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 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_kdtrace.h"
43 #include "opt_ktrace.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
49 #include <sys/capsicum.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
72 #include <sys/dirent.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
77 #include <sys/ktrace.h>
80 #include <machine/stdarg.h>
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
90 #include <ufs/ufs/quota.h>
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
94 SDT_PROVIDER_DEFINE(vfs);
95 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
96 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
98 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
99 static int kern_chflags(struct thread *td, const char *path,
100 enum uio_seg pathseg, u_long flags);
101 static int kern_chflagsat(struct thread *td, int fd, const char *path,
102 enum uio_seg pathseg, u_long flags, int atflag);
103 static int setfflags(struct thread *td, struct vnode *, u_long);
104 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
105 static int getutimens(const struct timespec *, enum uio_seg,
106 struct timespec *, int *);
107 static int setutimes(struct thread *td, struct vnode *,
108 const struct timespec *, int, int);
109 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
113 * The module initialization routine for POSIX asynchronous I/O will
114 * set this to the version of AIO that it implements. (Zero means
115 * that it is not implemented.) This value is used here by pathconf()
116 * and in kern_descrip.c by fpathconf().
118 int async_io_version;
121 * Sync each mounted filesystem.
123 #ifndef _SYS_SYSPROTO_H_
132 struct sync_args *uap;
134 struct mount *mp, *nmp;
137 mtx_lock(&mountlist_mtx);
138 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
139 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
140 nmp = TAILQ_NEXT(mp, mnt_list);
143 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
144 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
145 save = curthread_pflags_set(TDP_SYNCIO);
146 vfs_msync(mp, MNT_NOWAIT);
147 VFS_SYNC(mp, MNT_NOWAIT);
148 curthread_pflags_restore(save);
149 vn_finished_write(mp);
151 mtx_lock(&mountlist_mtx);
152 nmp = TAILQ_NEXT(mp, mnt_list);
155 mtx_unlock(&mountlist_mtx);
160 * Change filesystem quotas.
162 #ifndef _SYS_SYSPROTO_H_
163 struct quotactl_args {
171 sys_quotactl(td, uap)
173 register struct quotactl_args /* {
184 AUDIT_ARG_CMD(uap->cmd);
185 AUDIT_ARG_UID(uap->uid);
186 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
188 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
190 if ((error = namei(&nd)) != 0)
192 NDFREE(&nd, NDF_ONLY_PNBUF);
193 mp = nd.ni_vp->v_mount;
196 error = vfs_busy(mp, 0);
200 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
203 * Since quota on operation typically needs to open quota
204 * file, the Q_QUOTAON handler needs to unbusy the mount point
205 * before calling into namei. Otherwise, unmount might be
206 * started between two vfs_busy() invocations (first is our,
207 * second is from mount point cross-walk code in lookup()),
210 * Require that Q_QUOTAON handles the vfs_busy() reference on
211 * its own, always returning with ubusied mount point.
213 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
219 * Used by statfs conversion routines to scale the block size up if
220 * necessary so that all of the block counts are <= 'max_size'. Note
221 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
225 statfs_scale_blocks(struct statfs *sf, long max_size)
230 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
233 * Attempt to scale the block counts to give a more accurate
234 * overview to userland of the ratio of free space to used
235 * space. To do this, find the largest block count and compute
236 * a divisor that lets it fit into a signed integer <= max_size.
238 if (sf->f_bavail < 0)
239 count = -sf->f_bavail;
241 count = sf->f_bavail;
242 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
243 if (count <= max_size)
246 count >>= flsl(max_size);
253 sf->f_bsize <<= shift;
254 sf->f_blocks >>= shift;
255 sf->f_bfree >>= shift;
256 sf->f_bavail >>= shift;
260 * Get filesystem statistics.
262 #ifndef _SYS_SYSPROTO_H_
271 register struct statfs_args /* {
279 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
281 error = copyout(&sf, uap->buf, sizeof(sf));
286 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
290 struct statfs *sp, sb;
294 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
299 mp = nd.ni_vp->v_mount;
301 NDFREE(&nd, NDF_ONLY_PNBUF);
303 error = vfs_busy(mp, 0);
308 error = mac_mount_check_stat(td->td_ucred, mp);
313 * Set these in case the underlying filesystem fails to do so.
316 sp->f_version = STATFS_VERSION;
317 sp->f_namemax = NAME_MAX;
318 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
319 error = VFS_STATFS(mp, sp);
322 if (priv_check(td, PRIV_VFS_GENERATION)) {
323 bcopy(sp, &sb, sizeof(sb));
324 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
325 prison_enforce_statfs(td->td_ucred, mp, &sb);
335 * Get filesystem statistics.
337 #ifndef _SYS_SYSPROTO_H_
338 struct fstatfs_args {
346 register struct fstatfs_args /* {
354 error = kern_fstatfs(td, uap->fd, &sf);
356 error = copyout(&sf, uap->buf, sizeof(sf));
361 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
365 struct statfs *sp, sb;
371 error = getvnode(td->td_proc->p_fd, fd,
372 cap_rights_init(&rights, CAP_FSTATFS), &fp);
376 vn_lock(vp, LK_SHARED | LK_RETRY);
378 AUDIT_ARG_VNODE1(vp);
389 error = vfs_busy(mp, 0);
394 error = mac_mount_check_stat(td->td_ucred, mp);
399 * Set these in case the underlying filesystem fails to do so.
402 sp->f_version = STATFS_VERSION;
403 sp->f_namemax = NAME_MAX;
404 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
405 error = VFS_STATFS(mp, sp);
408 if (priv_check(td, PRIV_VFS_GENERATION)) {
409 bcopy(sp, &sb, sizeof(sb));
410 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
411 prison_enforce_statfs(td->td_ucred, mp, &sb);
422 * Get statistics on all filesystems.
424 #ifndef _SYS_SYSPROTO_H_
425 struct getfsstat_args {
432 sys_getfsstat(td, uap)
434 register struct getfsstat_args /* {
441 return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
446 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
447 * The caller is responsible for freeing memory which will be allocated
451 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
452 enum uio_seg bufseg, int flags)
454 struct mount *mp, *nmp;
455 struct statfs *sfsp, *sp, sb;
456 size_t count, maxcount;
459 maxcount = bufsize / sizeof(struct statfs);
462 else if (bufseg == UIO_USERSPACE)
464 else /* if (bufseg == UIO_SYSSPACE) */ {
466 mtx_lock(&mountlist_mtx);
467 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
470 mtx_unlock(&mountlist_mtx);
471 if (maxcount > count)
473 sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
477 mtx_lock(&mountlist_mtx);
478 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
479 if (prison_canseemount(td->td_ucred, mp) != 0) {
480 nmp = TAILQ_NEXT(mp, mnt_list);
484 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
485 nmp = TAILQ_NEXT(mp, mnt_list);
489 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
490 nmp = TAILQ_NEXT(mp, mnt_list);
493 if (sfsp && count < maxcount) {
496 * Set these in case the underlying filesystem
499 sp->f_version = STATFS_VERSION;
500 sp->f_namemax = NAME_MAX;
501 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
503 * If MNT_NOWAIT or MNT_LAZY is specified, do not
504 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
505 * overrides MNT_WAIT.
507 if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
508 (flags & MNT_WAIT)) &&
509 (error = VFS_STATFS(mp, sp))) {
510 mtx_lock(&mountlist_mtx);
511 nmp = TAILQ_NEXT(mp, mnt_list);
515 if (priv_check(td, PRIV_VFS_GENERATION)) {
516 bcopy(sp, &sb, sizeof(sb));
517 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
518 prison_enforce_statfs(td->td_ucred, mp, &sb);
521 if (bufseg == UIO_SYSSPACE)
522 bcopy(sp, sfsp, sizeof(*sp));
523 else /* if (bufseg == UIO_USERSPACE) */ {
524 error = copyout(sp, sfsp, sizeof(*sp));
533 mtx_lock(&mountlist_mtx);
534 nmp = TAILQ_NEXT(mp, mnt_list);
537 mtx_unlock(&mountlist_mtx);
538 if (sfsp && count > maxcount)
539 td->td_retval[0] = maxcount;
541 td->td_retval[0] = count;
545 #ifdef COMPAT_FREEBSD4
547 * Get old format filesystem statistics.
549 static void cvtstatfs(struct statfs *, struct ostatfs *);
551 #ifndef _SYS_SYSPROTO_H_
552 struct freebsd4_statfs_args {
558 freebsd4_statfs(td, uap)
560 struct freebsd4_statfs_args /* {
569 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
572 cvtstatfs(&sf, &osb);
573 return (copyout(&osb, uap->buf, sizeof(osb)));
577 * Get filesystem statistics.
579 #ifndef _SYS_SYSPROTO_H_
580 struct freebsd4_fstatfs_args {
586 freebsd4_fstatfs(td, uap)
588 struct freebsd4_fstatfs_args /* {
597 error = kern_fstatfs(td, uap->fd, &sf);
600 cvtstatfs(&sf, &osb);
601 return (copyout(&osb, uap->buf, sizeof(osb)));
605 * Get statistics on all filesystems.
607 #ifndef _SYS_SYSPROTO_H_
608 struct freebsd4_getfsstat_args {
615 freebsd4_getfsstat(td, uap)
617 register struct freebsd4_getfsstat_args /* {
623 struct statfs *buf, *sp;
628 count = uap->bufsize / sizeof(struct ostatfs);
629 size = count * sizeof(struct statfs);
630 error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
632 count = td->td_retval[0];
634 while (count > 0 && error == 0) {
636 error = copyout(&osb, uap->buf, sizeof(osb));
647 * Implement fstatfs() for (NFS) file handles.
649 #ifndef _SYS_SYSPROTO_H_
650 struct freebsd4_fhstatfs_args {
651 struct fhandle *u_fhp;
656 freebsd4_fhstatfs(td, uap)
658 struct freebsd4_fhstatfs_args /* {
659 struct fhandle *u_fhp;
668 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
671 error = kern_fhstatfs(td, fh, &sf);
674 cvtstatfs(&sf, &osb);
675 return (copyout(&osb, uap->buf, sizeof(osb)));
679 * Convert a new format statfs structure to an old format statfs structure.
687 statfs_scale_blocks(nsp, LONG_MAX);
688 bzero(osp, sizeof(*osp));
689 osp->f_bsize = nsp->f_bsize;
690 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
691 osp->f_blocks = nsp->f_blocks;
692 osp->f_bfree = nsp->f_bfree;
693 osp->f_bavail = nsp->f_bavail;
694 osp->f_files = MIN(nsp->f_files, LONG_MAX);
695 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
696 osp->f_owner = nsp->f_owner;
697 osp->f_type = nsp->f_type;
698 osp->f_flags = nsp->f_flags;
699 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
700 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
701 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
702 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
703 strlcpy(osp->f_fstypename, nsp->f_fstypename,
704 MIN(MFSNAMELEN, OMFSNAMELEN));
705 strlcpy(osp->f_mntonname, nsp->f_mntonname,
706 MIN(MNAMELEN, OMNAMELEN));
707 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
708 MIN(MNAMELEN, OMNAMELEN));
709 osp->f_fsid = nsp->f_fsid;
711 #endif /* COMPAT_FREEBSD4 */
714 * Change current working directory to a given file descriptor.
716 #ifndef _SYS_SYSPROTO_H_
724 struct fchdir_args /* {
728 register struct filedesc *fdp = td->td_proc->p_fd;
729 struct vnode *vp, *tdp, *vpold;
735 AUDIT_ARG_FD(uap->fd);
736 error = getvnode(fdp, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
743 vn_lock(vp, LK_SHARED | LK_RETRY);
744 AUDIT_ARG_VNODE1(vp);
745 error = change_dir(vp, td);
746 while (!error && (mp = vp->v_mountedhere) != NULL) {
749 error = VFS_ROOT(mp, LK_SHARED, &tdp);
762 vpold = fdp->fd_cdir;
764 FILEDESC_XUNLOCK(fdp);
770 * Change current working directory (``.'').
772 #ifndef _SYS_SYSPROTO_H_
780 struct chdir_args /* {
785 return (kern_chdir(td, uap->path, UIO_USERSPACE));
789 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
791 register struct filedesc *fdp = td->td_proc->p_fd;
796 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
798 if ((error = namei(&nd)) != 0)
800 if ((error = change_dir(nd.ni_vp, td)) != 0) {
802 NDFREE(&nd, NDF_ONLY_PNBUF);
805 VOP_UNLOCK(nd.ni_vp, 0);
806 NDFREE(&nd, NDF_ONLY_PNBUF);
809 fdp->fd_cdir = nd.ni_vp;
810 FILEDESC_XUNLOCK(fdp);
816 * Helper function for raised chroot(2) security function: Refuse if
817 * any filedescriptors are open directories.
820 chroot_refuse_vdir_fds(fdp)
821 struct filedesc *fdp;
827 FILEDESC_LOCK_ASSERT(fdp);
829 for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
830 fp = fget_locked(fdp, fd);
833 if (fp->f_type == DTYPE_VNODE) {
835 if (vp->v_type == VDIR)
843 * This sysctl determines if we will allow a process to chroot(2) if it
844 * has a directory open:
845 * 0: disallowed for all processes.
846 * 1: allowed for processes that were not already chroot(2)'ed.
847 * 2: allowed for all processes.
850 static int chroot_allow_open_directories = 1;
852 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
853 &chroot_allow_open_directories, 0,
854 "Allow a process to chroot(2) if it has a directory open");
857 * Change notion of root (``/'') directory.
859 #ifndef _SYS_SYSPROTO_H_
867 struct chroot_args /* {
874 error = priv_check(td, PRIV_VFS_CHROOT);
877 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
878 UIO_USERSPACE, uap->path, td);
882 error = change_dir(nd.ni_vp, td);
886 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
890 VOP_UNLOCK(nd.ni_vp, 0);
891 error = change_root(nd.ni_vp, td);
893 NDFREE(&nd, NDF_ONLY_PNBUF);
898 NDFREE(&nd, NDF_ONLY_PNBUF);
903 * Common routine for chroot and chdir. Callers must provide a locked vnode
915 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
916 if (vp->v_type != VDIR)
919 error = mac_vnode_check_chdir(td->td_ucred, vp);
923 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
927 * Common routine for kern_chroot() and jail_attach(). The caller is
928 * responsible for invoking priv_check() and mac_vnode_check_chroot() to
929 * authorize this operation.
936 struct filedesc *fdp;
940 fdp = td->td_proc->p_fd;
942 if (chroot_allow_open_directories == 0 ||
943 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
944 error = chroot_refuse_vdir_fds(fdp);
946 FILEDESC_XUNLOCK(fdp);
950 oldvp = fdp->fd_rdir;
957 FILEDESC_XUNLOCK(fdp);
963 flags_to_rights(int flags, cap_rights_t *rightsp)
966 if (flags & O_EXEC) {
967 cap_rights_set(rightsp, CAP_FEXECVE);
969 switch ((flags & O_ACCMODE)) {
971 cap_rights_set(rightsp, CAP_READ);
974 cap_rights_set(rightsp, CAP_READ);
977 cap_rights_set(rightsp, CAP_WRITE);
978 if (!(flags & (O_APPEND | O_TRUNC)))
979 cap_rights_set(rightsp, CAP_SEEK);
985 cap_rights_set(rightsp, CAP_CREATE);
988 cap_rights_set(rightsp, CAP_FTRUNCATE);
990 if (flags & (O_SYNC | O_FSYNC))
991 cap_rights_set(rightsp, CAP_FSYNC);
993 if (flags & (O_EXLOCK | O_SHLOCK))
994 cap_rights_set(rightsp, CAP_FLOCK);
998 * Check permissions, allocate an open file structure, and call the device
999 * open routine if any.
1001 #ifndef _SYS_SYSPROTO_H_
1011 register struct open_args /* {
1018 return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1021 #ifndef _SYS_SYSPROTO_H_
1022 struct openat_args {
1030 sys_openat(struct thread *td, struct openat_args *uap)
1033 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1038 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1042 return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1046 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1047 int flags, int mode)
1049 struct proc *p = td->td_proc;
1050 struct filedesc *fdp = p->p_fd;
1053 struct nameidata nd;
1054 cap_rights_t rights;
1055 int cmode, error, indx;
1059 AUDIT_ARG_FFLAGS(flags);
1060 AUDIT_ARG_MODE(mode);
1061 /* XXX: audit dirfd */
1062 cap_rights_init(&rights, CAP_LOOKUP);
1063 flags_to_rights(flags, &rights);
1065 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1068 if (flags & O_EXEC) {
1069 if (flags & O_ACCMODE)
1071 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1074 flags = FFLAGS(flags);
1078 * Allocate the file descriptor, but don't install a descriptor yet.
1080 error = falloc_noinstall(td, &fp);
1084 * An extra reference on `fp' has been held for us by
1085 * falloc_noinstall().
1087 /* Set the flags early so the finit in devfs can pick them up. */
1088 fp->f_flag = flags & FMASK;
1089 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1090 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1092 td->td_dupfd = -1; /* XXX check for fdopen */
1093 error = vn_open(&nd, &flags, cmode, fp);
1096 * If the vn_open replaced the method vector, something
1097 * wonderous happened deep below and we just pass it up
1098 * pretending we know what we do.
1100 if (error == ENXIO && fp->f_ops != &badfileops)
1104 * Handle special fdopen() case. bleh.
1106 * Don't do this for relative (capability) lookups; we don't
1107 * understand exactly what would happen, and we don't think
1108 * that it ever should.
1110 if (nd.ni_strictrelative == 0 &&
1111 (error == ENODEV || error == ENXIO) &&
1112 td->td_dupfd >= 0) {
1113 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1122 NDFREE(&nd, NDF_ONLY_PNBUF);
1126 * Store the vnode, for any f_type. Typically, the vnode use
1127 * count is decremented by direct call to vn_closefile() for
1128 * files that switched type in the cdevsw fdopen() method.
1132 * If the file wasn't claimed by devfs bind it to the normal
1133 * vnode operations here.
1135 if (fp->f_ops == &badfileops) {
1136 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1138 finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1139 DTYPE_VNODE, vp, &vnops);
1143 if (flags & O_TRUNC) {
1144 error = fo_truncate(fp, 0, td->td_ucred, td);
1150 * If we haven't already installed the FD (for dupfdopen), do so now.
1153 struct filecaps *fcaps;
1156 if (nd.ni_strictrelative == 1)
1157 fcaps = &nd.ni_filecaps;
1161 error = finstall(td, fp, &indx, flags, fcaps);
1162 /* On success finstall() consumes fcaps. */
1164 filecaps_free(&nd.ni_filecaps);
1168 filecaps_free(&nd.ni_filecaps);
1172 * Release our private reference, leaving the one associated with
1173 * the descriptor table intact.
1176 td->td_retval[0] = indx;
1179 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1188 #ifndef _SYS_SYSPROTO_H_
1189 struct ocreat_args {
1197 register struct ocreat_args /* {
1203 return (kern_open(td, uap->path, UIO_USERSPACE,
1204 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1206 #endif /* COMPAT_43 */
1209 * Create a special file.
1211 #ifndef _SYS_SYSPROTO_H_
1221 register struct mknod_args /* {
1228 return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1231 #ifndef _SYS_SYSPROTO_H_
1232 struct mknodat_args {
1240 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1243 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1248 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1252 return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1256 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1262 struct nameidata nd;
1263 cap_rights_t rights;
1264 int error, whiteout = 0;
1266 AUDIT_ARG_MODE(mode);
1268 switch (mode & S_IFMT) {
1271 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1274 error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1277 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1281 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1291 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1292 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1294 if ((error = namei(&nd)) != 0)
1298 NDFREE(&nd, NDF_ONLY_PNBUF);
1299 if (vp == nd.ni_dvp)
1307 vattr.va_mode = (mode & ALLPERMS) &
1308 ~td->td_proc->p_fd->fd_cmask;
1309 vattr.va_rdev = dev;
1312 switch (mode & S_IFMT) {
1313 case S_IFMT: /* used by badsect to flag bad sectors */
1314 vattr.va_type = VBAD;
1317 vattr.va_type = VCHR;
1320 vattr.va_type = VBLK;
1326 panic("kern_mknod: invalid mode");
1329 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1330 NDFREE(&nd, NDF_ONLY_PNBUF);
1332 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1337 if (error == 0 && !whiteout)
1338 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1339 &nd.ni_cnd, &vattr);
1343 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1345 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1346 &nd.ni_cnd, &vattr);
1351 NDFREE(&nd, NDF_ONLY_PNBUF);
1353 vn_finished_write(mp);
1358 * Create a named pipe.
1360 #ifndef _SYS_SYSPROTO_H_
1361 struct mkfifo_args {
1369 register struct mkfifo_args /* {
1375 return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1378 #ifndef _SYS_SYSPROTO_H_
1379 struct mkfifoat_args {
1386 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1389 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1394 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1397 return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1401 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1406 struct nameidata nd;
1407 cap_rights_t rights;
1410 AUDIT_ARG_MODE(mode);
1413 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1414 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1416 if ((error = namei(&nd)) != 0)
1418 if (nd.ni_vp != NULL) {
1419 NDFREE(&nd, NDF_ONLY_PNBUF);
1420 if (nd.ni_vp == nd.ni_dvp)
1427 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1428 NDFREE(&nd, NDF_ONLY_PNBUF);
1430 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1435 vattr.va_type = VFIFO;
1436 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1438 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1443 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1450 vn_finished_write(mp);
1451 NDFREE(&nd, NDF_ONLY_PNBUF);
1456 * Make a hard file link.
1458 #ifndef _SYS_SYSPROTO_H_
1467 register struct link_args /* {
1473 return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1476 #ifndef _SYS_SYSPROTO_H_
1477 struct linkat_args {
1486 sys_linkat(struct thread *td, struct linkat_args *uap)
1491 if (flag & ~AT_SYMLINK_FOLLOW)
1494 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1495 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1498 int hardlink_check_uid = 0;
1499 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1500 &hardlink_check_uid, 0,
1501 "Unprivileged processes cannot create hard links to files owned by other "
1503 static int hardlink_check_gid = 0;
1504 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1505 &hardlink_check_gid, 0,
1506 "Unprivileged processes cannot create hard links to files owned by other "
1510 can_hardlink(struct vnode *vp, struct ucred *cred)
1515 if (!hardlink_check_uid && !hardlink_check_gid)
1518 error = VOP_GETATTR(vp, &va, cred);
1522 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1523 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1528 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1529 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1538 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1541 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1545 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1546 enum uio_seg segflg, int follow)
1550 struct nameidata nd;
1551 cap_rights_t rights;
1556 NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1558 if ((error = namei(&nd)) != 0)
1560 NDFREE(&nd, NDF_ONLY_PNBUF);
1562 if (vp->v_type == VDIR) {
1564 return (EPERM); /* POSIX */
1566 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2 |
1567 NOCACHE, segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT),
1569 if ((error = namei(&nd)) == 0) {
1570 if (nd.ni_vp != NULL) {
1571 NDFREE(&nd, NDF_ONLY_PNBUF);
1572 if (nd.ni_dvp == nd.ni_vp)
1579 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1581 * Cross-device link. No need to recheck
1582 * vp->v_type, since it cannot change, except
1585 NDFREE(&nd, NDF_ONLY_PNBUF);
1589 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1590 error = can_hardlink(vp, td->td_ucred);
1593 error = mac_vnode_check_link(td->td_ucred,
1594 nd.ni_dvp, vp, &nd.ni_cnd);
1599 NDFREE(&nd, NDF_ONLY_PNBUF);
1602 error = vn_start_write(vp, &mp, V_NOWAIT);
1606 NDFREE(&nd, NDF_ONLY_PNBUF);
1607 error = vn_start_write(NULL, &mp,
1613 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1616 vn_finished_write(mp);
1617 NDFREE(&nd, NDF_ONLY_PNBUF);
1620 NDFREE(&nd, NDF_ONLY_PNBUF);
1630 * Make a symbolic link.
1632 #ifndef _SYS_SYSPROTO_H_
1633 struct symlink_args {
1639 sys_symlink(td, uap)
1641 register struct symlink_args /* {
1647 return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1650 #ifndef _SYS_SYSPROTO_H_
1651 struct symlinkat_args {
1658 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1661 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1666 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1669 return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1673 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1674 enum uio_seg segflg)
1679 struct nameidata nd;
1681 cap_rights_t rights;
1683 if (segflg == UIO_SYSSPACE) {
1686 syspath = uma_zalloc(namei_zone, M_WAITOK);
1687 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1690 AUDIT_ARG_TEXT(syspath);
1693 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1694 NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1696 if ((error = namei(&nd)) != 0)
1699 NDFREE(&nd, NDF_ONLY_PNBUF);
1700 if (nd.ni_vp == nd.ni_dvp)
1708 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1709 NDFREE(&nd, NDF_ONLY_PNBUF);
1711 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1716 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1718 vattr.va_type = VLNK;
1719 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1724 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1730 NDFREE(&nd, NDF_ONLY_PNBUF);
1732 vn_finished_write(mp);
1734 if (segflg != UIO_SYSSPACE)
1735 uma_zfree(namei_zone, syspath);
1740 * Delete a whiteout from the filesystem.
1743 sys_undelete(td, uap)
1745 register struct undelete_args /* {
1750 struct nameidata nd;
1755 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1756 UIO_USERSPACE, uap->path, td);
1761 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1762 NDFREE(&nd, NDF_ONLY_PNBUF);
1763 if (nd.ni_vp == nd.ni_dvp)
1771 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1772 NDFREE(&nd, NDF_ONLY_PNBUF);
1774 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1778 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1779 NDFREE(&nd, NDF_ONLY_PNBUF);
1781 vn_finished_write(mp);
1786 * Delete a name from the filesystem.
1788 #ifndef _SYS_SYSPROTO_H_
1789 struct unlink_args {
1796 struct unlink_args /* {
1801 return (kern_unlink(td, uap->path, UIO_USERSPACE));
1804 #ifndef _SYS_SYSPROTO_H_
1805 struct unlinkat_args {
1812 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1814 int flag = uap->flag;
1816 char *path = uap->path;
1818 if (flag & ~AT_REMOVEDIR)
1821 if (flag & AT_REMOVEDIR)
1822 return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1824 return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1828 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1831 return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1835 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1840 struct nameidata nd;
1842 cap_rights_t rights;
1847 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1848 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1849 if ((error = namei(&nd)) != 0)
1850 return (error == EINVAL ? EPERM : error);
1852 if (vp->v_type == VDIR && oldinum == 0) {
1853 error = EPERM; /* POSIX */
1854 } else if (oldinum != 0 &&
1855 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1856 sb.st_ino != oldinum) {
1857 error = EIDRM; /* Identifier removed */
1860 * The root of a mounted filesystem cannot be deleted.
1862 * XXX: can this only be a VDIR case?
1864 if (vp->v_vflag & VV_ROOT)
1868 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1869 NDFREE(&nd, NDF_ONLY_PNBUF);
1871 if (vp == nd.ni_dvp)
1875 if ((error = vn_start_write(NULL, &mp,
1876 V_XSLEEP | PCATCH)) != 0)
1881 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1886 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1887 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1891 vn_finished_write(mp);
1893 NDFREE(&nd, NDF_ONLY_PNBUF);
1895 if (vp == nd.ni_dvp)
1903 * Reposition read/write file offset.
1905 #ifndef _SYS_SYSPROTO_H_
1916 register struct lseek_args /* {
1924 cap_rights_t rights;
1927 AUDIT_ARG_FD(uap->fd);
1928 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1931 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1932 fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1937 #if defined(COMPAT_43)
1939 * Reposition read/write file offset.
1941 #ifndef _SYS_SYSPROTO_H_
1942 struct olseek_args {
1951 register struct olseek_args /* {
1957 struct lseek_args /* {
1965 nuap.offset = uap->offset;
1966 nuap.whence = uap->whence;
1967 return (sys_lseek(td, &nuap));
1969 #endif /* COMPAT_43 */
1971 /* Version with the 'pad' argument */
1973 freebsd6_lseek(td, uap)
1975 register struct freebsd6_lseek_args *uap;
1977 struct lseek_args ouap;
1980 ouap.offset = uap->offset;
1981 ouap.whence = uap->whence;
1982 return (sys_lseek(td, &ouap));
1986 * Check access permissions using passed credentials.
1989 vn_access(vp, user_flags, cred, td)
1998 /* Flags == 0 means only check for existence. */
2002 if (user_flags & R_OK)
2004 if (user_flags & W_OK)
2006 if (user_flags & X_OK)
2009 error = mac_vnode_check_access(cred, vp, accmode);
2013 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2014 error = VOP_ACCESS(vp, accmode, cred, td);
2020 * Check access permissions using "real" credentials.
2022 #ifndef _SYS_SYSPROTO_H_
2023 struct access_args {
2031 register struct access_args /* {
2037 return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2040 #ifndef _SYS_SYSPROTO_H_
2041 struct faccessat_args {
2049 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2052 if (uap->flag & ~AT_EACCESS)
2054 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2059 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2062 return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2066 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2067 int flag, int amode)
2069 struct ucred *cred, *tmpcred;
2071 struct nameidata nd;
2072 cap_rights_t rights;
2076 * Create and modify a temporary credential instead of one that
2077 * is potentially shared.
2079 if (!(flag & AT_EACCESS)) {
2080 cred = td->td_ucred;
2081 tmpcred = crdup(cred);
2082 tmpcred->cr_uid = cred->cr_ruid;
2083 tmpcred->cr_groups[0] = cred->cr_rgid;
2084 td->td_ucred = tmpcred;
2086 cred = tmpcred = td->td_ucred;
2087 AUDIT_ARG_VALUE(amode);
2088 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2089 AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
2091 if ((error = namei(&nd)) != 0)
2095 error = vn_access(vp, amode, tmpcred, td);
2096 NDFREE(&nd, NDF_ONLY_PNBUF);
2099 if (!(flag & AT_EACCESS)) {
2100 td->td_ucred = cred;
2107 * Check access permissions using "effective" credentials.
2109 #ifndef _SYS_SYSPROTO_H_
2110 struct eaccess_args {
2116 sys_eaccess(td, uap)
2118 register struct eaccess_args /* {
2124 return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2128 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2131 return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2134 #if defined(COMPAT_43)
2136 * Get file status; this version follows links.
2138 #ifndef _SYS_SYSPROTO_H_
2147 register struct ostat_args /* {
2156 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2160 return (copyout(&osb, uap->ub, sizeof (osb)));
2164 * Get file status; this version does not follow links.
2166 #ifndef _SYS_SYSPROTO_H_
2167 struct olstat_args {
2175 register struct olstat_args /* {
2184 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2188 return (copyout(&osb, uap->ub, sizeof (osb)));
2192 * Convert from an old to a new stat structure.
2200 ost->st_dev = st->st_dev;
2201 ost->st_ino = st->st_ino;
2202 ost->st_mode = st->st_mode;
2203 ost->st_nlink = st->st_nlink;
2204 ost->st_uid = st->st_uid;
2205 ost->st_gid = st->st_gid;
2206 ost->st_rdev = st->st_rdev;
2207 if (st->st_size < (quad_t)1 << 32)
2208 ost->st_size = st->st_size;
2211 ost->st_atim = st->st_atim;
2212 ost->st_mtim = st->st_mtim;
2213 ost->st_ctim = st->st_ctim;
2214 ost->st_blksize = st->st_blksize;
2215 ost->st_blocks = st->st_blocks;
2216 ost->st_flags = st->st_flags;
2217 ost->st_gen = st->st_gen;
2219 #endif /* COMPAT_43 */
2222 * Get file status; this version follows links.
2224 #ifndef _SYS_SYSPROTO_H_
2233 register struct stat_args /* {
2241 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2243 error = copyout(&sb, uap->ub, sizeof (sb));
2247 #ifndef _SYS_SYSPROTO_H_
2248 struct fstatat_args {
2256 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2261 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2262 UIO_USERSPACE, &sb);
2264 error = copyout(&sb, uap->buf, sizeof (sb));
2269 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2272 return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2276 kern_statat(struct thread *td, int flag, int fd, char *path,
2277 enum uio_seg pathseg, struct stat *sbp)
2280 return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2284 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2285 enum uio_seg pathseg, struct stat *sbp,
2286 void (*hook)(struct vnode *vp, struct stat *sbp))
2288 struct nameidata nd;
2290 cap_rights_t rights;
2293 if (flag & ~AT_SYMLINK_NOFOLLOW)
2296 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2297 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2298 cap_rights_init(&rights, CAP_FSTAT), td);
2300 if ((error = namei(&nd)) != 0)
2302 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2304 SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2305 if (S_ISREG(sb.st_mode))
2306 SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2307 if (__predict_false(hook != NULL))
2308 hook(nd.ni_vp, &sb);
2310 NDFREE(&nd, NDF_ONLY_PNBUF);
2316 if (KTRPOINT(td, KTR_STRUCT))
2323 * Get file status; this version does not follow links.
2325 #ifndef _SYS_SYSPROTO_H_
2334 register struct lstat_args /* {
2342 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2344 error = copyout(&sb, uap->ub, sizeof (sb));
2349 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2352 return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2357 * Implementation of the NetBSD [l]stat() functions.
2365 bzero(nsb, sizeof *nsb);
2366 nsb->st_dev = sb->st_dev;
2367 nsb->st_ino = sb->st_ino;
2368 nsb->st_mode = sb->st_mode;
2369 nsb->st_nlink = sb->st_nlink;
2370 nsb->st_uid = sb->st_uid;
2371 nsb->st_gid = sb->st_gid;
2372 nsb->st_rdev = sb->st_rdev;
2373 nsb->st_atim = sb->st_atim;
2374 nsb->st_mtim = sb->st_mtim;
2375 nsb->st_ctim = sb->st_ctim;
2376 nsb->st_size = sb->st_size;
2377 nsb->st_blocks = sb->st_blocks;
2378 nsb->st_blksize = sb->st_blksize;
2379 nsb->st_flags = sb->st_flags;
2380 nsb->st_gen = sb->st_gen;
2381 nsb->st_birthtim = sb->st_birthtim;
2384 #ifndef _SYS_SYSPROTO_H_
2393 register struct nstat_args /* {
2402 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2405 cvtnstat(&sb, &nsb);
2406 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2410 * NetBSD lstat. Get file status; this version does not follow links.
2412 #ifndef _SYS_SYSPROTO_H_
2421 register struct nlstat_args /* {
2430 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2433 cvtnstat(&sb, &nsb);
2434 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2438 * Get configurable pathname variables.
2440 #ifndef _SYS_SYSPROTO_H_
2441 struct pathconf_args {
2447 sys_pathconf(td, uap)
2449 register struct pathconf_args /* {
2455 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2458 #ifndef _SYS_SYSPROTO_H_
2459 struct lpathconf_args {
2465 sys_lpathconf(td, uap)
2467 register struct lpathconf_args /* {
2473 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2478 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2481 struct nameidata nd;
2484 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2486 if ((error = namei(&nd)) != 0)
2488 NDFREE(&nd, NDF_ONLY_PNBUF);
2490 /* If asynchronous I/O is available, it works for all files. */
2491 if (name == _PC_ASYNC_IO)
2492 td->td_retval[0] = async_io_version;
2494 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2500 * Return target name of a symbolic link.
2502 #ifndef _SYS_SYSPROTO_H_
2503 struct readlink_args {
2510 sys_readlink(td, uap)
2512 register struct readlink_args /* {
2519 return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2520 UIO_USERSPACE, uap->count));
2522 #ifndef _SYS_SYSPROTO_H_
2523 struct readlinkat_args {
2531 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2534 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2535 uap->buf, UIO_USERSPACE, uap->bufsize));
2539 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2540 enum uio_seg bufseg, size_t count)
2543 return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2548 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2549 char *buf, enum uio_seg bufseg, size_t count)
2554 struct nameidata nd;
2557 if (count > IOSIZE_MAX)
2560 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2561 pathseg, path, fd, td);
2563 if ((error = namei(&nd)) != 0)
2565 NDFREE(&nd, NDF_ONLY_PNBUF);
2568 error = mac_vnode_check_readlink(td->td_ucred, vp);
2574 if (vp->v_type != VLNK)
2577 aiov.iov_base = buf;
2578 aiov.iov_len = count;
2579 auio.uio_iov = &aiov;
2580 auio.uio_iovcnt = 1;
2581 auio.uio_offset = 0;
2582 auio.uio_rw = UIO_READ;
2583 auio.uio_segflg = bufseg;
2585 auio.uio_resid = count;
2586 error = VOP_READLINK(vp, &auio, td->td_ucred);
2587 td->td_retval[0] = count - auio.uio_resid;
2594 * Common implementation code for chflags() and fchflags().
2597 setfflags(td, vp, flags)
2606 /* We can't support the value matching VNOVAL. */
2607 if (flags == VNOVAL)
2608 return (EOPNOTSUPP);
2611 * Prevent non-root users from setting flags on devices. When
2612 * a device is reused, users can retain ownership of the device
2613 * if they are allowed to set flags and programs assume that
2614 * chown can't fail when done as root.
2616 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2617 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2622 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2625 vattr.va_flags = flags;
2626 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2628 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2631 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2633 vn_finished_write(mp);
2638 * Change flags of a file given a path name.
2640 #ifndef _SYS_SYSPROTO_H_
2641 struct chflags_args {
2647 sys_chflags(td, uap)
2649 register struct chflags_args /* {
2655 return (kern_chflags(td, uap->path, UIO_USERSPACE, uap->flags));
2658 #ifndef _SYS_SYSPROTO_H_
2659 struct chflagsat_args {
2667 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2670 const char *path = uap->path;
2671 u_long flags = uap->flags;
2672 int atflag = uap->atflag;
2674 if (atflag & ~AT_SYMLINK_NOFOLLOW)
2677 return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2681 kern_chflags(struct thread *td, const char *path, enum uio_seg pathseg,
2685 return (kern_chflagsat(td, AT_FDCWD, path, pathseg, flags, 0));
2689 * Same as chflags() but doesn't follow symlinks.
2692 sys_lchflags(td, uap)
2694 register struct lchflags_args /* {
2700 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2701 uap->flags, AT_SYMLINK_NOFOLLOW));
2705 kern_chflagsat(struct thread *td, int fd, const char *path,
2706 enum uio_seg pathseg, u_long flags, int atflag)
2708 struct nameidata nd;
2709 cap_rights_t rights;
2712 AUDIT_ARG_FFLAGS(flags);
2713 follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2714 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2715 cap_rights_init(&rights, CAP_FCHFLAGS), td);
2716 if ((error = namei(&nd)) != 0)
2718 NDFREE(&nd, NDF_ONLY_PNBUF);
2719 error = setfflags(td, nd.ni_vp, flags);
2725 * Change flags of a file given a file descriptor.
2727 #ifndef _SYS_SYSPROTO_H_
2728 struct fchflags_args {
2734 sys_fchflags(td, uap)
2736 register struct fchflags_args /* {
2742 cap_rights_t rights;
2745 AUDIT_ARG_FD(uap->fd);
2746 AUDIT_ARG_FFLAGS(uap->flags);
2747 error = getvnode(td->td_proc->p_fd, uap->fd,
2748 cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
2752 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2753 AUDIT_ARG_VNODE1(fp->f_vnode);
2754 VOP_UNLOCK(fp->f_vnode, 0);
2756 error = setfflags(td, fp->f_vnode, uap->flags);
2762 * Common implementation code for chmod(), lchmod() and fchmod().
2765 setfmode(td, cred, vp, mode)
2775 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2777 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2779 vattr.va_mode = mode & ALLPERMS;
2781 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2784 error = VOP_SETATTR(vp, &vattr, cred);
2786 vn_finished_write(mp);
2791 * Change mode of a file given path name.
2793 #ifndef _SYS_SYSPROTO_H_
2802 register struct chmod_args /* {
2808 return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2811 #ifndef _SYS_SYSPROTO_H_
2812 struct fchmodat_args {
2820 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2822 int flag = uap->flag;
2824 char *path = uap->path;
2825 mode_t mode = uap->mode;
2827 if (flag & ~AT_SYMLINK_NOFOLLOW)
2830 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2834 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2837 return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2841 * Change mode of a file given path name (don't follow links.)
2843 #ifndef _SYS_SYSPROTO_H_
2844 struct lchmod_args {
2852 register struct lchmod_args /* {
2858 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2859 uap->mode, AT_SYMLINK_NOFOLLOW));
2863 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2864 mode_t mode, int flag)
2866 struct nameidata nd;
2867 cap_rights_t rights;
2870 AUDIT_ARG_MODE(mode);
2871 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2872 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2873 cap_rights_init(&rights, CAP_FCHMOD), td);
2874 if ((error = namei(&nd)) != 0)
2876 NDFREE(&nd, NDF_ONLY_PNBUF);
2877 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2883 * Change mode of a file given a file descriptor.
2885 #ifndef _SYS_SYSPROTO_H_
2886 struct fchmod_args {
2892 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2895 cap_rights_t rights;
2898 AUDIT_ARG_FD(uap->fd);
2899 AUDIT_ARG_MODE(uap->mode);
2901 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2904 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2910 * Common implementation for chown(), lchown(), and fchown()
2913 setfown(td, cred, vp, uid, gid)
2924 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2926 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2931 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2935 error = VOP_SETATTR(vp, &vattr, cred);
2937 vn_finished_write(mp);
2942 * Set ownership given a path name.
2944 #ifndef _SYS_SYSPROTO_H_
2954 register struct chown_args /* {
2961 return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2964 #ifndef _SYS_SYSPROTO_H_
2965 struct fchownat_args {
2974 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2979 if (flag & ~AT_SYMLINK_NOFOLLOW)
2982 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2983 uap->gid, uap->flag));
2987 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2991 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2995 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2996 int uid, int gid, int flag)
2998 struct nameidata nd;
2999 cap_rights_t rights;
3002 AUDIT_ARG_OWNER(uid, gid);
3003 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3004 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
3005 cap_rights_init(&rights, CAP_FCHOWN), td);
3007 if ((error = namei(&nd)) != 0)
3009 NDFREE(&nd, NDF_ONLY_PNBUF);
3010 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3016 * Set ownership given a path name, do not cross symlinks.
3018 #ifndef _SYS_SYSPROTO_H_
3019 struct lchown_args {
3028 register struct lchown_args /* {
3035 return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3039 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3043 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3044 AT_SYMLINK_NOFOLLOW));
3048 * Set ownership given a file descriptor.
3050 #ifndef _SYS_SYSPROTO_H_
3051 struct fchown_args {
3060 register struct fchown_args /* {
3067 cap_rights_t rights;
3070 AUDIT_ARG_FD(uap->fd);
3071 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3072 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
3075 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3081 * Common implementation code for utimes(), lutimes(), and futimes().
3084 getutimes(usrtvp, tvpseg, tsp)
3085 const struct timeval *usrtvp;
3086 enum uio_seg tvpseg;
3087 struct timespec *tsp;
3089 struct timeval tv[2];
3090 const struct timeval *tvp;
3093 if (usrtvp == NULL) {
3094 vfs_timestamp(&tsp[0]);
3097 if (tvpseg == UIO_SYSSPACE) {
3100 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3105 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3106 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3108 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3109 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3115 * Common implementation code for futimens(), utimensat().
3117 #define UTIMENS_NULL 0x1
3118 #define UTIMENS_EXIT 0x2
3120 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3121 struct timespec *tsp, int *retflags)
3123 struct timespec tsnow;
3126 vfs_timestamp(&tsnow);
3128 if (usrtsp == NULL) {
3131 *retflags |= UTIMENS_NULL;
3134 if (tspseg == UIO_SYSSPACE) {
3137 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3139 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3140 *retflags |= UTIMENS_EXIT;
3141 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3142 *retflags |= UTIMENS_NULL;
3143 if (tsp[0].tv_nsec == UTIME_OMIT)
3144 tsp[0].tv_sec = VNOVAL;
3145 else if (tsp[0].tv_nsec == UTIME_NOW)
3147 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3149 if (tsp[1].tv_nsec == UTIME_OMIT)
3150 tsp[1].tv_sec = VNOVAL;
3151 else if (tsp[1].tv_nsec == UTIME_NOW)
3153 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3160 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3164 setutimes(td, vp, ts, numtimes, nullflag)
3167 const struct timespec *ts;
3173 int error, setbirthtime;
3175 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3177 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3179 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3180 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3183 vattr.va_atime = ts[0];
3184 vattr.va_mtime = ts[1];
3186 vattr.va_birthtime = ts[1];
3188 vattr.va_birthtime = ts[2];
3190 vattr.va_vaflags |= VA_UTIMES_NULL;
3192 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3196 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3198 vn_finished_write(mp);
3203 * Set the access and modification times of a file.
3205 #ifndef _SYS_SYSPROTO_H_
3206 struct utimes_args {
3208 struct timeval *tptr;
3214 register struct utimes_args /* {
3216 struct timeval *tptr;
3220 return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3224 #ifndef _SYS_SYSPROTO_H_
3225 struct futimesat_args {
3228 const struct timeval * times;
3232 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3235 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3236 uap->times, UIO_USERSPACE));
3240 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3241 struct timeval *tptr, enum uio_seg tptrseg)
3244 return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3248 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3249 struct timeval *tptr, enum uio_seg tptrseg)
3251 struct nameidata nd;
3252 struct timespec ts[2];
3253 cap_rights_t rights;
3256 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3258 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3259 cap_rights_init(&rights, CAP_FUTIMES), td);
3261 if ((error = namei(&nd)) != 0)
3263 NDFREE(&nd, NDF_ONLY_PNBUF);
3264 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3270 * Set the access and modification times of a file.
3272 #ifndef _SYS_SYSPROTO_H_
3273 struct lutimes_args {
3275 struct timeval *tptr;
3279 sys_lutimes(td, uap)
3281 register struct lutimes_args /* {
3283 struct timeval *tptr;
3287 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3292 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3293 struct timeval *tptr, enum uio_seg tptrseg)
3295 struct timespec ts[2];
3296 struct nameidata nd;
3299 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3301 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3302 if ((error = namei(&nd)) != 0)
3304 NDFREE(&nd, NDF_ONLY_PNBUF);
3305 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3311 * Set the access and modification times of a file.
3313 #ifndef _SYS_SYSPROTO_H_
3314 struct futimes_args {
3316 struct timeval *tptr;
3320 sys_futimes(td, uap)
3322 register struct futimes_args /* {
3324 struct timeval *tptr;
3328 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3332 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3333 enum uio_seg tptrseg)
3335 struct timespec ts[2];
3337 cap_rights_t rights;
3341 error = getutimes(tptr, tptrseg, ts);
3344 error = getvnode(td->td_proc->p_fd, fd,
3345 cap_rights_init(&rights, CAP_FUTIMES), &fp);
3349 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3350 AUDIT_ARG_VNODE1(fp->f_vnode);
3351 VOP_UNLOCK(fp->f_vnode, 0);
3353 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3359 sys_futimens(struct thread *td, struct futimens_args *uap)
3362 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3366 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3367 enum uio_seg tptrseg)
3369 struct timespec ts[2];
3371 cap_rights_t rights;
3375 error = getutimens(tptr, tptrseg, ts, &flags);
3378 if (flags & UTIMENS_EXIT)
3380 error = getvnode(td->td_proc->p_fd, fd,
3381 cap_rights_init(&rights, CAP_FUTIMES), &fp);
3385 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3386 AUDIT_ARG_VNODE1(fp->f_vnode);
3387 VOP_UNLOCK(fp->f_vnode, 0);
3389 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3395 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3398 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3399 uap->times, UIO_USERSPACE, uap->flag));
3403 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3404 struct timespec *tptr, enum uio_seg tptrseg, int flag)
3406 struct nameidata nd;
3407 struct timespec ts[2];
3410 if (flag & ~AT_SYMLINK_NOFOLLOW)
3413 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3415 NDINIT_AT(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3416 FOLLOW) | AUDITVNODE1, pathseg, path, fd, td);
3417 if ((error = namei(&nd)) != 0)
3420 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3422 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3423 * "Search permission is denied by a component of the path prefix."
3425 NDFREE(&nd, NDF_ONLY_PNBUF);
3426 if ((flags & UTIMENS_EXIT) == 0)
3427 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3433 * Truncate a file given its path name.
3435 #ifndef _SYS_SYSPROTO_H_
3436 struct truncate_args {
3443 sys_truncate(td, uap)
3445 register struct truncate_args /* {
3452 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3456 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3462 struct nameidata nd;
3467 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3468 if ((error = namei(&nd)) != 0)
3471 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3472 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3473 vn_rangelock_unlock(vp, rl_cookie);
3477 NDFREE(&nd, NDF_ONLY_PNBUF);
3478 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3479 if (vp->v_type == VDIR)
3482 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3485 else if ((error = vn_writechk(vp)) == 0 &&
3486 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3488 vattr.va_size = length;
3489 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3492 vn_finished_write(mp);
3493 vn_rangelock_unlock(vp, rl_cookie);
3498 #if defined(COMPAT_43)
3500 * Truncate a file given its path name.
3502 #ifndef _SYS_SYSPROTO_H_
3503 struct otruncate_args {
3511 register struct otruncate_args /* {
3516 struct truncate_args /* {
3522 nuap.path = uap->path;
3523 nuap.length = uap->length;
3524 return (sys_truncate(td, &nuap));
3526 #endif /* COMPAT_43 */
3528 /* Versions with the pad argument */
3530 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3532 struct truncate_args ouap;
3534 ouap.path = uap->path;
3535 ouap.length = uap->length;
3536 return (sys_truncate(td, &ouap));
3540 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3542 struct ftruncate_args ouap;
3545 ouap.length = uap->length;
3546 return (sys_ftruncate(td, &ouap));
3550 * Sync an open file.
3552 #ifndef _SYS_SYSPROTO_H_
3560 struct fsync_args /* {
3567 cap_rights_t rights;
3568 int error, lock_flags;
3570 AUDIT_ARG_FD(uap->fd);
3571 error = getvnode(td->td_proc->p_fd, uap->fd,
3572 cap_rights_init(&rights, CAP_FSYNC), &fp);
3576 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3579 if (MNT_SHARED_WRITES(mp) ||
3580 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3581 lock_flags = LK_SHARED;
3583 lock_flags = LK_EXCLUSIVE;
3585 vn_lock(vp, lock_flags | LK_RETRY);
3586 AUDIT_ARG_VNODE1(vp);
3587 if (vp->v_object != NULL) {
3588 VM_OBJECT_WLOCK(vp->v_object);
3589 vm_object_page_clean(vp->v_object, 0, 0, 0);
3590 VM_OBJECT_WUNLOCK(vp->v_object);
3592 error = VOP_FSYNC(vp, MNT_WAIT, td);
3595 vn_finished_write(mp);
3602 * Rename files. Source and destination must either both be directories, or
3603 * both not be directories. If target is a directory, it must be empty.
3605 #ifndef _SYS_SYSPROTO_H_
3606 struct rename_args {
3614 register struct rename_args /* {
3620 return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3623 #ifndef _SYS_SYSPROTO_H_
3624 struct renameat_args {
3632 sys_renameat(struct thread *td, struct renameat_args *uap)
3635 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3640 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3643 return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3647 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3648 enum uio_seg pathseg)
3650 struct mount *mp = NULL;
3651 struct vnode *tvp, *fvp, *tdvp;
3652 struct nameidata fromnd, tond;
3653 cap_rights_t rights;
3659 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3660 AUDITVNODE1, pathseg, old, oldfd,
3661 cap_rights_init(&rights, CAP_RENAMEAT), td);
3663 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3664 pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3667 if ((error = namei(&fromnd)) != 0)
3670 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3671 fromnd.ni_vp, &fromnd.ni_cnd);
3672 VOP_UNLOCK(fromnd.ni_dvp, 0);
3673 if (fromnd.ni_dvp != fromnd.ni_vp)
3674 VOP_UNLOCK(fromnd.ni_vp, 0);
3677 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3678 SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3679 cap_rights_init(&rights, CAP_LINKAT), td);
3680 if (fromnd.ni_vp->v_type == VDIR)
3681 tond.ni_cnd.cn_flags |= WILLBEDIR;
3682 if ((error = namei(&tond)) != 0) {
3683 /* Translate error code for rename("dir1", "dir2/."). */
3684 if (error == EISDIR && fvp->v_type == VDIR)
3686 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3687 vrele(fromnd.ni_dvp);
3693 error = vn_start_write(fvp, &mp, V_NOWAIT);
3695 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3696 NDFREE(&tond, NDF_ONLY_PNBUF);
3703 vrele(fromnd.ni_dvp);
3705 vrele(tond.ni_startdir);
3706 if (fromnd.ni_startdir != NULL)
3707 vrele(fromnd.ni_startdir);
3708 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3714 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3717 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3722 if (newfd != AT_FDCWD) {
3724 * If the target already exists we require CAP_UNLINKAT
3727 error = cap_check(&tond.ni_filecaps.fc_rights,
3728 cap_rights_init(&rights, CAP_UNLINKAT));
3739 * If the source is the same as the destination (that is, if they
3740 * are links to the same vnode), then there is nothing to do.
3746 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3747 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3751 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3752 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3753 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3754 NDFREE(&tond, NDF_ONLY_PNBUF);
3756 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3757 NDFREE(&tond, NDF_ONLY_PNBUF);
3764 vrele(fromnd.ni_dvp);
3767 vrele(tond.ni_startdir);
3768 vn_finished_write(mp);
3770 if (fromnd.ni_startdir)
3771 vrele(fromnd.ni_startdir);
3778 * Make a directory file.
3780 #ifndef _SYS_SYSPROTO_H_
3789 register struct mkdir_args /* {
3795 return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3798 #ifndef _SYS_SYSPROTO_H_
3799 struct mkdirat_args {
3806 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3809 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3813 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3816 return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3820 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3826 struct nameidata nd;
3827 cap_rights_t rights;
3830 AUDIT_ARG_MODE(mode);
3833 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3834 NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3836 nd.ni_cnd.cn_flags |= WILLBEDIR;
3837 if ((error = namei(&nd)) != 0)
3841 NDFREE(&nd, NDF_ONLY_PNBUF);
3843 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3844 * the strange behaviour of leaving the vnode unlocked
3845 * if the target is the same vnode as the parent.
3847 if (vp == nd.ni_dvp)
3854 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3855 NDFREE(&nd, NDF_ONLY_PNBUF);
3857 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3862 vattr.va_type = VDIR;
3863 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3865 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3870 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3874 NDFREE(&nd, NDF_ONLY_PNBUF);
3878 vn_finished_write(mp);
3883 * Remove a directory file.
3885 #ifndef _SYS_SYSPROTO_H_
3893 struct rmdir_args /* {
3898 return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3902 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3905 return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3909 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3913 struct nameidata nd;
3914 cap_rights_t rights;
3919 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3920 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3921 if ((error = namei(&nd)) != 0)
3924 if (vp->v_type != VDIR) {
3929 * No rmdir "." please.
3931 if (nd.ni_dvp == vp) {
3936 * The root of a mounted filesystem cannot be deleted.
3938 if (vp->v_vflag & VV_ROOT) {
3943 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3948 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3949 NDFREE(&nd, NDF_ONLY_PNBUF);
3951 if (nd.ni_dvp == vp)
3955 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3959 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3960 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3961 vn_finished_write(mp);
3963 NDFREE(&nd, NDF_ONLY_PNBUF);
3965 if (nd.ni_dvp == vp)
3974 * Read a block of directory entries in a filesystem independent format.
3976 #ifndef _SYS_SYSPROTO_H_
3977 struct ogetdirentries_args {
3985 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3990 error = kern_ogetdirentries(td, uap, &loff);
3992 error = copyout(&loff, uap->basep, sizeof(long));
3997 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4002 struct uio auio, kuio;
4003 struct iovec aiov, kiov;
4004 struct dirent *dp, *edp;
4005 cap_rights_t rights;
4007 int error, eofflag, readcnt;
4011 /* XXX arbitrary sanity limit on `count'. */
4012 if (uap->count > 64 * 1024)
4014 error = getvnode(td->td_proc->p_fd, uap->fd,
4015 cap_rights_init(&rights, CAP_READ), &fp);
4018 if ((fp->f_flag & FREAD) == 0) {
4023 foffset = foffset_lock(fp, 0);
4025 if (vp->v_type != VDIR) {
4026 foffset_unlock(fp, foffset, 0);
4030 aiov.iov_base = uap->buf;
4031 aiov.iov_len = uap->count;
4032 auio.uio_iov = &aiov;
4033 auio.uio_iovcnt = 1;
4034 auio.uio_rw = UIO_READ;
4035 auio.uio_segflg = UIO_USERSPACE;
4037 auio.uio_resid = uap->count;
4038 vn_lock(vp, LK_SHARED | LK_RETRY);
4039 loff = auio.uio_offset = foffset;
4041 error = mac_vnode_check_readdir(td->td_ucred, vp);
4044 foffset_unlock(fp, foffset, FOF_NOUPDATE);
4049 # if (BYTE_ORDER != LITTLE_ENDIAN)
4050 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
4051 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
4053 foffset = auio.uio_offset;
4058 kuio.uio_iov = &kiov;
4059 kuio.uio_segflg = UIO_SYSSPACE;
4060 kiov.iov_len = uap->count;
4061 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
4062 kiov.iov_base = dirbuf;
4063 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
4065 foffset = kuio.uio_offset;
4067 readcnt = uap->count - kuio.uio_resid;
4068 edp = (struct dirent *)&dirbuf[readcnt];
4069 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
4070 # if (BYTE_ORDER == LITTLE_ENDIAN)
4072 * The expected low byte of
4073 * dp->d_namlen is our dp->d_type.
4074 * The high MBZ byte of dp->d_namlen
4075 * is our dp->d_namlen.
4077 dp->d_type = dp->d_namlen;
4081 * The dp->d_type is the high byte
4082 * of the expected dp->d_namlen,
4083 * so must be zero'ed.
4087 if (dp->d_reclen > 0) {
4088 dp = (struct dirent *)
4089 ((char *)dp + dp->d_reclen);
4096 error = uiomove(dirbuf, readcnt, &auio);
4098 free(dirbuf, M_TEMP);
4102 foffset_unlock(fp, foffset, 0);
4106 if (uap->count == auio.uio_resid &&
4107 (vp->v_vflag & VV_ROOT) &&
4108 (vp->v_mount->mnt_flag & MNT_UNION)) {
4109 struct vnode *tvp = vp;
4110 vp = vp->v_mount->mnt_vnodecovered;
4119 foffset_unlock(fp, foffset, 0);
4121 td->td_retval[0] = uap->count - auio.uio_resid;
4126 #endif /* COMPAT_43 */
4129 * Read a block of directory entries in a filesystem independent format.
4131 #ifndef _SYS_SYSPROTO_H_
4132 struct getdirentries_args {
4140 sys_getdirentries(td, uap)
4142 register struct getdirentries_args /* {
4152 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4153 NULL, UIO_USERSPACE);
4156 if (uap->basep != NULL)
4157 error = copyout(&base, uap->basep, sizeof(long));
4162 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4163 long *basep, ssize_t *residp, enum uio_seg bufseg)
4169 cap_rights_t rights;
4175 if (count > IOSIZE_MAX)
4177 auio.uio_resid = count;
4178 error = getvnode(td->td_proc->p_fd, fd,
4179 cap_rights_init(&rights, CAP_READ), &fp);
4182 if ((fp->f_flag & FREAD) == 0) {
4187 foffset = foffset_lock(fp, 0);
4189 if (vp->v_type != VDIR) {
4193 aiov.iov_base = buf;
4194 aiov.iov_len = count;
4195 auio.uio_iov = &aiov;
4196 auio.uio_iovcnt = 1;
4197 auio.uio_rw = UIO_READ;
4198 auio.uio_segflg = bufseg;
4200 vn_lock(vp, LK_SHARED | LK_RETRY);
4201 AUDIT_ARG_VNODE1(vp);
4202 loff = auio.uio_offset = foffset;
4204 error = mac_vnode_check_readdir(td->td_ucred, vp);
4207 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4209 foffset = auio.uio_offset;
4214 if (count == auio.uio_resid &&
4215 (vp->v_vflag & VV_ROOT) &&
4216 (vp->v_mount->mnt_flag & MNT_UNION)) {
4217 struct vnode *tvp = vp;
4219 vp = vp->v_mount->mnt_vnodecovered;
4230 *residp = auio.uio_resid;
4231 td->td_retval[0] = count - auio.uio_resid;
4233 foffset_unlock(fp, foffset, 0);
4238 #ifndef _SYS_SYSPROTO_H_
4239 struct getdents_args {
4246 sys_getdents(td, uap)
4248 register struct getdents_args /* {
4254 struct getdirentries_args ap;
4258 ap.count = uap->count;
4260 return (sys_getdirentries(td, &ap));
4264 * Set the mode mask for creation of filesystem nodes.
4266 #ifndef _SYS_SYSPROTO_H_
4274 struct umask_args /* {
4278 register struct filedesc *fdp;
4280 FILEDESC_XLOCK(td->td_proc->p_fd);
4281 fdp = td->td_proc->p_fd;
4282 td->td_retval[0] = fdp->fd_cmask;
4283 fdp->fd_cmask = uap->newmask & ALLPERMS;
4284 FILEDESC_XUNLOCK(td->td_proc->p_fd);
4289 * Void all references to file by ripping underlying filesystem away from
4292 #ifndef _SYS_SYSPROTO_H_
4293 struct revoke_args {
4300 register struct revoke_args /* {
4306 struct nameidata nd;
4309 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4311 if ((error = namei(&nd)) != 0)
4314 NDFREE(&nd, NDF_ONLY_PNBUF);
4315 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4320 error = mac_vnode_check_revoke(td->td_ucred, vp);
4324 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4327 if (td->td_ucred->cr_uid != vattr.va_uid) {
4328 error = priv_check(td, PRIV_VFS_ADMIN);
4333 VOP_REVOKE(vp, REVOKEALL);
4340 * Convert a user file descriptor to a kernel file entry and check that, if it
4341 * is a capability, the correct rights are present. A reference on the file
4342 * entry is held upon returning.
4345 getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
4350 error = fget_unlocked(fdp, fd, rightsp, 0, &fp, NULL);
4355 * The file could be not of the vnode type, or it may be not
4356 * yet fully initialized, in which case the f_vnode pointer
4357 * may be set, but f_ops is still badfileops. E.g.,
4358 * devfs_open() transiently create such situation to
4359 * facilitate csw d_fdopen().
4361 * Dupfdopen() handling in kern_openat() installs the
4362 * half-baked file into the process descriptor table, allowing
4363 * other thread to dereference it. Guard against the race by
4366 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4367 fdrop(fp, curthread);
4376 * Get an (NFS) file handle.
4378 #ifndef _SYS_SYSPROTO_H_
4379 struct lgetfh_args {
4387 register struct lgetfh_args *uap;
4389 struct nameidata nd;
4391 register struct vnode *vp;
4394 error = priv_check(td, PRIV_VFS_GETFH);
4397 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4402 NDFREE(&nd, NDF_ONLY_PNBUF);
4404 bzero(&fh, sizeof(fh));
4405 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4406 error = VOP_VPTOFH(vp, &fh.fh_fid);
4409 error = copyout(&fh, uap->fhp, sizeof (fh));
4413 #ifndef _SYS_SYSPROTO_H_
4422 register struct getfh_args *uap;
4424 struct nameidata nd;
4426 register struct vnode *vp;
4429 error = priv_check(td, PRIV_VFS_GETFH);
4432 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4437 NDFREE(&nd, NDF_ONLY_PNBUF);
4439 bzero(&fh, sizeof(fh));
4440 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4441 error = VOP_VPTOFH(vp, &fh.fh_fid);
4444 error = copyout(&fh, uap->fhp, sizeof (fh));
4449 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4452 * warning: do not remove the priv_check() call or this becomes one giant
4455 #ifndef _SYS_SYSPROTO_H_
4456 struct fhopen_args {
4457 const struct fhandle *u_fhp;
4464 struct fhopen_args /* {
4465 const struct fhandle *u_fhp;
4476 error = priv_check(td, PRIV_VFS_FHOPEN);
4480 fmode = FFLAGS(uap->flags);
4481 /* why not allow a non-read/write open for our lockd? */
4482 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4484 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4487 /* find the mount point */
4488 mp = vfs_busyfs(&fhp.fh_fsid);
4491 /* now give me my vnode, it gets returned to me locked */
4492 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4497 error = falloc_noinstall(td, &fp);
4503 * An extra reference on `fp' has been held for us by
4504 * falloc_noinstall().
4510 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4512 KASSERT(fp->f_ops == &badfileops,
4513 ("VOP_OPEN in fhopen() set f_ops"));
4514 KASSERT(td->td_dupfd < 0,
4515 ("fhopen() encountered fdopen()"));
4525 finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4528 if ((fmode & O_TRUNC) != 0) {
4529 error = fo_truncate(fp, 0, td->td_ucred, td);
4534 error = finstall(td, fp, &indx, fmode, NULL);
4537 td->td_retval[0] = indx;
4542 * Stat an (NFS) file handle.
4544 #ifndef _SYS_SYSPROTO_H_
4545 struct fhstat_args {
4546 struct fhandle *u_fhp;
4553 register struct fhstat_args /* {
4554 struct fhandle *u_fhp;
4562 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4565 error = kern_fhstat(td, fh, &sb);
4567 error = copyout(&sb, uap->sb, sizeof(sb));
4572 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4578 error = priv_check(td, PRIV_VFS_FHSTAT);
4581 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4583 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4587 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4593 * Implement fstatfs() for (NFS) file handles.
4595 #ifndef _SYS_SYSPROTO_H_
4596 struct fhstatfs_args {
4597 struct fhandle *u_fhp;
4602 sys_fhstatfs(td, uap)
4604 struct fhstatfs_args /* {
4605 struct fhandle *u_fhp;
4613 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4616 error = kern_fhstatfs(td, fh, &sf);
4619 return (copyout(&sf, uap->buf, sizeof(sf)));
4623 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4630 error = priv_check(td, PRIV_VFS_FHSTATFS);
4633 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4635 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4641 error = prison_canseemount(td->td_ucred, mp);
4645 error = mac_mount_check_stat(td->td_ucred, mp);
4650 * Set these in case the underlying filesystem fails to do so.
4653 sp->f_version = STATFS_VERSION;
4654 sp->f_namemax = NAME_MAX;
4655 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4656 error = VFS_STATFS(mp, sp);
4665 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4670 cap_rights_t rights;
4671 off_t olen, ooffset;
4674 if (offset < 0 || len <= 0)
4676 /* Check for wrap. */
4677 if (offset > OFF_MAX - len)
4679 error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4682 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4686 if ((fp->f_flag & FWRITE) == 0) {
4690 if (fp->f_type != DTYPE_VNODE) {
4695 if (vp->v_type != VREG) {
4700 /* Allocating blocks may take a long time, so iterate. */
4707 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4710 error = vn_lock(vp, LK_EXCLUSIVE);
4712 vn_finished_write(mp);
4716 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4719 error = VOP_ALLOCATE(vp, &offset, &len);
4721 vn_finished_write(mp);
4723 if (olen + ooffset != offset + len) {
4724 panic("offset + len changed from %jx/%jx to %jx/%jx",
4725 ooffset, olen, offset, len);
4727 if (error != 0 || len == 0)
4729 KASSERT(olen > len, ("Iteration did not make progress?"));
4738 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4741 td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4747 * Unlike madvise(2), we do not make a best effort to remember every
4748 * possible caching hint. Instead, we remember the last setting with
4749 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4750 * region of any current setting.
4753 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4756 struct fadvise_info *fa, *new;
4759 cap_rights_t rights;
4763 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4766 case POSIX_FADV_SEQUENTIAL:
4767 case POSIX_FADV_RANDOM:
4768 case POSIX_FADV_NOREUSE:
4769 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4771 case POSIX_FADV_NORMAL:
4772 case POSIX_FADV_WILLNEED:
4773 case POSIX_FADV_DONTNEED:
4779 /* XXX: CAP_POSIX_FADVISE? */
4780 error = fget(td, fd, cap_rights_init(&rights), &fp);
4783 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4787 if (fp->f_type != DTYPE_VNODE) {
4792 if (vp->v_type != VREG) {
4799 end = offset + len - 1;
4801 case POSIX_FADV_SEQUENTIAL:
4802 case POSIX_FADV_RANDOM:
4803 case POSIX_FADV_NOREUSE:
4805 * Try to merge any existing non-standard region with
4806 * this new region if possible, otherwise create a new
4807 * non-standard region for this request.
4809 mtx_pool_lock(mtxpool_sleep, fp);
4811 if (fa != NULL && fa->fa_advice == advice &&
4812 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4813 (end != OFF_MAX && fa->fa_start == end + 1) ||
4814 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4815 if (offset < fa->fa_start)
4816 fa->fa_start = offset;
4817 if (end > fa->fa_end)
4820 new->fa_advice = advice;
4821 new->fa_start = offset;
4823 new->fa_prevstart = 0;
4824 new->fa_prevend = 0;
4828 mtx_pool_unlock(mtxpool_sleep, fp);
4830 case POSIX_FADV_NORMAL:
4832 * If a the "normal" region overlaps with an existing
4833 * non-standard region, trim or remove the
4834 * non-standard region.
4836 mtx_pool_lock(mtxpool_sleep, fp);
4839 if (offset <= fa->fa_start && end >= fa->fa_end) {
4841 fp->f_advice = NULL;
4842 } else if (offset <= fa->fa_start &&
4843 end >= fa->fa_start)
4844 fa->fa_start = end + 1;
4845 else if (offset <= fa->fa_end && end >= fa->fa_end)
4846 fa->fa_end = offset - 1;
4847 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4849 * If the "normal" region is a middle
4850 * portion of the existing
4851 * non-standard region, just remove
4852 * the whole thing rather than picking
4853 * one side or the other to
4857 fp->f_advice = NULL;
4860 mtx_pool_unlock(mtxpool_sleep, fp);
4862 case POSIX_FADV_WILLNEED:
4863 case POSIX_FADV_DONTNEED:
4864 error = VOP_ADVISE(vp, offset, end, advice);
4870 free(new, M_FADVISE);
4875 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4878 td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4879 uap->len, uap->advice);