/*
* ufs_link: create hard link.
*/
int
ufs_link(void *v)
{
struct vop_link_args /* {
struct vnode *a_dvp;
struct vnode *a_vp;
struct componentname *a_cnp;
} */ *ap = v;
struct vnode *dvp = ap->a_dvp;
struct vnode *vp = ap->a_vp;
struct componentname *cnp = ap->a_cnp;
struct inode *ip;
struct direct *newdir;
int error;
struct ufs_lookup_results *ulr;
KASSERT(dvp != vp);
KASSERT(vp->v_type != VDIR);
KASSERT(dvp->v_mount == vp->v_mount);
/* XXX should handle this material another way */
ulr = &VTOI(dvp)->i_crap;
UFS_CHECK_CRAPCOUNTER(VTOI(dvp));
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
error = vn_lock(vp, LK_EXCLUSIVE);
if (error) {
VOP_ABORTOP(dvp, cnp);
goto out2;
}
ip = VTOI(vp);
if ((nlink_t)ip->i_nlink >= LINK_MAX) {
VOP_ABORTOP(dvp, cnp);
error = EMLINK;
goto out1;
}
if (ip->i_flags & (IMMUTABLE | APPEND)) {
VOP_ABORTOP(dvp, cnp);
error = EPERM;
goto out1;
}
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error) {
VOP_ABORTOP(dvp, cnp);
goto out1;
}
ip->i_nlink++;
DIP_ASSIGN(ip, nlink, ip->i_nlink);
ip->i_flag |= IN_CHANGE;
error = UFS_UPDATE(vp, NULL, NULL, UPDATE_DIROP);
if (!error) {
newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
ufs_makedirentry(ip, cnp, newdir);
error = ufs_direnter(dvp, ulr, vp, newdir, cnp, NULL);
pool_cache_put(ufs_direct_cache, newdir);
}
if (error) {
ip->i_nlink--;
DIP_ASSIGN(ip, nlink, ip->i_nlink);
ip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(vp, NULL, NULL, UPDATE_DIROP);
}
UFS_WAPBL_END(vp->v_mount);
out1:
VOP_UNLOCK(vp);
out2:
VN_KNOTE(vp, NOTE_LINK);
VN_KNOTE(dvp, NOTE_WRITE);
vput(dvp);
fstrans_done(dvp->v_mount);
return (error);
}
int
ufs_rmdir(void *v)
{
struct vop_rmdir_args /* {
struct vnode *a_dvp;
struct vnode *a_vp;
struct componentname *a_cnp;
} */ *ap = v;
struct vnode *vp, *dvp;
struct componentname *cnp;
struct inode *ip, *dp;
int error;
struct ufs_lookup_results *ulr;
vp = ap->a_vp;
dvp = ap->a_dvp;
cnp = ap->a_cnp;
ip = VTOI(vp);
dp = VTOI(dvp);
/* XXX should handle this material another way */
ulr = &dp->i_crap;
UFS_CHECK_CRAPCOUNTER(dp);
/*
* No rmdir "." or of mounted directories please.
*/
if (dp == ip || vp->v_mountedhere != NULL) {
if (dp == ip)
vrele(dvp);
else
vput(dvp);
vput(vp);
return (EINVAL);
}
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
/*
* Do not remove a directory that is in the process of being renamed.
* Verify that the directory is empty (and valid). (Rmdir ".." won't
* be valid since ".." will contain a reference to the current
* directory and thus be non-empty.)
*/
error = 0;
if (ip->i_nlink != 2 ||
!ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
error = ENOTEMPTY;
goto out;
}
if ((dp->i_flags & APPEND) ||
(ip->i_flags & (IMMUTABLE | APPEND))) {
error = EPERM;
goto out;
}
error = UFS_WAPBL_BEGIN(dvp->v_mount);
if (error)
goto out;
/*
* Delete reference to directory before purging
* inode. If we crash in between, the directory
* will be reattached to lost+found,
*/
error = ufs_dirremove(dvp, ulr, ip, cnp->cn_flags, 1);
if (error) {
UFS_WAPBL_END(dvp->v_mount);
goto out;
}
VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
cache_purge(dvp);
/*
* Truncate inode. The only stuff left in the directory is "." and
* "..". The "." reference is inconsequential since we're quashing
* it.
*/
dp->i_nlink--;
DIP_ASSIGN(dp, nlink, dp->i_nlink);
dp->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP);
ip->i_nlink--;
DIP_ASSIGN(ip, nlink, ip->i_nlink);
ip->i_flag |= IN_CHANGE;
error = UFS_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred);
cache_purge(vp);
/*
* Unlock the log while we still have reference to unlinked
* directory vp so that it will not get locked for recycling
*/
UFS_WAPBL_END(dvp->v_mount);
#ifdef UFS_DIRHASH
if (ip->i_dirhash != NULL)
ufsdirhash_free(ip);
#endif
out:
VN_KNOTE(vp, NOTE_DELETE);
vput(vp);
fstrans_done(dvp->v_mount);
vput(dvp);
return (error);
}
/*
* Set attribute vnode op. called from several syscalls
*/
int
ufs_setattr(void *v)
{
struct vop_setattr_args /* {
struct vnode *a_vp;
struct vattr *a_vap;
kauth_cred_t a_cred;
} */ *ap = v;
struct vattr *vap;
struct vnode *vp;
struct inode *ip;
kauth_cred_t cred;
struct lwp *l;
int error;
kauth_action_t action;
bool changing_sysflags;
vap = ap->a_vap;
vp = ap->a_vp;
ip = VTOI(vp);
cred = ap->a_cred;
l = curlwp;
action = KAUTH_VNODE_WRITE_FLAGS;
changing_sysflags = false;
/*
* Check for unsettable attributes.
*/
if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
(vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
(vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
return (EINVAL);
}
fstrans_start(vp->v_mount, FSTRANS_SHARED);
if (vap->va_flags != VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
error = EROFS;
goto out;
}
/* Snapshot flag cannot be set or cleared */
if ((vap->va_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) !=
(ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL))) {
error = EPERM;
goto out;
}
if (ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) {
action |= KAUTH_VNODE_HAS_SYSFLAGS;
}
if ((vap->va_flags & SF_SETTABLE) !=
(ip->i_flags & SF_SETTABLE)) {
action |= KAUTH_VNODE_WRITE_SYSFLAGS;
changing_sysflags = true;
}
error = kauth_authorize_vnode(cred, action, vp, NULL,
genfs_can_chflags(cred, vp->v_type, ip->i_uid,
changing_sysflags));
if (error)
goto out;
if (changing_sysflags) {
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
ip->i_flags = vap->va_flags;
DIP_ASSIGN(ip, flags, ip->i_flags);
} else {
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
ip->i_flags &= SF_SETTABLE;
ip->i_flags |= (vap->va_flags & UF_SETTABLE);
DIP_ASSIGN(ip, flags, ip->i_flags);
}
ip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
UFS_WAPBL_END(vp->v_mount);
if (vap->va_flags & (IMMUTABLE | APPEND)) {
error = 0;
goto out;
}
}
if (ip->i_flags & (IMMUTABLE | APPEND)) {
error = EPERM;
goto out;
}
/*
* Go through the fields and update iff not VNOVAL.
*/
if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
error = EROFS;
goto out;
}
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, l);
UFS_WAPBL_END(vp->v_mount);
if (error)
goto out;
}
if (vap->va_size != VNOVAL) {
/*
* Disallow write attempts on read-only file systems;
* unless the file is a socket, fifo, or a block or
* character device resident on the file system.
*/
switch (vp->v_type) {
case VDIR:
error = EISDIR;
goto out;
case VCHR:
case VBLK:
case VFIFO:
break;
case VREG:
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
error = EROFS;
goto out;
}
if ((ip->i_flags & SF_SNAPSHOT) != 0) {
error = EPERM;
goto out;
}
error = ufs_truncate(vp, vap->va_size, cred);
if (error)
goto out;
break;
default:
error = EOPNOTSUPP;
goto out;
}
}
ip = VTOI(vp);
if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL ||
vap->va_birthtime.tv_sec != VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
error = EROFS;
goto out;
}
if ((ip->i_flags & SF_SNAPSHOT) != 0) {
error = EPERM;
goto out;
}
error = kauth_authorize_vnode(cred, KAUTH_VNODE_WRITE_TIMES, vp,
NULL, genfs_can_chtimes(vp, vap->va_vaflags, ip->i_uid, cred));
if (error)
goto out;
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
if (vap->va_atime.tv_sec != VNOVAL)
if (!(vp->v_mount->mnt_flag & MNT_NOATIME))
ip->i_flag |= IN_ACCESS;
if (vap->va_mtime.tv_sec != VNOVAL) {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (vp->v_mount->mnt_flag & MNT_RELATIME)
ip->i_flag |= IN_ACCESS;
}
if (vap->va_birthtime.tv_sec != VNOVAL &&
ip->i_ump->um_fstype == UFS2) {
ip->i_ffs2_birthtime = vap->va_birthtime.tv_sec;
ip->i_ffs2_birthnsec = vap->va_birthtime.tv_nsec;
}
error = UFS_UPDATE(vp, &vap->va_atime, &vap->va_mtime, 0);
UFS_WAPBL_END(vp->v_mount);
if (error)
goto out;
}
error = 0;
if (vap->va_mode != (mode_t)VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
error = EROFS;
goto out;
}
if ((ip->i_flags & SF_SNAPSHOT) != 0 &&
(vap->va_mode & (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP |
S_IXOTH | S_IWOTH))) {
error = EPERM;
goto out;
}
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
error = ufs_chmod(vp, (int)vap->va_mode, cred, l);
UFS_WAPBL_END(vp->v_mount);
}
VN_KNOTE(vp, NOTE_ATTRIB);
out:
fstrans_done(vp->v_mount);
return (error);
}
/*
* Last reference to an inode. If necessary, write or delete it.
*/
int
ufs_inactive(void *v)
{
struct vop_inactive_args /* {
struct vnode *a_vp;
struct bool *a_recycle;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct mount *transmp;
mode_t mode;
int error = 0;
int logged = 0;
UFS_WAPBL_JUNLOCK_ASSERT(vp->v_mount);
transmp = vp->v_mount;
fstrans_start(transmp, FSTRANS_SHARED);
/*
* Ignore inodes related to stale file handles.
*/
if (ip->i_mode == 0)
goto out;
if (ip->i_ffs_effnlink == 0 && DOINGSOFTDEP(vp))
softdep_releasefile(ip);
if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
logged = 1;
#ifdef QUOTA
(void)chkiq(ip, -1, NOCRED, 0);
#endif
#ifdef UFS_EXTATTR
ufs_extattr_vnode_inactive(vp, curlwp);
#endif
if (ip->i_size != 0) {
/*
* When journaling, only truncate one indirect block
* at a time
*/
if (vp->v_mount->mnt_wapbl) {
uint64_t incr = MNINDIR(ip->i_ump) <<
vp->v_mount->mnt_fs_bshift; /* Power of 2 */
uint64_t base = NDADDR <<
vp->v_mount->mnt_fs_bshift;
while (!error && ip->i_size > base + incr) {
/*
* round down to next full indirect
* block boundary.
*/
uint64_t nsize = base +
((ip->i_size - base - 1) &
~(incr - 1));
error = UFS_TRUNCATE(vp, nsize, 0,
NOCRED);
if (error)
break;
UFS_WAPBL_END(vp->v_mount);
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
}
}
if (!error)
error = UFS_TRUNCATE(vp, (off_t)0, 0, NOCRED);
}
/*
* Setting the mode to zero needs to wait for the inode
* to be written just as does a change to the link count.
* So, rather than creating a new entry point to do the
* same thing, we just use softdep_change_linkcnt().
*/
DIP_ASSIGN(ip, rdev, 0);
mode = ip->i_mode;
ip->i_mode = 0;
DIP_ASSIGN(ip, mode, 0);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
mutex_enter(&vp->v_interlock);
vp->v_iflag |= VI_FREEING;
mutex_exit(&vp->v_interlock);
if (DOINGSOFTDEP(vp))
softdep_change_linkcnt(ip);
UFS_VFREE(vp, ip->i_number, mode);
}
if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) {
if (!logged++) {
int err;
err = UFS_WAPBL_BEGIN(vp->v_mount);
if (err)
goto out;
}
UFS_UPDATE(vp, NULL, NULL, 0);
}
if (logged)
UFS_WAPBL_END(vp->v_mount);
out:
/*
* If we are done with the inode, reclaim it
* so that it can be reused immediately.
*/
*ap->a_recycle = (ip->i_mode == 0);
VOP_UNLOCK(vp, 0);
fstrans_done(transmp);
return (error);
}
int
ffs_wapbl_start(struct mount *mp)
{
struct ufsmount *ump = VFSTOUFS(mp);
struct fs *fs = ump->um_fs;
struct vnode *devvp = ump->um_devvp;
daddr_t off;
size_t count;
size_t blksize;
uint64_t extradata;
int error;
if (mp->mnt_wapbl == NULL) {
if (fs->fs_journal_flags & UFS_WAPBL_FLAGS_CLEAR_LOG) {
/* Clear out any existing journal file */
error = wapbl_remove_log(mp);
if (error != 0)
return error;
}
if (mp->mnt_flag & MNT_LOG) {
KDASSERT(fs->fs_ronly == 0);
/* WAPBL needs UFS2 format super block */
if (ffs_superblock_layout(fs) < 2) {
printf("%s fs superblock in old format, "
"not journaling\n",
VFSTOUFS(mp)->um_fs->fs_fsmnt);
mp->mnt_flag &= ~MNT_LOG;
return EINVAL;
}
error = wapbl_log_position(mp, fs, devvp, &off,
&count, &blksize, &extradata);
if (error)
return error;
error = wapbl_start(&mp->mnt_wapbl, mp, devvp, off,
count, blksize, mp->mnt_wapbl_replay,
ffs_wapbl_sync_metadata,
ffs_wapbl_abort_sync_metadata);
if (error)
return error;
mp->mnt_wapbl_op = &wapbl_ops;
#ifdef WAPBL_DEBUG
printf("%s: enabling logging\n", fs->fs_fsmnt);
#endif
if ((fs->fs_flags & FS_DOWAPBL) == 0) {
UFS_WAPBL_BEGIN(mp);
fs->fs_flags |= FS_DOWAPBL;
error = ffs_sbupdate(ump, MNT_WAIT);
if (error) {
UFS_WAPBL_END(mp);
ffs_wapbl_stop(mp, MNT_FORCE);
return error;
}
UFS_WAPBL_END(mp);
error = wapbl_flush(mp->mnt_wapbl, 1);
if (error) {
ffs_wapbl_stop(mp, MNT_FORCE);
return error;
}
}
} else if (fs->fs_flags & FS_DOWAPBL) {
fs->fs_fmod = 1;
fs->fs_flags &= ~FS_DOWAPBL;
}
}
/*
* It is recommended that you finish replay with logging enabled.
* However, even if logging is not enabled, the remaining log
* replay should be safely recoverable with an fsck, so perform
* it anyway.
*/
if ((fs->fs_ronly == 0) && mp->mnt_wapbl_replay) {
int saveflag = mp->mnt_flag & MNT_RDONLY;
/*
* Make sure MNT_RDONLY is not set so that the inode
* cleanup in ufs_inactive will actually do its work.
*/
mp->mnt_flag &= ~MNT_RDONLY;
ffs_wapbl_replay_finish(mp);
mp->mnt_flag |= saveflag;
KASSERT(fs->fs_ronly == 0);
}
return 0;
}
/*
* Last reference to an inode. If necessary, write or delete it.
*/
int
ufs_inactive(void *v)
{
struct vop_inactive_args *ap = v;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct fs *fs = ip->i_fs;
struct proc *p = curproc;
mode_t mode;
int error = 0, logged = 0, truncate_error = 0;
#ifdef DIAGNOSTIC
extern int prtactive;
if (prtactive && vp->v_usecount != 0)
vprint("ufs_inactive: pushing active", vp);
#endif
UFS_WAPBL_JUNLOCK_ASSERT(vp->v_mount);
/*
* Ignore inodes related to stale file handles.
*/
if (ip->i_din1 == NULL || DIP(ip, mode) == 0)
goto out;
if (DIP(ip, nlink) <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
logged = 1;
if (getinoquota(ip) == 0)
(void)ufs_quota_free_inode(ip, NOCRED);
if (DIP(ip, size) != 0 && vp->v_mount->mnt_wapbl) {
/*
* When journaling, only truncate one indirect block at
* a time.
*/
uint64_t incr = MNINDIR(ip->i_ump) << fs->fs_bshift;
uint64_t base = NDADDR << fs->fs_bshift;
while (!error && DIP(ip, size) > base + incr) {
/*
* round down to next full indirect block
* boundary.
*/
uint64_t nsize = base +
((DIP(ip, size) - base - 1) &
~(incr - 1));
error = UFS_TRUNCATE(ip, nsize, 0, NOCRED);
if (error)
break;
UFS_WAPBL_END(vp->v_mount);
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
}
}
if (error == 0) {
truncate_error = UFS_TRUNCATE(ip, (off_t)0, 0, NOCRED);
/* XXX pedro: remove me */
if (truncate_error)
printf("UFS_TRUNCATE()=%d\n", truncate_error);
}
DIP_ASSIGN(ip, rdev, 0);
mode = DIP(ip, mode);
DIP_ASSIGN(ip, mode, 0);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
/*
* Setting the mode to zero needs to wait for the inode to be
* written just as does a change to the link count. So, rather
* than creating a new entry point to do the same thing, we
* just use softdep_change_linkcnt(). Also, we can't let
* softdep co-opt us to help on its worklist, as we may end up
* trying to recycle vnodes and getting to this same point a
* couple of times, blowing the kernel stack. However, this
* could be optimized by checking if we are coming from
* vrele(), vput() or vclean() (by checking for VXLOCK) and
* just avoiding the co-opt to happen in the last case.
*/
if (DOINGSOFTDEP(vp))
softdep_change_linkcnt(ip, 1);
UFS_INODE_FREE(ip, ip->i_number, mode);
}
if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) {
if (!logged++) {
int err;
err = UFS_WAPBL_BEGIN(vp->v_mount);
if (err) {
error = err;
goto out;
}
}
UFS_UPDATE(ip, 0);
}
if (logged)
UFS_WAPBL_END(vp->v_mount);
out:
VOP_UNLOCK(vp, 0);
/*
* If we are done with the inode, reclaim it
* so that it can be reused immediately.
*/
if (error == 0 && truncate_error == 0 &&
(ip->i_din1 == NULL || DIP(ip, mode) == 0))
vrecycle(vp, p);
return (truncate_error ? truncate_error : error);
}
/* ARGSUSED */
int
ffs_full_fsync(struct vnode *vp, int flags)
{
int error, i, uflags;
struct mount *mp;
KASSERT(vp->v_tag == VT_UFS);
KASSERT(VTOI(vp) != NULL);
KASSERT(vp->v_type != VCHR && vp->v_type != VBLK);
error = 0;
uflags = UPDATE_CLOSE | ((flags & FSYNC_WAIT) ? UPDATE_WAIT : 0);
mp = vp->v_mount;
/*
* Flush all dirty data associated with the vnode.
*/
if (vp->v_type == VREG) {
int pflags = PGO_ALLPAGES | PGO_CLEANIT;
if ((flags & FSYNC_WAIT))
pflags |= PGO_SYNCIO;
if (fstrans_getstate(mp) == FSTRANS_SUSPENDING)
pflags |= PGO_FREE;
mutex_enter(vp->v_interlock);
error = VOP_PUTPAGES(vp, 0, 0, pflags);
if (error)
return error;
}
#ifdef WAPBL
if (mp && mp->mnt_wapbl) {
/*
* Don't bother writing out metadata if the syncer is
* making the request. We will let the sync vnode
* write it out in a single burst through a call to
* VFS_SYNC().
*/
if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY)) != 0)
return 0;
if ((VTOI(vp)->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE
| IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) != 0) {
error = UFS_WAPBL_BEGIN(mp);
if (error)
return error;
error = ffs_update(vp, NULL, NULL, uflags);
UFS_WAPBL_END(mp);
}
if (error || (flags & FSYNC_NOLOG) != 0)
return error;
/*
* Don't flush the log if the vnode being flushed
* contains no dirty buffers that could be in the log.
*/
if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
error = wapbl_flush(mp->mnt_wapbl, 0);
if (error)
return error;
}
if ((flags & FSYNC_WAIT) != 0) {
mutex_enter(vp->v_interlock);
while (vp->v_numoutput != 0)
cv_wait(&vp->v_cv, vp->v_interlock);
mutex_exit(vp->v_interlock);
}
return error;
}
#endif /* WAPBL */
error = vflushbuf(vp, (flags & FSYNC_WAIT) != 0);
if (error == 0)
error = ffs_update(vp, NULL, NULL, uflags);
if (error == 0 && (flags & FSYNC_CACHE) != 0) {
i = 1;
(void)VOP_IOCTL(VTOI(vp)->i_devvp, DIOCCACHESYNC, &i, FWRITE,
kauth_cred_get());
}
return error;
}
int
ffs_fsync(void *v)
{
struct vop_fsync_args /* {
struct vnode *a_vp;
kauth_cred_t a_cred;
int a_flags;
off_t a_offlo;
off_t a_offhi;
struct lwp *a_l;
} */ *ap = v;
struct buf *bp;
int num, error, i;
struct indir ia[NIADDR + 1];
int bsize;
daddr_t blk_high;
struct vnode *vp;
struct mount *mp;
vp = ap->a_vp;
mp = vp->v_mount;
fstrans_start(mp, FSTRANS_LAZY);
if ((ap->a_offlo == 0 && ap->a_offhi == 0) || (vp->v_type != VREG)) {
error = ffs_full_fsync(vp, ap->a_flags);
goto out;
}
bsize = mp->mnt_stat.f_iosize;
blk_high = ap->a_offhi / bsize;
if (ap->a_offhi % bsize != 0)
blk_high++;
/*
* First, flush all pages in range.
*/
mutex_enter(vp->v_interlock);
error = VOP_PUTPAGES(vp, trunc_page(ap->a_offlo),
round_page(ap->a_offhi), PGO_CLEANIT |
((ap->a_flags & FSYNC_WAIT) ? PGO_SYNCIO : 0));
if (error) {
goto out;
}
#ifdef WAPBL
KASSERT(vp->v_type == VREG);
if (mp->mnt_wapbl) {
/*
* Don't bother writing out metadata if the syncer is
* making the request. We will let the sync vnode
* write it out in a single burst through a call to
* VFS_SYNC().
*/
if ((ap->a_flags & (FSYNC_DATAONLY | FSYNC_LAZY)) != 0) {
fstrans_done(mp);
return 0;
}
error = 0;
if (vp->v_tag == VT_UFS && VTOI(vp)->i_flag &
(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY |
IN_MODIFIED | IN_ACCESSED)) {
error = UFS_WAPBL_BEGIN(mp);
if (error) {
fstrans_done(mp);
return error;
}
error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE |
((ap->a_flags & FSYNC_WAIT) ? UPDATE_WAIT : 0));
UFS_WAPBL_END(mp);
}
if (error || (ap->a_flags & FSYNC_NOLOG) != 0) {
fstrans_done(mp);
return error;
}
error = wapbl_flush(mp->mnt_wapbl, 0);
fstrans_done(mp);
return error;
}
#endif /* WAPBL */
/*
* Then, flush indirect blocks.
*/
if (blk_high >= NDADDR) {
error = ufs_getlbns(vp, blk_high, ia, &num);
if (error)
goto out;
mutex_enter(&bufcache_lock);
for (i = 0; i < num; i++) {
if ((bp = incore(vp, ia[i].in_lbn)) == NULL)
continue;
if ((bp->b_cflags & BC_BUSY) != 0 ||
(bp->b_oflags & BO_DELWRI) == 0)
continue;
bp->b_cflags |= BC_BUSY | BC_VFLUSH;
mutex_exit(&bufcache_lock);
bawrite(bp);
mutex_enter(&bufcache_lock);
}
mutex_exit(&bufcache_lock);
}
if (ap->a_flags & FSYNC_WAIT) {
mutex_enter(vp->v_interlock);
while (vp->v_numoutput > 0)
cv_wait(&vp->v_cv, vp->v_interlock);
mutex_exit(vp->v_interlock);
}
error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE |
(((ap->a_flags & (FSYNC_WAIT | FSYNC_DATAONLY)) == FSYNC_WAIT)
? UPDATE_WAIT : 0));
if (error == 0 && ap->a_flags & FSYNC_CACHE) {
int l = 0;
VOP_IOCTL(VTOI(vp)->i_devvp, DIOCCACHESYNC, &l, FWRITE,
curlwp->l_cred);
}
out:
fstrans_done(mp);
return error;
}
