Hello! > And by mentioning this you unearth the real bug, tp->backoff needs > to be reset here where tp->probes_out is reset. ... > Alexey, anything missed here? Maybe only reset tp->backoff when > tp->probes_out is non-zero at this point? No, it can grow together with probes_out, when window remains closed. zero window probing never stops, so we do not have any cutoff for probes_out. I think right solution would be limiting tp->backoff not to overflow TCP_MAX_RTO, like Martin proposed. His recipe does not look good though. It depends on value of tp->rto which can change in future. SOmething sort of? if (TCP_MAX_RTO < (0x7fffffff>>tp->backoff)) tp->backoff++ 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
