> I just checked the modified patch: INIT_LIST_HEAD *is* necessary. > subtract_locks() do a memcpy of the lock, so it copy list pointers, > but the new lock *is not* in the list. > So INIT_LIST_HEAD() must be called to initialize the list structure. > Else, this will just segfault (At least it did with my testcase). Where exactly did it segfault for you? The list head is never iterated over, and only list_add_tail'ed (which does not look at the stale pointers). Though INIT_LIST_HEAD'ing is a good practice, I'm curious to know where it segfaulted. Avati
