[prev in list] [next in list] [prev in thread] [next in thread] 

List:       e1000-devel
Subject:    Re: [E1000-devel] Strange behavior w/ e1000-7.2.7 + linux 2.6.10
From:       "Verbeeck Derek" <DVerbeeck () LIO ! AACISD ! com>
Date:       2007-08-23 12:37:24
Message-ID: BE8B3EECA608414FB163FBA4576FCEBD01650FC0 () EXNY ! us ! corp ! aacisd ! com
[Download RAW message or body]

Tried 2.6.22 on the diskless client. Same results. Slightly different
error output from the NFS, but same end result. The network service
dies, both slaves are lost. 

This is only happening on systems where the hardware clock is too far
off from the system time, so I guess I should find a workaround for that
issue rather than chase down what's happening to the bonding driver, as
the new kernel didn't fix that issue.   

-----Original Message-----
From: Kok, Auke [mailto:auke-jan.h.kok@intel.com] 
Sent: Wednesday, August 22, 2007 4:00 PM
To: Verbeeck Derek
Cc: e1000-devel@lists.sourceforge.net
Subject: Re: [E1000-devel] Strange behavior w/ e1000-7.2.7 + linux
2.6.10 + ntp

Verbeeck Derek wrote:
> I also forgot to mention that on the systems in question, the hardware

> clock winds up being about a day off from the system clock. The e1000 
> driver does not have any sort of interaction with the clock, right? 
> The readme doesn't say anything about sleep or suspend modes, and I'd 
> imagine the card would have to always remain in an awake state for WoL

> to function properly anyway. Thanks for your input again, Auke.

everything in a system that runs at kernel level can influence the
system time, so inherently the e1000 driver could as well. If the system
is running under heavy stress and doesn't yield enough to the processor
(slow systems and
HZ=1000) then this can easily be achieved. Try setting HZ=100 and
switching clocksource.

Auke

> 
> Regards,
> 
> Derek
> 
> -----Original Message-----
> From: Kok, Auke [mailto:auke-jan.h.kok@intel.com]
> Sent: Wednesday, August 22, 2007 2:30 PM
> To: Verbeeck Derek
> Cc: e1000-devel@lists.sourceforge.net
> Subject: Re: [E1000-devel] Strange behavior w/ e1000-7.2.7 + linux 
> 2.6.10 + ntp
> 
> Verbeeck Derek wrote:
>> Changing the kernel is not really an option for our production 
>> environment, but I will definitely try reproducing the failure with 
>> 7.6.5.
>>
>> 7.2.7 is not the stock driver for 2.6.10, either. We had to update 
>> from the stock driver to support newer revs of the card.
> 
> I would really suspect bonding to be the problem in your case, and 
> 2.6.10 is really old, many bugfixes have gone into the bonding code 
> since 2.6.10.
> 
> the e1000 driver doesn't really do much to affect bonding either.
> 
> I would at least attempt to run a newer kernel on maybe one of the 
> machines to see if this fixes your problem. At least then you can 
> investigate whether it is worth backporting bonding code to that 
> kernel version in case it fixes your problem.
> 
> Cheers,
> 
> Auke

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/
_______________________________________________
E1000-devel mailing list
E1000-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/e1000-devel
[prev in list] [next in list] [prev in thread] [next in thread]