CFT: msk(4) and Yukon FE+(88E8040, 88E8040T, 88E8048, 88E8070)
pyunyh at gmail.com
Wed May 27 01:50:09 UTC 2009
On Tue, May 26, 2009 at 09:40:10AM -0400, John Baldwin wrote:
> On Friday 22 May 2009 11:01:09 pm Pyun YongHyeon wrote:
> > Yukon Extreme uses the new descriptor format, the same as Yukon
> > FE+, so you should have 88E8072 use that. I guess it wouldn't be
> > hard to add support for Yukon Extreme once msk(4) got working code
> > for Yukon FE+.
> I saw you committed this over the weekend, so I will updated my patches to
> account for this. So far I have only had two issues:
> 1) TSO is broken. However, this is probably due to the descriptor format. I
> will try it again after merging in the FE+ changes.
> 2) I was getting RX overrun errors when pulling data over gige (at work)
> (pulling data == using csup to download the CVS repository for the first
> time). It would die with an RX overrun every few minutes. It required an
> ifconfig down/up to recover. (It would be nice if the driver recovered from
> an RX overrun better perhaps.) At home I only have a 10/100 network and the
Did you set GMF_RX_OVER_ON bit of RX_GMF_CTRL_T register? I guess
the GMF_RX_OVER_ON bit will activate flushing on Rx overrun.
Also have you ever checked the output hardware MAC statistics
(dev.msk.N.stats) when you notice Rx overruns?
> driver worked fine (finished full csup w/o any RX overruns) at 100. I will
> also see if I can still reproduce this after merging in the FE+ changes over
> the weekend.
I've updated Yukon Extreme patch again. Previous patch seems to
have bug in Tx checksum offloading. Also added code to disable ASF
for Yukon Extreme. You can get the patch at the following URL.
More information about the freebsd-current