truncate_inode_pages(inode->i_mapping, 0);
}
+ /* we can't invalide acl here: suppose we touch a new file
+ * under a dir, blocking ast on dir will lead to open failure
+ * on client, although succeed on mds. it's kind of weird,
+ * the real fix i think is improve client-vfs interaction.
+ */
+#if 0
+ ll_inode_invalidate_acl(inode);
+#endif
+
if (inode->i_sb->s_root &&
inode != inode->i_sb->s_root->d_inode &&
(bits & MDS_INODELOCK_LOOKUP))