After more testing, it appears that there is still a problem with the Intel e1000e driver and some – but not all – supported NICs..

I have several ex-HP systems that have Intel NICs, and some have more than one.   The HP Z820 has two (different models..):

$ lspci |grep Eth
00:19.0 Ethernet controller: Intel Corporation 82579LM Gigabit Network Connection (Lewisville) (rev 05)
01:00.0 Ethernet controller: Intel Corporation 82574L Gigabit Network Connection

I also have an HP Z220CMT, with the same two (different) NICs, and this exhibits the same behaviour..

With the latest kernels (tested with 5.5-rc6 and 5.4.11) the 82574L works correctly, but the 82579LM (eth0) fails, with constant up/down:

.................
e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: Rx/Tx
userif-2: sent link down event.
userif-2: sent link up event.
userif-2: sent link down event.
userif-2: sent link up event.
userif-2: sent link down event.
userif-2: sent link up event.
e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: Rx/Tx
userif-2: sent link down event.
userif-2: sent link up event.
userif-2: sent link down event.
userif-2: sent link up event.
userif-2: sent link down event.
userif-2: sent link up event.
........ repeated ..............

And, ethtool reports:

................
Speed: Unknown!
Duplex: Unknown! (255)
..................
Current message level: 0x00000007 (7)
drv probe link
Link detected: no

I found that – sometimes – after some minutes of up/down messages, the link eventually works..

I had also previously tested this with an official Fedora 31 kernel 5.4.5-300.fc31.x86_64, and the problem was the same.

The fix, is to revert the patch to netdev.c, as mentioned in a previous article:  http://rglinuxtech.com/?p=2632

Robert Gadsdon.  January 13th 2020.