release lock refernce only after add page to pages list.

Branch b_release_1_6_7
b=17972
i=shadow
i=johann

lustre/ChangeLog

@@ -44,6 +44,13 @@
 	  Lustre filesystem with 4K stack may cause a stack overflow. For more
 	  information, please refer to bugzilla 17630.
 
+Severity   : normal
+Frequency  : rarw
+Bugzilla   : 17972
+Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
+             freed look.
+Details    : release lock refernce only after add page to pages list.
+
 Severity   : normal
 Frequency  : always with long access acl
 Bugzilla   : 17636

lustre/osc/cache.c

@@ -86,6 +86,7 @@ int cache_add_extent(struct lustre_cache *cache, struct ldlm_res_id *res,
         struct lustre_handle tmplockh;
         ldlm_policy_data_t tmpex;
         struct ldlm_lock *lock = NULL;
+        int mode = 0;
         ENTRY;
 
         /* Don't add anything second time */
@@ -108,7 +109,6 @@ int cache_add_extent(struct lustre_cache *cache, struct ldlm_res_id *res,
                          lock->l_policy_data.l_extent.start,
                          lock->l_policy_data.l_extent.end, extent->oap_obj_off);
         } else {
-                int mode;
                 /* Real extent width calculation here once we have real
                  * extents
                  */
@@ -138,13 +138,16 @@ int cache_add_extent(struct lustre_cache *cache, struct ldlm_res_id *res,
                                "under us\n");
                         RETURN(-ENOLCK);
                 }
-                ldlm_lock_decref(&tmplockh, mode);
         }
 
         spin_lock(&lock->l_extents_list_lock);
         list_add_tail(&extent->oap_page_list, &lock->l_extents_list);
         spin_unlock(&lock->l_extents_list_lock);
         extent->oap_ldlm_lock = lock;
+        LASSERTF(!(lock->l_flags & LDLM_FL_CANCEL), "Adding a page to already "
+                 "cancelled lock %p", lock);
+        if (mode)
+                ldlm_lock_decref(&tmplockh, mode);
         LDLM_LOCK_PUT(lock);
 
         RETURN(0);