Realtek® 8111GR

[shep]

My Asus A68-HM K motherboard has an onboard Realtek® 8111GR that randomly drops a Static connection.

The manual re(4) indicates that it requires a twisted pair for 1000baseT.

My

PooBear$ cat /etc/hostname.re0
inet 192.168.2.2 255.255.255.0 192.168.2.254 media 100baseTX mediaopt full-duplex
mtu 1492

PooBear$ ifconfig re0 
re0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1492
        lladdr 70:4d:7b:26:b3:ed
        index 1 priority 0 llprio 3
        groups: egress
        media: Ethernet 100baseTX full-duplex (100baseTX full-duplex,rxpause,txpause)
        status: active
        inet 192.168.2.2 netmask 0xffffff00 broadcast 192.168.2.255

When I set my /etc/hostname.re0 to force 1000baseT, it will not connect.

I'm wondering if I my patch cables could be an issue. This particular NIC has been problematic in Linux and in Debian there is a prompt, on kernel upgrade, to install realtek firmware. The only realtek firmware I can find in OpenBSD is for wireless interfaces.

Are there some configuration tweaks I can do? Is a standard Cat5 cable considered a twister pair? I do not really need 1000baseT, my ADSL modem is 100baseTX. I do have an old Intel Pro/100+ pci adapter and I understand that Gigabyte Intel NIC pci are not that expensive.

[jggimi]

A standard Cat 5 cable contains 4 twisted pairs - 8 wires. Cat 6 cable, designed for 1000BaseT and much higher speed connections, includes a plastic core to keep the pairs separated.

With that out of the way...

I used to have an Asus netbook that contained a NIC which was 1000BaseT capable, but the media attachment (PHY) selected by Asus for the NIC was low-cost and only capable of 100BaseT. Welcome to low-cost equipment.

[shep]

I'm wondering if the PHY requires firmware for 1000BaseT? Debian considers the firmware "non-free" and if I recall correctly, Realtek will not work with developers without an NDA. I suspect, that unless reversed engineered, the 8111/8168 will never support 1000BaseT in OpenBSD.

I actually look for low-cost equipment that should be well supported by OpenBSD. As I mentioned, I do not need the 1000BaseT speed and would prefer to keep my system relatively simple and efficient. I'm fine with 100BaseTX if there would be some settings to make the connection more durable. Could it have something to do with power savings or "watchdogs"? I placed the autoselect settings into my /etc/hostname.re0.

The dropped connections occur 0 to 3x/day and are irritating. The last occurred while I was trying to generate a port diff.

[jggimi]

Originally Posted by shep

I'm wondering if the PHY ...

Does your dmesg discover the PHY?

The dropped connections occur 0 to 3x/day..

I've had this type of symptom when cable connectors go bad. Typically, by using a wired connection on a laptop that gets moved frequently. Consider swapping cables, and seeing if that solves the problem. In my case, I cut the failed connector away and add a new one to the existing cable. All it takes is a package of connectors, a crimping tool, and a large bucket of swear words.

[shep]

pci3 at ppb2 bus 3
re0 at pci3 dev 0 function 0 "Realtek 8168" rev 0x11: RTL8168G/8111G (0x4c00), msi, address 70:4d:7b:26:b3:ed
rgephy0 at re0 phy 7: RTL8251 PHY, rev. 0
pchb1 at pci0 dev 24 function 0 "AMD AMD64 15/1xh Link Cfg" rev 0x00

I'll try another cable.

[jggimi]

The rgephy(4) man page says it is Gigabit capable. So I'll assume my first assumption above that the PHY is only 100Mbps capable was wrong.

[shep]

It looks like the problem may not be OpenBSD implementation. The new cable made no difference. I'm using LEDE firmware on my router and randomly unresponsive static connections are an issue currently being discussed in the LEDE forums.