Bug 142273

Summary:	Periodic network breakdowns
Product:	[openSUSE] SUSE LINUX 10.0	Reporter:	Axel Braun <axel.braun>
Component:	Kernel	Assignee:	Karsten Keil <karsten.keil>
Status:	RESOLVED WORKSFORME	QA Contact:	E-mail List <qa-bugs>
Severity:	Critical
Priority:	P5 - None	CC:	suse-beta
Version:	Final
Target Milestone:	---
Hardware:	i586
OS:	Other
Whiteboard:
Found By:	Other	Services Priority:
Business Priority:		Blocker:	---
Marketing QA Status:	---	IT Deployment:	---
Attachments:	output of hwinfo

Description Axel Braun 2006-01-10 11:35:31 UTC

Hi,
I'm running a ThinkPad A30 since 4 years under various flavours of SuSE. Now, after installation of 10.0, I have regular, periodic breakdowns of the network. A ping to a machine reports in between:
ping: sendmsg: No buffer space available

From various sources I found that buffer memory might be an issue:

tp-a30:/home/axel # sysctl net.ipv4.tcp_mem
net.ipv4.tcp_mem = 12288        16384   24576
tp-a30:/home/axel # sysctl net.core.wmem_max
net.core.wmem_max = 113664
tp-a30:/home/axel # sysctl net.core.rmem_max
net.core.rmem_max = 113664

OK, but this did not really help in solving the issue - esp. as it seems to be new in 10.0

Comment 1 Olaf Kirch 2006-01-10 15:06:57 UTC

Are you running any other significant amount of network traffic while
pinging? What network card are you using? Any heavy memory usage?

Please provide hwinfo output and attach it to this report. Thanks!

Comment 2 Axel Braun 2006-01-10 16:35:14 UTC

Last time it occured I had only KDE and the mail program (Polarbar Mailer, a java mail client) up and running. No other memory or network usage.
Network card: Have a look at www.axxite.com/brn/en/a30.html , there I have all hardware information entered. output of hwinfo is attached

Comment 3 Axel Braun 2006-01-10 16:37:54 UTC

Created attachment 62701 [details]
output of hwinfo

Comment 5 Olaf Kirch 2006-01-20 20:01:40 UTC

The card is an e100 on board NIC. Never heard of trouble with these.

Please try using the eepro100 driver; just for testing whether
that makes a difference. If it does, it would point to a problem in
the e100 driver.

Comment 6 Axel Braun 2006-01-23 19:37:51 UTC

With the eepro100 it does not work very well. Looks the e100 is better.
I was trying to reproduce, my favorite for the moment to reproduce is:
- have an NFS-mount running
- have the automounter running
- send the laptop to hibernation
- wake it up again, and it is likely that the problem occurs. 
Seems it does not occur if the automounter is not running, but I cant finally confirm this yet.

Comment 7 Chris L Mason 2006-01-23 19:51:53 UTC

Can you trigger this without the hibernation step?  That would significantly reduce the scope of the debugging.

Comment 8 Axel Braun 2006-01-23 20:01:34 UTC

Just was thinking about giving you the update that I had the problem right after a reboot, without autofs or hibernation: The mail application logs in to one mail account (larger amount of mails in, failed last time due to the network down). Right after the login was successful, the 'no-buffer' message appeared.
Can this be a reason (Sorry, I'm a no-programmer)?

Comment 10 Karsten Keil 2006-01-24 11:43:22 UTC

Please try to increase the debug output for this card, before please post
'ethtool eth0' output.
You can set the messagelevel with
'ethtool -s eth0 msglvl 0xf4'
Look at /var/log/messages if a hang occurs. Please try to avoid hibernation.
And please check also the network cable, some days ago I was looking at a similar problem myself and it was a broken contact.
The ping output basically means, that the networkstack doesn't sent out packets and the queues are full at some point.

Comment 11 Axel Braun 2006-02-12 13:01:16 UTC

for the last time it was running like clockwork, today it appeared again:
Settings for eth0:
        Supported ports: [ TP MII ]
        Supported link modes:   10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
        Supports auto-negotiation: Yes
        Advertised link modes:  10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
        Advertised auto-negotiation: Yes
        Speed: 100Mb/s
        Duplex: Full
        Port: MII
        PHYAD: 1
        Transceiver: internal
        Auto-negotiation: on
        Supports Wake-on: g
        Wake-on: g
        Current message level: 0x000000f4 (244)
        Link detected: yes

Ater the hang, /var/log/messages displayed:
kernel: e100: eth0: e100_update_stats: exec cuc_dump_reset failed

Comment 12 Karsten Keil 2007-02-22 18:22:39 UTC

Does the problem still exist with more current versions ?

Comment 13 Karsten Keil 2007-02-22 20:33:10 UTC

Got the info from Axel offline: 10.2 dosn't show this problem anymore.