Hi Patrick, An EAGAIN leaks through to dlm_lock via dlm_write, when lock contention is local. Are you sure this is what you want? It would mean everybody has to handle EAGAIN in two places, once right after the call and again in the ast. I'd prefer the dlm_lock call just succeeds and delivers EGAIN to the ast, even if the lock is local. In dlm-kernel/src/locking.c, line 996 (dlm_convert_stage2): ret = lkb->lkb_retstatus = -EAGAIN; Regards, Daniel
