Intel 82567V-2 using em driver, no carrier or constant link-state changes with 1000BaseT

Thu Dec 1 20:08:23 UTC 2016

-----Original Message-----
From: owner-freebsd-net at freebsd.org [mailto:owner-freebsd-net at freebsd.org] On Behalf Of Sean Bruno
Sent: Thursday, December 01, 2016 9:53 AM
To: freebsd-net at freebsd.org
Subject: Re: Intel 82567V-2 using em driver, no carrier or constant link-state changes with 1000BaseT

On 11/27/16 12:40, Michael Knap wrote:
> Preface: I've been a long-time linux user, and I am enjoying the BSD
> experience. For the following scenario, I have reproduced these results
> with 3 or more cables which are known to be good.
> 
> Using FreeBSD 10.3 Stable :
> FreeBSD 10.3-STABLE #0 r295946+07c41cd(9.10-STABLE): Wed Nov  9 00:19:25
> UTC 2016     root at gauntlet
> :/freenas-9.10-releng/_BE/objs/freenas-9.10-releng/_BE/os/sys/FreeNAS.amd64
> 
> Also, this isn't a NIC, per se, as it is onboard, but I will call it a NIC
> or card or whatever.
> 
> Using em driver, though I don't know how to determine a driver version (if
> there is such a thing).
> 
> Basically, the card will not push gigabit, though it will work with
> 100BaseTX. However, there seems to be more than just that as the chip-set
> is advertised as a gigabit Ethernet controller.
> 
> When connected to a 100BaseTX switch, everything is good. Leaving it in
> autodetect mode, and connecting to a 1000baseTX switch (with good CAT6
> cable), the card's link state oscillates from link up and link down:
> 
> Nov 27 13:07:38 freenas kernel: em0: link state changed to DOWN
> Nov 27 13:07:38 freenas kernel: em0: link state changed to DOWN
> Nov 27 13:07:41 freenas kernel: em0: link state changed to UP
> Nov 27 13:07:41 freenas kernel: em0: link state changed to UP
> Nov 27 13:07:41 freenas kernel: em0: link state changed to DOWN
> Nov 27 13:07:41 freenas kernel: em0: link state changed to DOWN
> Nov 27 13:07:55 freenas kernel: em0: link state changed to UP
> Nov 27 13:07:55 freenas kernel: em0: link state changed to UP
> Nov 27 13:07:56 freenas kernel: em0: link state changed to DOWN
> Nov 27 13:07:56 freenas kernel: em0: link state changed to DOWN
> Nov 27 13:07:58 freenas kernel: em0: link state changed to UP
> Nov 27 13:07:58 freenas kernel: em0: link state changed to UP
> Nov 27 13:07:59 freenas kernel: em0: link state changed to DOWN
> Nov 27 13:07:59 freenas kernel: em0: link state changed to DOWN
> Nov 27 13:08:01 freenas kernel: em0: link state changed to UP
> Nov 27 13:08:01 freenas kernel: em0: link state changed to UP
> Nov 27 13:08:02 freenas kernel: em0: link state changed to DOWN
> Nov 27 13:08:02 freenas kernel: em0: link state changed to DOWN
> 
> If I issue  ifconfig em0 media 100baseTX mediaopt full-duplex, then I have
> no carrier.
> Also, ifconfig em0 media 1000baseT gives no carrier.
> However, if I ifconfig em0 media 10baseT, then I can get an active link.
> 
> Interestingly, man em(4) does not claim to support this chipset, though it
> seems that this chipset was once supported in either e1000 or igb for older
> freeBSD versions?
> 
> Now, at least from the driver's POV, I should be able to handle gigabit, as
> 
> ifconfig -m em0
> em0: flags=8c02<BROADCAST,OACTIVE,SIMPLEX,MULTICAST> metric 0 mtu 1500
> options=4219b<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM,TSO4,WOL_MAGIC,VLAN_HWTSO>
> capabilities=5399b<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM,TSO4,WOL_UCAST,WOL_MCAST,WOL_MAGIC,VLAN_HWFILTER,VLAN_HWTSO>
> ether 00:25:11:59:d4:be
> inet 10.42.0.111 netmask 0xffffff00 broadcast 10.42.0.255
> nd6 options=9<PERFORMNUD,IFDISABLED>
> media: Ethernet 10baseT/UTP (10baseT/UTP <half-duplex>)
> status: active
> supported media:
> media autoselect
> media 1000baseT
> media 1000baseT mediaopt full-duplex
> media 100baseTX mediaopt full-duplex
> media 100baseTX
> media 10baseT/UTP mediaopt full-duplex
> media 10baseT/UTP
> 
> However when connected to gigabit switch, any explicit declaration of media
> fails except for 10baseT, even the autoselect declaration fails with no
> carrier.
> 
> But remember, 100baseTX works out of the box on a 100Mbit switch.
> 
> So, those more experienced, can you shed some light on what is happening?
> I'm especially interested in the man em(4) result even though the chipset
> has been supported in the past, and intel does provide drivers for this
> chipset, though I'm not sure if I could compile them or not. BTW, my
> switches are only consumer grade, unmanaged. Is that where I'm stuck?
> 
> Thanks for such an awesome license, operating system, and community.
> _______________________________________________

>Just so I know which device this specifically is, can you post "pciconf
>-lvbc" to a pastebin or dpaste type site so we can review which driver
>this should be using.  Linux seems to use igb(4) for this instead of em(4).
>
>sean

Sean,

This is a client (em) part, and should be 10CE (Boazman/Boulder Creek).

Jeff