RE: ixp4xx ethernet driver issues

["Mo, Vincent" <vincent.mo@xxxxxxxxxxxxxxxxxx>]

I've resolved the issue, it was a simple problem: the phy was for some
reason configured for isolation mode.  I modified the driver to mask out
that bit in the BMCR register so it never sets it even if it was read
back in the first place.  Thanks for all your help Krzysztof.

>
>>>Even though MDIO doesn't use the firmware, the NPE needs it right?  I
>>>thought it can't work with out
>>>the microcode/firmware.
>
>Sure, BTW it needs the "basic" (non-VLAN) firmware file.
>
>>>Could you please explain what you mean by "platform" structure?
>
>I mean that thing in ... mach-ixp4xx/*.c, something like
>
>/* Built-in 10/100 Ethernet MAC interfaces */
>static struct eth_plat_info eth_plat[] = {
>        {
>                .phy            = 0,
>                .rxq            = 3,
>                .txreadyq       = 20,
>        }, {
>                .phy            = 1,
>                .rxq            = 4,
>                .txreadyq       = 21,
>        }
>};
>
>static struct platform_device device_eth_tab[] = {
>        {
>                .name                   = "ixp4xx_eth",
>                .id                     = IXP4XX_ETH_NPEB,
>                .dev.platform_data      = eth_plat,
>        }, {
>                .name                   = "ixp4xx_eth",
>                .id                     = IXP4XX_ETH_NPEC,
>                .dev.platform_data      = eth_plat + 1,
>        }
>};
>
>but I see in the log it's the same as mine.
>
>>>Another thing I want to point out, the IXP465 is not outputting the
>>>ETHB_RXCLK to the phy.  This is strange, the phy is clearly receiving
>>>packets and is toggling its' link/activity line, but it's not able to
>>>transmit the received packets to the CPU because there is no receive
>>>clock.
>>
>> I'm correcting myself, the phy should be driving the RXCLK line.  I'm
>> not sure why the phy is not driving it...
>
>Perhaps mii-tool (or with 2.6.29-rc* also ethtool) can tell us?
>
>Are you sure your board uses the right 25 MHz crystal/clock gen?
>Verifying with at least a frequency meter may be a good idea.
>OTOH since it works with Intel's driver...
>--
>Krzysztof Halasa

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php