Re: [Intel-wired-lan] [REGRESSION] e1000e probe/link detection fails since 6.2 kernel

[Takashi Iwai]

On Tue, 28 Mar 2023 16:39:01 +0200,
Paul Menzel wrote:
> 
> Dear Takashi,
> 
> 
> Am 28.03.23 um 14:40 schrieb Takashi Iwai:
> 
> > we've got a regression report for e1000e device on Lenovo T460p since
> > 6.2 kernel (with openSUSE Tumbleweed).  The details are found in
> >    https://bugzilla.opensuse.org/show_bug.cgi?id=1209254
> 
> Thank you for forwarding the report.
> 
> > It seems that the driver can't detect the 1000Mbps but only 10/100Mbps
> > link, eventually making the device unusable.
> > 
> > On 6.1.12:
> > [    5.119117] e1000e: Intel(R) PRO/1000 Network Driver
> > [    5.119120] e1000e: Copyright(c) 1999 - 2015 Intel Corporation.
> > [    5.121754] e1000e 0000:00:1f.6: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode
> > [    7.905526] e1000e 0000:00:1f.6 0000:00:1f.6 (uninitialized): Failed to disable ULP
> > [    7.988925] e1000e 0000:00:1f.6 0000:00:1f.6 (uninitialized): registered PHC clock
> > [    8.069935] e1000e 0000:00:1f.6 eth0: (PCI Express:2.5GT/s:Width x1) 50:7b:9d:cf:13:43
> > [    8.069942] e1000e 0000:00:1f.6 eth0: Intel(R) PRO/1000 Network Connection
> > [    8.072691] e1000e 0000:00:1f.6 eth0: MAC: 12, PHY: 12, PBA No: 1000FF-0FF
> > [   11.643919] e1000e 0000:00:1f.6 eth0: NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None
> > [   15.437437] e1000e 0000:00:1f.6 eth0: NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None
> > 
> > On 6.2.4:
> > [    4.344140] e1000e: Intel(R) PRO/1000 Network Driver
> > [    4.344143] e1000e: Copyright(c) 1999 - 2015 Intel Corporation.
> > [    4.344933] e1000e 0000:00:1f.6: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode
> > [    7.113334] e1000e 0000:00:1f.6 0000:00:1f.6 (uninitialized): Failed to disable ULP
> > [    7.201715] e1000e 0000:00:1f.6 0000:00:1f.6 (uninitialized): registered PHC clock
> > [    7.284038] e1000e 0000:00:1f.6 eth0: (PCI Express:2.5GT/s:Width x1) 50:7b:9d:cf:13:43
> > [    7.284044] e1000e 0000:00:1f.6 eth0: Intel(R) PRO/1000 Network Connection
> > [    7.284125] e1000e 0000:00:1f.6 eth0: MAC: 12, PHY: 12, PBA No: 1000FF-0FF
> > [   10.897973] e1000e 0000:00:1f.6 eth0: NIC Link is Up 10 Mbps Full Duplex, Flow Control: None
> > [   10.897977] e1000e 0000:00:1f.6 eth0: 10/100 speed: disabling TSO
> > [   14.710059] e1000e 0000:00:1f.6 eth0: NIC Link is Up 10 Mbps Full Duplex, Flow Control: None
> > [   14.710064] e1000e 0000:00:1f.6 eth0: 10/100 speed: disabling TSO
> > [   59.894807] e1000e 0000:00:1f.6 eth0: NIC Link is Up 10 Mbps Full Duplex, Flow Control: None
> > [   59.894812] e1000e 0000:00:1f.6 eth0: 10/100 speed: disabling TSO
> > [   63.808662] e1000e 0000:00:1f.6 eth0: NIC Link is Up 10 Mbps Full Duplex, Flow Control: None
> > [   63.808668] e1000e 0000:00:1f.6 eth0: 10/100 speed: disabling TSO
> > 
> > The same problem persists with 6.3-rc3.
> > 
> > Can you guys check what can go wrong, or if there is a fix?
> 
> Does openSUSE Tumbleweed make it easy to bisect the regression at
> least on “rc level”? It be great if narrow it more down, so we know
> it for example regressed in 6.2-rc7.

Not easy, unfortunately.  The official TW kernel moved to a newer
version only after the final release.  Although the rc kernel packages
are provided as unofficial packages, they are transient and go away at
each update, so we have no archives for testing the older ones.

Or, if you can suggest some commit to be tested for revert, I can
create a test kernel package at any time.


thanks,

Takashi