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_ktrace.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
48 #include <sys/capsicum.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
71 #include <sys/dirent.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
76 #include <sys/ktrace.h>
79 #include <machine/stdarg.h>
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
89 #include <ufs/ufs/quota.h>
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
97 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
98 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
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 setutimes(struct thread *td, struct vnode *,
105 const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
110 * The module initialization routine for POSIX asynchronous I/O will
111 * set this to the version of AIO that it implements. (Zero means
112 * that it is not implemented.) This value is used here by pathconf()
113 * and in kern_descrip.c by fpathconf().
115 int async_io_version;
118 static int syncprt = 0;
119 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
123 * Sync each mounted filesystem.
125 #ifndef _SYS_SYSPROTO_H_
134 struct sync_args *uap;
136 struct mount *mp, *nmp;
139 mtx_lock(&mountlist_mtx);
140 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
141 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
142 nmp = TAILQ_NEXT(mp, mnt_list);
145 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
146 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
147 save = curthread_pflags_set(TDP_SYNCIO);
148 vfs_msync(mp, MNT_NOWAIT);
149 VFS_SYNC(mp, MNT_NOWAIT);
150 curthread_pflags_restore(save);
151 vn_finished_write(mp);
153 mtx_lock(&mountlist_mtx);
154 nmp = TAILQ_NEXT(mp, mnt_list);
157 mtx_unlock(&mountlist_mtx);
162 * Change filesystem quotas.
164 #ifndef _SYS_SYSPROTO_H_
165 struct quotactl_args {
173 sys_quotactl(td, uap)
175 register struct quotactl_args /* {
186 AUDIT_ARG_CMD(uap->cmd);
187 AUDIT_ARG_UID(uap->uid);
188 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
190 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
192 if ((error = namei(&nd)) != 0)
194 NDFREE(&nd, NDF_ONLY_PNBUF);
195 mp = nd.ni_vp->v_mount;
198 error = vfs_busy(mp, 0);
202 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
205 * Since quota on operation typically needs to open quota
206 * file, the Q_QUOTAON handler needs to unbusy the mount point
207 * before calling into namei. Otherwise, unmount might be
208 * started between two vfs_busy() invocations (first is our,
209 * second is from mount point cross-walk code in lookup()),
212 * Require that Q_QUOTAON handles the vfs_busy() reference on
213 * its own, always returning with ubusied mount point.
215 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
221 * Used by statfs conversion routines to scale the block size up if
222 * necessary so that all of the block counts are <= 'max_size'. Note
223 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
227 statfs_scale_blocks(struct statfs *sf, long max_size)
232 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
235 * Attempt to scale the block counts to give a more accurate
236 * overview to userland of the ratio of free space to used
237 * space. To do this, find the largest block count and compute
238 * a divisor that lets it fit into a signed integer <= max_size.
240 if (sf->f_bavail < 0)
241 count = -sf->f_bavail;
243 count = sf->f_bavail;
244 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
245 if (count <= max_size)
248 count >>= flsl(max_size);
255 sf->f_bsize <<= shift;
256 sf->f_blocks >>= shift;
257 sf->f_bfree >>= shift;
258 sf->f_bavail >>= shift;
262 * Get filesystem statistics.
264 #ifndef _SYS_SYSPROTO_H_
273 register struct statfs_args /* {
281 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
283 error = copyout(&sf, uap->buf, sizeof(sf));
288 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
292 struct statfs *sp, sb;
296 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
301 mp = nd.ni_vp->v_mount;
303 NDFREE(&nd, NDF_ONLY_PNBUF);
305 error = vfs_busy(mp, 0);
310 error = mac_mount_check_stat(td->td_ucred, mp);
315 * Set these in case the underlying filesystem fails to do so.
318 sp->f_version = STATFS_VERSION;
319 sp->f_namemax = NAME_MAX;
320 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
321 error = VFS_STATFS(mp, sp);
324 if (priv_check(td, PRIV_VFS_GENERATION)) {
325 bcopy(sp, &sb, sizeof(sb));
326 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
327 prison_enforce_statfs(td->td_ucred, mp, &sb);
337 * Get filesystem statistics.
339 #ifndef _SYS_SYSPROTO_H_
340 struct fstatfs_args {
348 register struct fstatfs_args /* {
356 error = kern_fstatfs(td, uap->fd, &sf);
358 error = copyout(&sf, uap->buf, sizeof(sf));
363 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
367 struct statfs *sp, sb;
373 error = getvnode(td->td_proc->p_fd, fd,
374 cap_rights_init(&rights, CAP_FSTATFS), &fp);
378 vn_lock(vp, LK_SHARED | LK_RETRY);
380 AUDIT_ARG_VNODE1(vp);
391 error = vfs_busy(mp, 0);
396 error = mac_mount_check_stat(td->td_ucred, mp);
401 * Set these in case the underlying filesystem fails to do so.
404 sp->f_version = STATFS_VERSION;
405 sp->f_namemax = NAME_MAX;
406 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
407 error = VFS_STATFS(mp, sp);
410 if (priv_check(td, PRIV_VFS_GENERATION)) {
411 bcopy(sp, &sb, sizeof(sb));
412 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
413 prison_enforce_statfs(td->td_ucred, mp, &sb);
424 * Get statistics on all filesystems.
426 #ifndef _SYS_SYSPROTO_H_
427 struct getfsstat_args {
434 sys_getfsstat(td, uap)
436 register struct getfsstat_args /* {
443 return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
448 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
449 * The caller is responsible for freeing memory which will be allocated
453 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
454 enum uio_seg bufseg, int flags)
456 struct mount *mp, *nmp;
457 struct statfs *sfsp, *sp, sb;
458 size_t count, maxcount;
461 maxcount = bufsize / sizeof(struct statfs);
464 else if (bufseg == UIO_USERSPACE)
466 else /* if (bufseg == UIO_SYSSPACE) */ {
468 mtx_lock(&mountlist_mtx);
469 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
472 mtx_unlock(&mountlist_mtx);
473 if (maxcount > count)
475 sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
479 mtx_lock(&mountlist_mtx);
480 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
481 if (prison_canseemount(td->td_ucred, mp) != 0) {
482 nmp = TAILQ_NEXT(mp, mnt_list);
486 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
487 nmp = TAILQ_NEXT(mp, mnt_list);
491 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
492 nmp = TAILQ_NEXT(mp, mnt_list);
495 if (sfsp && count < maxcount) {
498 * Set these in case the underlying filesystem
501 sp->f_version = STATFS_VERSION;
502 sp->f_namemax = NAME_MAX;
503 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
505 * If MNT_NOWAIT or MNT_LAZY is specified, do not
506 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
507 * overrides MNT_WAIT.
509 if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
510 (flags & MNT_WAIT)) &&
511 (error = VFS_STATFS(mp, sp))) {
512 mtx_lock(&mountlist_mtx);
513 nmp = TAILQ_NEXT(mp, mnt_list);
517 if (priv_check(td, PRIV_VFS_GENERATION)) {
518 bcopy(sp, &sb, sizeof(sb));
519 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
520 prison_enforce_statfs(td->td_ucred, mp, &sb);
523 if (bufseg == UIO_SYSSPACE)
524 bcopy(sp, sfsp, sizeof(*sp));
525 else /* if (bufseg == UIO_USERSPACE) */ {
526 error = copyout(sp, sfsp, sizeof(*sp));
535 mtx_lock(&mountlist_mtx);
536 nmp = TAILQ_NEXT(mp, mnt_list);
539 mtx_unlock(&mountlist_mtx);
540 if (sfsp && count > maxcount)
541 td->td_retval[0] = maxcount;
543 td->td_retval[0] = count;
547 #ifdef COMPAT_FREEBSD4
549 * Get old format filesystem statistics.
551 static void cvtstatfs(struct statfs *, struct ostatfs *);
553 #ifndef _SYS_SYSPROTO_H_
554 struct freebsd4_statfs_args {
560 freebsd4_statfs(td, uap)
562 struct freebsd4_statfs_args /* {
571 error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
574 cvtstatfs(&sf, &osb);
575 return (copyout(&osb, uap->buf, sizeof(osb)));
579 * Get filesystem statistics.
581 #ifndef _SYS_SYSPROTO_H_
582 struct freebsd4_fstatfs_args {
588 freebsd4_fstatfs(td, uap)
590 struct freebsd4_fstatfs_args /* {
599 error = kern_fstatfs(td, uap->fd, &sf);
602 cvtstatfs(&sf, &osb);
603 return (copyout(&osb, uap->buf, sizeof(osb)));
607 * Get statistics on all filesystems.
609 #ifndef _SYS_SYSPROTO_H_
610 struct freebsd4_getfsstat_args {
617 freebsd4_getfsstat(td, uap)
619 register struct freebsd4_getfsstat_args /* {
625 struct statfs *buf, *sp;
630 count = uap->bufsize / sizeof(struct ostatfs);
631 size = count * sizeof(struct statfs);
632 error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
634 count = td->td_retval[0];
636 while (count > 0 && error == 0) {
638 error = copyout(&osb, uap->buf, sizeof(osb));
649 * Implement fstatfs() for (NFS) file handles.
651 #ifndef _SYS_SYSPROTO_H_
652 struct freebsd4_fhstatfs_args {
653 struct fhandle *u_fhp;
658 freebsd4_fhstatfs(td, uap)
660 struct freebsd4_fhstatfs_args /* {
661 struct fhandle *u_fhp;
670 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
673 error = kern_fhstatfs(td, fh, &sf);
676 cvtstatfs(&sf, &osb);
677 return (copyout(&osb, uap->buf, sizeof(osb)));
681 * Convert a new format statfs structure to an old format statfs structure.
689 statfs_scale_blocks(nsp, LONG_MAX);
690 bzero(osp, sizeof(*osp));
691 osp->f_bsize = nsp->f_bsize;
692 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
693 osp->f_blocks = nsp->f_blocks;
694 osp->f_bfree = nsp->f_bfree;
695 osp->f_bavail = nsp->f_bavail;
696 osp->f_files = MIN(nsp->f_files, LONG_MAX);
697 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
698 osp->f_owner = nsp->f_owner;
699 osp->f_type = nsp->f_type;
700 osp->f_flags = nsp->f_flags;
701 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
702 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
703 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
704 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
705 strlcpy(osp->f_fstypename, nsp->f_fstypename,
706 MIN(MFSNAMELEN, OMFSNAMELEN));
707 strlcpy(osp->f_mntonname, nsp->f_mntonname,
708 MIN(MNAMELEN, OMNAMELEN));
709 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
710 MIN(MNAMELEN, OMNAMELEN));
711 osp->f_fsid = nsp->f_fsid;
713 #endif /* COMPAT_FREEBSD4 */
716 * Change current working directory to a given file descriptor.
718 #ifndef _SYS_SYSPROTO_H_
726 struct fchdir_args /* {
730 register struct filedesc *fdp = td->td_proc->p_fd;
731 struct vnode *vp, *tdp, *vpold;
737 AUDIT_ARG_FD(uap->fd);
738 error = getvnode(fdp, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
745 vn_lock(vp, LK_SHARED | LK_RETRY);
746 AUDIT_ARG_VNODE1(vp);
747 error = change_dir(vp, td);
748 while (!error && (mp = vp->v_mountedhere) != NULL) {
751 error = VFS_ROOT(mp, LK_SHARED, &tdp);
764 vpold = fdp->fd_cdir;
766 FILEDESC_XUNLOCK(fdp);
772 * Change current working directory (``.'').
774 #ifndef _SYS_SYSPROTO_H_
782 struct chdir_args /* {
787 return (kern_chdir(td, uap->path, UIO_USERSPACE));
791 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
793 register struct filedesc *fdp = td->td_proc->p_fd;
798 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
800 if ((error = namei(&nd)) != 0)
802 if ((error = change_dir(nd.ni_vp, td)) != 0) {
804 NDFREE(&nd, NDF_ONLY_PNBUF);
807 VOP_UNLOCK(nd.ni_vp, 0);
808 NDFREE(&nd, NDF_ONLY_PNBUF);
811 fdp->fd_cdir = nd.ni_vp;
812 FILEDESC_XUNLOCK(fdp);
818 * Helper function for raised chroot(2) security function: Refuse if
819 * any filedescriptors are open directories.
822 chroot_refuse_vdir_fds(fdp)
823 struct filedesc *fdp;
829 FILEDESC_LOCK_ASSERT(fdp);
831 for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
832 fp = fget_locked(fdp, fd);
835 if (fp->f_type == DTYPE_VNODE) {
837 if (vp->v_type == VDIR)
845 * This sysctl determines if we will allow a process to chroot(2) if it
846 * has a directory open:
847 * 0: disallowed for all processes.
848 * 1: allowed for processes that were not already chroot(2)'ed.
849 * 2: allowed for all processes.
852 static int chroot_allow_open_directories = 1;
854 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
855 &chroot_allow_open_directories, 0,
856 "Allow a process to chroot(2) if it has a directory open");
859 * Change notion of root (``/'') directory.
861 #ifndef _SYS_SYSPROTO_H_
869 struct chroot_args /* {
876 error = priv_check(td, PRIV_VFS_CHROOT);
879 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
880 UIO_USERSPACE, uap->path, td);
884 error = change_dir(nd.ni_vp, td);
888 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
892 VOP_UNLOCK(nd.ni_vp, 0);
893 error = change_root(nd.ni_vp, td);
895 NDFREE(&nd, NDF_ONLY_PNBUF);
900 NDFREE(&nd, NDF_ONLY_PNBUF);
905 * Common routine for chroot and chdir. Callers must provide a locked vnode
917 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
918 if (vp->v_type != VDIR)
921 error = mac_vnode_check_chdir(td->td_ucred, vp);
925 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
929 * Common routine for kern_chroot() and jail_attach(). The caller is
930 * responsible for invoking priv_check() and mac_vnode_check_chroot() to
931 * authorize this operation.
938 struct filedesc *fdp;
942 fdp = td->td_proc->p_fd;
944 if (chroot_allow_open_directories == 0 ||
945 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
946 error = chroot_refuse_vdir_fds(fdp);
948 FILEDESC_XUNLOCK(fdp);
952 oldvp = fdp->fd_rdir;
959 FILEDESC_XUNLOCK(fdp);
965 flags_to_rights(int flags, cap_rights_t *rightsp)
968 if (flags & O_EXEC) {
969 cap_rights_set(rightsp, CAP_FEXECVE);
971 switch ((flags & O_ACCMODE)) {
973 cap_rights_set(rightsp, CAP_READ);
976 cap_rights_set(rightsp, CAP_READ);
979 cap_rights_set(rightsp, CAP_WRITE);
980 if (!(flags & (O_APPEND | O_TRUNC)))
981 cap_rights_set(rightsp, CAP_SEEK);
987 cap_rights_set(rightsp, CAP_CREATE);
990 cap_rights_set(rightsp, CAP_FTRUNCATE);
992 if (flags & (O_SYNC | O_FSYNC))
993 cap_rights_set(rightsp, CAP_FSYNC);
995 if (flags & (O_EXLOCK | O_SHLOCK))
996 cap_rights_set(rightsp, CAP_FLOCK);
1000 * Check permissions, allocate an open file structure, and call the device
1001 * open routine if any.
1003 #ifndef _SYS_SYSPROTO_H_
1013 register struct open_args /* {
1020 return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1023 #ifndef _SYS_SYSPROTO_H_
1024 struct openat_args {
1032 sys_openat(struct thread *td, struct openat_args *uap)
1035 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1040 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1044 return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1048 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1049 int flags, int mode)
1051 struct proc *p = td->td_proc;
1052 struct filedesc *fdp = p->p_fd;
1055 struct nameidata nd;
1056 cap_rights_t rights;
1057 int cmode, error, indx;
1061 AUDIT_ARG_FFLAGS(flags);
1062 AUDIT_ARG_MODE(mode);
1063 /* XXX: audit dirfd */
1064 cap_rights_init(&rights, CAP_LOOKUP);
1065 flags_to_rights(flags, &rights);
1067 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1070 if (flags & O_EXEC) {
1071 if (flags & O_ACCMODE)
1073 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
1076 flags = FFLAGS(flags);
1080 * Allocate the file descriptor, but don't install a descriptor yet.
1082 error = falloc_noinstall(td, &fp);
1086 * An extra reference on `fp' has been held for us by
1087 * falloc_noinstall().
1089 /* Set the flags early so the finit in devfs can pick them up. */
1090 fp->f_flag = flags & FMASK;
1091 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1092 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1094 td->td_dupfd = -1; /* XXX check for fdopen */
1095 error = vn_open(&nd, &flags, cmode, fp);
1098 * If the vn_open replaced the method vector, something
1099 * wonderous happened deep below and we just pass it up
1100 * pretending we know what we do.
1102 if (error == ENXIO && fp->f_ops != &badfileops)
1106 * Handle special fdopen() case. bleh.
1108 * Don't do this for relative (capability) lookups; we don't
1109 * understand exactly what would happen, and we don't think
1110 * that it ever should.
1112 if (nd.ni_strictrelative == 0 &&
1113 (error == ENODEV || error == ENXIO) &&
1114 td->td_dupfd >= 0) {
1115 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1124 NDFREE(&nd, NDF_ONLY_PNBUF);
1128 * Store the vnode, for any f_type. Typically, the vnode use
1129 * count is decremented by direct call to vn_closefile() for
1130 * files that switched type in the cdevsw fdopen() method.
1134 * If the file wasn't claimed by devfs bind it to the normal
1135 * vnode operations here.
1137 if (fp->f_ops == &badfileops) {
1138 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1140 finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1141 DTYPE_VNODE, vp, &vnops);
1145 if (flags & O_TRUNC) {
1146 error = fo_truncate(fp, 0, td->td_ucred, td);
1152 * If we haven't already installed the FD (for dupfdopen), do so now.
1155 struct filecaps *fcaps;
1158 if (nd.ni_strictrelative == 1)
1159 fcaps = &nd.ni_filecaps;
1163 error = finstall(td, fp, &indx, flags, fcaps);
1164 /* On success finstall() consumes fcaps. */
1166 filecaps_free(&nd.ni_filecaps);
1170 filecaps_free(&nd.ni_filecaps);
1174 * Release our private reference, leaving the one associated with
1175 * the descriptor table intact.
1178 td->td_retval[0] = indx;
1181 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1190 #ifndef _SYS_SYSPROTO_H_
1191 struct ocreat_args {
1199 register struct ocreat_args /* {
1205 return (kern_open(td, uap->path, UIO_USERSPACE,
1206 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1208 #endif /* COMPAT_43 */
1211 * Create a special file.
1213 #ifndef _SYS_SYSPROTO_H_
1223 register struct mknod_args /* {
1230 return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1233 #ifndef _SYS_SYSPROTO_H_
1234 struct mknodat_args {
1242 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1245 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1250 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1254 return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1258 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1264 struct nameidata nd;
1265 cap_rights_t rights;
1266 int error, whiteout = 0;
1268 AUDIT_ARG_MODE(mode);
1270 switch (mode & S_IFMT) {
1273 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1276 error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1279 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1283 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1293 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1294 pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT), td);
1295 if ((error = namei(&nd)) != 0)
1299 NDFREE(&nd, NDF_ONLY_PNBUF);
1300 if (vp == nd.ni_dvp)
1308 vattr.va_mode = (mode & ALLPERMS) &
1309 ~td->td_proc->p_fd->fd_cmask;
1310 vattr.va_rdev = dev;
1313 switch (mode & S_IFMT) {
1314 case S_IFMT: /* used by badsect to flag bad sectors */
1315 vattr.va_type = VBAD;
1318 vattr.va_type = VCHR;
1321 vattr.va_type = VBLK;
1327 panic("kern_mknod: invalid mode");
1330 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1331 NDFREE(&nd, NDF_ONLY_PNBUF);
1333 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1338 if (error == 0 && !whiteout)
1339 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1340 &nd.ni_cnd, &vattr);
1344 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1346 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1347 &nd.ni_cnd, &vattr);
1352 NDFREE(&nd, NDF_ONLY_PNBUF);
1354 vn_finished_write(mp);
1359 * Create a named pipe.
1361 #ifndef _SYS_SYSPROTO_H_
1362 struct mkfifo_args {
1370 register struct mkfifo_args /* {
1376 return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1379 #ifndef _SYS_SYSPROTO_H_
1380 struct mkfifoat_args {
1387 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1390 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1395 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1398 return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1402 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1407 struct nameidata nd;
1408 cap_rights_t rights;
1411 AUDIT_ARG_MODE(mode);
1414 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1415 pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT), td);
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 segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT), td);
1568 if ((error = namei(&nd)) == 0) {
1569 if (nd.ni_vp != NULL) {
1570 NDFREE(&nd, NDF_ONLY_PNBUF);
1571 if (nd.ni_dvp == nd.ni_vp)
1578 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1580 * Cross-device link. No need to recheck
1581 * vp->v_type, since it cannot change, except
1584 NDFREE(&nd, NDF_ONLY_PNBUF);
1588 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1589 error = can_hardlink(vp, td->td_ucred);
1592 error = mac_vnode_check_link(td->td_ucred,
1593 nd.ni_dvp, vp, &nd.ni_cnd);
1598 NDFREE(&nd, NDF_ONLY_PNBUF);
1601 error = vn_start_write(vp, &mp, V_NOWAIT);
1605 NDFREE(&nd, NDF_ONLY_PNBUF);
1606 error = vn_start_write(NULL, &mp,
1612 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1615 vn_finished_write(mp);
1616 NDFREE(&nd, NDF_ONLY_PNBUF);
1619 NDFREE(&nd, NDF_ONLY_PNBUF);
1629 * Make a symbolic link.
1631 #ifndef _SYS_SYSPROTO_H_
1632 struct symlink_args {
1638 sys_symlink(td, uap)
1640 register struct symlink_args /* {
1646 return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1649 #ifndef _SYS_SYSPROTO_H_
1650 struct symlinkat_args {
1657 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1660 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1665 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1668 return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1672 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1673 enum uio_seg segflg)
1678 struct nameidata nd;
1680 cap_rights_t rights;
1682 if (segflg == UIO_SYSSPACE) {
1685 syspath = uma_zalloc(namei_zone, M_WAITOK);
1686 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1689 AUDIT_ARG_TEXT(syspath);
1692 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1693 segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT), td);
1694 if ((error = namei(&nd)) != 0)
1697 NDFREE(&nd, NDF_ONLY_PNBUF);
1698 if (nd.ni_vp == nd.ni_dvp)
1706 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1707 NDFREE(&nd, NDF_ONLY_PNBUF);
1709 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1714 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1716 vattr.va_type = VLNK;
1717 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1722 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1728 NDFREE(&nd, NDF_ONLY_PNBUF);
1730 vn_finished_write(mp);
1732 if (segflg != UIO_SYSSPACE)
1733 uma_zfree(namei_zone, syspath);
1738 * Delete a whiteout from the filesystem.
1741 sys_undelete(td, uap)
1743 register struct undelete_args /* {
1748 struct nameidata nd;
1753 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1754 UIO_USERSPACE, uap->path, td);
1759 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1760 NDFREE(&nd, NDF_ONLY_PNBUF);
1761 if (nd.ni_vp == nd.ni_dvp)
1769 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1770 NDFREE(&nd, NDF_ONLY_PNBUF);
1772 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1776 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1777 NDFREE(&nd, NDF_ONLY_PNBUF);
1779 vn_finished_write(mp);
1784 * Delete a name from the filesystem.
1786 #ifndef _SYS_SYSPROTO_H_
1787 struct unlink_args {
1794 struct unlink_args /* {
1799 return (kern_unlink(td, uap->path, UIO_USERSPACE));
1802 #ifndef _SYS_SYSPROTO_H_
1803 struct unlinkat_args {
1810 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1812 int flag = uap->flag;
1814 char *path = uap->path;
1816 if (flag & ~AT_REMOVEDIR)
1819 if (flag & AT_REMOVEDIR)
1820 return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1822 return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1826 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1829 return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1833 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1838 struct nameidata nd;
1840 cap_rights_t rights;
1845 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1846 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1847 if ((error = namei(&nd)) != 0)
1848 return (error == EINVAL ? EPERM : error);
1850 if (vp->v_type == VDIR && oldinum == 0) {
1851 error = EPERM; /* POSIX */
1852 } else if (oldinum != 0 &&
1853 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1854 sb.st_ino != oldinum) {
1855 error = EIDRM; /* Identifier removed */
1858 * The root of a mounted filesystem cannot be deleted.
1860 * XXX: can this only be a VDIR case?
1862 if (vp->v_vflag & VV_ROOT)
1866 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1867 NDFREE(&nd, NDF_ONLY_PNBUF);
1869 if (vp == nd.ni_dvp)
1873 if ((error = vn_start_write(NULL, &mp,
1874 V_XSLEEP | PCATCH)) != 0)
1879 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1884 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1885 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1889 vn_finished_write(mp);
1891 NDFREE(&nd, NDF_ONLY_PNBUF);
1893 if (vp == nd.ni_dvp)
1901 * Reposition read/write file offset.
1903 #ifndef _SYS_SYSPROTO_H_
1914 register struct lseek_args /* {
1922 cap_rights_t rights;
1925 AUDIT_ARG_FD(uap->fd);
1926 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1929 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1930 fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1935 #if defined(COMPAT_43)
1937 * Reposition read/write file offset.
1939 #ifndef _SYS_SYSPROTO_H_
1940 struct olseek_args {
1949 register struct olseek_args /* {
1955 struct lseek_args /* {
1963 nuap.offset = uap->offset;
1964 nuap.whence = uap->whence;
1965 return (sys_lseek(td, &nuap));
1967 #endif /* COMPAT_43 */
1969 /* Version with the 'pad' argument */
1971 freebsd6_lseek(td, uap)
1973 register struct freebsd6_lseek_args *uap;
1975 struct lseek_args ouap;
1978 ouap.offset = uap->offset;
1979 ouap.whence = uap->whence;
1980 return (sys_lseek(td, &ouap));
1984 * Check access permissions using passed credentials.
1987 vn_access(vp, user_flags, cred, td)
1996 /* Flags == 0 means only check for existence. */
2000 if (user_flags & R_OK)
2002 if (user_flags & W_OK)
2004 if (user_flags & X_OK)
2007 error = mac_vnode_check_access(cred, vp, accmode);
2011 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2012 error = VOP_ACCESS(vp, accmode, cred, td);
2018 * Check access permissions using "real" credentials.
2020 #ifndef _SYS_SYSPROTO_H_
2021 struct access_args {
2029 register struct access_args /* {
2035 return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2038 #ifndef _SYS_SYSPROTO_H_
2039 struct faccessat_args {
2047 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2050 if (uap->flag & ~AT_EACCESS)
2052 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2057 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2060 return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2064 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2065 int flag, int amode)
2067 struct ucred *cred, *tmpcred;
2069 struct nameidata nd;
2070 cap_rights_t rights;
2073 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2077 * Create and modify a temporary credential instead of one that
2078 * is potentially shared.
2080 if (!(flag & AT_EACCESS)) {
2081 cred = td->td_ucred;
2082 tmpcred = crdup(cred);
2083 tmpcred->cr_uid = cred->cr_ruid;
2084 tmpcred->cr_groups[0] = cred->cr_rgid;
2085 td->td_ucred = tmpcred;
2087 cred = tmpcred = td->td_ucred;
2088 AUDIT_ARG_VALUE(amode);
2089 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2090 AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
2092 if ((error = namei(&nd)) != 0)
2096 error = vn_access(vp, amode, tmpcred, td);
2097 NDFREE(&nd, NDF_ONLY_PNBUF);
2100 if (!(flag & AT_EACCESS)) {
2101 td->td_ucred = cred;
2108 * Check access permissions using "effective" credentials.
2110 #ifndef _SYS_SYSPROTO_H_
2111 struct eaccess_args {
2117 sys_eaccess(td, uap)
2119 register struct eaccess_args /* {
2125 return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2129 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2132 return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2135 #if defined(COMPAT_43)
2137 * Get file status; this version follows links.
2139 #ifndef _SYS_SYSPROTO_H_
2148 register struct ostat_args /* {
2157 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2161 return (copyout(&osb, uap->ub, sizeof (osb)));
2165 * Get file status; this version does not follow links.
2167 #ifndef _SYS_SYSPROTO_H_
2168 struct olstat_args {
2176 register struct olstat_args /* {
2185 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2189 return (copyout(&osb, uap->ub, sizeof (osb)));
2193 * Convert from an old to a new stat structure.
2201 ost->st_dev = st->st_dev;
2202 ost->st_ino = st->st_ino;
2203 ost->st_mode = st->st_mode;
2204 ost->st_nlink = st->st_nlink;
2205 ost->st_uid = st->st_uid;
2206 ost->st_gid = st->st_gid;
2207 ost->st_rdev = st->st_rdev;
2208 if (st->st_size < (quad_t)1 << 32)
2209 ost->st_size = st->st_size;
2212 ost->st_atim = st->st_atim;
2213 ost->st_mtim = st->st_mtim;
2214 ost->st_ctim = st->st_ctim;
2215 ost->st_blksize = st->st_blksize;
2216 ost->st_blocks = st->st_blocks;
2217 ost->st_flags = st->st_flags;
2218 ost->st_gen = st->st_gen;
2220 #endif /* COMPAT_43 */
2223 * Get file status; this version follows links.
2225 #ifndef _SYS_SYSPROTO_H_
2234 register struct stat_args /* {
2242 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2244 error = copyout(&sb, uap->ub, sizeof (sb));
2248 #ifndef _SYS_SYSPROTO_H_
2249 struct fstatat_args {
2257 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2262 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2263 UIO_USERSPACE, &sb);
2265 error = copyout(&sb, uap->buf, sizeof (sb));
2270 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2273 return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2277 kern_statat(struct thread *td, int flag, int fd, char *path,
2278 enum uio_seg pathseg, struct stat *sbp)
2281 return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2285 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2286 enum uio_seg pathseg, struct stat *sbp,
2287 void (*hook)(struct vnode *vp, struct stat *sbp))
2289 struct nameidata nd;
2291 cap_rights_t rights;
2294 if (flag & ~AT_SYMLINK_NOFOLLOW)
2297 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2298 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2299 cap_rights_init(&rights, CAP_FSTAT), td);
2301 if ((error = namei(&nd)) != 0)
2303 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2305 SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2306 if (S_ISREG(sb.st_mode))
2307 SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2308 if (__predict_false(hook != NULL))
2309 hook(nd.ni_vp, &sb);
2311 NDFREE(&nd, NDF_ONLY_PNBUF);
2317 if (KTRPOINT(td, KTR_STRUCT))
2324 * Get file status; this version does not follow links.
2326 #ifndef _SYS_SYSPROTO_H_
2335 register struct lstat_args /* {
2343 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2345 error = copyout(&sb, uap->ub, sizeof (sb));
2350 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2353 return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2358 * Implementation of the NetBSD [l]stat() functions.
2366 bzero(nsb, sizeof *nsb);
2367 nsb->st_dev = sb->st_dev;
2368 nsb->st_ino = sb->st_ino;
2369 nsb->st_mode = sb->st_mode;
2370 nsb->st_nlink = sb->st_nlink;
2371 nsb->st_uid = sb->st_uid;
2372 nsb->st_gid = sb->st_gid;
2373 nsb->st_rdev = sb->st_rdev;
2374 nsb->st_atim = sb->st_atim;
2375 nsb->st_mtim = sb->st_mtim;
2376 nsb->st_ctim = sb->st_ctim;
2377 nsb->st_size = sb->st_size;
2378 nsb->st_blocks = sb->st_blocks;
2379 nsb->st_blksize = sb->st_blksize;
2380 nsb->st_flags = sb->st_flags;
2381 nsb->st_gen = sb->st_gen;
2382 nsb->st_birthtim = sb->st_birthtim;
2385 #ifndef _SYS_SYSPROTO_H_
2394 register struct nstat_args /* {
2403 error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2406 cvtnstat(&sb, &nsb);
2407 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2411 * NetBSD lstat. Get file status; this version does not follow links.
2413 #ifndef _SYS_SYSPROTO_H_
2422 register struct nlstat_args /* {
2431 error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2434 cvtnstat(&sb, &nsb);
2435 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2439 * Get configurable pathname variables.
2441 #ifndef _SYS_SYSPROTO_H_
2442 struct pathconf_args {
2448 sys_pathconf(td, uap)
2450 register struct pathconf_args /* {
2456 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2459 #ifndef _SYS_SYSPROTO_H_
2460 struct lpathconf_args {
2466 sys_lpathconf(td, uap)
2468 register struct lpathconf_args /* {
2474 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2479 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2482 struct nameidata nd;
2485 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2487 if ((error = namei(&nd)) != 0)
2489 NDFREE(&nd, NDF_ONLY_PNBUF);
2491 /* If asynchronous I/O is available, it works for all files. */
2492 if (name == _PC_ASYNC_IO)
2493 td->td_retval[0] = async_io_version;
2495 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2501 * Return target name of a symbolic link.
2503 #ifndef _SYS_SYSPROTO_H_
2504 struct readlink_args {
2511 sys_readlink(td, uap)
2513 register struct readlink_args /* {
2520 return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2521 UIO_USERSPACE, uap->count));
2523 #ifndef _SYS_SYSPROTO_H_
2524 struct readlinkat_args {
2532 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2535 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2536 uap->buf, UIO_USERSPACE, uap->bufsize));
2540 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2541 enum uio_seg bufseg, size_t count)
2544 return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2549 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2550 char *buf, enum uio_seg bufseg, size_t count)
2555 struct nameidata nd;
2558 if (count > IOSIZE_MAX)
2561 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2562 pathseg, path, fd, td);
2564 if ((error = namei(&nd)) != 0)
2566 NDFREE(&nd, NDF_ONLY_PNBUF);
2569 error = mac_vnode_check_readlink(td->td_ucred, vp);
2575 if (vp->v_type != VLNK)
2578 aiov.iov_base = buf;
2579 aiov.iov_len = count;
2580 auio.uio_iov = &aiov;
2581 auio.uio_iovcnt = 1;
2582 auio.uio_offset = 0;
2583 auio.uio_rw = UIO_READ;
2584 auio.uio_segflg = bufseg;
2586 auio.uio_resid = count;
2587 error = VOP_READLINK(vp, &auio, td->td_ucred);
2588 td->td_retval[0] = count - auio.uio_resid;
2595 * Common implementation code for chflags() and fchflags().
2598 setfflags(td, vp, flags)
2607 /* We can't support the value matching VNOVAL. */
2608 if (flags == VNOVAL)
2609 return (EOPNOTSUPP);
2612 * Prevent non-root users from setting flags on devices. When
2613 * a device is reused, users can retain ownership of the device
2614 * if they are allowed to set flags and programs assume that
2615 * chown can't fail when done as root.
2617 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2618 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2623 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2626 vattr.va_flags = flags;
2627 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2629 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2632 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2634 vn_finished_write(mp);
2639 * Change flags of a file given a path name.
2641 #ifndef _SYS_SYSPROTO_H_
2642 struct chflags_args {
2648 sys_chflags(td, uap)
2650 register struct chflags_args /* {
2656 return (kern_chflags(td, uap->path, UIO_USERSPACE, uap->flags));
2659 #ifndef _SYS_SYSPROTO_H_
2660 struct chflagsat_args {
2668 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2671 const char *path = uap->path;
2672 u_long flags = uap->flags;
2673 int atflag = uap->atflag;
2675 if (atflag & ~AT_SYMLINK_NOFOLLOW)
2678 return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2682 kern_chflags(struct thread *td, const char *path, enum uio_seg pathseg,
2686 return (kern_chflagsat(td, AT_FDCWD, path, pathseg, flags, 0));
2690 * Same as chflags() but doesn't follow symlinks.
2693 sys_lchflags(td, uap)
2695 register struct lchflags_args /* {
2701 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2702 uap->flags, AT_SYMLINK_NOFOLLOW));
2706 kern_chflagsat(struct thread *td, int fd, const char *path,
2707 enum uio_seg pathseg, u_long flags, int atflag)
2709 struct nameidata nd;
2710 cap_rights_t rights;
2713 AUDIT_ARG_FFLAGS(flags);
2714 follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2715 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2716 cap_rights_init(&rights, CAP_FCHFLAGS), td);
2717 if ((error = namei(&nd)) != 0)
2719 NDFREE(&nd, NDF_ONLY_PNBUF);
2720 error = setfflags(td, nd.ni_vp, flags);
2726 * Change flags of a file given a file descriptor.
2728 #ifndef _SYS_SYSPROTO_H_
2729 struct fchflags_args {
2735 sys_fchflags(td, uap)
2737 register struct fchflags_args /* {
2743 cap_rights_t rights;
2746 AUDIT_ARG_FD(uap->fd);
2747 AUDIT_ARG_FFLAGS(uap->flags);
2748 error = getvnode(td->td_proc->p_fd, uap->fd,
2749 cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
2753 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2754 AUDIT_ARG_VNODE1(fp->f_vnode);
2755 VOP_UNLOCK(fp->f_vnode, 0);
2757 error = setfflags(td, fp->f_vnode, uap->flags);
2763 * Common implementation code for chmod(), lchmod() and fchmod().
2766 setfmode(td, cred, vp, mode)
2776 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2778 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2780 vattr.va_mode = mode & ALLPERMS;
2782 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2785 error = VOP_SETATTR(vp, &vattr, cred);
2787 vn_finished_write(mp);
2792 * Change mode of a file given path name.
2794 #ifndef _SYS_SYSPROTO_H_
2803 register struct chmod_args /* {
2809 return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2812 #ifndef _SYS_SYSPROTO_H_
2813 struct fchmodat_args {
2821 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2823 int flag = uap->flag;
2825 char *path = uap->path;
2826 mode_t mode = uap->mode;
2828 if (flag & ~AT_SYMLINK_NOFOLLOW)
2831 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2835 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2838 return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2842 * Change mode of a file given path name (don't follow links.)
2844 #ifndef _SYS_SYSPROTO_H_
2845 struct lchmod_args {
2853 register struct lchmod_args /* {
2859 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2860 uap->mode, AT_SYMLINK_NOFOLLOW));
2864 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2865 mode_t mode, int flag)
2867 struct nameidata nd;
2868 cap_rights_t rights;
2871 AUDIT_ARG_MODE(mode);
2872 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2873 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2874 cap_rights_init(&rights, CAP_FCHMOD), td);
2875 if ((error = namei(&nd)) != 0)
2877 NDFREE(&nd, NDF_ONLY_PNBUF);
2878 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2884 * Change mode of a file given a file descriptor.
2886 #ifndef _SYS_SYSPROTO_H_
2887 struct fchmod_args {
2893 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2896 cap_rights_t rights;
2899 AUDIT_ARG_FD(uap->fd);
2900 AUDIT_ARG_MODE(uap->mode);
2902 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2905 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2911 * Common implementation for chown(), lchown(), and fchown()
2914 setfown(td, cred, vp, uid, gid)
2925 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2927 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2932 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2936 error = VOP_SETATTR(vp, &vattr, cred);
2938 vn_finished_write(mp);
2943 * Set ownership given a path name.
2945 #ifndef _SYS_SYSPROTO_H_
2955 register struct chown_args /* {
2962 return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2965 #ifndef _SYS_SYSPROTO_H_
2966 struct fchownat_args {
2975 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2980 if (flag & ~AT_SYMLINK_NOFOLLOW)
2983 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2984 uap->gid, uap->flag));
2988 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2992 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2996 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2997 int uid, int gid, int flag)
2999 struct nameidata nd;
3000 cap_rights_t rights;
3003 AUDIT_ARG_OWNER(uid, gid);
3004 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3005 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
3006 cap_rights_init(&rights, CAP_FCHOWN), td);
3008 if ((error = namei(&nd)) != 0)
3010 NDFREE(&nd, NDF_ONLY_PNBUF);
3011 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3017 * Set ownership given a path name, do not cross symlinks.
3019 #ifndef _SYS_SYSPROTO_H_
3020 struct lchown_args {
3029 register struct lchown_args /* {
3036 return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3040 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3044 return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3045 AT_SYMLINK_NOFOLLOW));
3049 * Set ownership given a file descriptor.
3051 #ifndef _SYS_SYSPROTO_H_
3052 struct fchown_args {
3061 register struct fchown_args /* {
3068 cap_rights_t rights;
3071 AUDIT_ARG_FD(uap->fd);
3072 AUDIT_ARG_OWNER(uap->uid, uap->gid);
3073 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
3076 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3082 * Common implementation code for utimes(), lutimes(), and futimes().
3085 getutimes(usrtvp, tvpseg, tsp)
3086 const struct timeval *usrtvp;
3087 enum uio_seg tvpseg;
3088 struct timespec *tsp;
3090 struct timeval tv[2];
3091 const struct timeval *tvp;
3094 if (usrtvp == NULL) {
3095 vfs_timestamp(&tsp[0]);
3098 if (tvpseg == UIO_SYSSPACE) {
3101 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3106 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3107 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3109 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3110 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3116 * Common implementation code for utimes(), lutimes(), and futimes().
3119 setutimes(td, vp, ts, numtimes, nullflag)
3122 const struct timespec *ts;
3128 int error, setbirthtime;
3130 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3132 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3134 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3135 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3138 vattr.va_atime = ts[0];
3139 vattr.va_mtime = ts[1];
3141 vattr.va_birthtime = ts[1];
3143 vattr.va_birthtime = ts[2];
3145 vattr.va_vaflags |= VA_UTIMES_NULL;
3147 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3151 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3153 vn_finished_write(mp);
3158 * Set the access and modification times of a file.
3160 #ifndef _SYS_SYSPROTO_H_
3161 struct utimes_args {
3163 struct timeval *tptr;
3169 register struct utimes_args /* {
3171 struct timeval *tptr;
3175 return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3179 #ifndef _SYS_SYSPROTO_H_
3180 struct futimesat_args {
3183 const struct timeval * times;
3187 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3190 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3191 uap->times, UIO_USERSPACE));
3195 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3196 struct timeval *tptr, enum uio_seg tptrseg)
3199 return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3203 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3204 struct timeval *tptr, enum uio_seg tptrseg)
3206 struct nameidata nd;
3207 struct timespec ts[2];
3208 cap_rights_t rights;
3211 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3213 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3214 cap_rights_init(&rights, CAP_FUTIMES), td);
3216 if ((error = namei(&nd)) != 0)
3218 NDFREE(&nd, NDF_ONLY_PNBUF);
3219 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3225 * Set the access and modification times of a file.
3227 #ifndef _SYS_SYSPROTO_H_
3228 struct lutimes_args {
3230 struct timeval *tptr;
3234 sys_lutimes(td, uap)
3236 register struct lutimes_args /* {
3238 struct timeval *tptr;
3242 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3247 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3248 struct timeval *tptr, enum uio_seg tptrseg)
3250 struct timespec ts[2];
3251 struct nameidata nd;
3254 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3256 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3257 if ((error = namei(&nd)) != 0)
3259 NDFREE(&nd, NDF_ONLY_PNBUF);
3260 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3266 * Set the access and modification times of a file.
3268 #ifndef _SYS_SYSPROTO_H_
3269 struct futimes_args {
3271 struct timeval *tptr;
3275 sys_futimes(td, uap)
3277 register struct futimes_args /* {
3279 struct timeval *tptr;
3283 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3287 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3288 enum uio_seg tptrseg)
3290 struct timespec ts[2];
3292 cap_rights_t rights;
3296 error = getutimes(tptr, tptrseg, ts);
3299 error = getvnode(td->td_proc->p_fd, fd,
3300 cap_rights_init(&rights, CAP_FUTIMES), &fp);
3304 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3305 AUDIT_ARG_VNODE1(fp->f_vnode);
3306 VOP_UNLOCK(fp->f_vnode, 0);
3308 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3314 * Truncate a file given its path name.
3316 #ifndef _SYS_SYSPROTO_H_
3317 struct truncate_args {
3324 sys_truncate(td, uap)
3326 register struct truncate_args /* {
3333 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3337 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3343 struct nameidata nd;
3348 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3349 if ((error = namei(&nd)) != 0)
3352 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3353 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3354 vn_rangelock_unlock(vp, rl_cookie);
3358 NDFREE(&nd, NDF_ONLY_PNBUF);
3359 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3360 if (vp->v_type == VDIR)
3363 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3366 else if ((error = vn_writechk(vp)) == 0 &&
3367 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3369 vattr.va_size = length;
3370 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3373 vn_finished_write(mp);
3374 vn_rangelock_unlock(vp, rl_cookie);
3379 #if defined(COMPAT_43)
3381 * Truncate a file given its path name.
3383 #ifndef _SYS_SYSPROTO_H_
3384 struct otruncate_args {
3392 register struct otruncate_args /* {
3397 struct truncate_args /* {
3403 nuap.path = uap->path;
3404 nuap.length = uap->length;
3405 return (sys_truncate(td, &nuap));
3407 #endif /* COMPAT_43 */
3409 /* Versions with the pad argument */
3411 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3413 struct truncate_args ouap;
3415 ouap.path = uap->path;
3416 ouap.length = uap->length;
3417 return (sys_truncate(td, &ouap));
3421 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3423 struct ftruncate_args ouap;
3426 ouap.length = uap->length;
3427 return (sys_ftruncate(td, &ouap));
3431 * Sync an open file.
3433 #ifndef _SYS_SYSPROTO_H_
3441 struct fsync_args /* {
3448 cap_rights_t rights;
3449 int error, lock_flags;
3451 AUDIT_ARG_FD(uap->fd);
3452 error = getvnode(td->td_proc->p_fd, uap->fd,
3453 cap_rights_init(&rights, CAP_FSYNC), &fp);
3457 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3460 if (MNT_SHARED_WRITES(mp) ||
3461 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3462 lock_flags = LK_SHARED;
3464 lock_flags = LK_EXCLUSIVE;
3466 vn_lock(vp, lock_flags | LK_RETRY);
3467 AUDIT_ARG_VNODE1(vp);
3468 if (vp->v_object != NULL) {
3469 VM_OBJECT_WLOCK(vp->v_object);
3470 vm_object_page_clean(vp->v_object, 0, 0, 0);
3471 VM_OBJECT_WUNLOCK(vp->v_object);
3473 error = VOP_FSYNC(vp, MNT_WAIT, td);
3476 vn_finished_write(mp);
3483 * Rename files. Source and destination must either both be directories, or
3484 * both not be directories. If target is a directory, it must be empty.
3486 #ifndef _SYS_SYSPROTO_H_
3487 struct rename_args {
3495 register struct rename_args /* {
3501 return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3504 #ifndef _SYS_SYSPROTO_H_
3505 struct renameat_args {
3513 sys_renameat(struct thread *td, struct renameat_args *uap)
3516 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3521 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3524 return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3528 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3529 enum uio_seg pathseg)
3531 struct mount *mp = NULL;
3532 struct vnode *tvp, *fvp, *tdvp;
3533 struct nameidata fromnd, tond;
3534 cap_rights_t rights;
3540 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3541 AUDITVNODE1, pathseg, old, oldfd,
3542 cap_rights_init(&rights, CAP_RENAMEAT), td);
3544 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3545 pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3548 if ((error = namei(&fromnd)) != 0)
3551 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3552 fromnd.ni_vp, &fromnd.ni_cnd);
3553 VOP_UNLOCK(fromnd.ni_dvp, 0);
3554 if (fromnd.ni_dvp != fromnd.ni_vp)
3555 VOP_UNLOCK(fromnd.ni_vp, 0);
3558 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3559 SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3560 cap_rights_init(&rights, CAP_LINKAT), td);
3561 if (fromnd.ni_vp->v_type == VDIR)
3562 tond.ni_cnd.cn_flags |= WILLBEDIR;
3563 if ((error = namei(&tond)) != 0) {
3564 /* Translate error code for rename("dir1", "dir2/."). */
3565 if (error == EISDIR && fvp->v_type == VDIR)
3567 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3568 vrele(fromnd.ni_dvp);
3574 error = vn_start_write(fvp, &mp, V_NOWAIT);
3576 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3577 NDFREE(&tond, NDF_ONLY_PNBUF);
3584 vrele(fromnd.ni_dvp);
3586 vrele(tond.ni_startdir);
3587 if (fromnd.ni_startdir != NULL)
3588 vrele(fromnd.ni_startdir);
3589 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3595 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3598 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3603 if (newfd != AT_FDCWD) {
3605 * If the target already exists we require CAP_UNLINKAT
3608 error = cap_check(&tond.ni_filecaps.fc_rights,
3609 cap_rights_init(&rights, CAP_UNLINKAT));
3620 * If the source is the same as the destination (that is, if they
3621 * are links to the same vnode), then there is nothing to do.
3627 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3628 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3632 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3633 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3634 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3635 NDFREE(&tond, NDF_ONLY_PNBUF);
3637 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3638 NDFREE(&tond, NDF_ONLY_PNBUF);
3645 vrele(fromnd.ni_dvp);
3648 vrele(tond.ni_startdir);
3649 vn_finished_write(mp);
3651 if (fromnd.ni_startdir)
3652 vrele(fromnd.ni_startdir);
3659 * Make a directory file.
3661 #ifndef _SYS_SYSPROTO_H_
3670 register struct mkdir_args /* {
3676 return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3679 #ifndef _SYS_SYSPROTO_H_
3680 struct mkdirat_args {
3687 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3690 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3694 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3697 return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3701 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3707 struct nameidata nd;
3708 cap_rights_t rights;
3711 AUDIT_ARG_MODE(mode);
3714 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
3715 segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT), td);
3716 nd.ni_cnd.cn_flags |= WILLBEDIR;
3717 if ((error = namei(&nd)) != 0)
3721 NDFREE(&nd, NDF_ONLY_PNBUF);
3723 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3724 * the strange behaviour of leaving the vnode unlocked
3725 * if the target is the same vnode as the parent.
3727 if (vp == nd.ni_dvp)
3734 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3735 NDFREE(&nd, NDF_ONLY_PNBUF);
3737 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3742 vattr.va_type = VDIR;
3743 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3745 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3750 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3754 NDFREE(&nd, NDF_ONLY_PNBUF);
3758 vn_finished_write(mp);
3763 * Remove a directory file.
3765 #ifndef _SYS_SYSPROTO_H_
3773 struct rmdir_args /* {
3778 return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3782 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3785 return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3789 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3793 struct nameidata nd;
3794 cap_rights_t rights;
3799 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3800 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3801 if ((error = namei(&nd)) != 0)
3804 if (vp->v_type != VDIR) {
3809 * No rmdir "." please.
3811 if (nd.ni_dvp == vp) {
3816 * The root of a mounted filesystem cannot be deleted.
3818 if (vp->v_vflag & VV_ROOT) {
3823 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3828 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3829 NDFREE(&nd, NDF_ONLY_PNBUF);
3831 if (nd.ni_dvp == vp)
3835 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3839 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3840 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3841 vn_finished_write(mp);
3843 NDFREE(&nd, NDF_ONLY_PNBUF);
3845 if (nd.ni_dvp == vp)
3854 * Read a block of directory entries in a filesystem independent format.
3856 #ifndef _SYS_SYSPROTO_H_
3857 struct ogetdirentries_args {
3865 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3870 error = kern_ogetdirentries(td, uap, &loff);
3872 error = copyout(&loff, uap->basep, sizeof(long));
3877 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3882 struct uio auio, kuio;
3883 struct iovec aiov, kiov;
3884 struct dirent *dp, *edp;
3885 cap_rights_t rights;
3887 int error, eofflag, readcnt;
3891 /* XXX arbitrary sanity limit on `count'. */
3892 if (uap->count > 64 * 1024)
3894 error = getvnode(td->td_proc->p_fd, uap->fd,
3895 cap_rights_init(&rights, CAP_READ), &fp);
3898 if ((fp->f_flag & FREAD) == 0) {
3903 foffset = foffset_lock(fp, 0);
3905 if (vp->v_type != VDIR) {
3906 foffset_unlock(fp, foffset, 0);
3910 aiov.iov_base = uap->buf;
3911 aiov.iov_len = uap->count;
3912 auio.uio_iov = &aiov;
3913 auio.uio_iovcnt = 1;
3914 auio.uio_rw = UIO_READ;
3915 auio.uio_segflg = UIO_USERSPACE;
3917 auio.uio_resid = uap->count;
3918 vn_lock(vp, LK_SHARED | LK_RETRY);
3919 loff = auio.uio_offset = foffset;
3921 error = mac_vnode_check_readdir(td->td_ucred, vp);
3924 foffset_unlock(fp, foffset, FOF_NOUPDATE);
3929 # if (BYTE_ORDER != LITTLE_ENDIAN)
3930 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3931 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3933 foffset = auio.uio_offset;
3938 kuio.uio_iov = &kiov;
3939 kuio.uio_segflg = UIO_SYSSPACE;
3940 kiov.iov_len = uap->count;
3941 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3942 kiov.iov_base = dirbuf;
3943 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3945 foffset = kuio.uio_offset;
3947 readcnt = uap->count - kuio.uio_resid;
3948 edp = (struct dirent *)&dirbuf[readcnt];
3949 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3950 # if (BYTE_ORDER == LITTLE_ENDIAN)
3952 * The expected low byte of
3953 * dp->d_namlen is our dp->d_type.
3954 * The high MBZ byte of dp->d_namlen
3955 * is our dp->d_namlen.
3957 dp->d_type = dp->d_namlen;
3961 * The dp->d_type is the high byte
3962 * of the expected dp->d_namlen,
3963 * so must be zero'ed.
3967 if (dp->d_reclen > 0) {
3968 dp = (struct dirent *)
3969 ((char *)dp + dp->d_reclen);
3976 error = uiomove(dirbuf, readcnt, &auio);
3978 free(dirbuf, M_TEMP);
3982 foffset_unlock(fp, foffset, 0);
3986 if (uap->count == auio.uio_resid &&
3987 (vp->v_vflag & VV_ROOT) &&
3988 (vp->v_mount->mnt_flag & MNT_UNION)) {
3989 struct vnode *tvp = vp;
3990 vp = vp->v_mount->mnt_vnodecovered;
3999 foffset_unlock(fp, foffset, 0);
4001 td->td_retval[0] = uap->count - auio.uio_resid;
4006 #endif /* COMPAT_43 */
4009 * Read a block of directory entries in a filesystem independent format.
4011 #ifndef _SYS_SYSPROTO_H_
4012 struct getdirentries_args {
4020 sys_getdirentries(td, uap)
4022 register struct getdirentries_args /* {
4032 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4033 NULL, UIO_USERSPACE);
4036 if (uap->basep != NULL)
4037 error = copyout(&base, uap->basep, sizeof(long));
4042 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4043 long *basep, ssize_t *residp, enum uio_seg bufseg)
4049 cap_rights_t rights;
4055 if (count > IOSIZE_MAX)
4057 auio.uio_resid = count;
4058 error = getvnode(td->td_proc->p_fd, fd,
4059 cap_rights_init(&rights, CAP_READ), &fp);
4062 if ((fp->f_flag & FREAD) == 0) {
4067 foffset = foffset_lock(fp, 0);
4069 if (vp->v_type != VDIR) {
4073 aiov.iov_base = buf;
4074 aiov.iov_len = count;
4075 auio.uio_iov = &aiov;
4076 auio.uio_iovcnt = 1;
4077 auio.uio_rw = UIO_READ;
4078 auio.uio_segflg = bufseg;
4080 vn_lock(vp, LK_SHARED | LK_RETRY);
4081 AUDIT_ARG_VNODE1(vp);
4082 loff = auio.uio_offset = foffset;
4084 error = mac_vnode_check_readdir(td->td_ucred, vp);
4087 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4089 foffset = auio.uio_offset;
4094 if (count == auio.uio_resid &&
4095 (vp->v_vflag & VV_ROOT) &&
4096 (vp->v_mount->mnt_flag & MNT_UNION)) {
4097 struct vnode *tvp = vp;
4099 vp = vp->v_mount->mnt_vnodecovered;
4110 *residp = auio.uio_resid;
4111 td->td_retval[0] = count - auio.uio_resid;
4113 foffset_unlock(fp, foffset, 0);
4118 #ifndef _SYS_SYSPROTO_H_
4119 struct getdents_args {
4126 sys_getdents(td, uap)
4128 register struct getdents_args /* {
4134 struct getdirentries_args ap;
4138 ap.count = uap->count;
4140 return (sys_getdirentries(td, &ap));
4144 * Set the mode mask for creation of filesystem nodes.
4146 #ifndef _SYS_SYSPROTO_H_
4154 struct umask_args /* {
4158 register struct filedesc *fdp;
4160 FILEDESC_XLOCK(td->td_proc->p_fd);
4161 fdp = td->td_proc->p_fd;
4162 td->td_retval[0] = fdp->fd_cmask;
4163 fdp->fd_cmask = uap->newmask & ALLPERMS;
4164 FILEDESC_XUNLOCK(td->td_proc->p_fd);
4169 * Void all references to file by ripping underlying filesystem away from
4172 #ifndef _SYS_SYSPROTO_H_
4173 struct revoke_args {
4180 register struct revoke_args /* {
4186 struct nameidata nd;
4189 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4191 if ((error = namei(&nd)) != 0)
4194 NDFREE(&nd, NDF_ONLY_PNBUF);
4195 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4200 error = mac_vnode_check_revoke(td->td_ucred, vp);
4204 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4207 if (td->td_ucred->cr_uid != vattr.va_uid) {
4208 error = priv_check(td, PRIV_VFS_ADMIN);
4213 VOP_REVOKE(vp, REVOKEALL);
4220 * Convert a user file descriptor to a kernel file entry and check that, if it
4221 * is a capability, the correct rights are present. A reference on the file
4222 * entry is held upon returning.
4225 getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
4230 error = fget_unlocked(fdp, fd, rightsp, 0, &fp, NULL);
4235 * The file could be not of the vnode type, or it may be not
4236 * yet fully initialized, in which case the f_vnode pointer
4237 * may be set, but f_ops is still badfileops. E.g.,
4238 * devfs_open() transiently create such situation to
4239 * facilitate csw d_fdopen().
4241 * Dupfdopen() handling in kern_openat() installs the
4242 * half-baked file into the process descriptor table, allowing
4243 * other thread to dereference it. Guard against the race by
4246 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4247 fdrop(fp, curthread);
4256 * Get an (NFS) file handle.
4258 #ifndef _SYS_SYSPROTO_H_
4259 struct lgetfh_args {
4267 register struct lgetfh_args *uap;
4269 struct nameidata nd;
4271 register struct vnode *vp;
4274 error = priv_check(td, PRIV_VFS_GETFH);
4277 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4282 NDFREE(&nd, NDF_ONLY_PNBUF);
4284 bzero(&fh, sizeof(fh));
4285 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4286 error = VOP_VPTOFH(vp, &fh.fh_fid);
4289 error = copyout(&fh, uap->fhp, sizeof (fh));
4293 #ifndef _SYS_SYSPROTO_H_
4302 register struct getfh_args *uap;
4304 struct nameidata nd;
4306 register struct vnode *vp;
4309 error = priv_check(td, PRIV_VFS_GETFH);
4312 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4317 NDFREE(&nd, NDF_ONLY_PNBUF);
4319 bzero(&fh, sizeof(fh));
4320 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4321 error = VOP_VPTOFH(vp, &fh.fh_fid);
4324 error = copyout(&fh, uap->fhp, sizeof (fh));
4329 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4332 * warning: do not remove the priv_check() call or this becomes one giant
4335 #ifndef _SYS_SYSPROTO_H_
4336 struct fhopen_args {
4337 const struct fhandle *u_fhp;
4344 struct fhopen_args /* {
4345 const struct fhandle *u_fhp;
4356 error = priv_check(td, PRIV_VFS_FHOPEN);
4360 fmode = FFLAGS(uap->flags);
4361 /* why not allow a non-read/write open for our lockd? */
4362 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4364 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4367 /* find the mount point */
4368 mp = vfs_busyfs(&fhp.fh_fsid);
4371 /* now give me my vnode, it gets returned to me locked */
4372 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4377 error = falloc_noinstall(td, &fp);
4383 * An extra reference on `fp' has been held for us by
4384 * falloc_noinstall().
4390 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4392 KASSERT(fp->f_ops == &badfileops,
4393 ("VOP_OPEN in fhopen() set f_ops"));
4394 KASSERT(td->td_dupfd < 0,
4395 ("fhopen() encountered fdopen()"));
4405 finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4408 if ((fmode & O_TRUNC) != 0) {
4409 error = fo_truncate(fp, 0, td->td_ucred, td);
4414 error = finstall(td, fp, &indx, fmode, NULL);
4417 td->td_retval[0] = indx;
4422 * Stat an (NFS) file handle.
4424 #ifndef _SYS_SYSPROTO_H_
4425 struct fhstat_args {
4426 struct fhandle *u_fhp;
4433 register struct fhstat_args /* {
4434 struct fhandle *u_fhp;
4442 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4445 error = kern_fhstat(td, fh, &sb);
4447 error = copyout(&sb, uap->sb, sizeof(sb));
4452 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4458 error = priv_check(td, PRIV_VFS_FHSTAT);
4461 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4463 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4467 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4473 * Implement fstatfs() for (NFS) file handles.
4475 #ifndef _SYS_SYSPROTO_H_
4476 struct fhstatfs_args {
4477 struct fhandle *u_fhp;
4482 sys_fhstatfs(td, uap)
4484 struct fhstatfs_args /* {
4485 struct fhandle *u_fhp;
4493 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4496 error = kern_fhstatfs(td, fh, &sf);
4499 return (copyout(&sf, uap->buf, sizeof(sf)));
4503 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4510 error = priv_check(td, PRIV_VFS_FHSTATFS);
4513 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4515 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4521 error = prison_canseemount(td->td_ucred, mp);
4525 error = mac_mount_check_stat(td->td_ucred, mp);
4530 * Set these in case the underlying filesystem fails to do so.
4533 sp->f_version = STATFS_VERSION;
4534 sp->f_namemax = NAME_MAX;
4535 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4536 error = VFS_STATFS(mp, sp);
4545 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4550 cap_rights_t rights;
4551 off_t olen, ooffset;
4555 error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4559 switch (fp->f_type) {
4570 if ((fp->f_flag & FWRITE) == 0) {
4575 if (vp->v_type != VREG) {
4579 if (offset < 0 || len <= 0) {
4583 /* Check for wrap. */
4584 if (offset > OFF_MAX - len) {
4589 /* Allocating blocks may take a long time, so iterate. */
4596 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4599 error = vn_lock(vp, LK_EXCLUSIVE);
4601 vn_finished_write(mp);
4605 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4608 error = VOP_ALLOCATE(vp, &offset, &len);
4610 vn_finished_write(mp);
4612 if (olen + ooffset != offset + len) {
4613 panic("offset + len changed from %jx/%jx to %jx/%jx",
4614 ooffset, olen, offset, len);
4616 if (error != 0 || len == 0)
4618 KASSERT(olen > len, ("Iteration did not make progress?"));
4628 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4631 td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4637 * Unlike madvise(2), we do not make a best effort to remember every
4638 * possible caching hint. Instead, we remember the last setting with
4639 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4640 * region of any current setting.
4643 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4646 struct fadvise_info *fa, *new;
4649 cap_rights_t rights;
4653 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4656 case POSIX_FADV_SEQUENTIAL:
4657 case POSIX_FADV_RANDOM:
4658 case POSIX_FADV_NOREUSE:
4659 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4661 case POSIX_FADV_NORMAL:
4662 case POSIX_FADV_WILLNEED:
4663 case POSIX_FADV_DONTNEED:
4669 /* XXX: CAP_POSIX_FADVISE? */
4670 error = fget(td, fd, cap_rights_init(&rights), &fp);
4674 switch (fp->f_type) {
4686 if (vp->v_type != VREG) {
4693 end = offset + len - 1;
4695 case POSIX_FADV_SEQUENTIAL:
4696 case POSIX_FADV_RANDOM:
4697 case POSIX_FADV_NOREUSE:
4699 * Try to merge any existing non-standard region with
4700 * this new region if possible, otherwise create a new
4701 * non-standard region for this request.
4703 mtx_pool_lock(mtxpool_sleep, fp);
4705 if (fa != NULL && fa->fa_advice == advice &&
4706 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4707 (end != OFF_MAX && fa->fa_start == end + 1) ||
4708 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4709 if (offset < fa->fa_start)
4710 fa->fa_start = offset;
4711 if (end > fa->fa_end)
4714 new->fa_advice = advice;
4715 new->fa_start = offset;
4717 new->fa_prevstart = 0;
4718 new->fa_prevend = 0;
4722 mtx_pool_unlock(mtxpool_sleep, fp);
4724 case POSIX_FADV_NORMAL:
4726 * If a the "normal" region overlaps with an existing
4727 * non-standard region, trim or remove the
4728 * non-standard region.
4730 mtx_pool_lock(mtxpool_sleep, fp);
4733 if (offset <= fa->fa_start && end >= fa->fa_end) {
4735 fp->f_advice = NULL;
4736 } else if (offset <= fa->fa_start &&
4737 end >= fa->fa_start)
4738 fa->fa_start = end + 1;
4739 else if (offset <= fa->fa_end && end >= fa->fa_end)
4740 fa->fa_end = offset - 1;
4741 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4743 * If the "normal" region is a middle
4744 * portion of the existing
4745 * non-standard region, just remove
4746 * the whole thing rather than picking
4747 * one side or the other to
4751 fp->f_advice = NULL;
4754 mtx_pool_unlock(mtxpool_sleep, fp);
4756 case POSIX_FADV_WILLNEED:
4757 case POSIX_FADV_DONTNEED:
4758 error = VOP_ADVISE(vp, offset, end, advice);
4764 free(new, M_FADVISE);
4769 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4772 td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4773 uap->len, uap->advice);