[realtek] rtl8139 and UltraSparc 5

Marcelo Roccasalva marcelo-rtl@ucongreso.edu.ar
Thu, 6 Sep 2001 20:01:39 -0300


Hello Donald,

On Thu, Sep 06, 2001 at 01:40:38AM -0400, Donald Becker wrote:
> On Mon, 3 Sep 2001, Marcelo Roccasalva wrote:
> 
> > I'm trying to use a rtl8139 in an UltraSparc 5 with little success... I have
> > RedHat 6.2 installed, using the stock kernel 2.2.19 rpm (64 bits).
> 
> What driver?  And what version number?
The relevant part of kernel messages:

sunhme.c:v1.10 27/Jan/99 David S. Miller (davem@caipfs.rutgers.edu)
eth0: HAPPY MEAL (PCI/CheerIO) 10/100BaseT Ethernet 08:00:20:ab:d2:96
  The PCI BIOS has not enabled the device at 2/8!  Updating PCI command 0003->0007.
rtl8139.c:v1.07 5/6/99 Donald Becker http://cesdis.gsfc.nasa.gov/linux/drivers/rtl8139.html
eth1: RealTek RTL8139 Fast Ethernet at 0xfffff9fe02000400, IRQ 6447936, 00:e0:7d:a8:b9:a2.
  PCI latency timer (CFLT) is unreasonably low at 0.  Setting to 64 clocks.
eth0: Link is up using internal transceiver at 100Mb/s, Half Duplex.

As you can see, there's an integrated sun device... I can't say what does the
message on the third line mean...

> > My dmesg told me about latency=0 so I tried several values, 255 being the
> > better, but when I ifup eth1, I get no more than 10 pings ok, and the iface
> > gets silent. With short latencies (<64) I get "packet too big" messages. I
> > don't have any MAC address problems...
> 
> A few possible issues:
>   Endian correction (unlikely, if a few packets are getting through)
>   Burst length problems (although I don't know of any specific Sparc bugs)
>   A problem with the ring wrap (we need to know the driver version).
>     This could be interacting with the DMA-able page allocation on the Sparc
I've done a few more tests. Sniffing the cable while pinging _from_ the sparc I
can see the initial arp request and answer, and the echo req and echo reply,
but by the time the other machine tries to re-check mac address, the echo
replies are not heard anymore by the sparc, but the packets come and go along
the cable...

> Donald Becker				becker@scyld.com
> Scyld Computing Corporation		http://www.scyld.com
> 410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
> Annapolis MD 21403			410-990-9993
> 

--
Marcelo