int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
sa_handle_t *sa_hdl, dmu_tx_t *tx,
struct lu_attr *la, uint64_t parent)
{
sa_bulk_attr_t *bulk = osd_oti_get(env)->oti_attr_bulk;
struct osa_attr *osa = &osd_oti_get(env)->oti_osa;
uint64_t gen;
uint64_t crtime[2];
timestruc_t now;
int cnt;
int rc;
LASSERT(sa_hdl);
gen = dmu_tx_get_txg(tx);
gethrestime(&now);
ZFS_TIME_ENCODE(&now, crtime);
osa->atime[0] = la->la_atime;
osa->ctime[0] = la->la_ctime;
osa->mtime[0] = la->la_mtime;
osa->mode = la->la_mode;
osa->uid = la->la_uid;
osa->gid = la->la_gid;
osa->rdev = la->la_rdev;
osa->nlink = la->la_nlink;
osa->flags = attrs_fs2zfs(la->la_flags);
osa->size = la->la_size;
/*
* we need to create all SA below upon object create.
*
* XXX The attribute order matters since the accounting callback relies
* on static offsets (i.e. SA_*_OFFSET, see zfs_space_delta_cb()) to
* look up the UID/GID attributes. Moreover, the callback does not seem
* to support the spill block.
* We define attributes in the same order as SA_*_OFFSET in order to
* work around the problem. See ORI-610.
*/
cnt = 0;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(osd), NULL, &osa->mode, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(osd), NULL, &osa->size, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GEN(osd), NULL, &gen, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(osd), NULL, &osa->uid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(osd), NULL, &osa->gid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_PARENT(osd), NULL, &parent, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(osd), NULL, &osa->flags, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(osd), NULL, osa->atime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(osd), NULL, osa->mtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(osd), NULL, osa->ctime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CRTIME(osd), NULL, crtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(osd), NULL, &osa->nlink, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_RDEV(osd), NULL, &osa->rdev, 8);
LASSERT(cnt <= ARRAY_SIZE(osd_oti_get(env)->oti_attr_bulk));
rc = -sa_replace_all_by_template(sa_hdl, bulk, cnt, tx);
return rc;
}
/*
* Update the embedded inode given the znode. We should work toward
* eliminating this function as soon as possible by removing values
* which are duplicated between the znode and inode. If the generic
* inode has the correct field it should be used, and the ZFS code
* updated to access the inode. This can be done incrementally.
*/
void
zfs_inode_update(znode_t *zp)
{
zfs_sb_t *zsb;
struct inode *ip;
uint32_t blksize;
uint64_t atime[2], mtime[2], ctime[2];
ASSERT(zp != NULL);
zsb = ZTOZSB(zp);
ip = ZTOI(zp);
/* Skip .zfs control nodes which do not exist on disk. */
if (zfsctl_is_node(ip))
return;
sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16);
sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16);
sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16);
spin_lock(&ip->i_lock);
ip->i_generation = zp->z_gen;
ip->i_uid = SUID_TO_KUID(zp->z_uid);
ip->i_gid = SGID_TO_KGID(zp->z_gid);
set_nlink(ip, zp->z_links);
ip->i_mode = zp->z_mode;
zfs_set_inode_flags(zp, ip);
ip->i_blkbits = SPA_MINBLOCKSHIFT;
dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
(u_longlong_t *)&ip->i_blocks);
ZFS_TIME_DECODE(&ip->i_atime, atime);
ZFS_TIME_DECODE(&ip->i_mtime, mtime);
ZFS_TIME_DECODE(&ip->i_ctime, ctime);
i_size_write(ip, zp->z_size);
spin_unlock(&ip->i_lock);
}
void
zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
{
dmu_buf_t *db = sa_get_db(hdl);
znode_t *zp = sa_get_userdata(hdl);
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
sa_bulk_attr_t bulk[20];
int count = 0;
sa_bulk_attr_t sa_attrs[20] = { { 0 } };
zfs_acl_locator_cb_t locate = { 0 };
uint64_t uid, gid, mode, rdev, xattr, parent;
uint64_t crtime[2], mtime[2], ctime[2];
zfs_acl_phys_t znode_acl;
char scanstamp[AV_SCANSTAMP_SZ];
boolean_t drop_lock = B_FALSE;
/*
* No upgrade if ACL isn't cached
* since we won't know which locks are held
* and ready the ACL would require special "locked"
* interfaces that would be messy
*/
if (zp->z_acl_cached == NULL || vnode_islnk(ZTOV(zp)))
return;
/*
* If the z_lock is held and we aren't the owner
* the just return since we don't want to deadlock
* trying to update the status of z_is_sa. This
* file can then be upgraded at a later time.
*
* Otherwise, we know we are doing the
* sa_update() that caused us to enter this function.
*/
if (mutex_owner(&zp->z_lock) != curthread) {
if (mutex_tryenter(&zp->z_lock) == 0)
return;
else
drop_lock = B_TRUE;
}
/* First do a bulk query of the attributes that aren't cached */
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, &xattr, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
&znode_acl, 88);
if (sa_bulk_lookup_locked(hdl, bulk, count) != 0)
goto done;
/*
* While the order here doesn't matter its best to try and organize
* it is such a way to pick up an already existing layout number
*/
count = 0;
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
&zp->z_size, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zfsvfs),
NULL, &zp->z_gen, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zfsvfs),
NULL, &parent, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zfsvfs), NULL,
zp->z_atime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zfsvfs), NULL,
&mtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zfsvfs), NULL,
&ctime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
&crtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL,
&zp->z_links, 8);
if (vnode_isblk(zp->z_vnode) || vnode_islnk(zp->z_vnode))
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
&rdev, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
&zp->z_acl_cached->z_acl_count, 8);
if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
zfs_acl_xform(zp, zp->z_acl_cached, CRED());
locate.cb_aclp = zp->z_acl_cached;
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
if (xattr)
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zfsvfs),
NULL, &xattr, 8);
/* if scanstamp then add scanstamp */
if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
scanstamp, AV_SCANSTAMP_SZ);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zfsvfs),
NULL, scanstamp, AV_SCANSTAMP_SZ);
zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
}
VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
count, tx) == 0);
if (znode_acl.z_acl_extern_obj)
VERIFY(0 == dmu_object_free(zfsvfs->z_os,
znode_acl.z_acl_extern_obj, tx));
zp->z_is_sa = B_TRUE;
done:
if (drop_lock)
mutex_exit(&zp->z_lock);
}
/*
* Set the attributes of an object
*
* The transaction passed to this routine must have
* dmu_tx_hold_bonus(tx, oid) called and then assigned
* to a transaction group.
*/
static int osd_attr_set(const struct lu_env *env, struct dt_object *dt,
const struct lu_attr *la, struct thandle *handle,
struct lustre_capa *capa)
{
struct osd_object *obj = osd_dt_obj(dt);
struct osd_device *osd = osd_obj2dev(obj);
udmu_objset_t *uos = &osd->od_objset;
struct osd_thandle *oh;
struct osa_attr *osa = &osd_oti_get(env)->oti_osa;
sa_bulk_attr_t *bulk;
int cnt;
int rc = 0;
ENTRY;
LASSERT(handle != NULL);
LASSERT(dt_object_exists(dt));
LASSERT(osd_invariant(obj));
LASSERT(obj->oo_sa_hdl);
oh = container_of0(handle, struct osd_thandle, ot_super);
/* Assert that the transaction has been assigned to a
transaction group. */
LASSERT(oh->ot_tx->tx_txg != 0);
if (la->la_valid == 0)
RETURN(0);
OBD_ALLOC(bulk, sizeof(sa_bulk_attr_t) * 10);
if (bulk == NULL)
RETURN(-ENOMEM);
/* do both accounting updates outside oo_attr_lock below */
if ((la->la_valid & LA_UID) && (la->la_uid != obj->oo_attr.la_uid)) {
/* Update user accounting. Failure isn't fatal, but we still
* log an error message */
rc = -zap_increment_int(osd->od_objset.os, osd->od_iusr_oid,
la->la_uid, 1, oh->ot_tx);
if (rc)
CERROR("%s: failed to update accounting ZAP for user "
"%d (%d)\n", osd->od_svname, la->la_uid, rc);
rc = -zap_increment_int(osd->od_objset.os, osd->od_iusr_oid,
obj->oo_attr.la_uid, -1, oh->ot_tx);
if (rc)
CERROR("%s: failed to update accounting ZAP for user "
"%d (%d)\n", osd->od_svname,
obj->oo_attr.la_uid, rc);
}
if ((la->la_valid & LA_GID) && (la->la_gid != obj->oo_attr.la_gid)) {
/* Update group accounting. Failure isn't fatal, but we still
* log an error message */
rc = -zap_increment_int(osd->od_objset.os, osd->od_igrp_oid,
la->la_gid, 1, oh->ot_tx);
if (rc)
CERROR("%s: failed to update accounting ZAP for user "
"%d (%d)\n", osd->od_svname, la->la_gid, rc);
rc = -zap_increment_int(osd->od_objset.os, osd->od_igrp_oid,
obj->oo_attr.la_gid, -1, oh->ot_tx);
if (rc)
CERROR("%s: failed to update accounting ZAP for user "
"%d (%d)\n", osd->od_svname,
obj->oo_attr.la_gid, rc);
}
write_lock(&obj->oo_attr_lock);
cnt = 0;
if (la->la_valid & LA_ATIME) {
osa->atime[0] = obj->oo_attr.la_atime = la->la_atime;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(uos), NULL,
osa->atime, 16);
}
if (la->la_valid & LA_MTIME) {
osa->mtime[0] = obj->oo_attr.la_mtime = la->la_mtime;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(uos), NULL,
osa->mtime, 16);
}
if (la->la_valid & LA_CTIME) {
osa->ctime[0] = obj->oo_attr.la_ctime = la->la_ctime;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(uos), NULL,
osa->ctime, 16);
}
if (la->la_valid & LA_MODE) {
/* mode is stored along with type, so read it first */
obj->oo_attr.la_mode = (obj->oo_attr.la_mode & S_IFMT) |
(la->la_mode & ~S_IFMT);
osa->mode = obj->oo_attr.la_mode;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(uos), NULL,
&osa->mode, 8);
}
if (la->la_valid & LA_SIZE) {
osa->size = obj->oo_attr.la_size = la->la_size;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(uos), NULL,
&osa->size, 8);
}
if (la->la_valid & LA_NLINK) {
osa->nlink = obj->oo_attr.la_nlink = la->la_nlink;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(uos), NULL,
&osa->nlink, 8);
}
if (la->la_valid & LA_RDEV) {
osa->rdev = obj->oo_attr.la_rdev = la->la_rdev;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_RDEV(uos), NULL,
&osa->rdev, 8);
}
if (la->la_valid & LA_FLAGS) {
osa->flags = obj->oo_attr.la_flags = la->la_flags;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(uos), NULL,
&osa->flags, 8);
}
if (la->la_valid & LA_UID) {
osa->uid = obj->oo_attr.la_uid = la->la_uid;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(uos), NULL,
&osa->uid, 8);
}
if (la->la_valid & LA_GID) {
osa->gid = obj->oo_attr.la_gid = la->la_gid;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(uos), NULL,
&osa->gid, 8);
}
obj->oo_attr.la_valid |= la->la_valid;
write_unlock(&obj->oo_attr_lock);
rc = osd_object_sa_bulk_update(obj, bulk, cnt, oh);
OBD_FREE(bulk, sizeof(sa_bulk_attr_t) * 10);
RETURN(rc);
}
/*
* Retrieve the attributes of a DMU object
*/
int __osd_object_attr_get(const struct lu_env *env, udmu_objset_t *uos,
struct osd_object *obj, struct lu_attr *la)
{
struct osa_attr *osa = &osd_oti_get(env)->oti_osa;
sa_handle_t *sa_hdl;
sa_bulk_attr_t *bulk;
int cnt = 0;
int rc;
ENTRY;
LASSERT(obj->oo_db != NULL);
rc = -sa_handle_get(uos->os, obj->oo_db->db_object, NULL,
SA_HDL_PRIVATE, &sa_hdl);
if (rc)
RETURN(rc);
OBD_ALLOC(bulk, sizeof(sa_bulk_attr_t) * 9);
if (bulk == NULL)
GOTO(out_sa, rc = -ENOMEM);
la->la_valid |= LA_ATIME | LA_MTIME | LA_CTIME | LA_MODE | LA_TYPE |
LA_SIZE | LA_UID | LA_GID | LA_FLAGS | LA_NLINK;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(uos), NULL, osa->atime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(uos), NULL, osa->mtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(uos), NULL, osa->ctime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(uos), NULL, &osa->mode, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(uos), NULL, &osa->size, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(uos), NULL, &osa->nlink, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(uos), NULL, &osa->uid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(uos), NULL, &osa->gid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(uos), NULL, &osa->flags, 8);
rc = -sa_bulk_lookup(sa_hdl, bulk, cnt);
if (rc)
GOTO(out_bulk, rc);
la->la_atime = osa->atime[0];
la->la_mtime = osa->mtime[0];
la->la_ctime = osa->ctime[0];
la->la_mode = osa->mode;
la->la_uid = osa->uid;
la->la_gid = osa->gid;
la->la_nlink = osa->nlink;
la->la_flags = osa->flags;
la->la_size = osa->size;
if (S_ISCHR(la->la_mode) || S_ISBLK(la->la_mode)) {
rc = -sa_lookup(sa_hdl, SA_ZPL_RDEV(uos), &osa->rdev, 8);
if (rc)
GOTO(out_bulk, rc);
la->la_rdev = osa->rdev;
la->la_valid |= LA_RDEV;
}
out_bulk:
OBD_FREE(bulk, sizeof(sa_bulk_attr_t) * 9);
out_sa:
sa_handle_destroy(sa_hdl);
RETURN(rc);
}
int __osd_attr_init(const struct lu_env *env, udmu_objset_t *uos,
uint64_t oid, dmu_tx_t *tx, struct lu_attr *la)
{
sa_bulk_attr_t *bulk;
sa_handle_t *sa_hdl;
struct osa_attr *osa = &osd_oti_get(env)->oti_osa;
uint64_t gen;
uint64_t parent;
uint64_t crtime[2];
timestruc_t now;
int cnt;
int rc;
gethrestime(&now);
gen = dmu_tx_get_txg(tx);
ZFS_TIME_ENCODE(&now, crtime);
/* XXX: this should be real id of parent for ZPL access, but we have no
* such info in OSD, probably it can be part of dt_object_format */
parent = 0;
osa->atime[0] = la->la_atime;
osa->ctime[0] = la->la_ctime;
osa->mtime[0] = la->la_mtime;
osa->mode = la->la_mode;
osa->uid = la->la_uid;
osa->gid = la->la_gid;
osa->rdev = la->la_rdev;
osa->nlink = la->la_nlink;
osa->flags = la->la_flags;
osa->size = la->la_size;
/* Now add in all of the "SA" attributes */
rc = -sa_handle_get(uos->os, oid, NULL, SA_HDL_PRIVATE, &sa_hdl);
if (rc)
return rc;
OBD_ALLOC(bulk, sizeof(sa_bulk_attr_t) * 13);
if (bulk == NULL) {
rc = -ENOMEM;
goto out;
}
/*
* we need to create all SA below upon object create.
*
* XXX The attribute order matters since the accounting callback relies
* on static offsets (i.e. SA_*_OFFSET, see zfs_space_delta_cb()) to
* look up the UID/GID attributes. Moreover, the callback does not seem
* to support the spill block.
* We define attributes in the same order as SA_*_OFFSET in order to
* work around the problem. See ORI-610.
*/
cnt = 0;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(uos), NULL, &osa->mode, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(uos), NULL, &osa->size, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GEN(uos), NULL, &gen, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(uos), NULL, &osa->uid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(uos), NULL, &osa->gid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_PARENT(uos), NULL, &parent, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(uos), NULL, &osa->flags, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(uos), NULL, osa->atime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(uos), NULL, osa->mtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(uos), NULL, osa->ctime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CRTIME(uos), NULL, crtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(uos), NULL, &osa->nlink, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_RDEV(uos), NULL, &osa->rdev, 8);
rc = -sa_replace_all_by_template(sa_hdl, bulk, cnt, tx);
OBD_FREE(bulk, sizeof(sa_bulk_attr_t) * 13);
out:
sa_handle_destroy(sa_hdl);
return rc;
}
/*
* Create a new DMU object to hold a zfs znode.
*
* IN: dzp - parent directory for new znode
* vap - file attributes for new znode
* tx - dmu transaction id for zap operations
* cr - credentials of caller
* flag - flags:
* IS_ROOT_NODE - new object will be root
* IS_XATTR - new object is an attribute
* bonuslen - length of bonus buffer
* setaclp - File/Dir initial ACL
* fuidp - Tracks fuid allocation.
*
* OUT: zpp - allocated znode
*
*/
void
zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
{
uint64_t crtime[2], atime[2], mtime[2], ctime[2];
uint64_t mode, size, links, parent, pflags;
uint64_t dzp_pflags = 0;
uint64_t rdev = 0;
zfs_sb_t *zsb = ZTOZSB(dzp);
dmu_buf_t *db;
timestruc_t now;
uint64_t gen, obj;
int bonuslen;
sa_handle_t *sa_hdl;
dmu_object_type_t obj_type;
sa_bulk_attr_t *sa_attrs;
int cnt = 0;
zfs_acl_locator_cb_t locate = { 0 };
if (zsb->z_replay) {
obj = vap->va_nodeid;
now = vap->va_ctime; /* see zfs_replay_create() */
gen = vap->va_nblocks; /* ditto */
} else {
obj = 0;
gethrestime(&now);
gen = dmu_tx_get_txg(tx);
}
obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
bonuslen = (obj_type == DMU_OT_SA) ?
DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
/*
* Create a new DMU object.
*/
/*
* There's currently no mechanism for pre-reading the blocks that will
* be needed to allocate a new object, so we accept the small chance
* that there will be an i/o error and we will fail one of the
* assertions below.
*/
if (S_ISDIR(vap->va_mode)) {
if (zsb->z_replay) {
VERIFY0(zap_create_claim_norm(zsb->z_os, obj,
zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
obj_type, bonuslen, tx));
} else {
obj = zap_create_norm(zsb->z_os,
zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
obj_type, bonuslen, tx);
}
} else {
if (zsb->z_replay) {
VERIFY0(dmu_object_claim(zsb->z_os, obj,
DMU_OT_PLAIN_FILE_CONTENTS, 0,
obj_type, bonuslen, tx));
} else {
obj = dmu_object_alloc(zsb->z_os,
DMU_OT_PLAIN_FILE_CONTENTS, 0,
obj_type, bonuslen, tx);
}
}
ZFS_OBJ_HOLD_ENTER(zsb, obj);
VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db));
/*
* If this is the root, fix up the half-initialized parent pointer
* to reference the just-allocated physical data area.
*/
if (flag & IS_ROOT_NODE) {
dzp->z_id = obj;
} else {
dzp_pflags = dzp->z_pflags;
}
/*
* If parent is an xattr, so am I.
*/
if (dzp_pflags & ZFS_XATTR) {
flag |= IS_XATTR;
}
if (zsb->z_use_fuids)
pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
else
pflags = 0;
if (S_ISDIR(vap->va_mode)) {
size = 2; /* contents ("." and "..") */
links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
} else {
size = links = 0;
}
if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))
rdev = vap->va_rdev;
parent = dzp->z_id;
mode = acl_ids->z_mode;
if (flag & IS_XATTR)
pflags |= ZFS_XATTR;
/*
* No execs denied will be deterimed when zfs_mode_compute() is called.
*/
pflags |= acl_ids->z_aclp->z_hints &
(ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
ZFS_TIME_ENCODE(&now, crtime);
ZFS_TIME_ENCODE(&now, ctime);
if (vap->va_mask & ATTR_ATIME) {
ZFS_TIME_ENCODE(&vap->va_atime, atime);
} else {
ZFS_TIME_ENCODE(&now, atime);
}
if (vap->va_mask & ATTR_MTIME) {
ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
} else {
ZFS_TIME_ENCODE(&now, mtime);
}
/* Now add in all of the "SA" attributes */
VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED,
&sa_hdl));
/*
* Setup the array of attributes to be replaced/set on the new file
*
* order for DMU_OT_ZNODE is critical since it needs to be constructed
* in the old znode_phys_t format. Don't change this ordering
*/
sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
if (obj_type == DMU_OT_ZNODE) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
NULL, &atime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
NULL, &mtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
NULL, &ctime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
NULL, &crtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
NULL, &gen, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
NULL, &mode, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
NULL, &size, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
NULL, &parent, 8);
} else {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
NULL, &mode, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
NULL, &size, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
NULL, &gen, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb),
NULL, &acl_ids->z_fuid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb),
NULL, &acl_ids->z_fgid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
NULL, &parent, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
NULL, &pflags, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
NULL, &atime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
NULL, &mtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
NULL, &ctime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
NULL, &crtime, 16);
}
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8);
if (obj_type == DMU_OT_ZNODE) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL,
&empty_xattr, 8);
}
if (obj_type == DMU_OT_ZNODE ||
(S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb),
NULL, &rdev, 8);
}
if (obj_type == DMU_OT_ZNODE) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
NULL, &pflags, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL,
&acl_ids->z_fuid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL,
&acl_ids->z_fgid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad,
sizeof (uint64_t) * 4);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL,
&acl_phys, sizeof (zfs_acl_phys_t));
} else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL,
&acl_ids->z_aclp->z_acl_count, 8);
locate.cb_aclp = acl_ids->z_aclp;
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb),
zfs_acl_data_locator, &locate,
acl_ids->z_aclp->z_acl_bytes);
mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
acl_ids->z_fuid, acl_ids->z_fgid);
}
VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
if (!(flag & IS_ROOT_NODE)) {
*zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl,
ZTOI(dzp));
VERIFY(*zpp != NULL);
VERIFY(dzp != NULL);
} else {
/*
* If we are creating the root node, the "parent" we
* passed in is the znode for the root.
*/
*zpp = dzp;
(*zpp)->z_sa_hdl = sa_hdl;
}
(*zpp)->z_pflags = pflags;
(*zpp)->z_mode = mode;
if (obj_type == DMU_OT_ZNODE ||
acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
}
kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
ZFS_OBJ_HOLD_EXIT(zsb, obj);
}
/*
* Construct a znode+inode and initialize.
*
* This does not do a call to dmu_set_user() that is
* up to the caller to do, in case you don't want to
* return the znode
*/
static znode_t *
zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl,
struct inode *dip)
{
znode_t *zp;
struct inode *ip;
uint64_t mode;
uint64_t parent;
sa_bulk_attr_t bulk[9];
int count = 0;
ASSERT(zsb != NULL);
ip = new_inode(zsb->z_sb);
if (ip == NULL)
return (NULL);
zp = ITOZ(ip);
ASSERT(zp->z_dirlocks == NULL);
ASSERT3P(zp->z_acl_cached, ==, NULL);
ASSERT3P(zp->z_xattr_cached, ==, NULL);
ASSERT3P(zp->z_xattr_parent, ==, NULL);
zp->z_moved = 0;
zp->z_sa_hdl = NULL;
zp->z_unlinked = 0;
zp->z_atime_dirty = 0;
zp->z_mapcnt = 0;
zp->z_id = db->db_object;
zp->z_blksz = blksz;
zp->z_seq = 0x7A4653;
zp->z_sync_cnt = 0;
zp->z_is_zvol = B_FALSE;
zp->z_is_mapped = B_FALSE;
zp->z_is_ctldir = B_FALSE;
zp->z_is_stale = B_FALSE;
zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
&zp->z_pflags, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
&parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
&zp->z_atime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
if (hdl == NULL)
sa_handle_destroy(zp->z_sa_hdl);
goto error;
}
zp->z_mode = mode;
/*
* xattr znodes hold a reference on their unique parent
*/
if (dip && zp->z_pflags & ZFS_XATTR) {
igrab(dip);
zp->z_xattr_parent = ITOZ(dip);
}
ip->i_ino = obj;
zfs_inode_update(zp);
zfs_inode_set_ops(zsb, ip);
/*
* The only way insert_inode_locked() can fail is if the ip->i_ino
* number is already hashed for this super block. This can never
* happen because the inode numbers map 1:1 with the object numbers.
*
* The one exception is rolling back a mounted file system, but in
* this case all the active inode are unhashed during the rollback.
*/
VERIFY3S(insert_inode_locked(ip), ==, 0);
mutex_enter(&zsb->z_znodes_lock);
list_insert_tail(&zsb->z_all_znodes, zp);
zsb->z_nr_znodes++;
membar_producer();
mutex_exit(&zsb->z_znodes_lock);
unlock_new_inode(ip);
return (zp);
error:
unlock_new_inode(ip);
iput(ip);
return (NULL);
}
int
zfs_rezget(znode_t *zp)
{
zfs_sb_t *zsb = ZTOZSB(zp);
dmu_object_info_t doi;
dmu_buf_t *db;
uint64_t obj_num = zp->z_id;
uint64_t mode;
sa_bulk_attr_t bulk[8];
int err;
int count = 0;
uint64_t gen;
ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
mutex_enter(&zp->z_acl_lock);
if (zp->z_acl_cached) {
zfs_acl_free(zp->z_acl_cached);
zp->z_acl_cached = NULL;
}
mutex_exit(&zp->z_acl_lock);
rw_enter(&zp->z_xattr_lock, RW_WRITER);
if (zp->z_xattr_cached) {
nvlist_free(zp->z_xattr_cached);
zp->z_xattr_cached = NULL;
}
if (zp->z_xattr_parent) {
iput(ZTOI(zp->z_xattr_parent));
zp->z_xattr_parent = NULL;
}
rw_exit(&zp->z_xattr_lock);
ASSERT(zp->z_sa_hdl == NULL);
err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
if (err) {
ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
return (err);
}
dmu_object_info_from_db(db, &doi);
if (doi.doi_bonus_type != DMU_OT_SA &&
(doi.doi_bonus_type != DMU_OT_ZNODE ||
(doi.doi_bonus_type == DMU_OT_ZNODE &&
doi.doi_bonus_size < sizeof (znode_phys_t)))) {
sa_buf_rele(db, NULL);
ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
return (SET_ERROR(EINVAL));
}
zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL);
/* reload cached values */
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL,
&gen, sizeof (gen));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL,
&zp->z_size, sizeof (zp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL,
&zp->z_links, sizeof (zp->z_links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
&zp->z_atime, sizeof (zp->z_atime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL,
&zp->z_uid, sizeof (zp->z_uid));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL,
&zp->z_gid, sizeof (zp->z_gid));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
&mode, sizeof (mode));
if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
zfs_znode_dmu_fini(zp);
ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
return (SET_ERROR(EIO));
}
zp->z_mode = mode;
if (gen != zp->z_gen) {
zfs_znode_dmu_fini(zp);
ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
return (SET_ERROR(EIO));
}
zp->z_unlinked = (zp->z_links == 0);
zp->z_blksz = doi.doi_data_block_size;
zfs_inode_update(zp);
ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
return (0);
}
void commonattrpack(attrinfo_t *aip, zfsvfs_t *zfsvfs, znode_t *zp,
const char *name, ino64_t objnum, enum vtype vtype,
boolean_t user64)
{
attrgroup_t commonattr = aip->ai_attrlist->commonattr;
void *attrbufptr = *aip->ai_attrbufpp;
void *varbufptr = *aip->ai_varbufpp;
struct mount *mp = zfsvfs->z_vfs;
cred_t *cr = (cred_t *)vfs_context_ucred(aip->ai_context);
finderinfo_t finderinfo;
/*
* We should probably combine all the sa_lookup into a bulk
* lookup operand.
*/
finderinfo.fi_flags = 0;
if (ATTR_CMN_NAME & commonattr) {
nameattrpack(aip, name, strlen(name));
attrbufptr = *aip->ai_attrbufpp;
varbufptr = *aip->ai_varbufpp;
}
if (ATTR_CMN_DEVID & commonattr) {
*((dev_t *)attrbufptr) = vfs_statfs(mp)->f_fsid.val[0];
attrbufptr = ((dev_t *)attrbufptr) + 1;
}
if (ATTR_CMN_FSID & commonattr) {
*((fsid_t *)attrbufptr) = vfs_statfs(mp)->f_fsid;
attrbufptr = ((fsid_t *)attrbufptr) + 1;
}
if (ATTR_CMN_OBJTYPE & commonattr) {
*((fsobj_type_t *)attrbufptr) = vtype;
attrbufptr = ((fsobj_type_t *)attrbufptr) + 1;
}
if (ATTR_CMN_OBJTAG & commonattr) {
*((fsobj_tag_t *)attrbufptr) = VT_ZFS;
attrbufptr = ((fsobj_tag_t *)attrbufptr) + 1;
}
/*
* Note: ATTR_CMN_OBJID is lossy (only 32 bits).
*/
if ((ATTR_CMN_OBJID | ATTR_CMN_OBJPERMANENTID) & commonattr) {
u_int32_t fileid;
/*
* On Mac OS X we always export the root directory id as 2
*/
fileid = (objnum == zfsvfs->z_root) ? 2 : objnum;
if (ATTR_CMN_OBJID & commonattr) {
((fsobj_id_t *)attrbufptr)->fid_objno = fileid;
((fsobj_id_t *)attrbufptr)->fid_generation = 0;
attrbufptr = ((fsobj_id_t *)attrbufptr) + 1;
}
if (ATTR_CMN_OBJPERMANENTID & commonattr) {
((fsobj_id_t *)attrbufptr)->fid_objno = fileid;
((fsobj_id_t *)attrbufptr)->fid_generation = 0;
attrbufptr = ((fsobj_id_t *)attrbufptr) + 1;
}
}
/*
* Note: ATTR_CMN_PAROBJID is lossy (only 32 bits).
*/
if (ATTR_CMN_PAROBJID & commonattr) {
uint64_t parentid;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&parentid, sizeof (parentid)) == 0);
/*
* On Mac OS X we always export the root
* directory id as 2 and its parent as 1
*/
if (zp && zp->z_id == zfsvfs->z_root)
parentid = 1;
else if (parentid == zfsvfs->z_root)
parentid = 2;
ASSERT(parentid != 0);
((fsobj_id_t *)attrbufptr)->fid_objno = (uint32_t)parentid;
((fsobj_id_t *)attrbufptr)->fid_generation = 0;
attrbufptr = ((fsobj_id_t *)attrbufptr) + 1;
}
if (ATTR_CMN_SCRIPT & commonattr) {
*((text_encoding_t *)attrbufptr) = kTextEncodingMacUnicode;
attrbufptr = ((text_encoding_t *)attrbufptr) + 1;
}
if (ATTR_CMN_CRTIME & commonattr) {
uint64_t times[2];
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs),
times, sizeof(times)) == 0);
if (user64) {
ZFS_TIME_DECODE((timespec_user64_t *)attrbufptr,
times);
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
ZFS_TIME_DECODE((timespec_user32_t *)attrbufptr,
times);
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_MODTIME & commonattr) {
uint64_t times[2];
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zfsvfs),
times, sizeof(times)) == 0);
if (user64) {
ZFS_TIME_DECODE((timespec_user64_t *)attrbufptr,
times);
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
ZFS_TIME_DECODE((timespec_user32_t *)attrbufptr,
times);
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_CHGTIME & commonattr) {
uint64_t times[2];
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zfsvfs),
times, sizeof(times)) == 0);
if (user64) {
ZFS_TIME_DECODE((timespec_user64_t *)attrbufptr,
times);
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
ZFS_TIME_DECODE((timespec_user32_t *)attrbufptr,
times);
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_ACCTIME & commonattr) {
uint64_t times[2];
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs),
times, sizeof(times)) == 0);
if (user64) {
ZFS_TIME_DECODE((timespec_user64_t *)attrbufptr,
times);
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
ZFS_TIME_DECODE((timespec_user32_t *)attrbufptr,
times);
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_BKUPTIME & commonattr) {
/* legacy attribute -- just pass zero */
if (user64) {
((timespec_user64_t *)attrbufptr)->tv_sec = 0;
((timespec_user64_t *)attrbufptr)->tv_nsec = 0;
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
((timespec_user32_t *)attrbufptr)->tv_sec = 0;
((timespec_user32_t *)attrbufptr)->tv_nsec = 0;
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_FNDRINFO & commonattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
&val, sizeof(val)) == 0);
getfinderinfo(zp, cr, &finderinfo);
/* Shadow ZFS_HIDDEN to Finder Info's invisible bit */
if (val & ZFS_HIDDEN) {
finderinfo.fi_flags |=
OSSwapHostToBigConstInt16(kIsInvisible);
}
bcopy(&finderinfo, attrbufptr, sizeof (finderinfo));
attrbufptr = (char *)attrbufptr + 32;
}
if (ATTR_CMN_OWNERID & commonattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_UID(zfsvfs),
&val, sizeof(val)) == 0);
*((uid_t *)attrbufptr) = val;
attrbufptr = ((uid_t *)attrbufptr) + 1;
}
if (ATTR_CMN_GRPID & commonattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_GID(zfsvfs),
&val, sizeof(val)) == 0);
*((gid_t *)attrbufptr) = val;
attrbufptr = ((gid_t *)attrbufptr) + 1;
}
if (ATTR_CMN_ACCESSMASK & commonattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs),
&val, sizeof(val)) == 0);
*((u_int32_t *)attrbufptr) = val;
attrbufptr = ((u_int32_t *)attrbufptr) + 1;
}
if (ATTR_CMN_FLAGS & commonattr) {
u_int32_t flags = zfs_getbsdflags(zp);
/* Shadow Finder Info's invisible bit to UF_HIDDEN */
if ((ATTR_CMN_FNDRINFO & commonattr) &&
(OSSwapBigToHostInt16(finderinfo.fi_flags) & kIsInvisible))
flags |= UF_HIDDEN;
*((u_int32_t *)attrbufptr) = flags;
attrbufptr = ((u_int32_t *)attrbufptr) + 1;
}
if (ATTR_CMN_USERACCESS & commonattr) {
u_int32_t user_access = 0;
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
&val, sizeof(val)) == 0);
user_access = getuseraccess(zp, aip->ai_context);
/* Also consider READ-ONLY file system. */
if (vfs_flags(mp) & MNT_RDONLY) {
user_access &= ~W_OK;
}
/* Locked objects are not writable either */
if ((val & ZFS_IMMUTABLE) &&
(vfs_context_suser(aip->ai_context) != 0)) {
user_access &= ~W_OK;
}
*((u_int32_t *)attrbufptr) = user_access;
attrbufptr = ((u_int32_t *)attrbufptr) + 1;
}
if (ATTR_CMN_FILEID & commonattr) {
/*
* On Mac OS X we always export the root directory id as 2
*/
if (objnum == zfsvfs->z_root)
objnum = 2;
*((u_int64_t *)attrbufptr) = objnum;
attrbufptr = ((u_int64_t *)attrbufptr) + 1;
}
if (ATTR_CMN_PARENTID & commonattr) {
uint64_t parentid;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&parentid, sizeof (parentid)) == 0);
/*
* On Mac OS X we always export the root
* directory id as 2 and its parent as 1
*/
if (zp && zp->z_id == zfsvfs->z_root)
parentid = 1;
else if (parentid == zfsvfs->z_root)
parentid = 2;
ASSERT(parentid != 0);
*((u_int64_t *)attrbufptr) = parentid;
attrbufptr = ((u_int64_t *)attrbufptr) + 1;
}
*aip->ai_attrbufpp = attrbufptr;
*aip->ai_varbufpp = varbufptr;
}
/* For part 1 of zfs_getattr() */
int
zfs_getattr_znode_locked(vattr_t *vap, znode_t *zp, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
int error;
uint64_t times[2];
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_mode = val & MODEMASK;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_UID(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_uid = val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_GID(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_gid = val;
//vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
/* On OS X, the root directory id is always 2 */
vap->va_fileid = (zp->z_id == zfsvfs->z_root) ? 2 : zp->z_id;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_nlink = val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_data_size = val;
vap->va_total_size = val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_rdev = val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_gen = val;
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs),
times, sizeof (times));
ZFS_TIME_DECODE(&vap->va_create_time, times);
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs),
times, sizeof (times));
ZFS_TIME_DECODE(&vap->va_access_time, times);
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zfsvfs),
times, sizeof (times));
ZFS_TIME_DECODE(&vap->va_modify_time, times);
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zfsvfs),
times, sizeof (times));
ZFS_TIME_DECODE(&vap->va_change_time, times);
if (VATTR_IS_ACTIVE(vap, va_backup_time)) {
vap->va_backup_time.tv_sec = 0;
vap->va_backup_time.tv_nsec = 0;
VATTR_SET_SUPPORTED(vap, va_backup_time);
}
vap->va_flags = zfs_getbsdflags(zp);
/* On OS X, the root directory id is always 2 and its parent is 1 */
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&val, sizeof (val)) == 0);
if (zp->z_id == zfsvfs->z_root)
vap->va_parentid = 1;
else if (val == zfsvfs->z_root)
vap->va_parentid = 2;
else
vap->va_parentid = val;
vap->va_iosize = zp->z_blksz ? zp->z_blksz : zfsvfs->z_max_blksz;
VATTR_SET_SUPPORTED(vap, va_iosize);
printf("stat blksize set to %d\n", vap->va_iosize);
vap->va_supported |= ZFS_SUPPORTED_VATTRS;
if (VATTR_IS_ACTIVE(vap, va_nchildren) && vnode_isdir(ZTOV(zp)))
VATTR_RETURN(vap, va_nchildren, vap->va_nlink - 2);
if (VATTR_IS_ACTIVE(vap, va_acl)) {
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
times, sizeof (times)))) {
// if (zp->z_phys->zp_acl.z_acl_count == 0) {
vap->va_acl = (kauth_acl_t) KAUTH_FILESEC_NONE;
} else {
error = zfs_getacl(zp, &vap->va_acl, B_TRUE, cr);
if (error)
return (error);
VATTR_SET_SUPPORTED(vap, va_acl);
/*
* va_acl implies that va_uuuid and va_guuid are
* also supported.
*/
VATTR_RETURN(vap, va_uuuid, kauth_null_guid);
VATTR_RETURN(vap, va_guuid, kauth_null_guid);
}
}
return (0);
}
/*
* Set the attributes of an object
*
* The transaction passed to this routine must have
* dmu_tx_hold_bonus(tx, oid) called and then assigned
* to a transaction group.
*/
static int osd_attr_set(const struct lu_env *env, struct dt_object *dt,
const struct lu_attr *la, struct thandle *handle)
{
struct osd_thread_info *info = osd_oti_get(env);
sa_bulk_attr_t *bulk = osd_oti_get(env)->oti_attr_bulk;
struct osd_object *obj = osd_dt_obj(dt);
struct osd_device *osd = osd_obj2dev(obj);
struct osd_thandle *oh;
struct osa_attr *osa = &info->oti_osa;
__u64 valid = la->la_valid;
int cnt;
int rc = 0;
ENTRY;
down_read(&obj->oo_guard);
if (unlikely(!dt_object_exists(dt) || obj->oo_destroyed))
GOTO(out, rc = -ENOENT);
LASSERT(handle != NULL);
LASSERT(osd_invariant(obj));
LASSERT(obj->oo_sa_hdl);
oh = container_of0(handle, struct osd_thandle, ot_super);
/* Assert that the transaction has been assigned to a
transaction group. */
LASSERT(oh->ot_tx->tx_txg != 0);
/* Only allow set size for regular file */
if (!S_ISREG(dt->do_lu.lo_header->loh_attr))
valid &= ~(LA_SIZE | LA_BLOCKS);
if (valid & LA_CTIME && la->la_ctime == obj->oo_attr.la_ctime)
valid &= ~LA_CTIME;
if (valid & LA_MTIME && la->la_mtime == obj->oo_attr.la_mtime)
valid &= ~LA_MTIME;
if (valid & LA_ATIME && la->la_atime == obj->oo_attr.la_atime)
valid &= ~LA_ATIME;
if (valid == 0)
GOTO(out, rc = 0);
if (valid & LA_FLAGS) {
struct lustre_mdt_attrs *lma;
struct lu_buf buf;
if (la->la_flags & LUSTRE_LMA_FL_MASKS) {
CLASSERT(sizeof(info->oti_buf) >= sizeof(*lma));
lma = (struct lustre_mdt_attrs *)&info->oti_buf;
buf.lb_buf = lma;
buf.lb_len = sizeof(info->oti_buf);
rc = osd_xattr_get(env, &obj->oo_dt, &buf,
XATTR_NAME_LMA);
if (rc > 0) {
lma->lma_incompat =
le32_to_cpu(lma->lma_incompat);
lma->lma_incompat |=
lustre_to_lma_flags(la->la_flags);
lma->lma_incompat =
cpu_to_le32(lma->lma_incompat);
buf.lb_buf = lma;
buf.lb_len = sizeof(*lma);
rc = osd_xattr_set_internal(env, obj, &buf,
XATTR_NAME_LMA,
LU_XATTR_REPLACE,
oh);
}
if (rc < 0) {
CWARN("%s: failed to set LMA flags: rc = %d\n",
osd->od_svname, rc);
RETURN(rc);
}
}
}
/* do both accounting updates outside oo_attr_lock below */
if ((valid & LA_UID) && (la->la_uid != obj->oo_attr.la_uid)) {
/* Update user accounting. Failure isn't fatal, but we still
* log an error message */
rc = -zap_increment_int(osd->od_os, osd->od_iusr_oid,
la->la_uid, 1, oh->ot_tx);
if (rc)
CERROR("%s: failed to update accounting ZAP for user "
"%d (%d)\n", osd->od_svname, la->la_uid, rc);
rc = -zap_increment_int(osd->od_os, osd->od_iusr_oid,
obj->oo_attr.la_uid, -1, oh->ot_tx);
if (rc)
CERROR("%s: failed to update accounting ZAP for user "
"%d (%d)\n", osd->od_svname,
obj->oo_attr.la_uid, rc);
}
if ((valid & LA_GID) && (la->la_gid != obj->oo_attr.la_gid)) {
/* Update group accounting. Failure isn't fatal, but we still
* log an error message */
rc = -zap_increment_int(osd->od_os, osd->od_igrp_oid,
la->la_gid, 1, oh->ot_tx);
if (rc)
CERROR("%s: failed to update accounting ZAP for user "
"%d (%d)\n", osd->od_svname, la->la_gid, rc);
rc = -zap_increment_int(osd->od_os, osd->od_igrp_oid,
obj->oo_attr.la_gid, -1, oh->ot_tx);
if (rc)
CERROR("%s: failed to update accounting ZAP for user "
"%d (%d)\n", osd->od_svname,
obj->oo_attr.la_gid, rc);
}
write_lock(&obj->oo_attr_lock);
cnt = 0;
if (valid & LA_ATIME) {
osa->atime[0] = obj->oo_attr.la_atime = la->la_atime;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(osd), NULL,
osa->atime, 16);
}
if (valid & LA_MTIME) {
osa->mtime[0] = obj->oo_attr.la_mtime = la->la_mtime;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(osd), NULL,
osa->mtime, 16);
}
if (valid & LA_CTIME) {
osa->ctime[0] = obj->oo_attr.la_ctime = la->la_ctime;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(osd), NULL,
osa->ctime, 16);
}
if (valid & LA_MODE) {
/* mode is stored along with type, so read it first */
obj->oo_attr.la_mode = (obj->oo_attr.la_mode & S_IFMT) |
(la->la_mode & ~S_IFMT);
osa->mode = obj->oo_attr.la_mode;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(osd), NULL,
&osa->mode, 8);
}
if (valid & LA_SIZE) {
osa->size = obj->oo_attr.la_size = la->la_size;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(osd), NULL,
&osa->size, 8);
}
if (valid & LA_NLINK) {
osa->nlink = obj->oo_attr.la_nlink = la->la_nlink;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(osd), NULL,
&osa->nlink, 8);
}
if (valid & LA_RDEV) {
osa->rdev = obj->oo_attr.la_rdev = la->la_rdev;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_RDEV(osd), NULL,
&osa->rdev, 8);
}
if (valid & LA_FLAGS) {
osa->flags = attrs_fs2zfs(la->la_flags);
/* many flags are not supported by zfs, so ensure a good cached
* copy */
obj->oo_attr.la_flags = attrs_zfs2fs(osa->flags);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(osd), NULL,
&osa->flags, 8);
}
if (valid & LA_UID) {
osa->uid = obj->oo_attr.la_uid = la->la_uid;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(osd), NULL,
&osa->uid, 8);
}
if (valid & LA_GID) {
osa->gid = obj->oo_attr.la_gid = la->la_gid;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(osd), NULL,
&osa->gid, 8);
}
obj->oo_attr.la_valid |= valid;
write_unlock(&obj->oo_attr_lock);
LASSERT(cnt <= ARRAY_SIZE(osd_oti_get(env)->oti_attr_bulk));
rc = osd_object_sa_bulk_update(obj, bulk, cnt, oh);
out:
up_read(&obj->oo_guard);
RETURN(rc);
}
/*
* Retrieve the attributes of a DMU object
*/
int __osd_object_attr_get(const struct lu_env *env, struct osd_device *o,
struct osd_object *obj, struct lu_attr *la)
{
struct osa_attr *osa = &osd_oti_get(env)->oti_osa;
sa_bulk_attr_t *bulk = osd_oti_get(env)->oti_attr_bulk;
sa_handle_t *sa_hdl;
int cnt = 0;
int rc;
ENTRY;
LASSERT(obj->oo_db != NULL);
rc = -sa_handle_get(o->od_os, obj->oo_db->db_object, NULL,
SA_HDL_PRIVATE, &sa_hdl);
if (rc)
RETURN(rc);
la->la_valid |= LA_ATIME | LA_MTIME | LA_CTIME | LA_MODE | LA_TYPE |
LA_SIZE | LA_UID | LA_GID | LA_FLAGS | LA_NLINK;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(o), NULL, osa->atime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(o), NULL, osa->mtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(o), NULL, osa->ctime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(o), NULL, &osa->mode, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(o), NULL, &osa->size, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(o), NULL, &osa->nlink, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(o), NULL, &osa->uid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(o), NULL, &osa->gid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(o), NULL, &osa->flags, 8);
LASSERT(cnt <= ARRAY_SIZE(osd_oti_get(env)->oti_attr_bulk));
rc = -sa_bulk_lookup(sa_hdl, bulk, cnt);
if (rc)
GOTO(out_sa, rc);
la->la_atime = osa->atime[0];
la->la_mtime = osa->mtime[0];
la->la_ctime = osa->ctime[0];
la->la_mode = osa->mode;
la->la_uid = osa->uid;
la->la_gid = osa->gid;
la->la_nlink = osa->nlink;
la->la_flags = attrs_zfs2fs(osa->flags);
la->la_size = osa->size;
/* Try to get extra flag from LMA. Right now, only LMAI_ORPHAN
* flags is stored in LMA, and it is only for orphan directory */
if (S_ISDIR(la->la_mode) && dt_object_exists(&obj->oo_dt)) {
struct osd_thread_info *info = osd_oti_get(env);
struct lustre_mdt_attrs *lma;
struct lu_buf buf;
lma = (struct lustre_mdt_attrs *)info->oti_buf;
buf.lb_buf = lma;
buf.lb_len = sizeof(info->oti_buf);
rc = osd_xattr_get(env, &obj->oo_dt, &buf, XATTR_NAME_LMA);
if (rc > 0) {
rc = 0;
lma->lma_incompat = le32_to_cpu(lma->lma_incompat);
obj->oo_lma_flags =
lma_to_lustre_flags(lma->lma_incompat);
} else if (rc == -ENODATA) {
rc = 0;
}
}
if (S_ISCHR(la->la_mode) || S_ISBLK(la->la_mode)) {
rc = -sa_lookup(sa_hdl, SA_ZPL_RDEV(o), &osa->rdev, 8);
if (rc)
GOTO(out_sa, rc);
la->la_rdev = osa->rdev;
la->la_valid |= LA_RDEV;
}
out_sa:
sa_handle_destroy(sa_hdl);
RETURN(rc);
}
/*
* Construct a znode+inode and initialize.
*
* This does not do a call to dmu_set_user() that is
* up to the caller to do, in case you don't want to
* return the znode
*/
static znode_t *
zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl,
struct dentry *dentry, struct inode *dip)
{
znode_t *zp;
struct inode *ip;
uint64_t parent;
sa_bulk_attr_t bulk[9];
int count = 0;
ASSERT(zsb != NULL);
ip = new_inode(zsb->z_sb);
if (ip == NULL)
return (NULL);
zp = ITOZ(ip);
ASSERT(zp->z_dirlocks == NULL);
ASSERT3P(zp->z_acl_cached, ==, NULL);
ASSERT3P(zp->z_xattr_cached, ==, NULL);
zp->z_moved = 0;
zp->z_sa_hdl = NULL;
zp->z_unlinked = 0;
zp->z_atime_dirty = 0;
zp->z_mapcnt = 0;
zp->z_id = db->db_object;
zp->z_blksz = blksz;
zp->z_seq = 0x7A4653;
zp->z_sync_cnt = 0;
zp->z_is_zvol = B_FALSE;
zp->z_is_mapped = B_FALSE;
zp->z_is_ctldir = B_FALSE;
zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
&zp->z_pflags, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
&parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
&zp->z_atime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
if (hdl == NULL)
sa_handle_destroy(zp->z_sa_hdl);
goto error;
}
ip->i_ino = obj;
zfs_inode_update(zp);
zfs_inode_set_ops(zsb, ip);
if (insert_inode_locked(ip))
goto error;
if (dentry) {
if (zpl_xattr_security_init(ip, dip, &dentry->d_name))
goto error;
d_instantiate(dentry, ip);
}
mutex_enter(&zsb->z_znodes_lock);
list_insert_tail(&zsb->z_all_znodes, zp);
zsb->z_nr_znodes++;
membar_producer();
mutex_exit(&zsb->z_znodes_lock);
unlock_new_inode(ip);
return (zp);
error:
unlock_new_inode(ip);
iput(ip);
return NULL;
}