It looks like setting a large rmem_default value doesn't necessarily waste memory. There doesn't seem to be a fixed, pre-allocated amount of socket receive buffer, it's just that an incoming sk_buff won't be placed on a socket's receive queue if there isn't space "reserved" via the per-socket rcvbuf setting. Is this correct? If you consider a UDP application that doesn't process requests fast enough, this does waste memory if it means packets are held in queue for too long, processed, but then the reply dropped due to protocol timeout. So I'm not necessarily saying that one is ok to set rmem_default very high and all will be fine. I just want to be sure that if I set it fairly high to accomodate one application, the mere existence of other socket-based applications doesn't eat all available memory. thanks /fc - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
