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

List:       e1000-devel
Subject:    [E1000-devel] Fwd: RE:  IGB driver 1.2.44.9 problem
From:       David Lawless <david_lawless () flumedata ! com>
Date:       2008-11-20 1:14:00
Message-ID: 6.2.5.6.2.20081119200117.01d61f78 () flumedata ! com
[Download RAW message or body]

Hi Alex,

I'm back from the mines, and can now spend some time on the 
'igb' driver and the 82575.

Since your message was written I did build and install the 
1.3.8.6 version with DCA enabled.  As with 1.2.44.9 only one 
interrupt is allocated instead of four as seen with RHEL5 and 
it is in 'PCI-MSI-edge' rather than in 'MSI-X' mode.  Haven't
tried 1.3.8.6 with the RHEL5 kernel 2.6.18-92.1.13.el5,
this was done with F9 and 2.6.27.1 kernel.org kernel running.

A positive result from switching to 1.3.8.6 was that some 
spurious packet loss was eliminated and now the 82575 and 82571 
systems are seeing identical UPD data streams.

So far DCA does not seem to make any difference.  The Opteron 
2354 system running 82571s+e1000e uses slightly less CPU to 
acquire packets than the Xeon E5430 running 82575s+DCA+igb 
despite running at a substantially slower clock speed.  However 
even if DCA is a bust the multiple queue support will be
attractive once it's working.

In no particular order

1) should the patch be applied to 'igb' 1.3.8.6 and tried out?

2) should a newer kernel.org kernel be built?

3) should any boot parameters be included?

Thanks,

David






> Date: Mon, 20 Oct 2008 13:33:57 -0400
> To: "Duyck, Alexander H" <alexander.h.duyck@intel.com>
> From: David Lawless <david_lawless@flumedata.com>
> Subject: RE: [E1000-devel] IGB driver 1.2.44.9 problem
> Cc: "e1000-devel@lists.sourceforge.net" <e1000-devel@lists.sourceforge.net>, \
> "Ronciak, John" <john.ronciak@intel.com> 
> Alex,
> 
> Thank you for the patch.  I've just hit an serious
> work crunch, so I have to set this project aside
> for about a month.  Sorry about that.
> 
> I will check out the patch once the other critical
> deadline is met.
> 
> David
> 
> 
> 
> 
> At 09:40 10/20/2008 -0700, Duyck, Alexander H wrote:
> > David,
> > 
> > The attached patch does two things.  One it adds a bit more to 
> > the dmesg output for the driver when the interface is brought up 
> > via ifconfig.  You should seem some messages about MSIXBM(x) in 
> > there.  Also there is a workaround included to bring the link up 
> > without need for the link status change interrupt, but it will 
> > only work on initial driver load if the interface is already 
> > connected.
> > 
> > My main concern at this point is that the interrupts don't seem 
> > to be making it from the adapter to the system.  This could 
> > either be due to a misrouting or dropping of the interrupt.  If 
> > you could try the attached patch and let me know how it works I 
> > would appreciate it.  Basically the two things I would be 
> > looking for would be the dmesg output after you configure the IP 
> > address of the interface, and a dump of /proc/interrupts after 
> > trying to pass some traffic.
> > 
> > Thanks,
> > 
> > Alex


-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
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]