After updating my main HP Z820 system to Kernel 5.4, I found that the previous Intel e1000e Ethernet problem I had encountered on my test system with Kernel 5.4-rc6 seemed to have reappeared. (see http://rglinuxtech.com/?p=2632 ).

But – this time, two of my systems with e1000e drivers were OK, but the third – the HP Z820 – was not, and the Ethernet connection kept going up and down, and was unuseable :

...............................
[ 193.640586] e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: Rx/Tx
[ 193.840732] userif-2: sent link down event.
[ 193.840733] userif-2: sent link up event.
[ 194.866096] userif-2: sent link down event.
[ 194.866098] userif-2: sent link up event.
[ 199.974917] userif-2: sent link down event.
[ 199.974919] userif-2: sent link up event.
[ 200.443424] e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: Rx/Tx
[ 200.643597] userif-2: sent link down event.
[ 200.643599] userif-2: sent link up event.
[ 201.650030] userif-2: sent link down event.
[ 201.650032] userif-2: sent link up event.
[ 206.779077] userif-2: sent link down event.
[ 206.779079] userif-2: sent link up event.
[ 207.231956] e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: Rx/Tx
..................  ( repeated ) ...............................

The problem is compounded, as two of the systems have apparently-identical 82574L Gigabit devices, and one (Asrock Xeon) works, while the other (HP Z820 Xeon) does not..    The third (working) system has an 82579LM.

On the Z820, I found that – as mentioned in the 5.4-rc6 article – reverting the patch fixed the problem.

More research is needed..

Robert Gadsdon.   November 24th 2019.