Hello! > Do you have any objections to making the transmission of an error > message reliable for NETLINK? Error is returned in any case, it is ENOBUFS returned by recvmsg() when we cannot return error specific for given request. I think, memory bounds are better to keep really hard. Reliable (in this sense) transmissions from kernel to user are impossible in any case. This policy can be relaxed to be more convenient: sk->rcvbuf is to be set high enough and some kinds of messages might be sent over this limit f.e. up to 2*sk->rcvbuf. But this cannot make the transport reliable anyway and therefore applications must implement recovery strategy. For netlink reliability used to be provided with the fact that when a message cannot be generated, netlink generates overrun: ENOBUFS. This means that synchronization between kernel and user views is lost and application must resync. This approach worked with gated very well. Alexey - : 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
