/*
* Perform chown operation on inode ip;
* inode must be locked prior to call.
*/
static int
ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, kauth_cred_t cred,
struct lwp *l)
{
struct inode *ip;
int error = 0;
#if defined(QUOTA) || defined(QUOTA2)
uid_t ouid;
gid_t ogid;
int64_t change;
#endif
ip = VTOI(vp);
error = 0;
if (uid == (uid_t)VNOVAL)
uid = ip->i_uid;
if (gid == (gid_t)VNOVAL)
gid = ip->i_gid;
error = kauth_authorize_vnode(cred, KAUTH_VNODE_CHANGE_OWNERSHIP, vp,
NULL, genfs_can_chown(cred, ip->i_uid, ip->i_gid, uid, gid));
if (error)
return (error);
fstrans_start(vp->v_mount, FSTRANS_SHARED);
#if defined(QUOTA) || defined(QUOTA2)
ogid = ip->i_gid;
ouid = ip->i_uid;
change = DIP(ip, blocks);
(void) chkdq(ip, -change, cred, 0);
(void) chkiq(ip, -1, cred, 0);
#endif
ip->i_gid = gid;
DIP_ASSIGN(ip, gid, gid);
ip->i_uid = uid;
DIP_ASSIGN(ip, uid, uid);
#if defined(QUOTA) || defined(QUOTA2)
if ((error = chkdq(ip, change, cred, 0)) == 0) {
if ((error = chkiq(ip, 1, cred, 0)) == 0)
goto good;
else
(void) chkdq(ip, -change, cred, FORCE);
}
ip->i_gid = ogid;
DIP_ASSIGN(ip, gid, ogid);
ip->i_uid = ouid;
DIP_ASSIGN(ip, uid, ouid);
(void) chkdq(ip, change, cred, FORCE);
(void) chkiq(ip, 1, cred, FORCE);
fstrans_done(vp->v_mount);
return (error);
good:
#endif /* QUOTA || QUOTA2 */
ip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
fstrans_done(vp->v_mount);
return (0);
}
/*
* Try to create a journal log inside the filesystem.
*/
int
wapbl_create_infs_log(struct mount *mp, struct fs *fs, struct vnode *devvp,
daddr_t *startp, size_t *countp, uint64_t *extradatap)
{
struct vnode *vp, *rvp;
struct inode *ip;
int error;
if ((error = VFS_ROOT(mp, &rvp)) != 0)
return error;
error = UFS_INODE_ALLOC(VTOI(rvp), 0 | S_IFREG, NOCRED, &vp);
if (mp->mnt_flag & MNT_UPDATE) {
vput(rvp);
} else {
VOP_UNLOCK(rvp, 0);
vgone(rvp);
}
if (error != 0)
return error;
vp->v_type = VREG;
ip = VTOI(vp);
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
/* ip->i_mode = 0 | IFREG; */
DIP_ASSIGN(ip, mode, 0 | IFREG);
/* ip->i_flags = SF_LOG; */
DIP_ASSIGN(ip, flags, SF_LOG);
ip->i_effnlink = 1;
DIP_ASSIGN(ip, nlink, 1);
ffs_update(ip, MNT_WAIT);
if ((error = wapbl_allocate_log_file(mp, vp,
startp, countp, extradatap)) != 0) {
/*
* If we couldn't allocate the space for the log file,
* remove the inode by setting its link count back to
* zero and bail.
*/
ip->i_effnlink = 0;
DIP_ASSIGN(ip, nlink, 0);
VOP_UNLOCK(vp, 0);
vgone(vp);
return error;
}
/*
* Now that we have the place-holder inode for the journal,
* we don't need the vnode ever again.
*/
VOP_UNLOCK(vp, 0);
vgone(vp);
return 0;
}
/*
* Last reference to an inode. If necessary, write or delete it.
*/
int
ulfs_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;
transmp = vp->v_mount;
fstrans_start(transmp, FSTRANS_LAZY);
/*
* Ignore inodes related to stale file handles.
*/
if (ip->i_mode == 0)
goto out;
if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
#ifdef LFS_EXTATTR
ulfs_extattr_vnode_inactive(vp, curlwp);
#endif
if (ip->i_size != 0) {
error = lfs_truncate(vp, (off_t)0, 0, NOCRED);
}
#if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
(void)lfs_chkiq(ip, -1, NOCRED, 0);
#endif
DIP_ASSIGN(ip, rdev, 0);
mode = ip->i_mode;
ip->i_mode = 0;
ip->i_omode = mode;
DIP_ASSIGN(ip, mode, 0);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
/*
* Defer final inode free and update to ulfs_reclaim().
*/
}
if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) {
lfs_update(vp, NULL, NULL, 0);
}
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);
fstrans_done(transmp);
return (error);
}
/*
* symlink -- make a symbolic link
*/
int
ufs_symlink(void *v)
{
struct vop_symlink_args *ap = v;
struct vnode *vp, **vpp = ap->a_vpp;
struct inode *ip;
int len, error;
error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
vpp, ap->a_cnp);
if (error)
return (error);
VN_KNOTE(ap->a_dvp, NOTE_WRITE);
vp = *vpp;
len = strlen(ap->a_target);
if (len < vp->v_mount->mnt_maxsymlinklen) {
ip = VTOI(vp);
bcopy(ap->a_target, (char *)SHORTLINK(ip), len);
DIP_ASSIGN(ip, size, len);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
} else
error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED, ap->a_cnp->cn_cred, NULL,
curproc);
vput(vp);
return (error);
}
/*
* Mknod vnode call
*/
int
ufs_mknod(void *v)
{
struct vop_mknod_args *ap = v;
struct vattr *vap = ap->a_vap;
struct vnode **vpp = ap->a_vpp;
struct inode *ip;
int error;
if ((error =
ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
ap->a_dvp, vpp, ap->a_cnp)) != 0)
return (error);
VN_KNOTE(ap->a_dvp, NOTE_WRITE);
ip = VTOI(*vpp);
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
if (vap->va_rdev != VNOVAL) {
/*
* Want to be able to use this to make badblock
* inodes, so don't truncate the dev number.
*/
DIP_ASSIGN(ip, rdev, vap->va_rdev);
}
/*
* Remove inode so that it will be reloaded by VFS_VGET and
* checked to see if it is an alias of an existing entry in
* the inode cache.
*/
vput(*vpp);
(*vpp)->v_type = VNON;
vgone(*vpp);
*vpp = NULL;
return (0);
}
/*
* symlink -- make a symbolic link
*/
int
ulfs_symlink(void *v)
{
struct vop_symlink_v3_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
struct vattr *a_vap;
char *a_target;
} */ *ap = v;
struct vnode *vp, **vpp;
struct inode *ip;
int len, error;
struct ulfs_lookup_results *ulr;
vpp = ap->a_vpp;
/* XXX should handle this material another way */
ulr = &VTOI(ap->a_dvp)->i_crap;
ULFS_CHECK_CRAPCOUNTER(VTOI(ap->a_dvp));
fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
error = ulfs_makeinode(LFS_IFLNK | ap->a_vap->va_mode, ap->a_dvp, ulr,
vpp, ap->a_cnp);
if (error)
goto out;
VN_KNOTE(ap->a_dvp, NOTE_WRITE);
vp = *vpp;
len = strlen(ap->a_target);
ip = VTOI(vp);
if (len < ip->i_lfs->um_maxsymlinklen) {
memcpy((char *)SHORTLINK(ip), ap->a_target, len);
ip->i_size = len;
DIP_ASSIGN(ip, size, len);
uvm_vnp_setsize(vp, ip->i_size);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (vp->v_mount->mnt_flag & MNT_RELATIME)
ip->i_flag |= IN_ACCESS;
} else
error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED | IO_JOURNALLOCKED,
ap->a_cnp->cn_cred, NULL, NULL);
VOP_UNLOCK(vp);
if (error)
vrele(vp);
out:
fstrans_done(ap->a_dvp->v_mount);
return (error);
}
/*
* Allocate len bytes at offset off.
*/
int
ufs_gop_alloc(struct vnode *vp, off_t off, off_t len, int flags,
kauth_cred_t cred)
{
struct inode *ip = VTOI(vp);
int error, delta, bshift, bsize;
UVMHIST_FUNC("ufs_gop_alloc"); UVMHIST_CALLED(ubchist);
error = 0;
bshift = vp->v_mount->mnt_fs_bshift;
bsize = 1 << bshift;
delta = off & (bsize - 1);
off -= delta;
len += delta;
while (len > 0) {
bsize = MIN(bsize, len);
error = UFS_BALLOC(vp, off, bsize, cred, flags, NULL);
if (error) {
goto out;
}
/*
* increase file size now, UFS_BALLOC() requires that
* EOF be up-to-date before each call.
*/
if (ip->i_size < off + bsize) {
UVMHIST_LOG(ubchist, "vp %p old 0x%x new 0x%x",
vp, ip->i_size, off + bsize, 0);
ip->i_size = off + bsize;
DIP_ASSIGN(ip, size, ip->i_size);
}
off += bsize;
len -= bsize;
}
out:
UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
return error;
}
/*
* Change the mode on a file.
* Inode must be locked before calling.
*/
static int
ulfs_chmod(struct vnode *vp, int mode, kauth_cred_t cred, struct lwp *l)
{
struct inode *ip;
int error;
ip = VTOI(vp);
error = kauth_authorize_vnode(cred, KAUTH_VNODE_WRITE_SECURITY, vp,
NULL, genfs_can_chmod(vp->v_type, cred, ip->i_uid, ip->i_gid, mode));
if (error)
return (error);
fstrans_start(vp->v_mount, FSTRANS_SHARED);
ip->i_mode &= ~ALLPERMS;
ip->i_mode |= (mode & ALLPERMS);
ip->i_flag |= IN_CHANGE;
DIP_ASSIGN(ip, mode, ip->i_mode);
fstrans_done(vp->v_mount);
return (0);
}
void
ffs_itimes(struct inode *ip, const struct timespec *acc,
const struct timespec *mod, const struct timespec *cre)
{
struct timespec now;
if (!(ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY))) {
return;
}
vfs_timestamp(&now);
if (ip->i_flag & IN_ACCESS) {
if (acc == NULL)
acc = &now;
DIP_ASSIGN(ip, atime, acc->tv_sec);
DIP_ASSIGN(ip, atimensec, acc->tv_nsec);
}
if (ip->i_flag & (IN_UPDATE | IN_MODIFY)) {
if ((ip->i_flags & SF_SNAPSHOT) == 0) {
if (mod == NULL)
mod = &now;
DIP_ASSIGN(ip, mtime, mod->tv_sec);
DIP_ASSIGN(ip, mtimensec, mod->tv_nsec);
}
ip->i_modrev++;
}
if (ip->i_flag & (IN_CHANGE | IN_MODIFY)) {
if (cre == NULL)
cre = &now;
DIP_ASSIGN(ip, ctime, cre->tv_sec);
DIP_ASSIGN(ip, ctimensec, cre->tv_nsec);
}
if (ip->i_flag & (IN_ACCESS | IN_MODIFY))
ip->i_flag |= IN_ACCESSED;
if (ip->i_flag & (IN_UPDATE | IN_CHANGE))
ip->i_flag |= IN_MODIFIED;
ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY);
}
/*
* Truncate the inode oip to at most length size, freeing the
* disk blocks.
*/
int
ffs_truncate(struct inode *oip, off_t length, int flags, struct ucred *cred)
{
struct vnode *ovp;
daddr64_t lastblock;
daddr64_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
daddr64_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
struct fs *fs;
struct buf *bp;
int offset, size, level;
long count, nblocks, vflags, blocksreleased = 0;
int i, aflags, error, allerror, indirect = 0;
off_t osize;
extern int num_indirdep;
extern int max_indirdep;
if (length < 0)
return (EINVAL);
ovp = ITOV(oip);
if (ovp->v_type != VREG &&
ovp->v_type != VDIR &&
ovp->v_type != VLNK)
return (0);
if (DIP(oip, size) == length)
return (0);
if (ovp->v_type == VLNK &&
(DIP(oip, size) < ovp->v_mount->mnt_maxsymlinklen ||
(ovp->v_mount->mnt_maxsymlinklen == 0 &&
oip->i_din1->di_blocks == 0))) {
#ifdef DIAGNOSTIC
if (length != 0)
panic("ffs_truncate: partial truncate of symlink");
#endif
memset(SHORTLINK(oip), 0, (size_t) DIP(oip, size));
DIP_ASSIGN(oip, size, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (UFS_UPDATE(oip, MNT_WAIT));
}
if ((error = getinoquota(oip)) != 0)
return (error);
uvm_vnp_setsize(ovp, length);
oip->i_ci.ci_lasta = oip->i_ci.ci_clen
= oip->i_ci.ci_cstart = oip->i_ci.ci_lastw = 0;
if (DOINGSOFTDEP(ovp)) {
if (length > 0 || softdep_slowdown(ovp)) {
/*
* If a file is only partially truncated, then
* we have to clean up the data structures
* describing the allocation past the truncation
* point. Finding and deallocating those structures
* is a lot of work. Since partial truncation occurs
* rarely, we solve the problem by syncing the file
* so that it will have no data structures left.
*/
if ((error = VOP_FSYNC(ovp, cred, MNT_WAIT)) != 0)
return (error);
} else {
(void)ufs_quota_free_blocks(oip, DIP(oip, blocks),
NOCRED);
softdep_setup_freeblocks(oip, length);
(void) vinvalbuf(ovp, 0, cred, curproc, 0, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (UFS_UPDATE(oip, 0));
}
}
fs = oip->i_fs;
osize = DIP(oip, size);
/*
* Lengthen the size of the file. We must ensure that the
* last byte of the file is allocated. Since the smallest
* value of osize is 0, length will be at least 1.
*/
if (osize < length) {
if (length > fs->fs_maxfilesize)
return (EFBIG);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
error = UFS_BUF_ALLOC(oip, length - 1, 1,
cred, aflags, &bp);
if (error)
return (error);
if (bp->b_lblkno >= NDADDR)
indirect = 1;
DIP_ASSIGN(oip, size, length);
uvm_vnp_setsize(ovp, length);
(void) uvm_vnp_uncache(ovp);
if (aflags & B_SYNC)
bwrite(bp);
else
bawrite(bp);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
error = UFS_UPDATE(oip, MNT_WAIT);
if (DOINGSOFTDEP(ovp) && num_indirdep > max_indirdep)
if (indirect) {
/*
* If the number of pending indirect block
* dependencies is sufficiently close to the
* maximum number of simultaneously mappable
* buffers force a sync on the vnode to prevent
* buffer cache exhaustion.
*/
VOP_FSYNC(ovp, curproc->p_ucred, MNT_WAIT);
}
return (error);
}
uvm_vnp_setsize(ovp, length);
/*
* Shorten the size of the file. If the file is not being
* truncated to a block boundary, the contents of the
* partial block following the end of the file must be
* zero'ed in case it ever becomes accessible again because
* of subsequent file growth. Directories however are not
* zero'ed as they should grow back initialized to empty.
*/
offset = blkoff(fs, length);
if (offset == 0) {
DIP_ASSIGN(oip, size, length);
} else {
lbn = lblkno(fs, length);
aflags = B_CLRBUF;
if (flags & IO_SYNC)
aflags |= B_SYNC;
error = UFS_BUF_ALLOC(oip, length - 1, 1,
cred, aflags, &bp);
if (error)
return (error);
/*
* When we are doing soft updates and the UFS_BALLOC
* above fills in a direct block hole with a full sized
* block that will be truncated down to a fragment below,
* we must flush out the block dependency with an FSYNC
* so that we do not get a soft updates inconsistency
* when we create the fragment below.
*/
if (DOINGSOFTDEP(ovp) && lbn < NDADDR &&
fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize &&
(error = VOP_FSYNC(ovp, cred, MNT_WAIT)) != 0)
return (error);
DIP_ASSIGN(oip, size, length);
size = blksize(fs, oip, lbn);
(void) uvm_vnp_uncache(ovp);
if (ovp->v_type != VDIR)
bzero((char *)bp->b_data + offset,
(u_int)(size - offset));
bp->b_bcount = size;
if (aflags & B_SYNC)
bwrite(bp);
else
bawrite(bp);
}
/*
* Calculate index into inode's block list of
* last direct and indirect blocks (if any)
* which we want to keep. Lastblock is -1 when
* the file is truncated to 0.
*/
lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
lastiblock[SINGLE] = lastblock - NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
nblocks = btodb(fs->fs_bsize);
/*
* Update file and block pointers on disk before we start freeing
* blocks. If we crash before free'ing blocks below, the blocks
* will be returned to the free list. lastiblock values are also
* normalized to -1 for calls to ffs_indirtrunc below.
*/
for (level = TRIPLE; level >= SINGLE; level--) {
oldblks[NDADDR + level] = DIP(oip, ib[level]);
if (lastiblock[level] < 0) {
DIP_ASSIGN(oip, ib[level], 0);
lastiblock[level] = -1;
}
}
for (i = 0; i < NDADDR; i++) {
oldblks[i] = DIP(oip, db[i]);
if (i > lastblock)
DIP_ASSIGN(oip, db[i], 0);
}
oip->i_flag |= IN_CHANGE | IN_UPDATE;
if ((error = UFS_UPDATE(oip, MNT_WAIT)) != 0)
allerror = error;
/*
* Having written the new inode to disk, save its new configuration
* and put back the old block pointers long enough to process them.
* Note that we save the new block configuration so we can check it
* when we are done.
*/
for (i = 0; i < NDADDR; i++) {
newblks[i] = DIP(oip, db[i]);
DIP_ASSIGN(oip, db[i], oldblks[i]);
}
for (i = 0; i < NIADDR; i++) {
newblks[NDADDR + i] = DIP(oip, ib[i]);
DIP_ASSIGN(oip, ib[i], oldblks[NDADDR + i]);
}
DIP_ASSIGN(oip, size, osize);
vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
allerror = vinvalbuf(ovp, vflags, cred, curproc, 0, 0);
/*
* Indirect blocks first.
*/
indir_lbn[SINGLE] = -NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
for (level = TRIPLE; level >= SINGLE; level--) {
bn = DIP(oip, ib[level]);
if (bn != 0) {
error = ffs_indirtrunc(oip, indir_lbn[level],
fsbtodb(fs, bn), lastiblock[level], level, &count);
if (error)
allerror = error;
blocksreleased += count;
if (lastiblock[level] < 0) {
DIP_ASSIGN(oip, ib[level], 0);
ffs_blkfree(oip, bn, fs->fs_bsize);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = NDADDR - 1; i > lastblock; i--) {
long bsize;
bn = DIP(oip, db[i]);
if (bn == 0)
continue;
DIP_ASSIGN(oip, db[i], 0);
bsize = blksize(fs, oip, i);
ffs_blkfree(oip, bn, bsize);
blocksreleased += btodb(bsize);
}
if (lastblock < 0)
goto done;
/*
* Finally, look for a change in size of the
* last direct block; release any frags.
*/
bn = DIP(oip, db[lastblock]);
if (bn != 0) {
long oldspace, newspace;
/*
* Calculate amount of space we're giving
* back as old block size minus new block size.
*/
oldspace = blksize(fs, oip, lastblock);
DIP_ASSIGN(oip, size, length);
newspace = blksize(fs, oip, lastblock);
if (newspace == 0)
panic("ffs_truncate: newspace");
if (oldspace - newspace > 0) {
/*
* Block number of space to be free'd is
* the old block # plus the number of frags
* required for the storage we're keeping.
*/
bn += numfrags(fs, newspace);
ffs_blkfree(oip, bn, oldspace - newspace);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#ifdef DIAGNOSTIC
for (level = SINGLE; level <= TRIPLE; level++)
if (newblks[NDADDR + level] != DIP(oip, ib[level]))
panic("ffs_truncate1");
for (i = 0; i < NDADDR; i++)
if (newblks[i] != DIP(oip, db[i]))
panic("ffs_truncate2");
#endif /* DIAGNOSTIC */
/*
* Put back the real size.
*/
DIP_ASSIGN(oip, size, length);
DIP_ADD(oip, blocks, -blocksreleased);
oip->i_flag |= IN_CHANGE;
(void)ufs_quota_free_blocks(oip, blocksreleased, NOCRED);
return (allerror);
}
/*
* Allocate a new inode.
*/
int
ufs_makeinode(int mode, struct vnode *dvp, const struct ufs_lookup_results *ulr,
struct vnode **vpp, struct componentname *cnp)
{
struct inode *ip, *pdir;
struct direct *newdir;
struct vnode *tvp;
int error;
UFS_WAPBL_JUNLOCK_ASSERT(dvp->v_mount);
pdir = VTOI(dvp);
if ((mode & IFMT) == 0)
mode |= IFREG;
if ((error = UFS_VALLOC(dvp, mode, cnp->cn_cred, vpp)) != 0) {
return (error);
}
tvp = *vpp;
ip = VTOI(tvp);
ip->i_gid = pdir->i_gid;
DIP_ASSIGN(ip, gid, ip->i_gid);
ip->i_uid = kauth_cred_geteuid(cnp->cn_cred);
DIP_ASSIGN(ip, uid, ip->i_uid);
error = UFS_WAPBL_BEGIN1(dvp->v_mount, dvp);
if (error) {
/*
* Note, we can't VOP_VFREE(tvp) here like we should
* because we can't write to the disk. Instead, we leave
* the vnode dangling from the journal.
*/
vput(tvp);
return (error);
}
#if defined(QUOTA) || defined(QUOTA2)
if ((error = chkiq(ip, 1, cnp->cn_cred, 0))) {
UFS_VFREE(tvp, ip->i_number, mode);
UFS_WAPBL_END1(dvp->v_mount, dvp);
vput(tvp);
return (error);
}
#endif
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
ip->i_mode = mode;
DIP_ASSIGN(ip, mode, mode);
tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
ip->i_nlink = 1;
DIP_ASSIGN(ip, nlink, 1);
/* Authorize setting SGID if needed. */
if (ip->i_mode & ISGID) {
error = kauth_authorize_vnode(cnp->cn_cred, KAUTH_VNODE_WRITE_SECURITY,
tvp, NULL, genfs_can_chmod(tvp->v_type, cnp->cn_cred, ip->i_uid,
ip->i_gid, mode));
if (error) {
ip->i_mode &= ~ISGID;
DIP_ASSIGN(ip, mode, ip->i_mode);
}
}
if (cnp->cn_flags & ISWHITEOUT) {
ip->i_flags |= UF_OPAQUE;
DIP_ASSIGN(ip, flags, ip->i_flags);
}
/*
* Make sure inode goes to disk before directory entry.
*/
if ((error = UFS_UPDATE(tvp, NULL, NULL, UPDATE_DIROP)) != 0)
goto bad;
newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
ufs_makedirentry(ip, cnp, newdir);
error = ufs_direnter(dvp, ulr, tvp, newdir, cnp, NULL);
pool_cache_put(ufs_direct_cache, newdir);
if (error)
goto bad;
*vpp = tvp;
return (0);
bad:
/*
* Write error occurred trying to update the inode
* or the directory so must deallocate the inode.
*/
ip->i_nlink = 0;
DIP_ASSIGN(ip, nlink, 0);
ip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(tvp, NULL, NULL, 0);
tvp->v_type = VNON; /* explodes later if VBLK */
UFS_WAPBL_END1(dvp->v_mount, dvp);
vput(tvp);
return (error);
}
int
wapbl_remove_log(struct mount *mp)
{
struct ufsmount *ump = VFSTOUFS(mp);
struct fs *fs = ump->um_fs;
struct vnode *vp;
struct inode *ip;
ino_t log_ino;
int error;
/* If super block layout is too old to support WAPBL, return */
if (ffs_superblock_layout(fs) < 2)
return 0;
/* If all the log locators are 0, just clean up */
if (fs->fs_journallocs[0] == 0 &&
fs->fs_journallocs[1] == 0 &&
fs->fs_journallocs[2] == 0 &&
fs->fs_journallocs[3] == 0) {
DPRINTF("empty locators, just clear\n");
goto done;
}
switch (fs->fs_journal_location) {
case UFS_WAPBL_JOURNALLOC_NONE:
/* nothing! */
DPRINTF("no log\n");
break;
case UFS_WAPBL_JOURNALLOC_IN_FILESYSTEM:
log_ino = fs->fs_journallocs[UFS_WAPBL_INFS_INO];
DPRINTF("in-fs log, ino = %lld\n",log_ino);
/* if no existing log inode, just clear all fields and bail */
if (log_ino == 0)
goto done;
error = VFS_VGET(mp, log_ino, &vp);
if (error != 0) {
printf("ffs_wapbl: vget failed %d\n",
error);
/* clear out log info on error */
goto done;
}
ip = VTOI(vp);
KASSERT(log_ino == ip->i_number);
if ((DIP(ip, flags) & SF_LOG) == 0) {
printf("ffs_wapbl: try to clear non-log inode "
"%lld\n", log_ino);
vput(vp);
/* clear out log info on error */
goto done;
}
/*
* remove the log inode by setting its link count back
* to zero and bail.
*/
ip->i_effnlink = 0;
DIP_ASSIGN(ip, nlink, 0);
vput(vp);
case UFS_WAPBL_JOURNALLOC_END_PARTITION:
DPRINTF("end-of-partition log\n");
/* no extra work required */
break;
default:
printf("ffs_wapbl: unknown journal type %d\n",
fs->fs_journal_location);
break;
}
done:
/* Clear out all previous knowledge of journal */
fs->fs_journal_version = 0;
fs->fs_journal_location = 0;
fs->fs_journal_flags = 0;
fs->fs_journallocs[0] = 0;
fs->fs_journallocs[1] = 0;
fs->fs_journallocs[2] = 0;
fs->fs_journallocs[3] = 0;
(void) ffs_sbupdate(ump, MNT_WAIT);
return 0;
}
/*
* Truncate the inode oip to at most length size, freeing the
* disk blocks.
*/
int
ffs_truncate(struct vnode *ovp, off_t length, int ioflag, kauth_cred_t cred)
{
daddr_t lastblock;
struct inode *oip = VTOI(ovp);
daddr_t bn, lastiblock[UFS_NIADDR], indir_lbn[UFS_NIADDR];
daddr_t blks[UFS_NDADDR + UFS_NIADDR];
struct fs *fs;
int offset, pgoffset, level;
int64_t count, blocksreleased = 0;
int i, aflag, nblocks;
int error, allerror = 0;
off_t osize;
int sync;
struct ufsmount *ump = oip->i_ump;
if (ovp->v_type == VCHR || ovp->v_type == VBLK ||
ovp->v_type == VFIFO || ovp->v_type == VSOCK) {
KASSERT(oip->i_size == 0);
return 0;
}
if (length < 0)
return (EINVAL);
if (ovp->v_type == VLNK &&
(oip->i_size < ump->um_maxsymlinklen ||
(ump->um_maxsymlinklen == 0 && DIP(oip, blocks) == 0))) {
KDASSERT(length == 0);
memset(SHORTLINK(oip), 0, (size_t)oip->i_size);
oip->i_size = 0;
DIP_ASSIGN(oip, size, 0);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ffs_update(ovp, NULL, NULL, 0));
}
if (oip->i_size == length) {
/* still do a uvm_vnp_setsize() as writesize may be larger */
uvm_vnp_setsize(ovp, length);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ffs_update(ovp, NULL, NULL, 0));
}
fs = oip->i_fs;
if (length > ump->um_maxfilesize)
return (EFBIG);
if ((oip->i_flags & SF_SNAPSHOT) != 0)
ffs_snapremove(ovp);
osize = oip->i_size;
aflag = ioflag & IO_SYNC ? B_SYNC : 0;
/*
* Lengthen the size of the file. We must ensure that the
* last byte of the file is allocated. Since the smallest
* value of osize is 0, length will be at least 1.
*/
if (osize < length) {
if (ffs_lblkno(fs, osize) < UFS_NDADDR &&
ffs_lblkno(fs, osize) != ffs_lblkno(fs, length) &&
ffs_blkroundup(fs, osize) != osize) {
off_t eob;
eob = ffs_blkroundup(fs, osize);
uvm_vnp_setwritesize(ovp, eob);
error = ufs_balloc_range(ovp, osize, eob - osize,
cred, aflag);
if (error) {
(void) ffs_truncate(ovp, osize,
ioflag & IO_SYNC, cred);
return error;
}
if (ioflag & IO_SYNC) {
mutex_enter(ovp->v_interlock);
VOP_PUTPAGES(ovp,
trunc_page(osize & fs->fs_bmask),
round_page(eob), PGO_CLEANIT | PGO_SYNCIO |
PGO_JOURNALLOCKED);
}
}
uvm_vnp_setwritesize(ovp, length);
error = ufs_balloc_range(ovp, length - 1, 1, cred, aflag);
if (error) {
(void) ffs_truncate(ovp, osize, ioflag & IO_SYNC, cred);
return (error);
}
uvm_vnp_setsize(ovp, length);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
KASSERT(ovp->v_size == oip->i_size);
return (ffs_update(ovp, NULL, NULL, 0));
}
/*
* When truncating a regular file down to a non-block-aligned size,
* we must zero the part of last block which is past the new EOF.
* We must synchronously flush the zeroed pages to disk
* since the new pages will be invalidated as soon as we
* inform the VM system of the new, smaller size.
* We must do this before acquiring the GLOCK, since fetching
* the pages will acquire the GLOCK internally.
* So there is a window where another thread could see a whole
* zeroed page past EOF, but that's life.
*/
offset = ffs_blkoff(fs, length);
pgoffset = length & PAGE_MASK;
if (ovp->v_type == VREG && (pgoffset != 0 || offset != 0) &&
osize > length) {
daddr_t lbn;
voff_t eoz;
int size;
if (offset != 0) {
error = ufs_balloc_range(ovp, length - 1, 1, cred,
aflag);
if (error)
return error;
}
lbn = ffs_lblkno(fs, length);
size = ffs_blksize(fs, oip, lbn);
eoz = MIN(MAX(ffs_lblktosize(fs, lbn) + size, round_page(pgoffset)),
osize);
ubc_zerorange(&ovp->v_uobj, length, eoz - length,
UBC_UNMAP_FLAG(ovp));
if (round_page(eoz) > round_page(length)) {
mutex_enter(ovp->v_interlock);
error = VOP_PUTPAGES(ovp, round_page(length),
round_page(eoz),
PGO_CLEANIT | PGO_DEACTIVATE | PGO_JOURNALLOCKED |
((ioflag & IO_SYNC) ? PGO_SYNCIO : 0));
if (error)
return error;
}
}
genfs_node_wrlock(ovp);
oip->i_size = length;
DIP_ASSIGN(oip, size, length);
uvm_vnp_setsize(ovp, length);
/*
* Calculate index into inode's block list of
* last direct and indirect blocks (if any)
* which we want to keep. Lastblock is -1 when
* the file is truncated to 0.
*/
lastblock = ffs_lblkno(fs, length + fs->fs_bsize - 1) - 1;
lastiblock[SINGLE] = lastblock - UFS_NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - FFS_NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs);
nblocks = btodb(fs->fs_bsize);
/*
* Update file and block pointers on disk before we start freeing
* blocks. If we crash before free'ing blocks below, the blocks
* will be returned to the free list. lastiblock values are also
* normalized to -1 for calls to ffs_indirtrunc below.
*/
sync = 0;
for (level = TRIPLE; level >= SINGLE; level--) {
blks[UFS_NDADDR + level] = DIP(oip, ib[level]);
if (lastiblock[level] < 0 && blks[UFS_NDADDR + level] != 0) {
sync = 1;
DIP_ASSIGN(oip, ib[level], 0);
lastiblock[level] = -1;
}
}
for (i = 0; i < UFS_NDADDR; i++) {
blks[i] = DIP(oip, db[i]);
if (i > lastblock && blks[i] != 0) {
sync = 1;
DIP_ASSIGN(oip, db[i], 0);
}
}
oip->i_flag |= IN_CHANGE | IN_UPDATE;
if (sync) {
error = ffs_update(ovp, NULL, NULL, UPDATE_WAIT);
if (error && !allerror)
allerror = error;
}
/*
* Having written the new inode to disk, save its new configuration
* and put back the old block pointers long enough to process them.
* Note that we save the new block configuration so we can check it
* when we are done.
*/
for (i = 0; i < UFS_NDADDR; i++) {
bn = DIP(oip, db[i]);
DIP_ASSIGN(oip, db[i], blks[i]);
blks[i] = bn;
}
for (i = 0; i < UFS_NIADDR; i++) {
bn = DIP(oip, ib[i]);
DIP_ASSIGN(oip, ib[i], blks[UFS_NDADDR + i]);
blks[UFS_NDADDR + i] = bn;
}
oip->i_size = osize;
DIP_ASSIGN(oip, size, osize);
error = vtruncbuf(ovp, lastblock + 1, 0, 0);
if (error && !allerror)
allerror = error;
/*
* Indirect blocks first.
*/
indir_lbn[SINGLE] = -UFS_NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - FFS_NINDIR(fs) - 1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - FFS_NINDIR(fs) * FFS_NINDIR(fs) - 1;
for (level = TRIPLE; level >= SINGLE; level--) {
if (oip->i_ump->um_fstype == UFS1)
bn = ufs_rw32(oip->i_ffs1_ib[level],UFS_FSNEEDSWAP(fs));
else
bn = ufs_rw64(oip->i_ffs2_ib[level],UFS_FSNEEDSWAP(fs));
if (bn != 0) {
error = ffs_indirtrunc(oip, indir_lbn[level],
FFS_FSBTODB(fs, bn), lastiblock[level], level, &count);
if (error)
allerror = error;
blocksreleased += count;
if (lastiblock[level] < 0) {
DIP_ASSIGN(oip, ib[level], 0);
if (oip->i_ump->um_mountp->mnt_wapbl) {
UFS_WAPBL_REGISTER_DEALLOCATION(
oip->i_ump->um_mountp,
FFS_FSBTODB(fs, bn), fs->fs_bsize);
} else
ffs_blkfree(fs, oip->i_devvp, bn,
fs->fs_bsize, oip->i_number);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = UFS_NDADDR - 1; i > lastblock; i--) {
long bsize;
if (oip->i_ump->um_fstype == UFS1)
bn = ufs_rw32(oip->i_ffs1_db[i], UFS_FSNEEDSWAP(fs));
else
bn = ufs_rw64(oip->i_ffs2_db[i], UFS_FSNEEDSWAP(fs));
if (bn == 0)
continue;
DIP_ASSIGN(oip, db[i], 0);
bsize = ffs_blksize(fs, oip, i);
if ((oip->i_ump->um_mountp->mnt_wapbl) &&
(ovp->v_type != VREG)) {
UFS_WAPBL_REGISTER_DEALLOCATION(oip->i_ump->um_mountp,
FFS_FSBTODB(fs, bn), bsize);
} else
ffs_blkfree(fs, oip->i_devvp, bn, bsize, oip->i_number);
blocksreleased += btodb(bsize);
}
if (lastblock < 0)
goto done;
/*
* Finally, look for a change in size of the
* last direct block; release any frags.
*/
if (oip->i_ump->um_fstype == UFS1)
bn = ufs_rw32(oip->i_ffs1_db[lastblock], UFS_FSNEEDSWAP(fs));
else
bn = ufs_rw64(oip->i_ffs2_db[lastblock], UFS_FSNEEDSWAP(fs));
if (bn != 0) {
long oldspace, newspace;
/*
* Calculate amount of space we're giving
* back as old block size minus new block size.
*/
oldspace = ffs_blksize(fs, oip, lastblock);
oip->i_size = length;
DIP_ASSIGN(oip, size, length);
newspace = ffs_blksize(fs, oip, lastblock);
if (newspace == 0)
panic("itrunc: newspace");
if (oldspace - newspace > 0) {
/*
* Block number of space to be free'd is
* the old block # plus the number of frags
* required for the storage we're keeping.
*/
bn += ffs_numfrags(fs, newspace);
if ((oip->i_ump->um_mountp->mnt_wapbl) &&
(ovp->v_type != VREG)) {
UFS_WAPBL_REGISTER_DEALLOCATION(
oip->i_ump->um_mountp, FFS_FSBTODB(fs, bn),
oldspace - newspace);
} else
ffs_blkfree(fs, oip->i_devvp, bn,
oldspace - newspace, oip->i_number);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#ifdef DIAGNOSTIC
for (level = SINGLE; level <= TRIPLE; level++)
if (blks[UFS_NDADDR + level] != DIP(oip, ib[level]))
panic("itrunc1");
for (i = 0; i < UFS_NDADDR; i++)
if (blks[i] != DIP(oip, db[i]))
panic("itrunc2");
if (length == 0 &&
(!LIST_EMPTY(&ovp->v_cleanblkhd) || !LIST_EMPTY(&ovp->v_dirtyblkhd)))
panic("itrunc3");
#endif /* DIAGNOSTIC */
/*
* Put back the real size.
*/
oip->i_size = length;
DIP_ASSIGN(oip, size, length);
DIP_ADD(oip, blocks, -blocksreleased);
genfs_node_unlock(ovp);
oip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(ovp, NULL, NULL, 0);
#if defined(QUOTA) || defined(QUOTA2)
(void) chkdq(oip, -blocksreleased, NOCRED, 0);
#endif
KASSERT(ovp->v_type != VREG || ovp->v_size == oip->i_size);
return (allerror);
}
/*
* symlink -- make a symbolic link
*/
int
ufs_symlink(void *v)
{
struct vop_symlink_v3_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
struct vattr *a_vap;
char *a_target;
} */ *ap = v;
struct vnode *vp, **vpp;
struct inode *ip;
int len, error;
struct ufs_lookup_results *ulr;
vpp = ap->a_vpp;
/* XXX should handle this material another way */
ulr = &VTOI(ap->a_dvp)->i_crap;
UFS_CHECK_CRAPCOUNTER(VTOI(ap->a_dvp));
/*
* UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful
* ufs_makeinode
*/
fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp, ulr,
vpp, ap->a_cnp);
if (error)
goto out;
VN_KNOTE(ap->a_dvp, NOTE_WRITE);
vp = *vpp;
len = strlen(ap->a_target);
ip = VTOI(vp);
/*
* This test is off by one. um_maxsymlinklen contains the
* number of bytes available, and we aren't storing a \0, so
* the test should properly be <=. However, it cannot be
* changed as this would break compatibility with existing fs
* images -- see the way ufs_readlink() works.
*/
if (len < ip->i_ump->um_maxsymlinklen) {
memcpy((char *)SHORTLINK(ip), ap->a_target, len);
ip->i_size = len;
DIP_ASSIGN(ip, size, len);
uvm_vnp_setsize(vp, ip->i_size);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (vp->v_mount->mnt_flag & MNT_RELATIME)
ip->i_flag |= IN_ACCESS;
UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
} else
error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED | IO_JOURNALLOCKED,
ap->a_cnp->cn_cred, NULL, NULL);
UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp);
VOP_UNLOCK(vp);
if (error)
vrele(vp);
out:
fstrans_done(ap->a_dvp->v_mount);
return (error);
}
/*
* Perform chown operation on inode ip;
* inode must be locked prior to call.
*/
int
ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
struct proc *p)
{
struct inode *ip = VTOI(vp);
uid_t ouid;
gid_t ogid;
int error = 0;
daddr64_t change;
enum ufs_quota_flags quota_flags = 0;
if (uid == (uid_t)VNOVAL)
uid = DIP(ip, uid);
if (gid == (gid_t)VNOVAL)
gid = DIP(ip, gid);
/*
* If we don't own the file, are trying to change the owner
* of the file, or are not a member of the target group,
* the caller must be superuser or the call fails.
*/
if ((cred->cr_uid != DIP(ip, uid) || uid != DIP(ip, uid) ||
(gid != DIP(ip, gid) && !groupmember((gid_t)gid, cred))) &&
(error = suser_ucred(cred)))
return (error);
ogid = DIP(ip, gid);
ouid = DIP(ip, uid);
change = DIP(ip, blocks);
if (ouid == uid)
quota_flags |= UFS_QUOTA_NOUID;
if (ogid == gid)
quota_flags |= UFS_QUOTA_NOGID;
if ((error = getinoquota(ip)) != 0)
return (error);
(void) ufs_quota_free_blocks2(ip, change, cred, quota_flags);
(void) ufs_quota_free_inode2(ip, cred, quota_flags);
(void) ufs_quota_delete(ip);
DIP_ASSIGN(ip, gid, gid);
DIP_ASSIGN(ip, uid, uid);
if ((error = getinoquota(ip)) != 0)
goto error;
if ((error = ufs_quota_alloc_blocks2(ip, change, cred,
quota_flags)) != 0)
goto error;
if ((error = ufs_quota_alloc_inode2(ip, cred ,
quota_flags)) != 0) {
(void)ufs_quota_free_blocks2(ip, change, cred,
quota_flags);
goto error;
}
if (getinoquota(ip))
panic("chown: lost quota");
if (ouid != uid || ogid != gid)
ip->i_flag |= IN_CHANGE;
if (ouid != uid && cred->cr_uid != 0)
DIP_AND(ip, mode, ~ISUID);
if (ogid != gid && cred->cr_uid != 0)
DIP_AND(ip, mode, ~ISGID);
return (0);
error:
(void) ufs_quota_delete(ip);
DIP_ASSIGN(ip, gid, ogid);
DIP_ASSIGN(ip, uid, ouid);
if (getinoquota(ip) == 0) {
(void) ufs_quota_alloc_blocks2(ip, change, cred,
quota_flags | UFS_QUOTA_FORCE);
(void) ufs_quota_alloc_inode2(ip, cred,
quota_flags | UFS_QUOTA_FORCE);
(void) getinoquota(ip);
}
return (error);
}
/*
* Vnode op for writing.
*/
int
ffs_write(void *v)
{
struct vop_write_args *ap = v;
struct vnode *vp;
struct uio *uio;
struct inode *ip;
struct fs *fs;
struct buf *bp;
struct proc *p;
daddr64_t lbn;
off_t osize;
int blkoffset, error, extended, flags, ioflag, resid, size, xfersize;
extended = 0;
ioflag = ap->a_ioflag;
uio = ap->a_uio;
vp = ap->a_vp;
ip = VTOI(vp);
#ifdef DIAGNOSTIC
if (uio->uio_rw != UIO_WRITE)
panic("ffs_write: mode");
#endif
/*
* If writing 0 bytes, succeed and do not change
* update time or file offset (standards compliance)
*/
if (uio->uio_resid == 0)
return (0);
switch (vp->v_type) {
case VREG:
if (ioflag & IO_APPEND)
uio->uio_offset = DIP(ip, size);
if ((DIP(ip, flags) & APPEND) && uio->uio_offset != DIP(ip, size))
return (EPERM);
/* FALLTHROUGH */
case VLNK:
break;
case VDIR:
if ((ioflag & IO_SYNC) == 0)
panic("ffs_write: nonsync dir write");
break;
default:
panic("ffs_write: type");
}
fs = ip->i_fs;
if (uio->uio_offset < 0 ||
(u_int64_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize)
return (EFBIG);
/*
* Maybe this should be above the vnode op call, but so long as
* file servers have no limits, I don't think it matters.
*/
p = uio->uio_procp;
if (vp->v_type == VREG && p && !(ioflag & IO_NOLIMIT) &&
uio->uio_offset + uio->uio_resid >
p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
psignal(p, SIGXFSZ);
return (EFBIG);
}
resid = uio->uio_resid;
osize = DIP(ip, size);
flags = ioflag & IO_SYNC ? B_SYNC : 0;
for (error = 0; uio->uio_resid > 0;) {
lbn = lblkno(fs, uio->uio_offset);
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->fs_bsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (fs->fs_bsize > xfersize)
flags |= B_CLRBUF;
else
flags &= ~B_CLRBUF;
if ((error = UFS_BUF_ALLOC(ip, uio->uio_offset, xfersize,
ap->a_cred, flags, &bp)) != 0)
break;
if (uio->uio_offset + xfersize > DIP(ip, size)) {
DIP_ASSIGN(ip, size, uio->uio_offset + xfersize);
uvm_vnp_setsize(vp, DIP(ip, size));
extended = 1;
}
(void)uvm_vnp_uncache(vp);
size = blksize(fs, ip, lbn) - bp->b_resid;
if (size < xfersize)
xfersize = size;
error =
uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
if (error != 0)
bzero((char *)bp->b_data + blkoffset, xfersize);
if (ioflag & IO_SYNC)
(void)bwrite(bp);
else if (xfersize + blkoffset == fs->fs_bsize) {
if (doclusterwrite)
cluster_write(bp, &ip->i_ci, DIP(ip, size));
else
bawrite(bp);
} else
bdwrite(bp);
if (error || xfersize == 0)
break;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* If we successfully wrote any data, and we are not the superuser
* we clear the setuid and setgid bits as a precaution against
* tampering.
*/
if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
DIP_ASSIGN(ip, mode, DIP(ip, mode) & ~(ISUID | ISGID));
if (resid > uio->uio_resid)
VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
if (error) {
if (ioflag & IO_UNIT) {
(void)UFS_TRUNCATE(ip, osize,
ioflag & IO_SYNC, ap->a_cred);
uio->uio_offset -= resid - uio->uio_resid;
uio->uio_resid = resid;
}
} else if (resid > uio->uio_resid && (ioflag & IO_SYNC)) {
error = UFS_UPDATE(ip, MNT_WAIT);
}
return (error);
}
/*
* Convert a component of a pathname into a pointer to a locked inode.
* This is a very central and rather complicated routine.
* If the file system is not maintained in a strict tree hierarchy,
* this can result in a deadlock situation (see comments in code below).
*
* The cnp->cn_nameiop argument is LOOKUP, CREATE, RENAME, or DELETE depending
* on whether the name is to be looked up, created, renamed, or deleted.
* When CREATE, RENAME, or DELETE is specified, information usable in
* creating, renaming, or deleting a directory entry may be calculated.
* If flag has LOCKPARENT or'ed into it and the target of the pathname
* exists, lookup returns both the target and its parent directory locked.
* When creating or renaming and LOCKPARENT is specified, the target may
* not be ".". When deleting and LOCKPARENT is specified, the target may
* be "."., but the caller must check to ensure it does an vrele and vput
* instead of two vputs.
*
* Overall outline of ufs_lookup:
*
* check accessibility of directory
* look for name in cache, if found, then if at end of path
* and deleting or creating, drop it, else return name
* search for name in directory, to found or notfound
* notfound:
* if creating, return locked directory, leaving info on available slots
* else return error
* found:
* if at end of path and deleting, return information to allow delete
* if at end of path and rewriting (RENAME and LOCKPARENT), lock target
* inode and return info to allow rewrite
* if not at end, add name to cache; if at end and neither creating
* nor deleting, add name to cache
*/
int
ufs_lookup(void *v)
{
struct vop_lookup_args *ap = v;
struct vnode *vdp; /* vnode for directory being searched */
struct inode *dp; /* inode for directory being searched */
struct buf *bp; /* a buffer of directory entries */
struct direct *ep; /* the current directory entry */
int entryoffsetinblock; /* offset of ep in bp's buffer */
enum {NONE, COMPACT, FOUND} slotstatus;
doff_t slotoffset; /* offset of area with free space */
int slotsize; /* size of area at slotoffset */
int slotfreespace; /* amount of space free in slot */
int slotneeded; /* size of the entry we're seeking */
int numdirpasses; /* strategy for directory search */
doff_t endsearch; /* offset to end directory search */
doff_t prevoff; /* prev entry dp->i_offset */
struct vnode *pdp; /* saved dp during symlink work */
struct vnode *tdp; /* returned by VFS_VGET */
doff_t enduseful; /* pointer past last used dir slot */
u_long bmask; /* block offset mask */
int lockparent; /* 1 => lockparent flag is set */
int wantparent; /* 1 => wantparent or lockparent flag */
int namlen, error;
struct vnode **vpp = ap->a_vpp;
struct componentname *cnp = ap->a_cnp;
struct ucred *cred = cnp->cn_cred;
int flags;
int nameiop = cnp->cn_nameiop;
struct proc *p = cnp->cn_proc;
cnp->cn_flags &= ~PDIRUNLOCK;
flags = cnp->cn_flags;
bp = NULL;
slotoffset = -1;
*vpp = NULL;
vdp = ap->a_dvp;
dp = VTOI(vdp);
lockparent = flags & LOCKPARENT;
wantparent = flags & (LOCKPARENT|WANTPARENT);
/*
* Check accessiblity of directory.
*/
if ((DIP(dp, mode) & IFMT) != IFDIR)
return (ENOTDIR);
if ((error = VOP_ACCESS(vdp, VEXEC, cred)) != 0)
return (error);
if ((flags & ISLASTCN) && (vdp->v_mount->mnt_flag & MNT_RDONLY) &&
(cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
return (EROFS);
/*
* We now have a segment name to search for, and a directory to search.
*
* Before tediously performing a linear scan of the directory,
* check the name cache to see if the directory/name pair
* we are looking for is known already.
*/
if ((error = cache_lookup(vdp, vpp, cnp)) >= 0)
return (error);
/*
* Suppress search for slots unless creating
* file and at end of pathname, in which case
* we watch for a place to put the new file in
* case it doesn't already exist.
*/
slotstatus = FOUND;
slotfreespace = slotsize = slotneeded = 0;
if ((nameiop == CREATE || nameiop == RENAME) &&
(flags & ISLASTCN)) {
slotstatus = NONE;
slotneeded = (sizeof(struct direct) - MAXNAMLEN +
cnp->cn_namelen + 3) &~ 3;
}
/*
* If there is cached information on a previous search of
* this directory, pick up where we last left off.
* We cache only lookups as these are the most common
* and have the greatest payoff. Caching CREATE has little
* benefit as it usually must search the entire directory
* to determine that the entry does not exist. Caching the
* location of the last DELETE or RENAME has not reduced
* profiling time and hence has been removed in the interest
* of simplicity.
*/
bmask = VFSTOUFS(vdp->v_mount)->um_mountp->mnt_stat.f_iosize - 1;
#ifdef UFS_DIRHASH
/*
* Use dirhash for fast operations on large directories. The logic
* to determine whether to hash the directory is contained within
* ufsdirhash_build(); a zero return means that it decided to hash
* this directory and it successfully built up the hash table.
*/
if (ufsdirhash_build(dp) == 0) {
/* Look for a free slot if needed. */
enduseful = DIP(dp, size);
if (slotstatus != FOUND) {
slotoffset = ufsdirhash_findfree(dp, slotneeded,
&slotsize);
if (slotoffset >= 0) {
slotstatus = COMPACT;
enduseful = ufsdirhash_enduseful(dp);
if (enduseful < 0)
enduseful = DIP(dp, size);
}
}
/* Look up the component. */
numdirpasses = 1;
entryoffsetinblock = 0; /* silence compiler warning */
switch (ufsdirhash_lookup(dp, cnp->cn_nameptr, cnp->cn_namelen,
&dp->i_offset, &bp, nameiop == DELETE ? &prevoff : NULL)) {
case 0:
ep = (struct direct *)((char *)bp->b_data +
(dp->i_offset & bmask));
goto foundentry;
case ENOENT:
#define roundup2(x, y) (((x)+((y)-1))&(~((y)-1))) /* if y is powers of two */
dp->i_offset = roundup2(DIP(dp, size), DIRBLKSIZ);
goto notfound;
default:
/* Something failed; just do a linear search. */
break;
}
}
#endif /* UFS_DIRHASH */
if (nameiop != LOOKUP || dp->i_diroff == 0 ||
dp->i_diroff >= DIP(dp, size)) {
entryoffsetinblock = 0;
dp->i_offset = 0;
numdirpasses = 1;
} else {
dp->i_offset = dp->i_diroff;
if ((entryoffsetinblock = dp->i_offset & bmask) &&
(error = UFS_BUFATOFF(dp, (off_t)dp->i_offset, NULL, &bp)))
return (error);
numdirpasses = 2;
nchstats.ncs_2passes++;
}
prevoff = dp->i_offset;
endsearch = roundup(DIP(dp, size), DIRBLKSIZ);
enduseful = 0;
searchloop:
while (dp->i_offset < endsearch) {
/*
* If necessary, get the next directory block.
*/
if ((dp->i_offset & bmask) == 0) {
if (bp != NULL)
brelse(bp);
error = UFS_BUFATOFF(dp, (off_t)dp->i_offset, NULL,
&bp);
if (error)
return (error);
entryoffsetinblock = 0;
}
/*
* If still looking for a slot, and at a DIRBLKSIZE
* boundary, have to start looking for free space again.
*/
if (slotstatus == NONE &&
(entryoffsetinblock & (DIRBLKSIZ - 1)) == 0) {
slotoffset = -1;
slotfreespace = 0;
}
/*
* Get pointer to next entry.
* Full validation checks are slow, so we only check
* enough to insure forward progress through the
* directory. Complete checks can be run by patching
* "dirchk" to be true.
*/
ep = (struct direct *)((char *)bp->b_data + entryoffsetinblock);
if (ep->d_reclen == 0 ||
(dirchk && ufs_dirbadentry(vdp, ep, entryoffsetinblock))) {
int i;
ufs_dirbad(dp, dp->i_offset, "mangled entry");
i = DIRBLKSIZ - (entryoffsetinblock & (DIRBLKSIZ - 1));
dp->i_offset += i;
entryoffsetinblock += i;
continue;
}
/*
* If an appropriate sized slot has not yet been found,
* check to see if one is available. Also accumulate space
* in the current block so that we can determine if
* compaction is viable.
*/
if (slotstatus != FOUND) {
int size = ep->d_reclen;
if (ep->d_ino != 0)
size -= DIRSIZ(FSFMT(vdp), ep);
if (size > 0) {
if (size >= slotneeded) {
slotstatus = FOUND;
slotoffset = dp->i_offset;
slotsize = ep->d_reclen;
} else if (slotstatus == NONE) {
slotfreespace += size;
if (slotoffset == -1)
slotoffset = dp->i_offset;
if (slotfreespace >= slotneeded) {
slotstatus = COMPACT;
slotsize = dp->i_offset +
ep->d_reclen - slotoffset;
}
}
}
}
/*
* Check for a name match.
*/
if (ep->d_ino) {
# if (BYTE_ORDER == LITTLE_ENDIAN)
if (vdp->v_mount->mnt_maxsymlinklen > 0)
namlen = ep->d_namlen;
else
namlen = ep->d_type;
# else
namlen = ep->d_namlen;
# endif
if (namlen == cnp->cn_namelen &&
!memcmp(cnp->cn_nameptr, ep->d_name, namlen)) {
#ifdef UFS_DIRHASH
foundentry:
#endif
/*
* Save directory entry's inode number and
* reclen in ndp->ni_ufs area, and release
* directory buffer.
*/
dp->i_ino = ep->d_ino;
dp->i_reclen = ep->d_reclen;
goto found;
}
}
prevoff = dp->i_offset;
dp->i_offset += ep->d_reclen;
entryoffsetinblock += ep->d_reclen;
if (ep->d_ino)
enduseful = dp->i_offset;
}
#ifdef UFS_DIRHASH
notfound:
#endif
/*
* If we started in the middle of the directory and failed
* to find our target, we must check the beginning as well.
*/
if (numdirpasses == 2) {
numdirpasses--;
dp->i_offset = 0;
endsearch = dp->i_diroff;
goto searchloop;
}
if (bp != NULL)
brelse(bp);
/*
* If creating, and at end of pathname and current
* directory has not been removed, then can consider
* allowing file to be created.
*/
if ((nameiop == CREATE || nameiop == RENAME) &&
(flags & ISLASTCN) && dp->i_effnlink != 0) {
/*
* Access for write is interpreted as allowing
* creation of files in the directory.
*/
error = VOP_ACCESS(vdp, VWRITE, cred);
if (error)
return (error);
/*
* Return an indication of where the new directory
* entry should be put. If we didn't find a slot,
* then set dp->i_count to 0 indicating
* that the new slot belongs at the end of the
* directory. If we found a slot, then the new entry
* can be put in the range from dp->i_offset to
* dp->i_offset + dp->i_count.
*/
if (slotstatus == NONE) {
dp->i_offset = roundup(DIP(dp, size), DIRBLKSIZ);
dp->i_count = 0;
enduseful = dp->i_offset;
} else if (nameiop == DELETE) {
dp->i_offset = slotoffset;
if ((dp->i_offset & (DIRBLKSIZ - 1)) == 0)
dp->i_count = 0;
else
dp->i_count = dp->i_offset - prevoff;
} else {
dp->i_offset = slotoffset;
dp->i_count = slotsize;
if (enduseful < slotoffset + slotsize)
enduseful = slotoffset + slotsize;
}
dp->i_endoff = roundup(enduseful, DIRBLKSIZ);
/*
* We return with the directory locked, so that
* the parameters we set up above will still be
* valid if we actually decide to do a direnter().
* We return ni_vp == NULL to indicate that the entry
* does not currently exist; we leave a pointer to
* the (locked) directory inode in ndp->ni_dvp.
* The pathname buffer is saved so that the name
* can be obtained later.
*
* NB - if the directory is unlocked, then this
* information cannot be used.
*/
cnp->cn_flags |= SAVENAME;
if (!lockparent) {
VOP_UNLOCK(vdp, 0);
cnp->cn_flags |= PDIRUNLOCK;
}
return (EJUSTRETURN);
}
/*
* Insert name into cache (as non-existent) if appropriate.
*/
if ((cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
cache_enter(vdp, *vpp, cnp);
return (ENOENT);
found:
if (numdirpasses == 2)
nchstats.ncs_pass2++;
/*
* Check that directory length properly reflects presence
* of this entry.
*/
if (dp->i_offset + DIRSIZ(FSFMT(vdp), ep) > DIP(dp, size)) {
ufs_dirbad(dp, dp->i_offset, "i_ffs_size too small");
DIP_ASSIGN(dp, size, dp->i_offset + DIRSIZ(FSFMT(vdp), ep));
dp->i_flag |= IN_CHANGE | IN_UPDATE;
UFS_WAPBL_UPDATE(dp, MNT_WAIT);
}
brelse(bp);
/*
* Found component in pathname.
* If the final component of path name, save information
* in the cache as to where the entry was found.
*/
if ((flags & ISLASTCN) && nameiop == LOOKUP)
dp->i_diroff = dp->i_offset &~ (DIRBLKSIZ - 1);
/*
* If deleting, and at end of pathname, return
* parameters which can be used to remove file.
* If the wantparent flag isn't set, we return only
* the directory (in ndp->ni_dvp), otherwise we go
* on and lock the inode, being careful with ".".
*/
if (nameiop == DELETE && (flags & ISLASTCN)) {
/*
* Write access to directory required to delete files.
*/
error = VOP_ACCESS(vdp, VWRITE, cred);
if (error)
return (error);
/*
* Return pointer to current entry in dp->i_offset,
* and distance past previous entry (if there
* is a previous entry in this block) in dp->i_count.
* Save directory inode pointer in ndp->ni_dvp for dirremove().
*/
if ((dp->i_offset & (DIRBLKSIZ - 1)) == 0)
dp->i_count = 0;
else
dp->i_count = dp->i_offset - prevoff;
if (dp->i_number == dp->i_ino) {
vref(vdp);
*vpp = vdp;
return (0);
}
error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp);
if (error)
return (error);
/*
* If directory is "sticky", then user must own
* the directory, or the file in it, else she
* may not delete it (unless she's root). This
* implements append-only directories.
*/
if ((DIP(dp, mode) & ISVTX) &&
cred->cr_uid != 0 &&
cred->cr_uid != DIP(dp, uid) &&
DIP(VTOI(tdp), uid) != cred->cr_uid) {
vput(tdp);
return (EPERM);
}
*vpp = tdp;
if (!lockparent) {
VOP_UNLOCK(vdp, 0);
cnp->cn_flags |= PDIRUNLOCK;
}
return (0);
}
/*
* If rewriting (RENAME), return the inode and the
* information required to rewrite the present directory
* Must get inode of directory entry to verify it's a
* regular file, or empty directory.
*/
if (nameiop == RENAME && wantparent &&
(flags & ISLASTCN)) {
error = VOP_ACCESS(vdp, VWRITE, cred);
if (error)
return (error);
/*
* Careful about locking second inode.
* This can only occur if the target is ".".
*/
if (dp->i_number == dp->i_ino)
return (EISDIR);
error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp);
if (error)
return (error);
*vpp = tdp;
cnp->cn_flags |= SAVENAME;
if (!lockparent) {
VOP_UNLOCK(vdp, 0);
cnp->cn_flags |= PDIRUNLOCK;
}
return (0);
}
/*
* Step through the translation in the name. We do not `vput' the
* directory because we may need it again if a symbolic link
* is relative to the current directory. Instead we save it
* unlocked as "pdp". We must get the target inode before unlocking
* the directory to insure that the inode will not be removed
* before we get it. We prevent deadlock by always fetching
* inodes from the root, moving down the directory tree. Thus
* when following backward pointers ".." we must unlock the
* parent directory before getting the requested directory.
* There is a potential race condition here if both the current
* and parent directories are removed before the VFS_VGET for the
* inode associated with ".." returns. We hope that this occurs
* infrequently since we cannot avoid this race condition without
* implementing a sophisticated deadlock detection algorithm.
* Note also that this simple deadlock detection scheme will not
* work if the file system has any hard links other than ".."
* that point backwards in the directory structure.
*/
pdp = vdp;
if (flags & ISDOTDOT) {
VOP_UNLOCK(pdp, 0); /* race to get the inode */
cnp->cn_flags |= PDIRUNLOCK;
error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp);
if (error) {
if (vn_lock(pdp, LK_EXCLUSIVE | LK_RETRY, p) == 0)
cnp->cn_flags &= ~PDIRUNLOCK;
return (error);
}
if (lockparent && (flags & ISLASTCN)) {
if ((error = vn_lock(pdp, LK_EXCLUSIVE, p))) {
vput(tdp);
return (error);
}
cnp->cn_flags &= ~PDIRUNLOCK;
}
*vpp = tdp;
} else if (dp->i_number == dp->i_ino) {
vref(vdp); /* we want ourself, ie "." */
*vpp = vdp;
} else {
error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp);
if (error)
return (error);
if (!lockparent || !(flags & ISLASTCN)) {
VOP_UNLOCK(pdp, 0);
cnp->cn_flags |= PDIRUNLOCK;
}
*vpp = tdp;
}
/*
* Insert name into cache if appropriate.
*/
if (cnp->cn_flags & MAKEENTRY)
cache_enter(vdp, *vpp, cnp);
return (0);
}
/*
* Mkdir system call
*/
int
ufs_mkdir(void *v)
{
struct vop_mkdir_args *ap = v;
struct vnode *dvp = ap->a_dvp;
struct vattr *vap = ap->a_vap;
struct componentname *cnp = ap->a_cnp;
struct inode *ip, *dp;
struct vnode *tvp;
struct buf *bp;
struct direct newdir;
struct dirtemplate dirtemplate, *dtp;
int error, dmode, blkoff;
#ifdef DIAGNOSTIC
if ((cnp->cn_flags & HASBUF) == 0)
panic("ufs_mkdir: no name");
#endif
dp = VTOI(dvp);
if ((nlink_t) DIP(dp, nlink) >= LINK_MAX) {
error = EMLINK;
goto out;
}
dmode = vap->va_mode & 0777;
dmode |= IFDIR;
/*
* Must simulate part of ufs_makeinode here to acquire the inode,
* but not have it entered in the parent directory. The entry is
* made later after writing "." and ".." entries.
*/
if ((error = UFS_INODE_ALLOC(dp, dmode, cnp->cn_cred, &tvp)) != 0)
goto out;
ip = VTOI(tvp);
DIP_ASSIGN(ip, uid, cnp->cn_cred->cr_uid);
DIP_ASSIGN(ip, gid, DIP(dp, gid));
if ((error = getinoquota(ip)) ||
(error = ufs_quota_alloc_inode(ip, cnp->cn_cred))) {
pool_put(&namei_pool, cnp->cn_pnbuf);
UFS_INODE_FREE(ip, ip->i_number, dmode);
vput(tvp);
vput(dvp);
return (error);
}
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
DIP_ASSIGN(ip, mode, dmode);
tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
ip->i_effnlink = 2;
DIP_ASSIGN(ip, nlink, 2);
if (DOINGSOFTDEP(tvp))
softdep_change_linkcnt(ip, 0);
/*
* Bump link count in parent directory to reflect work done below.
* Should be done before reference is create so cleanup is
* possible if we crash.
*/
dp->i_effnlink++;
DIP_ADD(dp, nlink, 1);
dp->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(dvp))
softdep_change_linkcnt(dp, 0);
if ((error = UFS_UPDATE(dp, !DOINGSOFTDEP(dvp))) != 0)
goto bad;
/*
* Initialize directory with "." and ".." from static template.
*/
if (dvp->v_mount->mnt_maxsymlinklen > 0)
dtp = &mastertemplate;
else
dtp = (struct dirtemplate *)&omastertemplate;
dirtemplate = *dtp;
dirtemplate.dot_ino = ip->i_number;
dirtemplate.dotdot_ino = dp->i_number;
if ((error = UFS_BUF_ALLOC(ip, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
B_CLRBUF, &bp)) != 0)
goto bad;
DIP_ASSIGN(ip, size, DIRBLKSIZ);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
uvm_vnp_setsize(tvp, DIP(ip, size));
bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
if (DOINGSOFTDEP(tvp)) {
/*
* Ensure that the entire newly allocated block is a
* valid directory so that future growth within the
* block does not have to ensure that the block is
* written before the inode
*/
blkoff = DIRBLKSIZ;
while (blkoff < bp->b_bcount) {
((struct direct *)
(bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
blkoff += DIRBLKSIZ;
}
}
if ((error = UFS_UPDATE(ip, !DOINGSOFTDEP(tvp))) != 0) {
(void)VOP_BWRITE(bp);
goto bad;
}
/*
* Directory set up, now install its entry in the parent directory.
*
* If we are not doing soft dependencies, then we must write out the
* buffer containing the new directory body before entering the new
* name in the parent. If we are doing soft dependencies, then the
* buffer containing the new directory body will be passed to and
* released in the soft dependency code after the code has attached
* an appropriate ordering dependency to the buffer which ensures that
* the buffer is written before the new name is written in the parent.
*/
if (!DOINGSOFTDEP(dvp) && ((error = VOP_BWRITE(bp)) != 0))
goto bad;
ufs_makedirentry(ip, cnp, &newdir);
error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
bad:
if (error == 0) {
VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
*ap->a_vpp = tvp;
} else {
dp->i_effnlink--;
DIP_ADD(dp, nlink, -1);
dp->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(dvp))
softdep_change_linkcnt(dp, 0);
/*
* No need to do an explicit VOP_TRUNCATE here, vrele will
* do this for us because we set the link count to 0.
*/
ip->i_effnlink = 0;
DIP_ASSIGN(ip, nlink, 0);
ip->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(tvp))
softdep_change_linkcnt(ip, 0);
vput(tvp);
}
out:
pool_put(&namei_pool, cnp->cn_pnbuf);
vput(dvp);
return (error);
}
/*
* Update the access, modified, and inode change times as specified by the
* IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. The IN_MODIFIED
* flag is used to specify that the inode needs to be updated but that the
* times have already been set. The access and modified times are taken from
* the second and third parameters; the inode change time is always taken
* from the current time. If waitfor is set, then wait for the disk write
* of the inode to complete.
*/
int
ffs_update(struct inode *ip, struct timespec *atime,
struct timespec *mtime, int waitfor)
{
struct vnode *vp;
struct fs *fs;
struct buf *bp;
int error;
struct timespec ts;
vp = ITOV(ip);
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
ip->i_flag &=
~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE);
return (0);
}
if ((ip->i_flag &
(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
waitfor != MNT_WAIT)
return (0);
getnanotime(&ts);
if (ip->i_flag & IN_ACCESS) {
DIP_ASSIGN(ip, atime, atime ? atime->tv_sec : ts.tv_sec);
DIP_ASSIGN(ip, atimensec, atime ? atime->tv_nsec : ts.tv_nsec);
}
if (ip->i_flag & IN_UPDATE) {
DIP_ASSIGN(ip, mtime, mtime ? mtime->tv_sec : ts.tv_sec);
DIP_ASSIGN(ip, mtimensec, mtime ? mtime->tv_nsec : ts.tv_nsec);
ip->i_modrev++;
}
if (ip->i_flag & IN_CHANGE) {
DIP_ASSIGN(ip, ctime, ts.tv_sec);
DIP_ASSIGN(ip, ctimensec, ts.tv_nsec);
}
ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE);
fs = ip->i_fs;
/*
* Ensure that uid and gid are correct. This is a temporary
* fix until fsck has been changed to do the update.
*/
if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_inodefmt < FS_44INODEFMT) {
ip->i_din1->di_ouid = ip->i_ffs1_uid;
ip->i_din1->di_ogid = ip->i_ffs1_gid;
}
error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->fs_bsize, &bp);
if (error) {
brelse(bp);
return (error);
}
if (DOINGSOFTDEP(vp))
softdep_update_inodeblock(ip, bp, waitfor);
else if (ip->i_effnlink != DIP(ip, nlink))
panic("ffs_update: bad link cnt");
#ifdef FFS2
if (ip->i_ump->um_fstype == UM_UFS2)
*((struct ufs2_dinode *)bp->b_data +
ino_to_fsbo(fs, ip->i_number)) = *ip->i_din2;
else
#endif
*((struct ufs1_dinode *)bp->b_data +
ino_to_fsbo(fs, ip->i_number)) = *ip->i_din1;
if (waitfor && !DOINGASYNC(vp)) {
return (bwrite(bp));
} else {
bdwrite(bp);
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 proc *p = ap->a_p;
mode_t mode;
int error = 0;
#ifdef DIAGNOSTIC
extern int prtactive;
if (prtactive && vp->v_usecount != 0)
vprint("ffs_inactive: pushing active", vp);
#endif
/*
* 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) {
if (getinoquota(ip) == 0)
(void)ufs_quota_free_inode(ip, NOCRED);
error = UFS_TRUNCATE(ip, (off_t)0, IO_EXT | IO_NORMAL, NOCRED);
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)) {
UFS_UPDATE(ip, 0);
}
out:
VOP_UNLOCK(vp, 0, p);
/*
* If we are done with the inode, reclaim it
* so that it can be reused immediately.
*/
if (ip->i_din1 == NULL || DIP(ip, mode) == 0)
vrecycle(vp, p);
return (error);
}
int
ufs_mkdir(void *v)
{
struct vop_mkdir_v3_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
struct vattr *a_vap;
} */ *ap = v;
struct vnode *dvp = ap->a_dvp, *tvp;
struct vattr *vap = ap->a_vap;
struct componentname *cnp = ap->a_cnp;
struct inode *ip, *dp = VTOI(dvp);
struct buf *bp;
struct dirtemplate dirtemplate;
struct direct *newdir;
int error, dmode;
struct ufsmount *ump = dp->i_ump;
int dirblksiz = ump->um_dirblksiz;
struct ufs_lookup_results *ulr;
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
/* XXX should handle this material another way */
ulr = &dp->i_crap;
UFS_CHECK_CRAPCOUNTER(dp);
if ((nlink_t)dp->i_nlink >= LINK_MAX) {
error = EMLINK;
goto out;
}
dmode = vap->va_mode & ACCESSPERMS;
dmode |= IFDIR;
/*
* Must simulate part of ufs_makeinode here to acquire the inode,
* but not have it entered in the parent directory. The entry is
* made later after writing "." and ".." entries.
*/
if ((error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, ap->a_vpp)) != 0)
goto out;
tvp = *ap->a_vpp;
ip = VTOI(tvp);
error = UFS_WAPBL_BEGIN(ap->a_dvp->v_mount);
if (error) {
UFS_VFREE(tvp, ip->i_number, dmode);
vput(tvp);
goto out;
}
ip->i_uid = kauth_cred_geteuid(cnp->cn_cred);
DIP_ASSIGN(ip, uid, ip->i_uid);
ip->i_gid = dp->i_gid;
DIP_ASSIGN(ip, gid, ip->i_gid);
#if defined(QUOTA) || defined(QUOTA2)
if ((error = chkiq(ip, 1, cnp->cn_cred, 0))) {
UFS_VFREE(tvp, ip->i_number, dmode);
UFS_WAPBL_END(dvp->v_mount);
fstrans_done(dvp->v_mount);
vput(tvp);
return (error);
}
#endif
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
ip->i_mode = dmode;
DIP_ASSIGN(ip, mode, dmode);
tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
ip->i_nlink = 2;
DIP_ASSIGN(ip, nlink, 2);
if (cnp->cn_flags & ISWHITEOUT) {
ip->i_flags |= UF_OPAQUE;
DIP_ASSIGN(ip, flags, ip->i_flags);
}
/*
* Bump link count in parent directory to reflect work done below.
* Should be done before reference is created so cleanup is
* possible if we crash.
*/
dp->i_nlink++;
DIP_ASSIGN(dp, nlink, dp->i_nlink);
dp->i_flag |= IN_CHANGE;
if ((error = UFS_UPDATE(dvp, NULL, NULL, UPDATE_DIROP)) != 0)
goto bad;
/*
* Initialize directory with "." and ".." from static template.
*/
dirtemplate = mastertemplate;
dirtemplate.dotdot_reclen = dirblksiz - dirtemplate.dot_reclen;
dirtemplate.dot_ino = ufs_rw32(ip->i_number, UFS_MPNEEDSWAP(ump));
dirtemplate.dotdot_ino = ufs_rw32(dp->i_number, UFS_MPNEEDSWAP(ump));
dirtemplate.dot_reclen = ufs_rw16(dirtemplate.dot_reclen,
UFS_MPNEEDSWAP(ump));
dirtemplate.dotdot_reclen = ufs_rw16(dirtemplate.dotdot_reclen,
UFS_MPNEEDSWAP(ump));
if (ump->um_maxsymlinklen <= 0) {
#if BYTE_ORDER == LITTLE_ENDIAN
if (UFS_MPNEEDSWAP(ump) == 0)
#else
if (UFS_MPNEEDSWAP(ump) != 0)
#endif
{
dirtemplate.dot_type = dirtemplate.dot_namlen;
dirtemplate.dotdot_type = dirtemplate.dotdot_namlen;
dirtemplate.dot_namlen = dirtemplate.dotdot_namlen = 0;
} else
dirtemplate.dot_type = dirtemplate.dotdot_type = 0;
}
if ((error = UFS_BALLOC(tvp, (off_t)0, dirblksiz, cnp->cn_cred,
B_CLRBUF, &bp)) != 0)
goto bad;
ip->i_size = dirblksiz;
DIP_ASSIGN(ip, size, dirblksiz);
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
uvm_vnp_setsize(tvp, ip->i_size);
memcpy((void *)bp->b_data, (void *)&dirtemplate, sizeof dirtemplate);
/*
* Directory set up, now install its entry in the parent directory.
* We must write out the buffer containing the new directory body
* before entering the new name in the parent.
*/
if ((error = VOP_BWRITE(bp->b_vp, bp)) != 0)
goto bad;
if ((error = UFS_UPDATE(tvp, NULL, NULL, UPDATE_DIROP)) != 0) {
goto bad;
}
newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
ufs_makedirentry(ip, cnp, newdir);
error = ufs_direnter(dvp, ulr, tvp, newdir, cnp, bp);
pool_cache_put(ufs_direct_cache, newdir);
bad:
if (error == 0) {
VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
VOP_UNLOCK(tvp);
UFS_WAPBL_END(dvp->v_mount);
} else {
dp->i_nlink--;
DIP_ASSIGN(dp, nlink, dp->i_nlink);
dp->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP);
/*
* No need to do an explicit UFS_TRUNCATE here, vrele will
* do this for us because we set the link count to 0.
*/
ip->i_nlink = 0;
DIP_ASSIGN(ip, nlink, 0);
ip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(tvp, NULL, NULL, UPDATE_DIROP);
UFS_WAPBL_END(dvp->v_mount);
vput(tvp);
}
out:
fstrans_done(dvp->v_mount);
return (error);
}
/*
* 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);
}
/*
* 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);
}
/*
* 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);
}
/*
* Look up a FFS dinode number to find its incore vnode, otherwise read it
* in from disk. If it is in core, wait for the lock bit to clear, then
* return the inode locked. Detection and handling of mount points must be
* done by the calling routine.
*/
int
ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
{
struct fs *fs;
struct inode *ip;
struct ufs1_dinode *dp1;
#ifdef FFS2
struct ufs2_dinode *dp2;
#endif
struct ufsmount *ump;
struct buf *bp;
struct vnode *vp;
dev_t dev;
int error;
ump = VFSTOUFS(mp);
dev = ump->um_dev;
retry:
if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
return (0);
/* Allocate a new vnode/inode. */
if ((error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) != 0) {
*vpp = NULL;
return (error);
}
#ifdef VFSDEBUG
vp->v_flag |= VLOCKSWORK;
#endif
/* XXX - we use the same pool for ffs and mfs */
ip = pool_get(&ffs_ino_pool, PR_WAITOK);
bzero((caddr_t)ip, sizeof(struct inode));
lockinit(&ip->i_lock, PINOD, "inode", 0, 0);
ip->i_ump = ump;
VREF(ip->i_devvp);
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_fs = fs = ump->um_fs;
ip->i_dev = dev;
ip->i_number = ino;
ip->i_vtbl = &ffs_vtbl;
/*
* Put it onto its hash chain and lock it so that other requests for
* this inode will block if they arrive while we are sleeping waiting
* for old data structures to be purged or for the contents of the
* disk portion of this inode to be read.
*/
error = ufs_ihashins(ip);
if (error) {
/*
* VOP_INACTIVE will treat this as a stale file
* and recycle it quickly
*/
vrele(vp);
if (error == EEXIST)
goto retry;
return (error);
}
/* Read in the disk contents for the inode, copy into the inode. */
error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
(int)fs->fs_bsize, NOCRED, &bp);
if (error) {
/*
* The inode does not contain anything useful, so it would
* be misleading to leave it on its hash chain. With mode
* still zero, it will be unlinked and returned to the free
* list by vput().
*/
vput(vp);
brelse(bp);
*vpp = NULL;
return (error);
}
#ifdef FFS2
if (ip->i_ump->um_fstype == UM_UFS2) {
ip->i_din2 = pool_get(&ffs_dinode2_pool, PR_WAITOK);
dp2 = (struct ufs2_dinode *) bp->b_data + ino_to_fsbo(fs, ino);
*ip->i_din2 = *dp2;
} else
#endif
{
ip->i_din1 = pool_get(&ffs_dinode1_pool, PR_WAITOK);
dp1 = (struct ufs1_dinode *) bp->b_data + ino_to_fsbo(fs, ino);
*ip->i_din1 = *dp1;
}
brelse(bp);
if (DOINGSOFTDEP(vp))
softdep_load_inodeblock(ip);
else
ip->i_effnlink = DIP(ip, nlink);
/*
* Initialize the vnode from the inode, check for aliases.
* Note that the underlying vnode may have changed.
*/
error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp);
if (error) {
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Set up a generation number for this inode if it does not
* already have one. This should only happen on old filesystems.
*/
if (DIP(ip, gen) == 0) {
DIP_ASSIGN(ip, gen, arc4random() & INT_MAX);
if (DIP(ip, gen) == 0 || DIP(ip, gen) == -1)
DIP_ASSIGN(ip, gen, 1); /* Shouldn't happen */
if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
/*
* Ensure that uid and gid are correct. This is a temporary
* fix until fsck has been changed to do the update.
*/
if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_inodefmt < FS_44INODEFMT) {
ip->i_ffs1_uid = ip->i_din1->di_ouid;
ip->i_ffs1_gid = ip->i_din1->di_ogid;
}
*vpp = vp;
return (0);
}
/*
* Allocate a new inode.
*/
int
ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
struct componentname *cnp)
{
struct inode *ip, *pdir;
struct direct newdir;
struct vnode *tvp;
int error;
pdir = VTOI(dvp);
#ifdef DIAGNOSTIC
if ((cnp->cn_flags & HASBUF) == 0)
panic("ufs_makeinode: no name");
#endif
*vpp = NULL;
if ((mode & IFMT) == 0)
mode |= IFREG;
if ((error = UFS_INODE_ALLOC(pdir, mode, cnp->cn_cred, &tvp)) != 0) {
pool_put(&namei_pool, cnp->cn_pnbuf);
vput(dvp);
return (error);
}
ip = VTOI(tvp);
DIP_ASSIGN(ip, gid, DIP(pdir, gid));
DIP_ASSIGN(ip, uid, cnp->cn_cred->cr_uid);
if ((error = getinoquota(ip)) ||
(error = ufs_quota_alloc_inode(ip, cnp->cn_cred))) {
pool_put(&namei_pool, cnp->cn_pnbuf);
UFS_INODE_FREE(ip, ip->i_number, mode);
vput(tvp);
vput(dvp);
return (error);
}
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
DIP_ASSIGN(ip, mode, mode);
tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
ip->i_effnlink = 1;
DIP_ASSIGN(ip, nlink, 1);
if (DOINGSOFTDEP(tvp))
softdep_change_linkcnt(ip, 0);
if ((DIP(ip, mode) & ISGID) &&
!groupmember(DIP(ip, gid), cnp->cn_cred) &&
suser_ucred(cnp->cn_cred))
DIP_AND(ip, mode, ~ISGID);
/*
* Make sure inode goes to disk before directory entry.
*/
if ((error = UFS_UPDATE(ip, !DOINGSOFTDEP(tvp))) != 0)
goto bad;
ufs_makedirentry(ip, cnp, &newdir);
if ((error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL)) != 0)
goto bad;
if ((cnp->cn_flags & SAVESTART) == 0)
pool_put(&namei_pool, cnp->cn_pnbuf);
vput(dvp);
*vpp = tvp;
return (0);
bad:
/*
* Write error occurred trying to update the inode
* or the directory so must deallocate the inode.
*/
pool_put(&namei_pool, cnp->cn_pnbuf);
vput(dvp);
ip->i_effnlink = 0;
DIP_ASSIGN(ip, nlink, 0);
ip->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(tvp))
softdep_change_linkcnt(ip, 0);
tvp->v_type = VNON;
vput(tvp);
return (error);
}
// ffs文件系统的写入操作
int
ffs_write(void *v)
{
struct vop_write_args *ap = v;
struct vnode *vp;
struct uio *uio;
struct inode *ip;
struct fs *fs;
struct buf *bp;
daddr_t lbn;
off_t osize;
int blkoffset, error, extended, flags, ioflag, size, xfersize;
ssize_t resid, overrun;
extended = 0;
ioflag = ap->a_ioflag;
uio = ap->a_uio;
vp = ap->a_vp;
ip = VTOI(vp);
#ifdef DIAGNOSTIC
if (uio->uio_rw != UIO_WRITE)
panic("ffs_write: mode");
#endif
/*
* If writing 0 bytes, succeed and do not change
* update time or file offset (standards compliance)
*/
if (uio->uio_resid == 0)
return (0);
switch (vp->v_type) {
case VREG:
if (ioflag & IO_APPEND)
uio->uio_offset = DIP(ip, size);
if ((DIP(ip, flags) & APPEND) && uio->uio_offset != DIP(ip, size))
return (EPERM);
/* FALLTHROUGH */
case VLNK:
break;
case VDIR:
if ((ioflag & IO_SYNC) == 0)
panic("ffs_write: nonsync dir write");
break;
default:
panic("ffs_write: type");
}
fs = ip->i_fs;
if (uio->uio_offset < 0 ||
(u_int64_t)uio->uio_offset + uio->uio_resid > fs->fs_maxfilesize)
return (EFBIG);
/* do the filesize rlimit check */
if ((error = vn_fsizechk(vp, uio, ioflag, &overrun)))
return (error);
resid = uio->uio_resid;
osize = DIP(ip, size);
flags = ioflag & IO_SYNC ? B_SYNC : 0;
for (error = 0; uio->uio_resid > 0;) {
lbn = lblkno(fs, uio->uio_offset);
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->fs_bsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (fs->fs_bsize > xfersize)
flags |= B_CLRBUF;
else
flags &= ~B_CLRBUF;
if ((error = UFS_BUF_ALLOC(ip, uio->uio_offset, xfersize,
ap->a_cred, flags, &bp)) != 0)
break;
if (uio->uio_offset + xfersize > DIP(ip, size)) {
DIP_ASSIGN(ip, size, uio->uio_offset + xfersize);
uvm_vnp_setsize(vp, DIP(ip, size));
extended = 1;
}
(void)uvm_vnp_uncache(vp);
size = blksize(fs, ip, lbn) - bp->b_resid;
if (size < xfersize)
xfersize = size;
error =
uiomovei(bp->b_data + blkoffset, xfersize, uio);
if (error != 0)
memset(bp->b_data + blkoffset, 0, xfersize);
#if 0
if (ioflag & IO_NOCACHE)
bp->b_flags |= B_NOCACHE;
#endif
if (ioflag & IO_SYNC)
(void)bwrite(bp);
else if (xfersize + blkoffset == fs->fs_bsize) {
if (doclusterwrite)
cluster_write(bp, &ip->i_ci, DIP(ip, size));
else
bawrite(bp);
} else
bdwrite(bp);
if (error || xfersize == 0)
break;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* If we successfully wrote any data, and we are not the superuser
* we clear the setuid and setgid bits as a precaution against
* tampering.
*/
if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
DIP_ASSIGN(ip, mode, DIP(ip, mode) & ~(ISUID | ISGID));
if (resid > uio->uio_resid)
VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0));
if (error) {
if (ioflag & IO_UNIT) {
(void)UFS_TRUNCATE(ip, osize,
ioflag & IO_SYNC, ap->a_cred);
uio->uio_offset -= resid - uio->uio_resid;
uio->uio_resid = resid;
}
} else if (resid > uio->uio_resid && (ioflag & IO_SYNC)) {
error = UFS_UPDATE(ip, 1);
}
/* correct the result for writes clamped by vn_fsizechk() */
uio->uio_resid += overrun;
return (error);
}
/*
* Set attribute vnode op. called from several syscalls
*/
int
ufs_setattr(void *v)
{
struct vop_setattr_args *ap = v;
struct vattr *vap = ap->a_vap;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct ucred *cred = ap->a_cred;
struct proc *p = curproc;
int error;
long hint = NOTE_ATTRIB;
u_quad_t oldsize;
/*
* 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);
}
if (vap->va_flags != VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
if (cred->cr_uid != DIP(ip, uid) &&
(error = suser_ucred(cred)))
return (error);
if (cred->cr_uid == 0) {
if ((DIP(ip, flags) & (SF_IMMUTABLE | SF_APPEND)) &&
securelevel > 0)
return (EPERM);
DIP_ASSIGN(ip, flags, vap->va_flags);
} else {
if (DIP(ip, flags) & (SF_IMMUTABLE | SF_APPEND) ||
(vap->va_flags & UF_SETTABLE) != vap->va_flags)
return (EPERM);
DIP_AND(ip, flags, SF_SETTABLE);
DIP_OR(ip, flags, vap->va_flags & UF_SETTABLE);
}
ip->i_flag |= IN_CHANGE;
if (vap->va_flags & (IMMUTABLE | APPEND))
return (0);
}
if (DIP(ip, flags) & (IMMUTABLE | APPEND))
return (EPERM);
/*
* Go through the fields and update if not VNOVAL.
*/
if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, p);
if (error)
return (error);
}
if (vap->va_size != VNOVAL) {
oldsize = DIP(ip, size);
/*
* 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:
return (EISDIR);
case VLNK:
case VREG:
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
break;
default:
break;
}
if ((error = UFS_TRUNCATE(ip, vap->va_size, 0, cred)) != 0)
return (error);
if (vap->va_size < oldsize)
hint |= NOTE_TRUNCATE;
}
if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
if (cred->cr_uid != DIP(ip, uid) &&
(error = suser_ucred(cred)) &&
((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
(error = VOP_ACCESS(vp, VWRITE, cred))))
return (error);
if (vap->va_mtime.tv_sec != VNOVAL)
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (vap->va_atime.tv_sec != VNOVAL) {
if (!(vp->v_mount->mnt_flag & MNT_NOATIME) ||
(ip->i_flag & (IN_CHANGE | IN_UPDATE)))
ip->i_flag |= IN_ACCESS;
}
error = UFS_UPDATE2(ip, &vap->va_atime, &vap->va_mtime, 0);
if (error)
return (error);
}
error = 0;
if (vap->va_mode != (mode_t)VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
error = ufs_chmod(vp, (int)vap->va_mode, cred, p);
}
VN_KNOTE(vp, hint);
return (error);
}
/*
* Write a directory entry after a call to namei, using the parameters
* that it left in nameidata. The argument dirp is the new directory
* entry contents. Dvp is a pointer to the directory to be written,
* which was left locked by namei. Remaining parameters (dp->i_offset,
* dp->i_count) indicate how the space for the new entry is to be obtained.
* Non-null bp indicates that a directory is being created (for the
* soft dependency code).
*/
int
ufs_direnter(struct vnode *dvp, struct vnode *tvp, struct direct *dirp,
struct componentname *cnp, struct buf *newdirbp)
{
struct ucred *cr;
struct proc *p;
int newentrysize;
struct inode *dp;
struct buf *bp;
u_int dsize;
struct direct *ep, *nep;
int error, ret, blkoff, loc, spacefree, flags;
char *dirbuf;
UFS_WAPBL_JLOCK_ASSERT(dvp->v_mount);
error = 0;
cr = cnp->cn_cred;
p = cnp->cn_proc;
dp = VTOI(dvp);
newentrysize = DIRSIZ(FSFMT(dvp), dirp);
if (dp->i_count == 0) {
/*
* If dp->i_count is 0, then namei could find no
* space in the directory. Here, dp->i_offset will
* be on a directory block boundary and we will write the
* new entry into a fresh block.
*/
if (dp->i_offset & (DIRBLKSIZ - 1))
panic("ufs_direnter: newblk");
flags = B_CLRBUF;
if (!DOINGSOFTDEP(dvp))
flags |= B_SYNC;
if ((error = UFS_BUF_ALLOC(dp, (off_t)dp->i_offset, DIRBLKSIZ,
cr, flags, &bp)) != 0) {
if (DOINGSOFTDEP(dvp) && newdirbp != NULL)
bdwrite(newdirbp);
return (error);
}
DIP_ASSIGN(dp, size, dp->i_offset + DIRBLKSIZ);
dp->i_flag |= IN_CHANGE | IN_UPDATE;
uvm_vnp_setsize(dvp, DIP(dp, size));
dirp->d_reclen = DIRBLKSIZ;
blkoff = dp->i_offset &
(VFSTOUFS(dvp->v_mount)->um_mountp->mnt_stat.f_iosize - 1);
memcpy(bp->b_data + blkoff, dirp, newentrysize);
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL) {
ufsdirhash_newblk(dp, dp->i_offset);
ufsdirhash_add(dp, dirp, dp->i_offset);
ufsdirhash_checkblock(dp, (char *)bp->b_data + blkoff,
dp->i_offset);
}
#endif
if (DOINGSOFTDEP(dvp)) {
/*
* Ensure that the entire newly allocated block is a
* valid directory so that future growth within the
* block does not have to ensure that the block is
* written before the inode.
*/
blkoff += DIRBLKSIZ;
while (blkoff < bp->b_bcount) {
((struct direct *)
(bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
blkoff += DIRBLKSIZ;
}
if (softdep_setup_directory_add(bp, dp, dp->i_offset,
dirp->d_ino, newdirbp, 1) == 0) {
bdwrite(bp);
return (UFS_UPDATE(dp, 0));
}
/* We have just allocated a directory block in an
* indirect block. Rather than tracking when it gets
* claimed by the inode, we simply do a VOP_FSYNC
* now to ensure that it is there (in case the user
* does a future fsync). Note that we have to unlock
* the inode for the entry that we just entered, as
* the VOP_FSYNC may need to lock other inodes which
* can lead to deadlock if we also hold a lock on
* the newly entered node.
*/
if ((error = VOP_BWRITE(bp)))
return (error);
if (tvp != NULL)
VOP_UNLOCK(tvp, 0);
error = VOP_FSYNC(dvp, p->p_ucred, MNT_WAIT);
if (tvp != NULL)
vn_lock(tvp, LK_EXCLUSIVE | LK_RETRY, p);
return (error);
}
error = VOP_BWRITE(bp);
ret = UFS_UPDATE(dp, !DOINGSOFTDEP(dvp));
if (error == 0)
return (ret);
return (error);
}
/*
* If dp->i_count is non-zero, then namei found space for the new
* entry in the range dp->i_offset to dp->i_offset + dp->i_count
* in the directory. To use this space, we may have to compact
* the entries located there, by copying them together towards the
* beginning of the block, leaving the free space in one usable
* chunk at the end.
*/
/*
* Increase size of directory if entry eats into new space.
* This should never push the size past a new multiple of
* DIRBLKSIZE.
*
* N.B. - THIS IS AN ARTIFACT OF 4.2 AND SHOULD NEVER HAPPEN.
*/
if (dp->i_offset + dp->i_count > DIP(dp, size)) {
DIP_ASSIGN(dp, size, dp->i_offset + dp->i_count);
dp->i_flag |= IN_CHANGE | IN_UPDATE;
UFS_WAPBL_UPDATE(dp, MNT_WAIT);
}
/*
* Get the block containing the space for the new directory entry.
*/
if ((error = UFS_BUFATOFF(dp, (off_t)dp->i_offset, &dirbuf, &bp))
!= 0) {
if (DOINGSOFTDEP(dvp) && newdirbp != NULL)
bdwrite(newdirbp);
return (error);
}
/*
* Find space for the new entry. In the simple case, the entry at
* offset base will have the space. If it does not, then namei
* arranged that compacting the region dp->i_offset to
* dp->i_offset + dp->i_count would yield the space.
*/
ep = (struct direct *)dirbuf;
dsize = ep->d_ino ? DIRSIZ(FSFMT(dvp), ep) : 0;
spacefree = ep->d_reclen - dsize;
for (loc = ep->d_reclen; loc < dp->i_count; ) {
nep = (struct direct *)(dirbuf + loc);
/* Trim the existing slot (NB: dsize may be zero). */
ep->d_reclen = dsize;
ep = (struct direct *)((char *)ep + dsize);
/* Read nep->d_reclen now as the memmove() may clobber it. */
loc += nep->d_reclen;
if (nep->d_ino == 0) {
/*
* A mid-block unused entry. Such entries are
* never created by the kernel, but fsck_ffs
* can create them (and it doesn't fix them).
*
* Add up the free space, and initialise the
* relocated entry since we don't memmove it.
*/
spacefree += nep->d_reclen;
ep->d_ino = 0;
dsize = 0;
continue;
}
dsize = DIRSIZ(FSFMT(dvp), nep);
spacefree += nep->d_reclen - dsize;
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL)
ufsdirhash_move(dp, nep,
dp->i_offset + ((char *)nep - dirbuf),
dp->i_offset + ((char *)ep - dirbuf));
#endif
if (DOINGSOFTDEP(dvp))
softdep_change_directoryentry_offset(dp, dirbuf,
(caddr_t)nep, (caddr_t)ep, dsize);
else
memmove(ep, nep, dsize);
}
/*
* Here, `ep' points to a directory entry containing `dsize' in-use
* bytes followed by `spacefree' unused bytes. If ep->d_ino == 0,
* then the entry is completely unused (dsize == 0). The value
* of ep->d_reclen is always indeterminate.
*
* Update the pointer fields in the previous entry (if any),
* copy in the new entry, and write out the block.
*/
if (ep->d_ino == 0) {
if (spacefree + dsize < newentrysize)
panic("ufs_direnter: compact1");
dirp->d_reclen = spacefree + dsize;
} else {
if (spacefree < newentrysize)
panic("ufs_direnter: compact2");
dirp->d_reclen = spacefree;
ep->d_reclen = dsize;
ep = (struct direct *)((char *)ep + dsize);
}
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL && (ep->d_ino == 0 ||
dirp->d_reclen == spacefree))
ufsdirhash_add(dp, dirp, dp->i_offset + ((char *)ep - dirbuf));
#endif
memcpy(ep, dirp, newentrysize);
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL)
ufsdirhash_checkblock(dp, dirbuf -
(dp->i_offset & (DIRBLKSIZ - 1)),
dp->i_offset & ~(DIRBLKSIZ - 1));
#endif
if (DOINGSOFTDEP(dvp)) {
(void)softdep_setup_directory_add(bp, dp,
dp->i_offset + (caddr_t)ep - dirbuf,
dirp->d_ino, newdirbp, 0);
bdwrite(bp);
} else {
error = VOP_BWRITE(bp);
}
dp->i_flag |= IN_CHANGE | IN_UPDATE;
/*
* If all went well, and the directory can be shortened, proceed
* with the truncation. Note that we have to unlock the inode for
* the entry that we just entered, as the truncation may need to
* lock other inodes which can lead to deadlock if we also hold a
* lock on the newly entered node.
*/
if (error == 0 && dp->i_endoff && dp->i_endoff < DIP(dp, size)) {
if (tvp != NULL)
VOP_UNLOCK(tvp, 0);
#ifdef UFS_DIRHASH
if (dp->i_dirhash != NULL)
ufsdirhash_dirtrunc(dp, dp->i_endoff);
#endif
error = UFS_TRUNCATE(dp, (off_t)dp->i_endoff, IO_SYNC, cr);
if (tvp != NULL)
vn_lock(tvp, LK_EXCLUSIVE | LK_RETRY, p);
}
UFS_WAPBL_UPDATE(dp, MNT_WAIT);
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);
}
