Re: Linux on 440GP Networking issues

Re: Linux on 440GP Networking issues

[Todd Poynor]

>>Yep, the kernel tries to setup the second controller. But there is someting
>>very strange. Eventhough there is no network cable connected to the second
>>ethernet port , the OCP ENET driver says that it is 10Mbps, Link up and half
>>duplex.

The above is the default provided when autonegotiation doesn't turn up
anything useful, at least for the AMD Am79C875* PHYs (see ibm_ocp_phy.c:
mii_parse_Am79C875_pcr), which probably isn't the way to go.  And
handling of the "no link" case does seem to need some work.

Could verify whether the lockup is related to configuring 2 EMACs by a
hack to ppc405_enet_open to return -ENODEV if fep->emac_num > 0.  If
that boots fine then see if ppc405_eth_mal_dump shows registers such as
EMAC0's TXCTP0R or RXCTP0R changing values after EMAC1 is configured.
If so, that FIXME in the ppc405_enet_open code probably needs revisiting.


--
Todd


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Linux on 440GP Networking issues

[Roland Dreier]

>>>>> "Todd" == Todd Poynor <tpoynor@mvista.com> writes:

    Todd> Could verify whether the lockup is related to configuring 2
    Todd> EMACs by a hack to ppc405_enet_open to return -ENODEV if
    Todd> fep->emac_num > 0.

Here's another data point.  I just got my ebony today, and in my
initial testing running nothing but the kernel and busybox, I got an
oops if I did "ifconfig" (looked like it came from somewhere near
register_gifconf).  I changed EMAC_NUMS in ibm440.h to 1 (from 2) and
now ifconfig works fine.

However, I didn't experience any crashes, even with EMAC_NUMS equal to
2, running with NFS root.

Best,
  Roland

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/