Bug 1742

Summary:	e1000e: no ethernet network, kworker eating CPU
Product:	Mageia	Reporter:	Olivier FAURAX <olivier>
Component:	RPM Packages	Assignee:	Thomas Backlund <tmb>
Status:	RESOLVED FIXED	QA Contact:
Severity:	major
Priority:	Normal	CC:	tmb
Version:	1
Target Milestone:	---
Hardware:	i586
OS:	Linux
Whiteboard:
Source RPM:	kernel-desktop586-latest-2.6.38.7-1.mga1	CVE:
Status comment:
Attachments:	Dmesg output after boot dmidecode output (with pcie_ports=compat)

Description Olivier FAURAX 2011-06-11 03:39:49 CEST

Even if the card is correctly detected (I see it the wizard), there is no detection when I plug it.
In the messages I see that "the link is not ready".

e1000e          : Intel Corporation|82566MC Gigabit Network Connection [NETWORK_ETHERNET] (vendor:8086 device:104d subv:17aa subd:20ba) (rev: 03)

I rank it as critical as ethernet is basic use of a computer, and one might discard mageia for this.

Comment 1 Olivier FAURAX 2011-06-11 18:59:05 CEST

Also, there is a kworker process that eats CPU "by peak" at least, half of the time.
On rmmod e1000e, there is no such CPU consumption.
This is related to bug 1743.

So, the e1000e module is suspect, as:
 * ethernet doesn't work as expected
 * removing the module solve the kworker problem

Summary: No network with ethernet card => e1000e: no ethernet network, kworker eating CPU
Source RPM: kernel => kernel-desktop586-latest-2.6.38.7-1.mga1

Comment 2 Dick Gevers 2011-06-11 19:10:43 CEST

Is it a card or onboard integrated port(s). If a card could it have become slightly unseated?

CC: (none) => dvgevers

Comment 3 Thomas Backlund 2011-06-11 19:58:45 CEST

Can you attach output of dmesg after boot.

Does it help if you disable link detection in drakconf / draknet ?

CC: (none) => tmb

Thomas Backlund 2011-06-11 19:59:15 CEST

Assignee: bugsquad => tmb

Comment 4 Olivier FAURAX 2011-06-14 09:16:12 CEST

Created attachment 565 [details]
Dmesg output after boot

There is Call Trace in the dmesg....

Comment 5 Olivier FAURAX 2011-06-14 09:21:36 CEST

I don't know if it's related, but I just removed dkms-nvidia96xx, and it seems that the link is detected now.
I don't know where to disable link detection.

I'll try pcie_ports=compat ASAP.

Comment 6 Olivier FAURAX 2011-06-15 21:12:26 CEST

pcie_ports=compat solves the issue.

tmb, do you need something else to handle this bug ?

Comment 7 Thomas Backlund 2011-06-15 21:22:12 CEST

 attach output of dmidecode

Comment 8 Olivier FAURAX 2011-06-15 21:27:51 CEST

Created attachment 574 [details]
dmidecode output (with pcie_ports=compat)

Here is the output of dmidecode.
I booted with pcie_ports=compat (if it changes anything).

Comment 9 Olivier FAURAX 2011-06-25 12:42:53 CEST

Any progress on this bug? Any info missing?

Comment 10 Dick Gevers 2011-06-25 12:57:22 CEST

What please is the result of (as root) of the command:

ifconfig

# and if eth0 does not show, of the command:

ifup eth0

?
Can you attach the output of:

service --status-all >/tmp/ssa.txt 2>&1

You probably have to change the ownership of /tmp/ssa.txt to your own (not root) before you can upload it.

Comment 11 Olivier FAURAX 2011-06-25 13:03:31 CEST

Do you need this with or without booting with pcie_ports=compat (which make the card works)?

Comment 12 Dick Gevers 2011-06-25 16:02:54 CEST

With this parameters it's solved, so I'd say without. (But I don't *need* anything, I'm just trying to get some data to enable someone to analyze it further).

Comment 13 Olivier FAURAX 2011-06-27 21:59:39 CEST

OK. On IRC one week ago, tmb said he didn't need anything else, so...

Comment 14 Dick Gevers 2011-06-27 22:40:16 CEST

Okay. Then my help is not required. Sorry for the noise.

CC: dvgevers => (none)

Comment 15 Olivier FAURAX 2011-09-03 22:41:27 CEST

Fixed in a kernel update (don't remember which one).

Status: NEW => RESOLVED
Resolution: (none) => FIXED

Comment 16 Olivier FAURAX 2011-10-17 22:50:27 CEST

The problem reappeared in 2.6.38.8-desktop586-6.mga

Status: RESOLVED => REOPENED
Resolution: FIXED => (none)

Comment 17 Olivier FAURAX 2011-10-17 23:54:13 CEST

I tried to add pcie_ports=compat and this doesn't work :(

Comment 18 Olivier FAURAX 2011-10-17 23:56:40 CEST

However, if I "sudo rmmod e1000e", there is no CPU problem.
So the bug might be related to e1000e but not with pcie_ports....

Comment 19 Olivier FAURAX 2011-10-23 11:01:15 CEST

I don't know why, but this bug disappeared.
I didn't change the command line to boot in grub, but perhaps the regeneration of grub config in the MCC did something....

Status: REOPENED => RESOLVED
Resolution: (none) => FIXED