net: Incorrect logic for TCP "ackerr" statistics

[pfalcon]

In https://github.com/zephyrproject-rtos/zephyr/blob/master/subsys/net/ip/tcp.c#L907, we have the code like:

        if (IS_ENABLED(CONFIG_NET_STATISTICS_TCP) &&
            sys_slist_is_empty(list)) {
                net_stats_update_tcp_seg_ackerr();
        }

It means that if retransmit list is empty, then any ACK received will be considered "error". But that doesn't make sense, because TCP mandates ACK to be set for any packet of the established connection - if there's no advancement in the ack number, the last value will be used, and ACK bit still set.

Thus, the above condition should be removed. The (only?) condition for ACK being invalid is described in #4653 (comment) .

[pfalcon]

Going further, we have code at https://github.com/zephyrproject-rtos/zephyr/blob/master/subsys/net/ip/tcp.c#L931 again treating something as "ACK error":

                if (!net_tcp_seq_greater(ack, seq)) {
                        net_stats_update_tcp_seg_ackerr();
                        break;
                }

Likewise, that's not correct - there's nothing wrong if only the head part of retransmit list is being acknowledged (instead of the entire list). So, the above is the list processing loop termination condition, but it's not an error.