http://news.gmane.org/find-root.php?message_id=%3c4C9A4A48.7060501%40cbnco.com%3e
As a workaround you can apply this patch: http://jrz.cbnco.com/git/?p=toastix/toastix.git;a=blob_plain;f=public/recipes/linux/reinstate-elan-pit-special-case.patch;h=b08c41c2f19bf57ce07e1ab90ec5ead49200424b;hb=5dbf845a7f2d8a2d1efb007a66d8971cfcc85bbc Mike On Mon, 27 Sep 2010, Kyle Brantley wrote: > I've had my 4521 for a long time, and it has served me well. However, the > clock has never been the most accurate thing in the world, and when I say > that, I mean it drifts more than five minutes per day. > > This is actually a larger drift rate that ntpd can manage. With ntpd, it > drifts about two and a half minutes per day and eventually falls out of sync > with the various configured peers (one of which is <1ms away with a very > stable clock, synched to a stratum two server). I've resorted to ntpdate via > cron. > > Considering that the net4501 and the net4521 are said to be some of the single > most accurate pieces of time keeping hardware out there[1], and considering > that I just reconfigured it to use some cryptographic software that requires > <5 minutes of clock skew (not asking much!) I'm curious what the list thinks > the problem could be. I'd like to use this 4521 as my primary timesource for > my network, and kill the cron job. > > Attached is my kernel config. It is a hand-rolled 2.6.34 built for i486 from > vanilla sources, running a relatively old debian userspace (I don't have the > desire to update the userspace given the CPU, RAM, and mandatory package > signature validation required). If I'm missing something blindingly obvious > from my kernel config that would be great, but given the many kernels this has > ran I'm leaning towards a hardware issue. > > Input appreciated! > > --Kyle > > [1] http://www.febo.com/pages/soekris/ > > _______________________________________________ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech