[Intel-wired-lan] Slow e1000e Network Interface with different kernel

[Intel-wired-lan] Slow e1000e Network Interface with different kernel

[Marco Witte]

Scenario: Plugin into network, establishes link and IP is assigned.
scp from another machine in the same subnet gives approx. 1kb/sec.
Cross check with an usb 3 dock with integrated network adapter. Full speed.

Kai Heng-Feng build a patch for tgp https://lkml.org/lkml/2021/7/12/2927
His deb kernel packages with that patch based on 5.10.x kernel ran full 
throttle on this machine.
However this was the only successfull attempt to get full speed.

The Kernel Mainline 5.13. contains visual similar patches, but they 
still have the slow down on the same machine.
Some people suspect pcie changes at that point.

Some curious fact I found out while testing: I booted up a manjaro 
(current) usb stick and had full speed on the e1000e. Then I installed 
manjaro from that stick to the harddrive. Booted that and it didn't 
perform well.

As this is my first rodeo on this mailinglist, I need to reference some 
information:

Referencing: https://bugzilla.kernel.org/show_bug.cgi?id=213377
Here is the launchpad: 
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1927925
e1000e Sourceforge bugreport(s): https://sourceforge.net/p/e1000/bugs/681/
This is still an issue with all current mainline kernels. Last test was 
5.14 RC7

I can provide any output you need.

[Intel-wired-lan] Slow e1000e Network Interface with different kernel

[Paul Menzel]

Dear Marco,


Am 02.09.21 um 12:01 schrieb Marco Witte:
> Scenario: Plugin into network, establishes link and IP is assigned.
> scp from another machine in the same subnet gives approx. 1kb/sec.
> Cross check with an usb 3 dock with integrated network adapter. Full speed.

Just to be clear. Did it ever work in the past?

> Kai Heng-Feng build a patch for tgp https://lkml.org/lkml/2021/7/12/2927
> His deb kernel packages with that patch based on 5.10.x kernel ran full 
> throttle on this machine.
> However this was the only successful attempt to get full speed.

That sounds more like a network device problem.

> The Kernel Mainline 5.13. contains visual similar patches, but they 
> still have the slow down on the same machine.
> Some people suspect pcie changes at that point.
> 
> Some curious fact I found out while testing: I booted up a manjaro 
> (current) usb stick and had full speed on the e1000e. Then I installed 
> manjaro from that stick to the harddrive. Booted that and it didn't 
> perform well.
> 
> As this is my first rodeo on this mailinglist, I need to reference some 
> information:
> 
> Referencing: https://bugzilla.kernel.org/show_bug.cgi?id=213377
> Here is the launchpad: 
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1927925
> e1000e Sourceforge bugreport(s): https://sourceforge.net/p/e1000/bugs/681/
> This is still an issue with all current mainline kernels. Last test was 
> 5.14 RC7
> 
> I can provide any output you need.

What devices is it exactly? Is the link speed each time 1 Gb/s? Linux 
logs that like

     [   10.871450] e1000e 0000:00:19.0 eno1: NIC Link is Up 1000 Mbps 
Full Duplex, Flow Control: Rx/Tx

You can also check that with `ethtool`, for example, `/usr/sbin/ethtool 
eno1`.


Kind regards,

Paul

[Intel-wired-lan] Slow e1000e Network Interface with different kernel

[Sasha Neftin]

On 9/2/2021 13:01, Marco Witte wrote:
> Scenario: Plugin into network, establishes link and IP is assigned.
> scp from another machine in the same subnet gives approx. 1kb/sec.
> Cross check with an usb 3 dock with integrated network adapter. Full speed.
> 
> Kai Heng-Feng build a patch for tgp https://lkml.org/lkml/2021/7/12/2927
> His deb kernel packages with that patch based on 5.10.x kernel ran full 
> throttle on this machine.
> However this was the only successfull attempt to get full speed.
> 
> The Kernel Mainline 5.13. contains visual similar patches, but they 
> still have the slow down on the same machine.
> Some people suspect pcie changes at that point.
> 
> Some curious fact I found out while testing: I booted up a manjaro 
> (current) usb stick and had full speed on the e1000e. Then I installed 
> manjaro from that stick to the harddrive. Booted that and it didn't 
> perform well.
> 
> As this is my first rodeo on this mailinglist, I need to reference some 
> information:
> 
> Referencing: https://bugzilla.kernel.org/show_bug.cgi?id=213377
> Here is the launchpad: 
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1927925
> e1000e Sourceforge bugreport(s): https://sourceforge.net/p/e1000/bugs/681/
> This is still an issue with all current mainline kernels. Last test was 
> 5.14 RC7
> 
> I can provide any output you need.
> 
Marco,
Run echo on > /sys/bus/pci/devices/0000\:00\:16.0/power/control command 
on your machine.
Please, let us know the result.

> 
> _______________________________________________
> Intel-wired-lan mailing list
> Intel-wired-lan at osuosl.org
> https://lists.osuosl.org/mailman/listinfo/intel-wired-lan
Sasha