Re: alc(4) link autoselect problem

On Fri, Oct 09, 2009 at 07:29:08AM +0200, Ian FREISLICH wrote:
> Pyun YongHyeon wrote:
> > On Thu, Oct 08, 2009 at 12:38:45PM +0200, Ian FREISLICH wrote:
> > > wgsw-24010# sh interfaces status ethernet g14
> > >                                              Flow Link          Back   Mdix
> > > Port     Type         Duplex  Speed Neg      ctrl State       Pressure Mode
> > > -------- ------------ ------  ----- -------- ---- ----------- -------- ----
> ---
> > > g14      1G-Copper    Full    100   Enabled  Off  Up          Disabled Off 
> > > 
> > 
> > Hmm, does your switch have an option to enable/disable downshifting
> > feature? If so, how about toggling the option?
> 
> I'm not sure exactly what you mean.  There's no configuration options
> that obviously match.
> 

Modern PHYs have ability to correct several cabling problems, for
example,
 - pair swaps
 - pair skews
 - pair polarity
 - automatic MDI/MDIX
 - downshift to enable 10/100Mps link establishment with two
   pairs by downgrading the link speed to 10/100Mbps during
   auto-negotiation process when it can't establish 1000Mbps link.
   Existing cables used to connect 10/100Mbps link partners may
   have only two pairs. With two pairs PHYs can announce 1000Mbps
   capability to link partner but it can't establish the link as
   1000Mbps link requires 4 pairs.

Since F1 PHY seems to have above capability I just wanted to see
whether disabling downshifting on link partner makes any
difference.

> > > > How about checking MIB statistics of controller?
> > > > (sysctl dev.alc.0.stats)
> > > 
> > > This is after a reboot, but the link was up for a short while to
> > > do the above testing.  rx.good_frames looks a bit high for "up 21
> > > mins".
> > > 
> > > dev.alc.0.stats.rx.good_frames: 3348588249
> >   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> > > dev.alc.0.stats.rx.good_octets: 0
> >   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> > 
> > Yeah, that looks odd, it seemed controller received a lot of good
> > frames but none were meaningful packets so I guess there are link
> > establishment issues. I still have no clue but if I manage to find
> > more test case I will let you know.
> 
> That number is bogus.  It's currently incrementing by approximately
> 33435108 per second without a cable plugged in.
> 

Hmm, I have to test this case.

Thanks.

> Ian
> 
> --
> Ian Freislich