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;
}
/*
* 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);
}
/*
* add/remove/replace a single attribute and then rewrite the entire set
* of attributes.
*/
static int
sa_modify_attrs(sa_handle_t *hdl, sa_attr_type_t newattr,
sa_data_op_t action, sa_data_locator_t *locator, void *datastart,
uint16_t buflen, dmu_tx_t *tx)
{
sa_os_t *sa = hdl->sa_os->os_sa;
dmu_buf_impl_t *db = (dmu_buf_impl_t *)hdl->sa_bonus;
dnode_t *dn;
sa_bulk_attr_t *attr_desc;
void *old_data[2];
int bonus_attr_count = 0;
int bonus_data_size, spill_data_size;
int spill_attr_count = 0;
int error;
uint16_t length;
int i, j, k, length_idx;
sa_hdr_phys_t *hdr;
sa_idx_tab_t *idx_tab;
int attr_count;
int count;
ASSERT(MUTEX_HELD(&hdl->sa_lock));
/* First make of copy of the old data */
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
if (dn->dn_bonuslen != 0) {
bonus_data_size = hdl->sa_bonus->db_size;
old_data[0] = kmem_alloc(bonus_data_size, KM_SLEEP);
bcopy(hdl->sa_bonus->db_data, old_data[0],
hdl->sa_bonus->db_size);
bonus_attr_count = hdl->sa_bonus_tab->sa_layout->lot_attr_count;
} else {
old_data[0] = NULL;
}
DB_DNODE_EXIT(db);
/* Bring spill buffer online if it isn't currently */
if ((error = sa_get_spill(hdl)) == 0) {
spill_data_size = hdl->sa_spill->db_size;
old_data[1] = kmem_alloc(spill_data_size, KM_SLEEP);
bcopy(hdl->sa_spill->db_data, old_data[1],
hdl->sa_spill->db_size);
spill_attr_count =
hdl->sa_spill_tab->sa_layout->lot_attr_count;
} else if (error && error != ENOENT) {
if (old_data[0])
kmem_free(old_data[0], bonus_data_size);
return (error);
} else {
old_data[1] = NULL;
}
/* build descriptor of all attributes */
attr_count = bonus_attr_count + spill_attr_count;
if (action == SA_ADD)
attr_count++;
else if (action == SA_REMOVE)
attr_count--;
attr_desc = kmem_zalloc(sizeof (sa_bulk_attr_t) * attr_count, KM_SLEEP);
/*
* loop through bonus and spill buffer if it exists, and
* build up new attr_descriptor to reset the attributes
*/
k = j = 0;
count = bonus_attr_count;
hdr = SA_GET_HDR(hdl, SA_BONUS);
idx_tab = SA_IDX_TAB_GET(hdl, SA_BONUS);
for (; k != 2; k++) {
/* iterate over each attribute in layout */
for (i = 0, length_idx = 0; i != count; i++) {
sa_attr_type_t attr;
attr = idx_tab->sa_layout->lot_attrs[i];
if (attr == newattr) {
if (action == SA_REMOVE) {
j++;
continue;
}
ASSERT(SA_REGISTERED_LEN(sa, attr) == 0);
ASSERT(action == SA_REPLACE);
SA_ADD_BULK_ATTR(attr_desc, j, attr,
locator, datastart, buflen);
} else {
length = SA_REGISTERED_LEN(sa, attr);
if (length == 0) {
length = hdr->sa_lengths[length_idx++];
}
SA_ADD_BULK_ATTR(attr_desc, j, attr,
NULL, (void *)
(TOC_OFF(idx_tab->sa_idx_tab[attr]) +
(uintptr_t)old_data[k]), length);
}
}
if (k == 0 && hdl->sa_spill) {
hdr = SA_GET_HDR(hdl, SA_SPILL);
idx_tab = SA_IDX_TAB_GET(hdl, SA_SPILL);
count = spill_attr_count;
} else {
break;
}
}
if (action == SA_ADD) {
length = SA_REGISTERED_LEN(sa, newattr);
if (length == 0) {
length = buflen;
}
SA_ADD_BULK_ATTR(attr_desc, j, newattr, locator,
datastart, buflen);
}
error = sa_build_layouts(hdl, attr_desc, attr_count, tx);
if (old_data[0])
kmem_free(old_data[0], bonus_data_size);
if (old_data[1])
kmem_free(old_data[1], spill_data_size);
kmem_free(attr_desc, sizeof (sa_bulk_attr_t) * attr_count);
return (error);
}
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 || ZTOV(zp)->v_type == VLNK)
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 (zp->z_vnode->v_type == VBLK || zp->z_vnode->v_type == VCHR)
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);
}
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;
}
/*
* 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);
}
/*
* Free space in a file
*
* IN: zp - znode of file to free data in.
* off - start of range
* len - end of range (0 => EOF)
* flag - current file open mode flags.
* log - TRUE if this action should be logged
*
* RETURN: 0 on success, error code on failure
*/
int
zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
{
struct inode *ip = ZTOI(zp);
dmu_tx_t *tx;
zfs_sb_t *zsb = ZTOZSB(zp);
zilog_t *zilog = zsb->z_log;
uint64_t mode;
uint64_t mtime[2], ctime[2];
sa_bulk_attr_t bulk[3];
int count = 0;
int error;
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode,
sizeof (mode))) != 0)
return (error);
if (off > zp->z_size) {
error = zfs_extend(zp, off+len);
if (error == 0 && log)
goto log;
else
return (error);
}
/*
* Check for any locks in the region to be freed.
*/
if (ip->i_flock && mandatory_lock(ip)) {
uint64_t length = (len ? len : zp->z_size - off);
if (!lock_may_write(ip, off, length))
return (SET_ERROR(EAGAIN));
}
if (len == 0) {
error = zfs_trunc(zp, off);
} else {
if ((error = zfs_free_range(zp, off, len)) == 0 &&
off + len > zp->z_size)
error = zfs_extend(zp, off+len);
}
if (error || !log)
return (error);
log:
tx = dmu_tx_create(zsb->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (error);
}
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
NULL, &zp->z_pflags, 8);
zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
dmu_tx_commit(tx);
zfs_inode_update(zp);
return (0);
}
/*
* Unlink zp from dl, and mark zp for deletion if this was the last link. Can
* fail if zp is a mount point (EBUSY) or a non-empty directory (ENOTEMPTY).
* If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
* If it's non-NULL, we use it to indicate whether the znode needs deletion,
* and it's the caller's job to do it.
*/
int
zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
boolean_t *unlinkedp)
{
znode_t *dzp = dl->dl_dzp;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode);
boolean_t unlinked = B_FALSE;
sa_bulk_attr_t bulk[5];
uint64_t mtime[2], ctime[2];
uint64_t links;
int count = 0;
int error;
#ifdef HAVE_DNLC
dnlc_remove(ZTOI(dzp), dl->dl_name);
#endif /* HAVE_DNLC */
if (!(flag & ZRENAMING)) {
mutex_enter(&zp->z_lock);
if (zp_is_dir && !zfs_dirempty(zp)) {
mutex_exit(&zp->z_lock);
return (SET_ERROR(ENOTEMPTY));
}
/*
* If we get here, we are going to try to remove the object.
* First try removing the name from the directory; if that
* fails, return the error.
*/
error = zfs_dropname(dl, zp, dzp, tx, flag);
if (error != 0) {
mutex_exit(&zp->z_lock);
return (error);
}
if (ZTOI(zp)->i_nlink <= zp_is_dir) {
zfs_panic_recover("zfs: link count on %lu is %u, "
"should be at least %u", zp->z_id,
(int)ZTOI(zp)->i_nlink, zp_is_dir + 1);
set_nlink(ZTOI(zp), zp_is_dir + 1);
}
drop_nlink(ZTOI(zp));
if (ZTOI(zp)->i_nlink == zp_is_dir) {
zp->z_unlinked = B_TRUE;
clear_nlink(ZTOI(zp));
unlinked = B_TRUE;
} else {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
NULL, &ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
NULL, &zp->z_pflags, sizeof (zp->z_pflags));
zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
ctime);
}
links = ZTOI(zp)->i_nlink;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
NULL, &links, sizeof (links));
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
count = 0;
ASSERT(error == 0);
mutex_exit(&zp->z_lock);
} else {
error = zfs_dropname(dl, zp, dzp, tx, flag);
if (error != 0)
return (error);
}
mutex_enter(&dzp->z_lock);
dzp->z_size--; /* one dirent removed */
if (zp_is_dir)
drop_nlink(ZTOI(dzp)); /* ".." link from zp */
links = ZTOI(dzp)->i_nlink;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
NULL, &links, sizeof (links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
NULL, &dzp->z_size, sizeof (dzp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
NULL, ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
NULL, mtime, sizeof (mtime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
NULL, &dzp->z_pflags, sizeof (dzp->z_pflags));
zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime);
error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&dzp->z_lock);
if (unlinkedp != NULL)
*unlinkedp = unlinked;
else if (unlinked)
zfs_unlinked_add(zp, tx);
return (0);
}
/*
* Unlink zp from dl, and mark zp for deletion if this was the last link.
* Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
* If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
* If it's non-NULL, we use it to indicate whether the znode needs deletion,
* and it's the caller's job to do it.
*/
int
zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
boolean_t *unlinkedp)
{
znode_t *dzp = dl->dl_dzp;
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
vnode_t *vp = ZTOV(zp);
#ifdef __APPLE__
//int zp_is_dir = S_ISDIR(zp->z_phys->zp_mode);
int zp_is_dir = S_ISDIR(zp->z_mode);
#else
int zp_is_dir = (vp->v_type == VDIR);
#endif
boolean_t unlinked = B_FALSE;
sa_bulk_attr_t bulk[5];
uint64_t mtime[2], ctime[2];
int count = 0;
int error;
#ifndef __APPLE__
dnlc_remove(ZTOV(dzp), dl->dl_name);
#endif
if (!(flag & ZRENAMING)) {
if (vp) {
if (vn_vfswlock(vp)) /* prevent new mounts on zp */
return (EBUSY);
if (vn_ismntpt(vp)) { /* don't remove mount point */
vn_vfsunlock(vp);
return (EBUSY);
}
} /* if (vp) */
mutex_enter(&zp->z_lock);
if (zp_is_dir && !zfs_dirempty(zp)) { /* dir not empty */
mutex_exit(&zp->z_lock);
#ifdef __APPLE__
return (ENOTEMPTY);
#else
vn_vfsunlock(vp);
return (EEXIST);
#endif
}
/*
* If we get here, we are going to try to remove the object.
* First try removing the name from the directory; if that
* fails, return the error.
*/
error = zfs_dropname(dl, zp, dzp, tx, flag);
if (error != 0) {
mutex_exit(&zp->z_lock);
//vn_vfsunlock(vp);
return (error);
}
if (zp->z_links <= zp_is_dir) {
#ifndef __APPLE__
zfs_panic_recover("zfs: link count on %s is %u, "
"should be at least %u",
zp->z_vnode->v_path ? zp->z_vnode->v_path :
"<unknown>", (int)zp->z_links,
zp_is_dir + 1);
#endif
zp->z_links = zp_is_dir + 1;
}
if (--zp->z_links == zp_is_dir) {
zp->z_unlinked = B_TRUE;
zp->z_links = 0;
unlinked = B_TRUE;
} else {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
NULL, &ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
NULL, &zp->z_pflags, sizeof (zp->z_pflags));
zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime,
B_TRUE);
}
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
NULL, &zp->z_links, sizeof (zp->z_links));
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
count = 0;
ASSERT(error == 0);
mutex_exit(&zp->z_lock);
#ifndef __APPLE__
vn_vfsunlock(vp);
#endif
} else {
error = zfs_dropname(dl, zp, dzp, tx, flag);
if (error != 0)
return (error);
}
mutex_enter(&dzp->z_lock);
dzp->z_size--; /* one dirent removed */
dzp->z_links -= zp_is_dir; /* ".." link from zp */
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
NULL, &dzp->z_links, sizeof (dzp->z_links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
NULL, &dzp->z_size, sizeof (dzp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
NULL, ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
NULL, mtime, sizeof (mtime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
NULL, &dzp->z_pflags, sizeof (dzp->z_pflags));
zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&dzp->z_lock);
/* Inform VFS to dump this znode as soon as possible... */
if (unlinked && ZTOV(zp))
vnode_recycle(ZTOV(zp));
//error = zap_remove(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name, tx);
//ASSERT(error == 0);
if (unlinkedp != NULL)
*unlinkedp = unlinked;
else if (unlinked)
zfs_unlinked_add(zp, tx);
return (0);
}
/*
* Truncate a file
*
* IN: zp - znode of file to free data in.
* end - new end-of-file.
*
* RETURN: 0 on success, error code on failure
*/
static int
zfs_trunc(znode_t *zp, uint64_t end)
{
zfs_sb_t *zsb = ZTOZSB(zp);
dmu_tx_t *tx;
rl_t *rl;
int error;
sa_bulk_attr_t bulk[2];
int count = 0;
/*
* We will change zp_size, lock the whole file.
*/
rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
/*
* Nothing to do if file already at desired length.
*/
if (end >= zp->z_size) {
zfs_range_unlock(rl);
return (0);
}
error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1);
if (error) {
zfs_range_unlock(rl);
return (error);
}
top:
tx = dmu_tx_create(zsb->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_NOWAIT);
if (error) {
if (error == ERESTART) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
zfs_range_unlock(rl);
return (error);
}
zp->z_size = end;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb),
NULL, &zp->z_size, sizeof (zp->z_size));
if (end == 0) {
zp->z_pflags &= ~ZFS_SPARSE;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
NULL, &zp->z_pflags, 8);
}
VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
dmu_tx_commit(tx);
zfs_range_unlock(rl);
return (0);
}
/*
* Link zp into dl. Can only fail if zp has been unlinked.
*/
int
zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
{
znode_t *dzp = dl->dl_dzp;
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
#ifdef __APPLE__
uint64_t value;
/* OSX - don't access the vnode here since it might not be attached yet. */
//int zp_is_dir = S_ISDIR(zp->z_phys->zp_mode);
int zp_is_dir = S_ISDIR(zp->z_mode);
#else
vnode_t *vp = ZTOV(zp);
uint64_t value;
int zp_is_dir = (vp->v_type == VDIR);
#endif
sa_bulk_attr_t bulk[5];
uint64_t mtime[2], ctime[2];
int count = 0;
int error;
mutex_enter(&zp->z_lock);
if (!(flag & ZRENAMING)) {
if (zp->z_unlinked) { /* no new links to unlinked zp */
ASSERT(!(flag & (ZNEW | ZEXISTS)));
mutex_exit(&zp->z_lock);
return (ENOENT);
}
zp->z_links++;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&zp->z_links, sizeof (zp->z_links));
}
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
&dzp->z_id, sizeof (dzp->z_id));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
if (!(flag & ZNEW)) {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
ctime, B_TRUE);
}
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
// Needed?
#ifdef __APPLE__
zp->z_parent = dzp->z_id;
#endif
mutex_exit(&zp->z_lock);
mutex_enter(&dzp->z_lock);
dzp->z_size++;
dzp->z_links += zp_is_dir;
count = 0;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
&dzp->z_size, sizeof (dzp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&dzp->z_links, sizeof (dzp->z_links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
mtime, sizeof (mtime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&dzp->z_pflags, sizeof (dzp->z_pflags));
zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
mutex_exit(&dzp->z_lock);
value = zfs_dirent(zp, zp->z_mode);
error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name,
8, 1, &value, tx);
ASSERT(error == 0);
#ifndef __APPLE__
/* On Mac OS X, this is done up in VFS layer. */
dnlc_update(ZTOV(dzp), dl->dl_name, vp);
#endif
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);
}
/*
* 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);
}
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);
}
/*
* Link zp into dl. Can only fail if zp has been unlinked.
*/
int
zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
{
znode_t *dzp = dl->dl_dzp;
zfsvfs_t *zfsvfs = ZTOZSB(zp);
uint64_t value;
int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode);
sa_bulk_attr_t bulk[5];
uint64_t mtime[2], ctime[2];
uint64_t links;
int count = 0;
int error;
mutex_enter(&zp->z_lock);
if (!(flag & ZRENAMING)) {
if (zp->z_unlinked) { /* no new links to unlinked zp */
ASSERT(!(flag & (ZNEW | ZEXISTS)));
mutex_exit(&zp->z_lock);
return (SET_ERROR(ENOENT));
}
if (!(flag & ZNEW)) {
/*
* ZNEW nodes come from zfs_mknode() where the link
* count has already been initialised
*/
inc_nlink(ZTOI(zp));
links = ZTOI(zp)->i_nlink;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
NULL, &links, sizeof (links));
}
}
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
&dzp->z_id, sizeof (dzp->z_id));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
if (!(flag & ZNEW)) {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
ctime);
}
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&zp->z_lock);
mutex_enter(&dzp->z_lock);
dzp->z_size++;
if (zp_is_dir)
inc_nlink(ZTOI(dzp));
links = ZTOI(dzp)->i_nlink;
count = 0;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
&dzp->z_size, sizeof (dzp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&links, sizeof (links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
mtime, sizeof (mtime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&dzp->z_pflags, sizeof (dzp->z_pflags));
zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime);
error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&dzp->z_lock);
value = zfs_dirent(zp, zp->z_mode);
error = zap_add(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name,
8, 1, &value, tx);
ASSERT(error == 0);
return (0);
}
/*
* 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);
}
/*
* Link zp into dl. Can only fail if zp has been unlinked.
*/
int
zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
{
znode_t *dzp = dl->dl_dzp;
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
vnode_t *vp = ZTOV(zp);
uint64_t value;
int zp_is_dir = (vnode_isdir(vp));
sa_bulk_attr_t bulk[5];
uint64_t mtime[2], ctime[2];
int count = 0;
int error;
mutex_enter(&zp->z_lock);
if (!(flag & ZRENAMING)) {
if (zp->z_unlinked) { /* no new links to unlinked zp */
ASSERT(!(flag & (ZNEW | ZEXISTS)));
mutex_exit(&zp->z_lock);
return (SET_ERROR(ENOENT));
}
zp->z_links++;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&zp->z_links, sizeof (zp->z_links));
}
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
&dzp->z_id, sizeof (dzp->z_id));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
if (!(flag & ZNEW)) {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
ctime, B_TRUE);
}
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&zp->z_lock);
mutex_enter(&dzp->z_lock);
dzp->z_size++;
dzp->z_links += zp_is_dir;
count = 0;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
&dzp->z_size, sizeof (dzp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&dzp->z_links, sizeof (dzp->z_links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
mtime, sizeof (mtime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&dzp->z_pflags, sizeof (dzp->z_pflags));
zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&dzp->z_lock);
value = zfs_dirent(zp, zp->z_mode);
error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name,
8, 1, &value, tx);
ASSERT(error == 0);
dnlc_update(ZTOV(dzp), dl->dl_name, vp);
return (0);
}
/*
* 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;
}