On Wed, 2007-08-15 at 12:48 -0400, sowmini.varadhan at sun.com wrote:
> On (08/15/07 08:57), Garrett D'Amore wrote:
> >
> > rem_fault has only three values:
> >
> > none
> > we-detected-and-sent-a-remote-fault
> > we-received-remove-fault-from-peer
>
> so, in that case, what's the man page talking about? :-?
I have no idea. But there are only two bits in the MII register. One
for sending a remote fault indiciation, and one for receiving it.
Maybe they are talking about parallel detection faults?
>
> > If it is detected, I'd just mark the whole row "faulted" or somesuch,
> > and rely on FMA to offline the driver. :-)
>
> Is it fair to say 0 => "good", non-zero => "faulted"?
>
> > I'm a little uncomfortable with separate columns for speed and duplex.
> > While most drivers implement both duplex modes for all speeds, the specs
> > do not require this. And today its easy to arrange using ndd to
> > advertise something like "all modes except 10-Half".
> >
> > Also, the specs for 10G do not support half duplex for 10GBase-T.
>
> we could combine them to be (just tossing out the thought.. the "/"
> looks odd, maybe there's a better char we can use here)
>
> LINK PARAM STATE AUTONEG SPEED/DUPLEX PAUSE
> bge0 current up on 1G/f 1
> capable -- on 1G/f/h,100M/f/h,10M/f/h 1
> adv -- on 1G/f 1
> peeradv -- on 1G/f 1
Yeah, its kind of hard to represent this cleanly.
Maybe it doesn't make sense to include the STATE field here. There are
other ways to get this information. (dladm show-dev reports it, for
example.)
For AUTONEG, I'd prefer "yes/no" to "on/off". (Maybe just a personal
preference.) You might be able to get some more horizontal room by
shortening AUTONEG to just AUTO.
Maybe show speed/duplex as:
1000FH, 100FH, 10F <-- 1G, 100, 10, all duplex modes)
1000F <-- 1G full
10H <-- 10 half
10000F <-- 10G
I don't know if that is any easier to parse, or not.
-- Garrett
>
> --Sowmini
>
