Netra T1 105 hme0 woes ?

[Marius Strobl]

On Fri, Oct 22, 2004 at 02:54:02PM +0200, Xavier Beaudouin wrote:
> Hi there,
> 
> I am experimenting some woes about hme0 and Netra T1 model 105.
> 
> It is mostly working like a charm, but when I setup in /etc/rc.conf 100 
> Full-Duplex operation mode (/sbin/ifconfig hme0 media 100BaseTX 
> mediaopt full-duplex) the interface seems to not send any packets on 
> the network.
> 
> But if I set as autodiscovery mode and after (in rc.local for example), 
> the interface works like a charm...
> 
> It doesn't seems to be a bug inside hme drivers, but some strange arch 
> on theses boxes :
> 
> ---//--- (extract of /var/run/dmesg.boot)---
> hme0: <Sun HME 10/100 Ethernet> mem 0xe0000000-0xe0007fff at device 1.1 
> on pci1
> miibus0: <MII bus> on hme0
> ukphy0: <Generic IEEE 802.3u media interface> on miibus0
> ukphy0:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
> ukphy1: <Generic IEEE 802.3u media interface> on miibus0
> ukphy1:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
> hme0: Ethernet address: 08:00:20:d9:d3:b0
> hme0: if_start running deferred for Giant
> hme0: [GIANT-LOCKED]
> sym0: <875> port 0xc00000-0xc000ff mem 
> 0xe000a000-0xe000afff,0xe0008000-0xe00080
> ff at device 2.0 on pci1
> sym0: No NVRAM, ID 7, Fast-20, SE, parity checking
> sym0: [GIANT-LOCKED]
> hme1: <Sun HME 10/100 Ethernet> mem 0xe0010000-0xe0017fff at device 3.1 
> on pci1
> miibus1: <MII bus> on hme1
> ukphy2: <Generic IEEE 802.3u media interface> on miibus1
> ukphy2:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
> hme1: Ethernet address: 08:00:20:d9:d3:b0
> hme1: if_start running deferred for Giant
> hme1: [GIANT-LOCKED]
> ---//-----
> 
> As you can see there is *two* media interfaces on hme0 (ukphy0 and 
> ukphy1)... I think that the system may take the first media interface 
> that is not connected (or unexistant) on this platform.
> 
> Note that when opening this box, I can find a ethernet port near the 

Yes, that's ukphy1. The HME chips can handle a maximum of two PHYs,
an internal and an external one. On machines and cards equipped with
a single PHY per HME chip this is usually the internal one (with a
MII interface on some machines and cards to connect an external one.)
The Netra T1 105 really is a Sun CP1500 CompactPCI board with one
single PHY for the first HME chip plus an extension card with the two
external PHYs (ukphy0 and ukphy2) connected to the ethernet ports on
the rear.

> CPU... Maybe this port is the ukphy0 and it is desactivated per default 
> on Solaris ?

I don't think that Solaris disbales the internal PHY in this case.
The problem is some strangeness that when the unused PHY of the first
HME chip (either ukphy0 or ukphy1) is isolated from the MII bus the
whole bus wedges.
Could you give the patch at:
http://alchemy.franken.de/~marius/hme_luphy.diff
a try? Please start with an empty kernel build directory.
Thanks.