rge: add support for RTL8127ATF with SFP+ interface

[brad_comstyle.com]

Summary

Add support for SFP+ interfaces and fix a bug with the original code from
NetBSD that broke the copper interfaces as the SFP mode detection was
broken.
9895cbfad2f202d3087bd4973516d6e90c6af4f7

Tested with a copper interface and SFP+ with a DAC.

rge0: flags=1008843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST,LOWER_UP> metric 0 mtu 1500
        options=9b<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM>
        media: Ethernet autoselect (10Gbase-T <full-duplex>)
        status: active
        nd6 options=829<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL,STABLEADDR>
rge1: flags=1008843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST,LOWER_UP> metric 0 mtu 1500
        options=9b<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM>
        media: Ethernet autoselect (10Gbase-LR <full-duplex>)
        status: active
        nd6 options=829<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL,STABLEADDR>

[zlei]

Tested with a copper interface and SFP+ with a DAC.
media: Ethernet autoselect (10Gbase-LR <full-duplex>)

With a DAC module the media shall show 10Gbase-Twinax ? The LR / SR is for optical modules if I understand right.

[bz]

In D55976#1280369, @zlei wrote:

Tested with a copper interface and SFP+ with a DAC.
media: Ethernet autoselect (10Gbase-LR <full-duplex>)

With a DAC module the media shall show 10Gbase-Twinax ? The LR / SR is for optical modules if I understand right.

To my understanding, that all depends on how they lie about themselves. Sometimes there are ways to infer the correct.
Normally for SFP we provide an ioctl to get access to I2C and libifconfig decodes the eeproms. But that's a per-driver thing still. I have a WIP to make that different but it needs more work and testing and some time from me again.