[U-Boot-Users] miiphy_speed() broken after latest patch...

[Travis Sawyer]

Tolunay, Wolfgang:

See below:

On Thu, 2004-03-18 at 19:32, Tolunay Orkun wrote:
> >> Perhaps we need to make the detection of 1000Mbps PHY included
> >> optionally
> >> based on a CONFIG_ macro and 1000MBps PHY boards define that variable. I
> >> have no use for extra code.
> >
Ouch!

> > OK - can you please submit a patch, then?
> 
> I can but I would like to hear from Travis first.
> 
PING!

> Another thing to consider is incompatible GigE PHY register sets. So,
> perhaps we should have someting have CONFIG_PHY_1000_TYPE as well and use
> it like:
> 
Incompatible!?!  Then they do not follow the IEEE 802.3 standard.  The
only reason I felt comfortable putting the code in was I read the .3
standard wrt unsupported registers.

>According to datasheet of Intel LXT971A, PHY, PHY_1000BTSR register is
>listed among registers as 1000BASE-T/100BASE-T2 Status Register but it
>is
>marked as not implemented and no further data is available. So, the
>check
>against 0xFFFF is not working properly.

So the LXT971A doesn't follow the 802.3 standard.  Odd in that the intel
lxt9762 does...

> #if defined(CONFIG_PHY_1000_TYPE)
> #if CONFIG_PHY_1000_TYPE == CONFIG_PHY_MARVELL_88E1XXX
> ... marvell specific code ...
> #elif CONFIG_PHY_1000_TYPE == CONFIG_PHY_1000_GENERIC...
> ... assumed generic gigE PHY code ...
> #endif
> #endif
I deeply regret having broken the code!

By all means patch away!  I do like the CONFIG_PHY_1000_TYPE suggestion.

Darn.  And I was feeling like a hero too ;)

AFAIK, the only two boards that _*I*_ use GbE on is the XPedite1K and
Ocotea boards, both having phys that behave (Xpedite has BCom phys, and
they behave :)

Thanx,

Travis
(I wonder what other phys don't return 0xFFFF, I only have 4 different
types on hand ;)