Hi, still having this problem, even with default sysctl values. Just not as often and not as fast (after a reboot).
According to (hope thats the right doc) http://www.sun.com/processors/manuals/cs_plus.pdf it sounds like this should not happen at all. At least to me it sounds like a bug. Would be nice if someone could look into that. On Wed, 23 Jul 2008 13:06:05 +0200, Michael <[EMAIL PROTECTED]> wrote: > adding some information... > > Michael schrieb: >> sometimes I get this on my SUN Fire v440 when there is some traffic: >> >> cas0: status=7889090<RXDONE,RX_COMP_FULL,RXMAC> >> >> cas0 stops working then. Using "ifconfig cas0 down/up" only helps for a >> very short time. Only way to get it to work again for some time is a >> full reboot. > > Played a bit around and noticed that this only happens when increasing > net.inet.tcp.sendspace and/or net.inet.tcp.recvspace. > > Increasing net.inet.tcp.sendspace is suggested by dhartmei@ inside his > undeadly.org CGI scripts. > > > 4) General system tuning > > > > /etc/sysctl.conf > > > > kern.somaxconn=1024 # max. listen(2) backlog > > net.inet.tcp.sendspace=65536 # TCP send buffer size > > Changing those values works just fine on all other systems I am running > OpenBSD on with vr, sis, bnx, bge, em and ral interfaces, just not with > cas. > > So, bug in the cas driver?