On Fri, Jan 22, 2016 at 03:12:08PM -0800, William Tu wrote: > I'm reading through the lib/cmap.c and I wonder if we need an extra lock at > cmap_insert_bfs(). > > When cmap_insert_bfs() found a path and started to move all elements > backward, it goes through the for loop below and calls cmap_set_bucket() > one-by-one, and eventually place the new_node. However, is it possible that > another writer at the same time inserts an element, which is among one of > the elements in the path, as a result, making the path no longer valid?
No, cmap is single-writer, any writer has to hold an exclusive lock anyway. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
