2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. 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 kern_chflagsat(struct thread *td, int fd, const char *path,
98 enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102 struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104 const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
109 * Sync each mounted filesystem.
111 #ifndef _SYS_SYSPROTO_H_
120 struct sync_args *uap;
122 struct mount *mp, *nmp;
125 mtx_lock(&mountlist_mtx);
126 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
127 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
128 nmp = TAILQ_NEXT(mp, mnt_list);
131 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
132 vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
133 save = curthread_pflags_set(TDP_SYNCIO);
134 vfs_msync(mp, MNT_NOWAIT);
135 VFS_SYNC(mp, MNT_NOWAIT);
136 curthread_pflags_restore(save);
137 vn_finished_write(mp);
139 mtx_lock(&mountlist_mtx);
140 nmp = TAILQ_NEXT(mp, mnt_list);
143 mtx_unlock(&mountlist_mtx);
148 * Change filesystem quotas.
150 #ifndef _SYS_SYSPROTO_H_
151 struct quotactl_args {
159 sys_quotactl(td, uap)
161 register struct quotactl_args /* {
172 AUDIT_ARG_CMD(uap->cmd);
173 AUDIT_ARG_UID(uap->uid);
174 if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
176 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
178 if ((error = namei(&nd)) != 0)
180 NDFREE(&nd, NDF_ONLY_PNBUF);
181 mp = nd.ni_vp->v_mount;
184 error = vfs_busy(mp, 0);
188 error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
191 * Since quota on operation typically needs to open quota
192 * file, the Q_QUOTAON handler needs to unbusy the mount point
193 * before calling into namei. Otherwise, unmount might be
194 * started between two vfs_busy() invocations (first is our,
195 * second is from mount point cross-walk code in lookup()),
198 * Require that Q_QUOTAON handles the vfs_busy() reference on
199 * its own, always returning with ubusied mount point.
201 if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
207 * Used by statfs conversion routines to scale the block size up if
208 * necessary so that all of the block counts are <= 'max_size'. Note
209 * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
213 statfs_scale_blocks(struct statfs *sf, long max_size)
218 KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
221 * Attempt to scale the block counts to give a more accurate
222 * overview to userland of the ratio of free space to used
223 * space. To do this, find the largest block count and compute
224 * a divisor that lets it fit into a signed integer <= max_size.
226 if (sf->f_bavail < 0)
227 count = -sf->f_bavail;
229 count = sf->f_bavail;
230 count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
231 if (count <= max_size)
234 count >>= flsl(max_size);
241 sf->f_bsize <<= shift;
242 sf->f_blocks >>= shift;
243 sf->f_bfree >>= shift;
244 sf->f_bavail >>= shift;
248 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
255 error = vfs_busy(mp, 0);
260 error = mac_mount_check_stat(td->td_ucred, mp);
265 * Set these in case the underlying filesystem fails to do so.
268 sp->f_version = STATFS_VERSION;
269 sp->f_namemax = NAME_MAX;
270 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
271 error = VFS_STATFS(mp, sp);
275 if (priv_check(td, PRIV_VFS_GENERATION)) {
276 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
277 prison_enforce_statfs(td->td_ucred, mp, buf);
285 * Get filesystem statistics.
287 #ifndef _SYS_SYSPROTO_H_
296 register struct statfs_args /* {
304 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
305 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
307 error = copyout(sfp, uap->buf, sizeof(struct statfs));
313 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
320 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
325 mp = nd.ni_vp->v_mount;
327 NDFREE(&nd, NDF_ONLY_PNBUF);
329 return (kern_do_statfs(td, mp, buf));
333 * Get filesystem statistics.
335 #ifndef _SYS_SYSPROTO_H_
336 struct fstatfs_args {
344 register struct fstatfs_args /* {
352 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
353 error = kern_fstatfs(td, uap->fd, sfp);
355 error = copyout(sfp, uap->buf, sizeof(struct statfs));
361 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
370 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
374 vn_lock(vp, LK_SHARED | LK_RETRY);
376 AUDIT_ARG_VNODE1(vp);
383 return (kern_do_statfs(td, mp, buf));
387 * Get statistics on all filesystems.
389 #ifndef _SYS_SYSPROTO_H_
390 struct getfsstat_args {
397 sys_getfsstat(td, uap)
399 register struct getfsstat_args /* {
408 if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
410 error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
411 UIO_USERSPACE, uap->mode);
413 td->td_retval[0] = count;
418 * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
419 * The caller is responsible for freeing memory which will be allocated
423 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
424 size_t *countp, enum uio_seg bufseg, int mode)
426 struct mount *mp, *nmp;
427 struct statfs *sfsp, *sp, *sptmp, *tofree;
428 size_t count, maxcount;
436 if (bufseg == UIO_SYSSPACE)
441 maxcount = bufsize / sizeof(struct statfs);
445 } else if (bufseg == UIO_USERSPACE) {
448 } else /* if (bufseg == UIO_SYSSPACE) */ {
450 mtx_lock(&mountlist_mtx);
451 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
454 mtx_unlock(&mountlist_mtx);
455 if (maxcount > count)
457 tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
461 mtx_lock(&mountlist_mtx);
462 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
463 if (prison_canseemount(td->td_ucred, mp) != 0) {
464 nmp = TAILQ_NEXT(mp, mnt_list);
468 if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
469 nmp = TAILQ_NEXT(mp, mnt_list);
473 if (mode == MNT_WAIT) {
474 if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
476 * If vfs_busy() failed, and MBF_NOWAIT
477 * wasn't passed, then the mp is gone.
478 * Furthermore, because of MBF_MNTLSTLOCK,
479 * the mountlist_mtx was dropped. We have
480 * no other choice than to start over.
482 mtx_unlock(&mountlist_mtx);
483 free(tofree, M_STATFS);
487 if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
488 nmp = TAILQ_NEXT(mp, mnt_list);
492 if (sfsp != NULL && count < maxcount) {
495 * Set these in case the underlying filesystem
498 sp->f_version = STATFS_VERSION;
499 sp->f_namemax = NAME_MAX;
500 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
502 * If MNT_NOWAIT is specified, do not refresh
505 if (mode != MNT_NOWAIT) {
506 error = VFS_STATFS(mp, sp);
508 mtx_lock(&mountlist_mtx);
509 nmp = TAILQ_NEXT(mp, mnt_list);
514 if (priv_check(td, PRIV_VFS_GENERATION)) {
515 sptmp = malloc(sizeof(struct statfs), M_STATFS,
518 sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
519 prison_enforce_statfs(td->td_ucred, mp, sptmp);
523 if (bufseg == UIO_SYSSPACE) {
524 bcopy(sp, sfsp, sizeof(*sp));
525 free(sptmp, M_STATFS);
526 } else /* if (bufseg == UIO_USERSPACE) */ {
527 error = copyout(sp, sfsp, sizeof(*sp));
528 free(sptmp, M_STATFS);
537 mtx_lock(&mountlist_mtx);
538 nmp = TAILQ_NEXT(mp, mnt_list);
541 mtx_unlock(&mountlist_mtx);
542 if (sfsp != NULL && count > maxcount)
549 #ifdef COMPAT_FREEBSD4
551 * Get old format filesystem statistics.
553 static void cvtstatfs(struct statfs *, struct ostatfs *);
555 #ifndef _SYS_SYSPROTO_H_
556 struct freebsd4_statfs_args {
562 freebsd4_statfs(td, uap)
564 struct freebsd4_statfs_args /* {
573 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
574 error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
576 cvtstatfs(sfp, &osb);
577 error = copyout(&osb, uap->buf, sizeof(osb));
584 * Get filesystem statistics.
586 #ifndef _SYS_SYSPROTO_H_
587 struct freebsd4_fstatfs_args {
593 freebsd4_fstatfs(td, uap)
595 struct freebsd4_fstatfs_args /* {
604 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
605 error = kern_fstatfs(td, uap->fd, sfp);
607 cvtstatfs(sfp, &osb);
608 error = copyout(&osb, uap->buf, sizeof(osb));
615 * Get statistics on all filesystems.
617 #ifndef _SYS_SYSPROTO_H_
618 struct freebsd4_getfsstat_args {
625 freebsd4_getfsstat(td, uap)
627 register struct freebsd4_getfsstat_args /* {
633 struct statfs *buf, *sp;
638 if (uap->bufsize < 0)
640 count = uap->bufsize / sizeof(struct ostatfs);
641 if (count > SIZE_MAX / sizeof(struct statfs))
643 size = count * sizeof(struct statfs);
644 error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
646 td->td_retval[0] = count;
649 while (count != 0 && error == 0) {
651 error = copyout(&osb, uap->buf, sizeof(osb));
662 * Implement fstatfs() for (NFS) file handles.
664 #ifndef _SYS_SYSPROTO_H_
665 struct freebsd4_fhstatfs_args {
666 struct fhandle *u_fhp;
671 freebsd4_fhstatfs(td, uap)
673 struct freebsd4_fhstatfs_args /* {
674 struct fhandle *u_fhp;
683 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
686 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
687 error = kern_fhstatfs(td, fh, sfp);
689 cvtstatfs(sfp, &osb);
690 error = copyout(&osb, uap->buf, sizeof(osb));
697 * Convert a new format statfs structure to an old format statfs structure.
705 statfs_scale_blocks(nsp, LONG_MAX);
706 bzero(osp, sizeof(*osp));
707 osp->f_bsize = nsp->f_bsize;
708 osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
709 osp->f_blocks = nsp->f_blocks;
710 osp->f_bfree = nsp->f_bfree;
711 osp->f_bavail = nsp->f_bavail;
712 osp->f_files = MIN(nsp->f_files, LONG_MAX);
713 osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
714 osp->f_owner = nsp->f_owner;
715 osp->f_type = nsp->f_type;
716 osp->f_flags = nsp->f_flags;
717 osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
718 osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
719 osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
720 osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
721 strlcpy(osp->f_fstypename, nsp->f_fstypename,
722 MIN(MFSNAMELEN, OMFSNAMELEN));
723 strlcpy(osp->f_mntonname, nsp->f_mntonname,
724 MIN(MNAMELEN, OMNAMELEN));
725 strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
726 MIN(MNAMELEN, OMNAMELEN));
727 osp->f_fsid = nsp->f_fsid;
729 #endif /* COMPAT_FREEBSD4 */
732 * Change current working directory to a given file descriptor.
734 #ifndef _SYS_SYSPROTO_H_
742 struct fchdir_args /* {
746 struct vnode *vp, *tdp;
752 AUDIT_ARG_FD(uap->fd);
753 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
760 vn_lock(vp, LK_SHARED | LK_RETRY);
761 AUDIT_ARG_VNODE1(vp);
762 error = change_dir(vp, td);
763 while (!error && (mp = vp->v_mountedhere) != NULL) {
766 error = VFS_ROOT(mp, LK_SHARED, &tdp);
783 * Change current working directory (``.'').
785 #ifndef _SYS_SYSPROTO_H_
793 struct chdir_args /* {
798 return (kern_chdir(td, uap->path, UIO_USERSPACE));
802 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
807 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
809 if ((error = namei(&nd)) != 0)
811 if ((error = change_dir(nd.ni_vp, td)) != 0) {
813 NDFREE(&nd, NDF_ONLY_PNBUF);
816 VOP_UNLOCK(nd.ni_vp, 0);
817 NDFREE(&nd, NDF_ONLY_PNBUF);
818 pwd_chdir(td, nd.ni_vp);
823 * Change notion of root (``/'') directory.
825 #ifndef _SYS_SYSPROTO_H_
833 struct chroot_args /* {
840 error = priv_check(td, PRIV_VFS_CHROOT);
843 NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
844 UIO_USERSPACE, uap->path, td);
848 error = change_dir(nd.ni_vp, td);
852 error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
856 VOP_UNLOCK(nd.ni_vp, 0);
857 error = pwd_chroot(td, nd.ni_vp);
859 NDFREE(&nd, NDF_ONLY_PNBUF);
864 NDFREE(&nd, NDF_ONLY_PNBUF);
869 * Common routine for chroot and chdir. Callers must provide a locked vnode
881 ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
882 if (vp->v_type != VDIR)
885 error = mac_vnode_check_chdir(td->td_ucred, vp);
889 return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
893 flags_to_rights(int flags, cap_rights_t *rightsp)
896 if (flags & O_EXEC) {
897 cap_rights_set(rightsp, CAP_FEXECVE);
899 switch ((flags & O_ACCMODE)) {
901 cap_rights_set(rightsp, CAP_READ);
904 cap_rights_set(rightsp, CAP_READ);
907 cap_rights_set(rightsp, CAP_WRITE);
908 if (!(flags & (O_APPEND | O_TRUNC)))
909 cap_rights_set(rightsp, CAP_SEEK);
915 cap_rights_set(rightsp, CAP_CREATE);
918 cap_rights_set(rightsp, CAP_FTRUNCATE);
920 if (flags & (O_SYNC | O_FSYNC))
921 cap_rights_set(rightsp, CAP_FSYNC);
923 if (flags & (O_EXLOCK | O_SHLOCK))
924 cap_rights_set(rightsp, CAP_FLOCK);
928 * Check permissions, allocate an open file structure, and call the device
929 * open routine if any.
931 #ifndef _SYS_SYSPROTO_H_
941 register struct open_args /* {
948 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
949 uap->flags, uap->mode));
952 #ifndef _SYS_SYSPROTO_H_
961 sys_openat(struct thread *td, struct openat_args *uap)
964 AUDIT_ARG_FD(uap->fd);
965 return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
970 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
973 struct proc *p = td->td_proc;
974 struct filedesc *fdp = p->p_fd;
979 int cmode, error, indx;
983 AUDIT_ARG_FFLAGS(flags);
984 AUDIT_ARG_MODE(mode);
985 cap_rights_init(&rights, CAP_LOOKUP);
986 flags_to_rights(flags, &rights);
988 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
991 if (flags & O_EXEC) {
992 if (flags & O_ACCMODE)
994 } else if ((flags & O_ACCMODE) == O_ACCMODE) {
997 flags = FFLAGS(flags);
1001 * Allocate a file structure. The descriptor to reference it
1002 * is allocated and set by finstall() below.
1004 error = falloc_noinstall(td, &fp);
1008 * An extra reference on `fp' has been held for us by
1009 * falloc_noinstall().
1011 /* Set the flags early so the finit in devfs can pick them up. */
1012 fp->f_flag = flags & FMASK;
1013 cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1014 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1016 td->td_dupfd = -1; /* XXX check for fdopen */
1017 error = vn_open(&nd, &flags, cmode, fp);
1020 * If the vn_open replaced the method vector, something
1021 * wonderous happened deep below and we just pass it up
1022 * pretending we know what we do.
1024 if (error == ENXIO && fp->f_ops != &badfileops)
1028 * Handle special fdopen() case. bleh.
1030 * Don't do this for relative (capability) lookups; we don't
1031 * understand exactly what would happen, and we don't think
1032 * that it ever should.
1034 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1035 (error == ENODEV || error == ENXIO) &&
1036 td->td_dupfd >= 0) {
1037 error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1046 NDFREE(&nd, NDF_ONLY_PNBUF);
1050 * Store the vnode, for any f_type. Typically, the vnode use
1051 * count is decremented by direct call to vn_closefile() for
1052 * files that switched type in the cdevsw fdopen() method.
1056 * If the file wasn't claimed by devfs bind it to the normal
1057 * vnode operations here.
1059 if (fp->f_ops == &badfileops) {
1060 KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1062 finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1063 DTYPE_VNODE, vp, &vnops);
1067 if (flags & O_TRUNC) {
1068 error = fo_truncate(fp, 0, td->td_ucred, td);
1074 * If we haven't already installed the FD (for dupfdopen), do so now.
1077 struct filecaps *fcaps;
1080 if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1081 fcaps = &nd.ni_filecaps;
1085 error = finstall(td, fp, &indx, flags, fcaps);
1086 /* On success finstall() consumes fcaps. */
1088 filecaps_free(&nd.ni_filecaps);
1092 filecaps_free(&nd.ni_filecaps);
1096 * Release our private reference, leaving the one associated with
1097 * the descriptor table intact.
1100 td->td_retval[0] = indx;
1103 KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1112 #ifndef _SYS_SYSPROTO_H_
1113 struct ocreat_args {
1121 register struct ocreat_args /* {
1127 return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1128 O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1130 #endif /* COMPAT_43 */
1133 * Create a special file.
1135 #ifndef _SYS_SYSPROTO_H_
1145 register struct mknod_args /* {
1152 return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1153 uap->mode, uap->dev));
1156 #ifndef _SYS_SYSPROTO_H_
1157 struct mknodat_args {
1165 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1168 return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1173 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1179 struct nameidata nd;
1180 cap_rights_t rights;
1181 int error, whiteout = 0;
1183 AUDIT_ARG_MODE(mode);
1185 switch (mode & S_IFMT) {
1188 error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1189 if (error == 0 && dev == VNOVAL)
1193 error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1196 error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1200 return (kern_mkfifoat(td, fd, path, pathseg, mode));
1210 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1211 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1213 if ((error = namei(&nd)) != 0)
1217 NDFREE(&nd, NDF_ONLY_PNBUF);
1218 if (vp == nd.ni_dvp)
1226 vattr.va_mode = (mode & ALLPERMS) &
1227 ~td->td_proc->p_fd->fd_cmask;
1228 vattr.va_rdev = dev;
1231 switch (mode & S_IFMT) {
1232 case S_IFMT: /* used by badsect to flag bad sectors */
1233 vattr.va_type = VBAD;
1236 vattr.va_type = VCHR;
1239 vattr.va_type = VBLK;
1245 panic("kern_mknod: invalid mode");
1248 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1249 NDFREE(&nd, NDF_ONLY_PNBUF);
1251 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1256 if (error == 0 && !whiteout)
1257 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1258 &nd.ni_cnd, &vattr);
1262 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1264 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1265 &nd.ni_cnd, &vattr);
1270 NDFREE(&nd, NDF_ONLY_PNBUF);
1272 vn_finished_write(mp);
1277 * Create a named pipe.
1279 #ifndef _SYS_SYSPROTO_H_
1280 struct mkfifo_args {
1288 register struct mkfifo_args /* {
1294 return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1298 #ifndef _SYS_SYSPROTO_H_
1299 struct mkfifoat_args {
1306 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1309 return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1314 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1319 struct nameidata nd;
1320 cap_rights_t rights;
1323 AUDIT_ARG_MODE(mode);
1326 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1327 NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1329 if ((error = namei(&nd)) != 0)
1331 if (nd.ni_vp != NULL) {
1332 NDFREE(&nd, NDF_ONLY_PNBUF);
1333 if (nd.ni_vp == nd.ni_dvp)
1340 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1341 NDFREE(&nd, NDF_ONLY_PNBUF);
1343 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1348 vattr.va_type = VFIFO;
1349 vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1351 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1356 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1363 vn_finished_write(mp);
1364 NDFREE(&nd, NDF_ONLY_PNBUF);
1369 * Make a hard file link.
1371 #ifndef _SYS_SYSPROTO_H_
1380 register struct link_args /* {
1386 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1387 UIO_USERSPACE, FOLLOW));
1390 #ifndef _SYS_SYSPROTO_H_
1391 struct linkat_args {
1400 sys_linkat(struct thread *td, struct linkat_args *uap)
1405 if (flag & ~AT_SYMLINK_FOLLOW)
1408 return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1409 UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1412 int hardlink_check_uid = 0;
1413 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1414 &hardlink_check_uid, 0,
1415 "Unprivileged processes cannot create hard links to files owned by other "
1417 static int hardlink_check_gid = 0;
1418 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1419 &hardlink_check_gid, 0,
1420 "Unprivileged processes cannot create hard links to files owned by other "
1424 can_hardlink(struct vnode *vp, struct ucred *cred)
1429 if (!hardlink_check_uid && !hardlink_check_gid)
1432 error = VOP_GETATTR(vp, &va, cred);
1436 if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1437 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1442 if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1443 error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1452 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1453 enum uio_seg segflg, int follow)
1457 struct nameidata nd;
1458 cap_rights_t rights;
1463 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1464 cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1466 if ((error = namei(&nd)) != 0)
1468 NDFREE(&nd, NDF_ONLY_PNBUF);
1470 if (vp->v_type == VDIR) {
1472 return (EPERM); /* POSIX */
1474 NDINIT_ATRIGHTS(&nd, CREATE,
1475 LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1476 cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1477 if ((error = namei(&nd)) == 0) {
1478 if (nd.ni_vp != NULL) {
1479 NDFREE(&nd, NDF_ONLY_PNBUF);
1480 if (nd.ni_dvp == nd.ni_vp)
1487 } else if (nd.ni_dvp->v_mount != vp->v_mount) {
1489 * Cross-device link. No need to recheck
1490 * vp->v_type, since it cannot change, except
1493 NDFREE(&nd, NDF_ONLY_PNBUF);
1497 } else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1498 error = can_hardlink(vp, td->td_ucred);
1501 error = mac_vnode_check_link(td->td_ucred,
1502 nd.ni_dvp, vp, &nd.ni_cnd);
1507 NDFREE(&nd, NDF_ONLY_PNBUF);
1510 error = vn_start_write(vp, &mp, V_NOWAIT);
1514 NDFREE(&nd, NDF_ONLY_PNBUF);
1515 error = vn_start_write(NULL, &mp,
1521 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1524 vn_finished_write(mp);
1525 NDFREE(&nd, NDF_ONLY_PNBUF);
1528 NDFREE(&nd, NDF_ONLY_PNBUF);
1538 * Make a symbolic link.
1540 #ifndef _SYS_SYSPROTO_H_
1541 struct symlink_args {
1547 sys_symlink(td, uap)
1549 register struct symlink_args /* {
1555 return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1559 #ifndef _SYS_SYSPROTO_H_
1560 struct symlinkat_args {
1567 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1570 return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1575 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1576 enum uio_seg segflg)
1581 struct nameidata nd;
1583 cap_rights_t rights;
1585 if (segflg == UIO_SYSSPACE) {
1588 syspath = uma_zalloc(namei_zone, M_WAITOK);
1589 if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1592 AUDIT_ARG_TEXT(syspath);
1595 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1596 NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1598 if ((error = namei(&nd)) != 0)
1601 NDFREE(&nd, NDF_ONLY_PNBUF);
1602 if (nd.ni_vp == nd.ni_dvp)
1610 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1611 NDFREE(&nd, NDF_ONLY_PNBUF);
1613 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1618 vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1620 vattr.va_type = VLNK;
1621 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1626 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1632 NDFREE(&nd, NDF_ONLY_PNBUF);
1634 vn_finished_write(mp);
1636 if (segflg != UIO_SYSSPACE)
1637 uma_zfree(namei_zone, syspath);
1642 * Delete a whiteout from the filesystem.
1645 sys_undelete(td, uap)
1647 register struct undelete_args /* {
1652 struct nameidata nd;
1657 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1658 UIO_USERSPACE, uap->path, td);
1663 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1664 NDFREE(&nd, NDF_ONLY_PNBUF);
1665 if (nd.ni_vp == nd.ni_dvp)
1673 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1674 NDFREE(&nd, NDF_ONLY_PNBUF);
1676 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1680 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1681 NDFREE(&nd, NDF_ONLY_PNBUF);
1683 vn_finished_write(mp);
1688 * Delete a name from the filesystem.
1690 #ifndef _SYS_SYSPROTO_H_
1691 struct unlink_args {
1698 struct unlink_args /* {
1703 return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1706 #ifndef _SYS_SYSPROTO_H_
1707 struct unlinkat_args {
1714 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1716 int flag = uap->flag;
1718 char *path = uap->path;
1720 if (flag & ~AT_REMOVEDIR)
1723 if (flag & AT_REMOVEDIR)
1724 return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1726 return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1730 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1735 struct nameidata nd;
1737 cap_rights_t rights;
1742 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1743 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1744 if ((error = namei(&nd)) != 0)
1745 return (error == EINVAL ? EPERM : error);
1747 if (vp->v_type == VDIR && oldinum == 0) {
1748 error = EPERM; /* POSIX */
1749 } else if (oldinum != 0 &&
1750 ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1751 sb.st_ino != oldinum) {
1752 error = EIDRM; /* Identifier removed */
1755 * The root of a mounted filesystem cannot be deleted.
1757 * XXX: can this only be a VDIR case?
1759 if (vp->v_vflag & VV_ROOT)
1763 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1764 NDFREE(&nd, NDF_ONLY_PNBUF);
1766 if (vp == nd.ni_dvp)
1770 if ((error = vn_start_write(NULL, &mp,
1771 V_XSLEEP | PCATCH)) != 0)
1776 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1781 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1782 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1786 vn_finished_write(mp);
1788 NDFREE(&nd, NDF_ONLY_PNBUF);
1790 if (vp == nd.ni_dvp)
1798 * Reposition read/write file offset.
1800 #ifndef _SYS_SYSPROTO_H_
1809 sys_lseek(struct thread *td, struct lseek_args *uap)
1812 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1816 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1819 cap_rights_t rights;
1823 error = fget(td, fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1826 error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1827 fo_seek(fp, offset, whence, td) : ESPIPE;
1832 #if defined(COMPAT_43)
1834 * Reposition read/write file offset.
1836 #ifndef _SYS_SYSPROTO_H_
1837 struct olseek_args {
1844 olseek(struct thread *td, struct olseek_args *uap)
1847 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1849 #endif /* COMPAT_43 */
1851 #if defined(COMPAT_FREEBSD6)
1852 /* Version with the 'pad' argument */
1854 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1857 return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1862 * Check access permissions using passed credentials.
1865 vn_access(vp, user_flags, cred, td)
1874 /* Flags == 0 means only check for existence. */
1875 if (user_flags == 0)
1879 if (user_flags & R_OK)
1881 if (user_flags & W_OK)
1883 if (user_flags & X_OK)
1886 error = mac_vnode_check_access(cred, vp, accmode);
1890 if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1891 error = VOP_ACCESS(vp, accmode, cred, td);
1896 * Check access permissions using "real" credentials.
1898 #ifndef _SYS_SYSPROTO_H_
1899 struct access_args {
1907 register struct access_args /* {
1913 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1917 #ifndef _SYS_SYSPROTO_H_
1918 struct faccessat_args {
1926 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1929 return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1934 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1935 int flag, int amode)
1937 struct ucred *cred, *usecred;
1939 struct nameidata nd;
1940 cap_rights_t rights;
1943 if (flag & ~AT_EACCESS)
1945 if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1949 * Create and modify a temporary credential instead of one that
1950 * is potentially shared (if we need one).
1952 cred = td->td_ucred;
1953 if ((flag & AT_EACCESS) == 0 &&
1954 ((cred->cr_uid != cred->cr_ruid ||
1955 cred->cr_rgid != cred->cr_groups[0]))) {
1956 usecred = crdup(cred);
1957 usecred->cr_uid = cred->cr_ruid;
1958 usecred->cr_groups[0] = cred->cr_rgid;
1959 td->td_ucred = usecred;
1962 AUDIT_ARG_VALUE(amode);
1963 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1964 AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1966 if ((error = namei(&nd)) != 0)
1970 error = vn_access(vp, amode, usecred, td);
1971 NDFREE(&nd, NDF_ONLY_PNBUF);
1974 if (usecred != cred) {
1975 td->td_ucred = cred;
1982 * Check access permissions using "effective" credentials.
1984 #ifndef _SYS_SYSPROTO_H_
1985 struct eaccess_args {
1991 sys_eaccess(td, uap)
1993 register struct eaccess_args /* {
1999 return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2000 AT_EACCESS, uap->amode));
2003 #if defined(COMPAT_43)
2005 * Get file status; this version follows links.
2007 #ifndef _SYS_SYSPROTO_H_
2016 register struct ostat_args /* {
2025 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2030 return (copyout(&osb, uap->ub, sizeof (osb)));
2034 * Get file status; this version does not follow links.
2036 #ifndef _SYS_SYSPROTO_H_
2037 struct olstat_args {
2045 register struct olstat_args /* {
2054 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2055 UIO_USERSPACE, &sb, NULL);
2059 return (copyout(&osb, uap->ub, sizeof (osb)));
2063 * Convert from an old to a new stat structure.
2071 bzero(ost, sizeof(*ost));
2072 ost->st_dev = st->st_dev;
2073 ost->st_ino = st->st_ino;
2074 ost->st_mode = st->st_mode;
2075 ost->st_nlink = st->st_nlink;
2076 ost->st_uid = st->st_uid;
2077 ost->st_gid = st->st_gid;
2078 ost->st_rdev = st->st_rdev;
2079 if (st->st_size < (quad_t)1 << 32)
2080 ost->st_size = st->st_size;
2083 ost->st_atim = st->st_atim;
2084 ost->st_mtim = st->st_mtim;
2085 ost->st_ctim = st->st_ctim;
2086 ost->st_blksize = st->st_blksize;
2087 ost->st_blocks = st->st_blocks;
2088 ost->st_flags = st->st_flags;
2089 ost->st_gen = st->st_gen;
2091 #endif /* COMPAT_43 */
2094 * Get file status; this version follows links.
2096 #ifndef _SYS_SYSPROTO_H_
2105 register struct stat_args /* {
2113 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2116 error = copyout(&sb, uap->ub, sizeof (sb));
2120 #ifndef _SYS_SYSPROTO_H_
2121 struct fstatat_args {
2129 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2134 error = kern_statat(td, uap->flag, uap->fd, uap->path,
2135 UIO_USERSPACE, &sb, NULL);
2137 error = copyout(&sb, uap->buf, sizeof (sb));
2142 kern_statat(struct thread *td, int flag, int fd, char *path,
2143 enum uio_seg pathseg, struct stat *sbp,
2144 void (*hook)(struct vnode *vp, struct stat *sbp))
2146 struct nameidata nd;
2148 cap_rights_t rights;
2151 if (flag & ~AT_SYMLINK_NOFOLLOW)
2154 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2155 FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2156 cap_rights_init(&rights, CAP_FSTAT), td);
2158 if ((error = namei(&nd)) != 0)
2160 error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2162 SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2163 if (S_ISREG(sb.st_mode))
2164 SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2165 if (__predict_false(hook != NULL))
2166 hook(nd.ni_vp, &sb);
2168 NDFREE(&nd, NDF_ONLY_PNBUF);
2174 if (KTRPOINT(td, KTR_STRUCT))
2181 * Get file status; this version does not follow links.
2183 #ifndef _SYS_SYSPROTO_H_
2192 register struct lstat_args /* {
2200 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2201 UIO_USERSPACE, &sb, NULL);
2203 error = copyout(&sb, uap->ub, sizeof (sb));
2208 * Implementation of the NetBSD [l]stat() functions.
2216 bzero(nsb, sizeof *nsb);
2217 nsb->st_dev = sb->st_dev;
2218 nsb->st_ino = sb->st_ino;
2219 nsb->st_mode = sb->st_mode;
2220 nsb->st_nlink = sb->st_nlink;
2221 nsb->st_uid = sb->st_uid;
2222 nsb->st_gid = sb->st_gid;
2223 nsb->st_rdev = sb->st_rdev;
2224 nsb->st_atim = sb->st_atim;
2225 nsb->st_mtim = sb->st_mtim;
2226 nsb->st_ctim = sb->st_ctim;
2227 nsb->st_size = sb->st_size;
2228 nsb->st_blocks = sb->st_blocks;
2229 nsb->st_blksize = sb->st_blksize;
2230 nsb->st_flags = sb->st_flags;
2231 nsb->st_gen = sb->st_gen;
2232 nsb->st_birthtim = sb->st_birthtim;
2235 #ifndef _SYS_SYSPROTO_H_
2244 register struct nstat_args /* {
2253 error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2257 cvtnstat(&sb, &nsb);
2258 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2262 * NetBSD lstat. Get file status; this version does not follow links.
2264 #ifndef _SYS_SYSPROTO_H_
2273 register struct nlstat_args /* {
2282 error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2283 UIO_USERSPACE, &sb, NULL);
2286 cvtnstat(&sb, &nsb);
2287 return (copyout(&nsb, uap->ub, sizeof (nsb)));
2291 * Get configurable pathname variables.
2293 #ifndef _SYS_SYSPROTO_H_
2294 struct pathconf_args {
2300 sys_pathconf(td, uap)
2302 register struct pathconf_args /* {
2308 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2311 #ifndef _SYS_SYSPROTO_H_
2312 struct lpathconf_args {
2318 sys_lpathconf(td, uap)
2320 register struct lpathconf_args /* {
2326 return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2331 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2334 struct nameidata nd;
2337 NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2339 if ((error = namei(&nd)) != 0)
2341 NDFREE(&nd, NDF_ONLY_PNBUF);
2343 error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2349 * Return target name of a symbolic link.
2351 #ifndef _SYS_SYSPROTO_H_
2352 struct readlink_args {
2359 sys_readlink(td, uap)
2361 register struct readlink_args /* {
2368 return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2369 uap->buf, UIO_USERSPACE, uap->count));
2371 #ifndef _SYS_SYSPROTO_H_
2372 struct readlinkat_args {
2380 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2383 return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2384 uap->buf, UIO_USERSPACE, uap->bufsize));
2388 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2389 char *buf, enum uio_seg bufseg, size_t count)
2394 struct nameidata nd;
2397 if (count > IOSIZE_MAX)
2400 NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2401 pathseg, path, fd, td);
2403 if ((error = namei(&nd)) != 0)
2405 NDFREE(&nd, NDF_ONLY_PNBUF);
2408 error = mac_vnode_check_readlink(td->td_ucred, vp);
2414 if (vp->v_type != VLNK)
2417 aiov.iov_base = buf;
2418 aiov.iov_len = count;
2419 auio.uio_iov = &aiov;
2420 auio.uio_iovcnt = 1;
2421 auio.uio_offset = 0;
2422 auio.uio_rw = UIO_READ;
2423 auio.uio_segflg = bufseg;
2425 auio.uio_resid = count;
2426 error = VOP_READLINK(vp, &auio, td->td_ucred);
2427 td->td_retval[0] = count - auio.uio_resid;
2434 * Common implementation code for chflags() and fchflags().
2437 setfflags(td, vp, flags)
2446 /* We can't support the value matching VNOVAL. */
2447 if (flags == VNOVAL)
2448 return (EOPNOTSUPP);
2451 * Prevent non-root users from setting flags on devices. When
2452 * a device is reused, users can retain ownership of the device
2453 * if they are allowed to set flags and programs assume that
2454 * chown can't fail when done as root.
2456 if (vp->v_type == VCHR || vp->v_type == VBLK) {
2457 error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2462 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2465 vattr.va_flags = flags;
2466 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2468 error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2471 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2473 vn_finished_write(mp);
2478 * Change flags of a file given a path name.
2480 #ifndef _SYS_SYSPROTO_H_
2481 struct chflags_args {
2487 sys_chflags(td, uap)
2489 register struct chflags_args /* {
2495 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2499 #ifndef _SYS_SYSPROTO_H_
2500 struct chflagsat_args {
2508 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2511 const char *path = uap->path;
2512 u_long flags = uap->flags;
2513 int atflag = uap->atflag;
2515 if (atflag & ~AT_SYMLINK_NOFOLLOW)
2518 return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2522 * Same as chflags() but doesn't follow symlinks.
2525 sys_lchflags(td, uap)
2527 register struct lchflags_args /* {
2533 return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2534 uap->flags, AT_SYMLINK_NOFOLLOW));
2538 kern_chflagsat(struct thread *td, int fd, const char *path,
2539 enum uio_seg pathseg, u_long flags, int atflag)
2541 struct nameidata nd;
2542 cap_rights_t rights;
2545 AUDIT_ARG_FFLAGS(flags);
2546 follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2547 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2548 cap_rights_init(&rights, CAP_FCHFLAGS), td);
2549 if ((error = namei(&nd)) != 0)
2551 NDFREE(&nd, NDF_ONLY_PNBUF);
2552 error = setfflags(td, nd.ni_vp, flags);
2558 * Change flags of a file given a file descriptor.
2560 #ifndef _SYS_SYSPROTO_H_
2561 struct fchflags_args {
2567 sys_fchflags(td, uap)
2569 register struct fchflags_args /* {
2575 cap_rights_t rights;
2578 AUDIT_ARG_FD(uap->fd);
2579 AUDIT_ARG_FFLAGS(uap->flags);
2580 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2585 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2586 AUDIT_ARG_VNODE1(fp->f_vnode);
2587 VOP_UNLOCK(fp->f_vnode, 0);
2589 error = setfflags(td, fp->f_vnode, uap->flags);
2595 * Common implementation code for chmod(), lchmod() and fchmod().
2598 setfmode(td, cred, vp, mode)
2608 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2610 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2612 vattr.va_mode = mode & ALLPERMS;
2614 error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2617 error = VOP_SETATTR(vp, &vattr, cred);
2619 vn_finished_write(mp);
2624 * Change mode of a file given path name.
2626 #ifndef _SYS_SYSPROTO_H_
2635 register struct chmod_args /* {
2641 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2645 #ifndef _SYS_SYSPROTO_H_
2646 struct fchmodat_args {
2654 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2656 int flag = uap->flag;
2658 char *path = uap->path;
2659 mode_t mode = uap->mode;
2661 if (flag & ~AT_SYMLINK_NOFOLLOW)
2664 return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2668 * Change mode of a file given path name (don't follow links.)
2670 #ifndef _SYS_SYSPROTO_H_
2671 struct lchmod_args {
2679 register struct lchmod_args /* {
2685 return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2686 uap->mode, AT_SYMLINK_NOFOLLOW));
2690 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2691 mode_t mode, int flag)
2693 struct nameidata nd;
2694 cap_rights_t rights;
2697 AUDIT_ARG_MODE(mode);
2698 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2699 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2700 cap_rights_init(&rights, CAP_FCHMOD), td);
2701 if ((error = namei(&nd)) != 0)
2703 NDFREE(&nd, NDF_ONLY_PNBUF);
2704 error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2710 * Change mode of a file given a file descriptor.
2712 #ifndef _SYS_SYSPROTO_H_
2713 struct fchmod_args {
2719 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2722 cap_rights_t rights;
2725 AUDIT_ARG_FD(uap->fd);
2726 AUDIT_ARG_MODE(uap->mode);
2728 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2731 error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2737 * Common implementation for chown(), lchown(), and fchown()
2740 setfown(td, cred, vp, uid, gid)
2751 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2753 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2758 error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2762 error = VOP_SETATTR(vp, &vattr, cred);
2764 vn_finished_write(mp);
2769 * Set ownership given a path name.
2771 #ifndef _SYS_SYSPROTO_H_
2781 register struct chown_args /* {
2788 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2792 #ifndef _SYS_SYSPROTO_H_
2793 struct fchownat_args {
2802 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2807 if (flag & ~AT_SYMLINK_NOFOLLOW)
2810 return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2811 uap->gid, uap->flag));
2815 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2816 int uid, int gid, int flag)
2818 struct nameidata nd;
2819 cap_rights_t rights;
2822 AUDIT_ARG_OWNER(uid, gid);
2823 follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2824 NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2825 cap_rights_init(&rights, CAP_FCHOWN), td);
2827 if ((error = namei(&nd)) != 0)
2829 NDFREE(&nd, NDF_ONLY_PNBUF);
2830 error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2836 * Set ownership given a path name, do not cross symlinks.
2838 #ifndef _SYS_SYSPROTO_H_
2839 struct lchown_args {
2848 register struct lchown_args /* {
2855 return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2856 uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2860 * Set ownership given a file descriptor.
2862 #ifndef _SYS_SYSPROTO_H_
2863 struct fchown_args {
2872 register struct fchown_args /* {
2879 cap_rights_t rights;
2882 AUDIT_ARG_FD(uap->fd);
2883 AUDIT_ARG_OWNER(uap->uid, uap->gid);
2884 error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2887 error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2893 * Common implementation code for utimes(), lutimes(), and futimes().
2896 getutimes(usrtvp, tvpseg, tsp)
2897 const struct timeval *usrtvp;
2898 enum uio_seg tvpseg;
2899 struct timespec *tsp;
2901 struct timeval tv[2];
2902 const struct timeval *tvp;
2905 if (usrtvp == NULL) {
2906 vfs_timestamp(&tsp[0]);
2909 if (tvpseg == UIO_SYSSPACE) {
2912 if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2917 if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2918 tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2920 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2921 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2927 * Common implementation code for futimens(), utimensat().
2929 #define UTIMENS_NULL 0x1
2930 #define UTIMENS_EXIT 0x2
2932 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2933 struct timespec *tsp, int *retflags)
2935 struct timespec tsnow;
2938 vfs_timestamp(&tsnow);
2940 if (usrtsp == NULL) {
2943 *retflags |= UTIMENS_NULL;
2946 if (tspseg == UIO_SYSSPACE) {
2949 } else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2951 if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2952 *retflags |= UTIMENS_EXIT;
2953 if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2954 *retflags |= UTIMENS_NULL;
2955 if (tsp[0].tv_nsec == UTIME_OMIT)
2956 tsp[0].tv_sec = VNOVAL;
2957 else if (tsp[0].tv_nsec == UTIME_NOW)
2959 else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2961 if (tsp[1].tv_nsec == UTIME_OMIT)
2962 tsp[1].tv_sec = VNOVAL;
2963 else if (tsp[1].tv_nsec == UTIME_NOW)
2965 else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2972 * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2976 setutimes(td, vp, ts, numtimes, nullflag)
2979 const struct timespec *ts;
2985 int error, setbirthtime;
2987 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2989 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2991 if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
2992 timespeccmp(&ts[1], &vattr.va_birthtime, < ))
2995 vattr.va_atime = ts[0];
2996 vattr.va_mtime = ts[1];
2998 vattr.va_birthtime = ts[1];
3000 vattr.va_birthtime = ts[2];
3002 vattr.va_vaflags |= VA_UTIMES_NULL;
3004 error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3008 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3010 vn_finished_write(mp);
3015 * Set the access and modification times of a file.
3017 #ifndef _SYS_SYSPROTO_H_
3018 struct utimes_args {
3020 struct timeval *tptr;
3026 register struct utimes_args /* {
3028 struct timeval *tptr;
3032 return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3033 uap->tptr, UIO_USERSPACE));
3036 #ifndef _SYS_SYSPROTO_H_
3037 struct futimesat_args {
3040 const struct timeval * times;
3044 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3047 return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3048 uap->times, UIO_USERSPACE));
3052 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3053 struct timeval *tptr, enum uio_seg tptrseg)
3055 struct nameidata nd;
3056 struct timespec ts[2];
3057 cap_rights_t rights;
3060 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3062 NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3063 cap_rights_init(&rights, CAP_FUTIMES), td);
3065 if ((error = namei(&nd)) != 0)
3067 NDFREE(&nd, NDF_ONLY_PNBUF);
3068 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3074 * Set the access and modification times of a file.
3076 #ifndef _SYS_SYSPROTO_H_
3077 struct lutimes_args {
3079 struct timeval *tptr;
3083 sys_lutimes(td, uap)
3085 register struct lutimes_args /* {
3087 struct timeval *tptr;
3091 return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3096 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3097 struct timeval *tptr, enum uio_seg tptrseg)
3099 struct timespec ts[2];
3100 struct nameidata nd;
3103 if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3105 NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3106 if ((error = namei(&nd)) != 0)
3108 NDFREE(&nd, NDF_ONLY_PNBUF);
3109 error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3115 * Set the access and modification times of a file.
3117 #ifndef _SYS_SYSPROTO_H_
3118 struct futimes_args {
3120 struct timeval *tptr;
3124 sys_futimes(td, uap)
3126 register struct futimes_args /* {
3128 struct timeval *tptr;
3132 return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3136 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3137 enum uio_seg tptrseg)
3139 struct timespec ts[2];
3141 cap_rights_t rights;
3145 error = getutimes(tptr, tptrseg, ts);
3148 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3152 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3153 AUDIT_ARG_VNODE1(fp->f_vnode);
3154 VOP_UNLOCK(fp->f_vnode, 0);
3156 error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3162 sys_futimens(struct thread *td, struct futimens_args *uap)
3165 return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3169 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3170 enum uio_seg tptrseg)
3172 struct timespec ts[2];
3174 cap_rights_t rights;
3178 error = getutimens(tptr, tptrseg, ts, &flags);
3181 if (flags & UTIMENS_EXIT)
3183 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3187 vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3188 AUDIT_ARG_VNODE1(fp->f_vnode);
3189 VOP_UNLOCK(fp->f_vnode, 0);
3191 error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3197 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3200 return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3201 uap->times, UIO_USERSPACE, uap->flag));
3205 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3206 struct timespec *tptr, enum uio_seg tptrseg, int flag)
3208 struct nameidata nd;
3209 struct timespec ts[2];
3210 cap_rights_t rights;
3213 if (flag & ~AT_SYMLINK_NOFOLLOW)
3216 if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3218 NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3219 FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3220 cap_rights_init(&rights, CAP_FUTIMES), td);
3221 if ((error = namei(&nd)) != 0)
3224 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3226 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3227 * "Search permission is denied by a component of the path prefix."
3229 NDFREE(&nd, NDF_ONLY_PNBUF);
3230 if ((flags & UTIMENS_EXIT) == 0)
3231 error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3237 * Truncate a file given its path name.
3239 #ifndef _SYS_SYSPROTO_H_
3240 struct truncate_args {
3247 sys_truncate(td, uap)
3249 register struct truncate_args /* {
3256 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3260 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3266 struct nameidata nd;
3271 NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3272 if ((error = namei(&nd)) != 0)
3275 rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3276 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3277 vn_rangelock_unlock(vp, rl_cookie);
3281 NDFREE(&nd, NDF_ONLY_PNBUF);
3282 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3283 if (vp->v_type == VDIR)
3286 else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3289 else if ((error = vn_writechk(vp)) == 0 &&
3290 (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3292 vattr.va_size = length;
3293 error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3296 vn_finished_write(mp);
3297 vn_rangelock_unlock(vp, rl_cookie);
3302 #if defined(COMPAT_43)
3304 * Truncate a file given its path name.
3306 #ifndef _SYS_SYSPROTO_H_
3307 struct otruncate_args {
3313 otruncate(struct thread *td, struct otruncate_args *uap)
3316 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3318 #endif /* COMPAT_43 */
3320 #if defined(COMPAT_FREEBSD6)
3321 /* Versions with the pad argument */
3323 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3326 return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3330 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3333 return (kern_ftruncate(td, uap->fd, uap->length));
3338 kern_fsync(struct thread *td, int fd, bool fullsync)
3343 cap_rights_t rights;
3344 int error, lock_flags;
3347 error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3353 /* XXXKIB: compete outstanding aio writes */;
3355 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3358 if (MNT_SHARED_WRITES(mp) ||
3359 ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3360 lock_flags = LK_SHARED;
3362 lock_flags = LK_EXCLUSIVE;
3364 vn_lock(vp, lock_flags | LK_RETRY);
3365 AUDIT_ARG_VNODE1(vp);
3366 if (vp->v_object != NULL) {
3367 VM_OBJECT_WLOCK(vp->v_object);
3368 vm_object_page_clean(vp->v_object, 0, 0, 0);
3369 VM_OBJECT_WUNLOCK(vp->v_object);
3371 error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3373 vn_finished_write(mp);
3380 * Sync an open file.
3382 #ifndef _SYS_SYSPROTO_H_
3388 sys_fsync(struct thread *td, struct fsync_args *uap)
3391 return (kern_fsync(td, uap->fd, true));
3395 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3398 return (kern_fsync(td, uap->fd, false));
3402 * Rename files. Source and destination must either both be directories, or
3403 * both not be directories. If target is a directory, it must be empty.
3405 #ifndef _SYS_SYSPROTO_H_
3406 struct rename_args {
3414 register struct rename_args /* {
3420 return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3421 uap->to, UIO_USERSPACE));
3424 #ifndef _SYS_SYSPROTO_H_
3425 struct renameat_args {
3433 sys_renameat(struct thread *td, struct renameat_args *uap)
3436 return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3441 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3442 enum uio_seg pathseg)
3444 struct mount *mp = NULL;
3445 struct vnode *tvp, *fvp, *tdvp;
3446 struct nameidata fromnd, tond;
3447 cap_rights_t rights;
3453 NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3454 AUDITVNODE1, pathseg, old, oldfd,
3455 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3457 NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3458 pathseg, old, oldfd,
3459 cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3462 if ((error = namei(&fromnd)) != 0)
3465 error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3466 fromnd.ni_vp, &fromnd.ni_cnd);
3467 VOP_UNLOCK(fromnd.ni_dvp, 0);
3468 if (fromnd.ni_dvp != fromnd.ni_vp)
3469 VOP_UNLOCK(fromnd.ni_vp, 0);
3472 NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3473 SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3474 cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3475 if (fromnd.ni_vp->v_type == VDIR)
3476 tond.ni_cnd.cn_flags |= WILLBEDIR;
3477 if ((error = namei(&tond)) != 0) {
3478 /* Translate error code for rename("dir1", "dir2/."). */
3479 if (error == EISDIR && fvp->v_type == VDIR)
3481 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3482 vrele(fromnd.ni_dvp);
3488 error = vn_start_write(fvp, &mp, V_NOWAIT);
3490 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3491 NDFREE(&tond, NDF_ONLY_PNBUF);
3498 vrele(fromnd.ni_dvp);
3500 vrele(tond.ni_startdir);
3501 if (fromnd.ni_startdir != NULL)
3502 vrele(fromnd.ni_startdir);
3503 error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3509 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3512 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3517 if (newfd != AT_FDCWD) {
3519 * If the target already exists we require CAP_UNLINKAT
3522 error = cap_check(&tond.ni_filecaps.fc_rights,
3523 cap_rights_init(&rights, CAP_UNLINKAT));
3534 * If the source is the same as the destination (that is, if they
3535 * are links to the same vnode), then there is nothing to do.
3541 error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3542 tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3546 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3547 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3548 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3549 NDFREE(&tond, NDF_ONLY_PNBUF);
3551 NDFREE(&fromnd, NDF_ONLY_PNBUF);
3552 NDFREE(&tond, NDF_ONLY_PNBUF);
3559 vrele(fromnd.ni_dvp);
3562 vrele(tond.ni_startdir);
3563 vn_finished_write(mp);
3565 if (fromnd.ni_startdir)
3566 vrele(fromnd.ni_startdir);
3573 * Make a directory file.
3575 #ifndef _SYS_SYSPROTO_H_
3584 register struct mkdir_args /* {
3590 return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3594 #ifndef _SYS_SYSPROTO_H_
3595 struct mkdirat_args {
3602 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3605 return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3609 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3615 struct nameidata nd;
3616 cap_rights_t rights;
3619 AUDIT_ARG_MODE(mode);
3622 NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3623 NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3625 nd.ni_cnd.cn_flags |= WILLBEDIR;
3626 if ((error = namei(&nd)) != 0)
3630 NDFREE(&nd, NDF_ONLY_PNBUF);
3632 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3633 * the strange behaviour of leaving the vnode unlocked
3634 * if the target is the same vnode as the parent.
3636 if (vp == nd.ni_dvp)
3643 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3644 NDFREE(&nd, NDF_ONLY_PNBUF);
3646 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3651 vattr.va_type = VDIR;
3652 vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3654 error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3659 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3663 NDFREE(&nd, NDF_ONLY_PNBUF);
3667 vn_finished_write(mp);
3672 * Remove a directory file.
3674 #ifndef _SYS_SYSPROTO_H_
3682 struct rmdir_args /* {
3687 return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3691 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3695 struct nameidata nd;
3696 cap_rights_t rights;
3701 NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3702 pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3703 if ((error = namei(&nd)) != 0)
3706 if (vp->v_type != VDIR) {
3711 * No rmdir "." please.
3713 if (nd.ni_dvp == vp) {
3718 * The root of a mounted filesystem cannot be deleted.
3720 if (vp->v_vflag & VV_ROOT) {
3725 error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3730 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3731 NDFREE(&nd, NDF_ONLY_PNBUF);
3733 if (nd.ni_dvp == vp)
3737 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3741 vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3742 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3743 vn_finished_write(mp);
3745 NDFREE(&nd, NDF_ONLY_PNBUF);
3747 if (nd.ni_dvp == vp)
3756 * Read a block of directory entries in a filesystem independent format.
3758 #ifndef _SYS_SYSPROTO_H_
3759 struct ogetdirentries_args {
3767 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3772 error = kern_ogetdirentries(td, uap, &loff);
3774 error = copyout(&loff, uap->basep, sizeof(long));
3779 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3784 struct uio auio, kuio;
3785 struct iovec aiov, kiov;
3786 struct dirent *dp, *edp;
3787 cap_rights_t rights;
3789 int error, eofflag, readcnt;
3793 /* XXX arbitrary sanity limit on `count'. */
3794 if (uap->count > 64 * 1024)
3796 error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp);
3799 if ((fp->f_flag & FREAD) == 0) {
3804 foffset = foffset_lock(fp, 0);
3806 if (vp->v_type != VDIR) {
3807 foffset_unlock(fp, foffset, 0);
3811 aiov.iov_base = uap->buf;
3812 aiov.iov_len = uap->count;
3813 auio.uio_iov = &aiov;
3814 auio.uio_iovcnt = 1;
3815 auio.uio_rw = UIO_READ;
3816 auio.uio_segflg = UIO_USERSPACE;
3818 auio.uio_resid = uap->count;
3819 vn_lock(vp, LK_SHARED | LK_RETRY);
3820 loff = auio.uio_offset = foffset;
3822 error = mac_vnode_check_readdir(td->td_ucred, vp);
3825 foffset_unlock(fp, foffset, FOF_NOUPDATE);
3830 # if (BYTE_ORDER != LITTLE_ENDIAN)
3831 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3832 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3834 foffset = auio.uio_offset;
3839 kuio.uio_iov = &kiov;
3840 kuio.uio_segflg = UIO_SYSSPACE;
3841 kiov.iov_len = uap->count;
3842 dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3843 kiov.iov_base = dirbuf;
3844 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3846 foffset = kuio.uio_offset;
3848 readcnt = uap->count - kuio.uio_resid;
3849 edp = (struct dirent *)&dirbuf[readcnt];
3850 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3851 # if (BYTE_ORDER == LITTLE_ENDIAN)
3853 * The expected low byte of
3854 * dp->d_namlen is our dp->d_type.
3855 * The high MBZ byte of dp->d_namlen
3856 * is our dp->d_namlen.
3858 dp->d_type = dp->d_namlen;
3862 * The dp->d_type is the high byte
3863 * of the expected dp->d_namlen,
3864 * so must be zero'ed.
3868 if (dp->d_reclen > 0) {
3869 dp = (struct dirent *)
3870 ((char *)dp + dp->d_reclen);
3877 error = uiomove(dirbuf, readcnt, &auio);
3879 free(dirbuf, M_TEMP);
3883 foffset_unlock(fp, foffset, 0);
3887 if (uap->count == auio.uio_resid &&
3888 (vp->v_vflag & VV_ROOT) &&
3889 (vp->v_mount->mnt_flag & MNT_UNION)) {
3890 struct vnode *tvp = vp;
3891 vp = vp->v_mount->mnt_vnodecovered;
3900 foffset_unlock(fp, foffset, 0);
3902 td->td_retval[0] = uap->count - auio.uio_resid;
3907 #endif /* COMPAT_43 */
3910 * Read a block of directory entries in a filesystem independent format.
3912 #ifndef _SYS_SYSPROTO_H_
3913 struct getdirentries_args {
3921 sys_getdirentries(td, uap)
3923 register struct getdirentries_args /* {
3933 error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3934 NULL, UIO_USERSPACE);
3937 if (uap->basep != NULL)
3938 error = copyout(&base, uap->basep, sizeof(long));
3943 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3944 long *basep, ssize_t *residp, enum uio_seg bufseg)
3950 cap_rights_t rights;
3956 if (count > IOSIZE_MAX)
3958 auio.uio_resid = count;
3959 error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3962 if ((fp->f_flag & FREAD) == 0) {
3967 foffset = foffset_lock(fp, 0);
3969 if (vp->v_type != VDIR) {
3973 aiov.iov_base = buf;
3974 aiov.iov_len = count;
3975 auio.uio_iov = &aiov;
3976 auio.uio_iovcnt = 1;
3977 auio.uio_rw = UIO_READ;
3978 auio.uio_segflg = bufseg;
3980 vn_lock(vp, LK_SHARED | LK_RETRY);
3981 AUDIT_ARG_VNODE1(vp);
3982 loff = auio.uio_offset = foffset;
3984 error = mac_vnode_check_readdir(td->td_ucred, vp);
3987 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3989 foffset = auio.uio_offset;
3994 if (count == auio.uio_resid &&
3995 (vp->v_vflag & VV_ROOT) &&
3996 (vp->v_mount->mnt_flag & MNT_UNION)) {
3997 struct vnode *tvp = vp;
3999 vp = vp->v_mount->mnt_vnodecovered;
4010 *residp = auio.uio_resid;
4011 td->td_retval[0] = count - auio.uio_resid;
4013 foffset_unlock(fp, foffset, 0);
4018 #ifndef _SYS_SYSPROTO_H_
4019 struct getdents_args {
4026 sys_getdents(td, uap)
4028 register struct getdents_args /* {
4034 struct getdirentries_args ap;
4038 ap.count = uap->count;
4040 return (sys_getdirentries(td, &ap));
4044 * Set the mode mask for creation of filesystem nodes.
4046 #ifndef _SYS_SYSPROTO_H_
4054 struct umask_args /* {
4058 struct filedesc *fdp;
4060 fdp = td->td_proc->p_fd;
4061 FILEDESC_XLOCK(fdp);
4062 td->td_retval[0] = fdp->fd_cmask;
4063 fdp->fd_cmask = uap->newmask & ALLPERMS;
4064 FILEDESC_XUNLOCK(fdp);
4069 * Void all references to file by ripping underlying filesystem away from
4072 #ifndef _SYS_SYSPROTO_H_
4073 struct revoke_args {
4080 register struct revoke_args /* {
4086 struct nameidata nd;
4089 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4091 if ((error = namei(&nd)) != 0)
4094 NDFREE(&nd, NDF_ONLY_PNBUF);
4095 if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4100 error = mac_vnode_check_revoke(td->td_ucred, vp);
4104 error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4107 if (td->td_ucred->cr_uid != vattr.va_uid) {
4108 error = priv_check(td, PRIV_VFS_ADMIN);
4113 VOP_REVOKE(vp, REVOKEALL);
4120 * Convert a user file descriptor to a kernel file entry and check that, if it
4121 * is a capability, the correct rights are present. A reference on the file
4122 * entry is held upon returning.
4125 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4130 error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4135 * The file could be not of the vnode type, or it may be not
4136 * yet fully initialized, in which case the f_vnode pointer
4137 * may be set, but f_ops is still badfileops. E.g.,
4138 * devfs_open() transiently create such situation to
4139 * facilitate csw d_fdopen().
4141 * Dupfdopen() handling in kern_openat() installs the
4142 * half-baked file into the process descriptor table, allowing
4143 * other thread to dereference it. Guard against the race by
4146 if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4156 * Get an (NFS) file handle.
4158 #ifndef _SYS_SYSPROTO_H_
4159 struct lgetfh_args {
4167 register struct lgetfh_args *uap;
4169 struct nameidata nd;
4171 register struct vnode *vp;
4174 error = priv_check(td, PRIV_VFS_GETFH);
4177 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4182 NDFREE(&nd, NDF_ONLY_PNBUF);
4184 bzero(&fh, sizeof(fh));
4185 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4186 error = VOP_VPTOFH(vp, &fh.fh_fid);
4189 error = copyout(&fh, uap->fhp, sizeof (fh));
4193 #ifndef _SYS_SYSPROTO_H_
4202 register struct getfh_args *uap;
4204 struct nameidata nd;
4206 register struct vnode *vp;
4209 error = priv_check(td, PRIV_VFS_GETFH);
4212 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4217 NDFREE(&nd, NDF_ONLY_PNBUF);
4219 bzero(&fh, sizeof(fh));
4220 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4221 error = VOP_VPTOFH(vp, &fh.fh_fid);
4224 error = copyout(&fh, uap->fhp, sizeof (fh));
4229 * syscall for the rpc.lockd to use to translate a NFS file handle into an
4232 * warning: do not remove the priv_check() call or this becomes one giant
4235 #ifndef _SYS_SYSPROTO_H_
4236 struct fhopen_args {
4237 const struct fhandle *u_fhp;
4244 struct fhopen_args /* {
4245 const struct fhandle *u_fhp;
4256 error = priv_check(td, PRIV_VFS_FHOPEN);
4260 fmode = FFLAGS(uap->flags);
4261 /* why not allow a non-read/write open for our lockd? */
4262 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4264 error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4267 /* find the mount point */
4268 mp = vfs_busyfs(&fhp.fh_fsid);
4271 /* now give me my vnode, it gets returned to me locked */
4272 error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4277 error = falloc_noinstall(td, &fp);
4283 * An extra reference on `fp' has been held for us by
4284 * falloc_noinstall().
4290 error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4292 KASSERT(fp->f_ops == &badfileops,
4293 ("VOP_OPEN in fhopen() set f_ops"));
4294 KASSERT(td->td_dupfd < 0,
4295 ("fhopen() encountered fdopen()"));
4305 finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4308 if ((fmode & O_TRUNC) != 0) {
4309 error = fo_truncate(fp, 0, td->td_ucred, td);
4314 error = finstall(td, fp, &indx, fmode, NULL);
4317 td->td_retval[0] = indx;
4322 * Stat an (NFS) file handle.
4324 #ifndef _SYS_SYSPROTO_H_
4325 struct fhstat_args {
4326 struct fhandle *u_fhp;
4333 register struct fhstat_args /* {
4334 struct fhandle *u_fhp;
4342 error = copyin(uap->u_fhp, &fh, sizeof(fh));
4345 error = kern_fhstat(td, fh, &sb);
4347 error = copyout(&sb, uap->sb, sizeof(sb));
4352 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4358 error = priv_check(td, PRIV_VFS_FHSTAT);
4361 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4363 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4367 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4373 * Implement fstatfs() for (NFS) file handles.
4375 #ifndef _SYS_SYSPROTO_H_
4376 struct fhstatfs_args {
4377 struct fhandle *u_fhp;
4382 sys_fhstatfs(td, uap)
4384 struct fhstatfs_args /* {
4385 struct fhandle *u_fhp;
4393 error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4396 sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4397 error = kern_fhstatfs(td, fh, sfp);
4399 error = copyout(sfp, uap->buf, sizeof(*sfp));
4400 free(sfp, M_STATFS);
4405 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4412 error = priv_check(td, PRIV_VFS_FHSTATFS);
4415 if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4417 error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4423 error = prison_canseemount(td->td_ucred, mp);
4427 error = mac_mount_check_stat(td->td_ucred, mp);
4432 * Set these in case the underlying filesystem fails to do so.
4435 sp->f_version = STATFS_VERSION;
4436 sp->f_namemax = NAME_MAX;
4437 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4438 error = VFS_STATFS(mp, sp);
4447 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4452 cap_rights_t rights;
4453 off_t olen, ooffset;
4456 if (offset < 0 || len <= 0)
4458 /* Check for wrap. */
4459 if (offset > OFF_MAX - len)
4461 error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4464 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4468 if ((fp->f_flag & FWRITE) == 0) {
4472 if (fp->f_type != DTYPE_VNODE) {
4477 if (vp->v_type != VREG) {
4482 /* Allocating blocks may take a long time, so iterate. */
4489 error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4492 error = vn_lock(vp, LK_EXCLUSIVE);
4494 vn_finished_write(mp);
4498 error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4501 error = VOP_ALLOCATE(vp, &offset, &len);
4503 vn_finished_write(mp);
4505 if (olen + ooffset != offset + len) {
4506 panic("offset + len changed from %jx/%jx to %jx/%jx",
4507 ooffset, olen, offset, len);
4509 if (error != 0 || len == 0)
4511 KASSERT(olen > len, ("Iteration did not make progress?"));
4520 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4524 error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4525 return (kern_posix_error(td, error));
4529 * Unlike madvise(2), we do not make a best effort to remember every
4530 * possible caching hint. Instead, we remember the last setting with
4531 * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4532 * region of any current setting.
4535 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4538 struct fadvise_info *fa, *new;
4541 cap_rights_t rights;
4545 if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4548 case POSIX_FADV_SEQUENTIAL:
4549 case POSIX_FADV_RANDOM:
4550 case POSIX_FADV_NOREUSE:
4551 new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4553 case POSIX_FADV_NORMAL:
4554 case POSIX_FADV_WILLNEED:
4555 case POSIX_FADV_DONTNEED:
4561 /* XXX: CAP_POSIX_FADVISE? */
4562 error = fget(td, fd, cap_rights_init(&rights), &fp);
4565 if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4569 if (fp->f_type != DTYPE_VNODE) {
4574 if (vp->v_type != VREG) {
4581 end = offset + len - 1;
4583 case POSIX_FADV_SEQUENTIAL:
4584 case POSIX_FADV_RANDOM:
4585 case POSIX_FADV_NOREUSE:
4587 * Try to merge any existing non-standard region with
4588 * this new region if possible, otherwise create a new
4589 * non-standard region for this request.
4591 mtx_pool_lock(mtxpool_sleep, fp);
4593 if (fa != NULL && fa->fa_advice == advice &&
4594 ((fa->fa_start <= end && fa->fa_end >= offset) ||
4595 (end != OFF_MAX && fa->fa_start == end + 1) ||
4596 (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4597 if (offset < fa->fa_start)
4598 fa->fa_start = offset;
4599 if (end > fa->fa_end)
4602 new->fa_advice = advice;
4603 new->fa_start = offset;
4608 mtx_pool_unlock(mtxpool_sleep, fp);
4610 case POSIX_FADV_NORMAL:
4612 * If a the "normal" region overlaps with an existing
4613 * non-standard region, trim or remove the
4614 * non-standard region.
4616 mtx_pool_lock(mtxpool_sleep, fp);
4619 if (offset <= fa->fa_start && end >= fa->fa_end) {
4621 fp->f_advice = NULL;
4622 } else if (offset <= fa->fa_start &&
4623 end >= fa->fa_start)
4624 fa->fa_start = end + 1;
4625 else if (offset <= fa->fa_end && end >= fa->fa_end)
4626 fa->fa_end = offset - 1;
4627 else if (offset >= fa->fa_start && end <= fa->fa_end) {
4629 * If the "normal" region is a middle
4630 * portion of the existing
4631 * non-standard region, just remove
4632 * the whole thing rather than picking
4633 * one side or the other to
4637 fp->f_advice = NULL;
4640 mtx_pool_unlock(mtxpool_sleep, fp);
4642 case POSIX_FADV_WILLNEED:
4643 case POSIX_FADV_DONTNEED:
4644 error = VOP_ADVISE(vp, offset, end, advice);
4650 free(new, M_FADVISE);
4655 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4659 error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4661 return (kern_posix_error(td, error));