IIRC this is actually a hardware bug in some revisions of e1000e compatible chipsets (i218/i219). I researched this when one of my customers had daily hangs because of this.
The solution? Swap out your NICs. Even cheapo Realtek ones are better than hangs. This problem has been present for a loooong time (not only in IPfire of course) and it must be fixed upstream.
unfortunately the bug is still present in the core 160
after about 3 days after a restart after installing the 160 version, the green network card has hung up again
the usual command lets the network card continue to run without problems until the restart, then I have to enter the command again so that it runs until the next restart
maybe one day we could report the bug
e1000e 0000:00:19.0 green0: Detected Hardware Unit Hang:
00:03:48 kernel: TDH <c8>
00:03:48 kernel: TDT <e9>
00:03:48 kernel: next_to_use <e9>
00:03:48 kernel: next_to_clean <c8>
00:03:48 kernel: buffer_info[next_to_clean]:
00:03:48 kernel: time_stamp <1017eeb1c>
00:03:48 kernel: next_to_watch <c8>
00:03:48 kernel: jiffies <1017eebc0>
00:03:48 kernel: next_to_watch.status <0>
00:03:48 kernel: MAC Status <40080083>
00:03:48 kernel: PHY Status <796d>
00:03:48 kernel: PHY 1000BASE-T Status <3800>
00:03:48 kernel: PHY Extended Status <3000>
00:03:48 kernel: PCI Status <10>
00:03:50 kernel: e1000e 0000:00:19.0 green0: Detected Hardware Unit Hang:
00:03:50 kernel: TDH <c8>
00:03:50 kernel: TDT <e9>
00:03:50 kernel: next_to_use <e9>
00:03:50 kernel: next_to_clean <c8>
00:03:50 kernel: buffer_info[next_to_clean]:
00:03:50 kernel: time_stamp <1017eeb1c>
00:03:50 kernel: next_to_watch <c8>
00:03:50 kernel: jiffies <1017eec88>
00:03:50 kernel: next_to_watch.status <0>
00:03:50 kernel: MAC Status <40080083>
00:03:50 kernel: PHY Status <796d>
00:03:50 kernel: PHY 1000BASE-T Status <3800>
00:03:50 kernel: PHY Extended Status <3000>
00:03:50 kernel: PCI Status <10>
00:03:52 kernel: e1000e 0000:00:19.0 green0: Reset adapter unexpectedly
00:03:56 kernel: e1000e 0000:00:19.0 green0: NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None
00:03:58 kernel: e1000e 0000:00:19.0 green0: Detected Hardware Unit Hang:
00:03:58 kernel: TDH <0>
00:03:58 kernel: TDT <9>
00:03:58 kernel: next_to_use <9>
00:03:58 kernel: next_to_clean <0>
00:03:58 kernel: buffer_info[next_to_clean]:
00:03:58 kernel: time_stamp <1017eeee9>
00:03:58 kernel: next_to_watch <0>
00:03:58 kernel: jiffies <1017eefa8>
00:03:58 kernel: next_to_watch.status <0>
00:03:58 kernel: MAC Status <40080083>
00:03:58 kernel: PHY Status <796d>
00:03:58 kernel: PHY 1000BASE-T Status <3800>
00:03:58 kernel: PHY Extended Status <3000>
00:03:58 kernel: PCI Status <10>
00:04:00 kernel: e1000e 0000:00:19.0 green0: Detected Hardware Unit Hang:
I donât think that this is a bug. I run systems from old Intel 82576 till X740 and donât have unit hangs at all. Basic infos are missing here. What NIC are we talking about? Is that NIC onboard or an addon card? Is the systems firmware up to date?
If you all google âe1000e hangsâ, you will see that this is an extremely common problem in the Linux world for many years. It must be fixed upstream. Discussion here wonât help. If you have the problem, swap out your NICs.
The hardware bug I was referring too was likely this:
Look for â82573(V/L/E) TX Unit Hang Messagesâ (the driver is old, the error messages have changed)
That fix didnât help in my case, though. Ethtool commands didnât help, either. Itâs either a hardware or a driver bug (or a combination). Swap out your NICs!
The cards you have mentioned donât use the e1000e driver, they use the igb driver. This is a driver/hardware bug, and OP has indicated in the title that itâs about the e1000e driver.
The e1000e, or at least some hardware revisions using that driver, is buggy under Linux since, like, 10 years. It wonât be fixed in ipfire, it canât be fixed in ipfire, it must be fixed upstream. Swap out your network cards.
(btw if this reappeared in 158 itâs likely because of the linux-firmware 20210511 that was in that update, but it doesnât matter, as e1000e has been more-or-less broken since ten years, and hunting this bug downstream is a foolâs errand. It must be fixed upstream. It hasnât reliably been fixed upstream for ten years. Do the math)
Atm I run multiple 82572EI cards in 5 different systems and no NIC hangs at all @24/7.
In Private I had problems in the past with cheap chinese clone cards that died or chrashed because of thermal issues. Exspecially with X540/550 cards. When I found out, I improved the cooling. Since then no probs any more.
