = auio.uio_resid;
3717 error = VOP_SETEXTATTR(vp, attrname, &auio, nd.nl_cred);
3718 cnt -= auio.uio_resid;
3719 uap->sysmsg_result = cnt;
3724 FREE(needfree, M_IOV);
3729 * Syscall to get a named extended attribute on a file or directory.
3730 * Accepts attribute name, and a uio structure pointing to a buffer for the
3731 * data. The uio is consumed in the style of readv(). The real work
3732 * happens in VOP_GETEXTATTR();
3735 sys_extattr_get_file(struct extattr_get_file_args *uap)
3737 char attrname[EXTATTR_MAXNAMELEN];
3738 struct iovec aiov[UIO_SMALLIOV];
3739 struct iovec *needfree;
3740 struct nlookupdata nd;
3749 error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
3754 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3756 error = nlookup(&nd);
3758 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
3764 iovlen = uap->iovcnt * sizeof (struct iovec);
3766 if (uap->iovcnt > UIO_SMALLIOV) {
3767 if (uap->iovcnt > UIO_MAXIOV) {
3771 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
3777 auio.uio_iovcnt = uap->iovcnt;
3778 auio.uio_rw = UIO_READ;
3779 auio.uio_segflg = UIO_USERSPACE;
3780 auio.uio_td = nd.nl_td;
3781 auio.uio_offset = 0;
3782 if ((error = copyin(uap->iovp, iov, iovlen)))
3785 for (i = 0; i < uap->iovcnt; i++) {
3786 if (iov->iov_len > INT_MAX - auio.uio_resid) {
3790 auio.uio_resid += iov->iov_len;
3793 cnt = auio.uio_resid;
3794 error = VOP_GETEXTATTR(vp, attrname, &auio, nd.nl_cred);
3795 cnt -= auio.uio_resid;
3796 uap->sysmsg_result = cnt;
3801 FREE(needfree, M_IOV);
3806 * Syscall to delete a named extended attribute from a file or directory.
3807 * Accepts attribute name. The real work happens in VOP_SETEXTATTR().
3810 sys_extattr_delete_file(struct extattr_delete_file_args *uap)
3812 char attrname[EXTATTR_MAXNAMELEN];
3813 struct nlookupdata nd;
3817 error = copyin(uap->attrname, attrname, EXTATTR_MAXNAMELEN);
3822 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
3824 error = nlookup(&nd);
3826 error = ncp_writechk(&nd.nl_nch);
3828 error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
3834 error = VOP_SETEXTATTR(vp, attrname, NULL, nd.nl_cred);
3841 * Determine if the mount is visible to the process.
3844 chroot_visible_mnt(struct mount *mp, struct proc *p)
3846 struct nchandle nch;
3849 * Traverse from the mount point upwards. If we hit the process
3850 * root then the mount point is visible to the process.
3852 nch = mp->mnt_ncmountpt;
3854 if (nch.mount == p->p_fd->fd_nrdir.mount &&
3855 nch.ncp == p->p_fd->fd_nrdir.ncp) {
3858 if (nch.ncp == nch.mount->mnt_ncmountpt.ncp) {
3859 nch = nch.mount->mnt_ncmounton;
3861 nch.ncp = nch.ncp->nc_parent;
3866 * If the mount point is not visible to the process, but the
3867 * process root is in a subdirectory of the mount, return
3870 if (p->p_fd->fd_nrdir.mount == mp)