On Wednesday 26 January 2005 17:59, Greg Troxel wrote:
> So should tcp_xmit_timer do sanity checking, or should we check
> opti.ts_ecr against TCP_TIMESTAMP(tp) while unpacking, and unset
> opti.ts_present if it is bogus?

Neither.  Once we get into tcp_xmit_timer() it's too late to fall back to 
old-style RTT calculation, and unsetting ts_present would also cause us to 
not update the time stamp *we* echo.  See the diff(s) I sent.