On Sunday 27 November 2005 01:17, Patrick McHardy wrote: > Michael Frank wrote: > > Current Netpoll init does not trust carrier detected > > right upon bringing the NIC up and waits for a delay > > before allowing boot to continue. > > > > The delay varied with the kernel version: > > 10 seconds in 2.6.11 > > 3 seconds in 2.6.12 > > 4 seconds in 2.6.13 and later > > > > Any significant delay is undesired in production > > environments as 3 seconds boot the entire kernel. > > > > The patch reduces the delay to about 300msec + > > NIC startup time. > > > > The adjusted timeout mechanism polls the carrier every > > 100ms and polls for a good carrier for N_CARRIER_POLL > > (3) _consecutive_ cycles or 10s maximum. > > Why do you increase the maximum again?
In the _current_ implementation, the timeout is effectively a fixed delay as the carrier for most NIC's is very fast and "not to be trusted" In _this_ implementation this is a true timeout which will only be reached in case of NIC failure. To clarify the practical delay of this implementation N_CARRIER_POLL * 0.1s + a few 100 to few 1000 us , 0.3s - 10 times as fast as current implementation ;) - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html