On Fri, 29 Jul 2005 13:47:36 +0200, Andreas Bihlmaier wrote: > I later on found out (after many trips down the stairs ) that a ram was going > really bad (as in seqfault in random places), but I also changed my NIC during > my search for the bad hardware in my box.
Not much chance here on bad hardware: It had been up 172 days; got 3.7, and was up some 50 days. > My GUESS is that the problem is with the 'xl' 3.7 release driver because > I didn't have the problem since using a real NIC (fxp) an not a piece of > junk (sorry but did some testing with xl -> crap ). Not so sure about this either. Never had problems with xln before. At least, nothing specific to Intel. > As written above dump the nic or if that is not a choice leave the media > setting at 'auto'. This would never work when the switch is forced to a media and rate (100BASE-T Full Duplex). Any box would see this as 100BASE-T Half Duplex. This happened to a colleague with Windows-something on that switch; to my other box there (RHEL with RTL8139) and this box (OpenBSD 3.7 with xln): All and any would - when set to auto - assume a 100BASE-T Half Duplex and produce horrible amounts of errors. No clue why. In any case: autonegotiation on that switch is out. Thanks anyway, Uwe
