Created attachment 166765 [details]
Proposed patch

I'm not 100% sure this is a bug, but the ixv driver does not fill in the link_speed field during initialization. Also the call to check_link during initialization does not actually check the link (these behaviors are related). You can see this difference in a minor change in the "media" field displayed by ifconfig.

Before applying my patch (media does not show "full-duplex"):

ixv0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
        options=405bb<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU,VLAN_HWCSUM,TSO4,LRO,VLAN_HWTSO>
        ether 02:cf:31:b5:e9:19
        inet 10.50.3.59 netmask 0xffffffc0 broadcast 10.50.3.63 
        nd6 options=29<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL>
        media: Ethernet autoselect
        status: active
---

After applying my patch (media shows some additional values):

ixv0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
        options=405bb<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU,VLAN_HWCSUM,TSO4,LRO,VLAN_HWTSO>
        ether 02:cf:31:b5:e9:19
        inet 10.50.3.59 netmask 0xffffffc0 broadcast 10.50.3.63 
        nd6 options=29<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL>
        media: Ethernet autoselect (autoselect <full-duplex>)
        status: active
---

It's not clear to me if the lack of check_link and adapter_speed being filled in matters in any meaningful way. In particular, I'm not sure if it would be a problem if the link started as "down", since check_link is not called during init. I do my testing in EC2 and the link is always up there, so I can't easily test that. Despite potentially not being a bug in functionality, the current behavior of the code was counter to my expectation, so I'm filing this so the maintainer can make a decision if the attached patch is worthwhile and/or correct.