Hi Eric,

Thanks for your hint on softnet_stat.  After reading up a little
bit of kernel networking code, I know were my packet is getting
dropped.
Thanks
Ching

--- On Fri, 5/20/11, Eric Dumazet <[email protected]> wrote:

> From: Eric Dumazet <[email protected]>
> Subject: Re: [E1000-devel] Question on net_stats->rx_dropped  setting to "0"
> To: "Filo FeFi" <[email protected]>
> Cc: [email protected]
> Date: Friday, May 20, 2011, 4:45 AM
> Le vendredi 20 mai 2011 à 03:30
> -0700, Filo FeFi a écrit :
> > 
> > --- On Thu, 5/19/11, Eric Dumazet <[email protected]>
> wrote:
> > 
> > > From: Eric Dumazet <[email protected]>
> > > Subject: Re: [E1000-devel] Question on
> net_stats->rx_dropped  setting to "0"
> > > To: "Filo FeFi" <[email protected]>
> > > Cc: [email protected]
> > > Date: Thursday, May 19, 2011, 9:15 PM
> > > Le jeudi 19 mai 2011 à 19:19 -0700,
> > > Filo FeFi a écrit :
> > > > Dear ixgbe developers:
> > > > 
> > > > I'm debugging a problem where some frames
> get dropped
> > > by the ixgbe
> > > > driver (version 2.0.44-k2), i.e.
> /proc/net/dev "drop"
> > > is not 0.
> > > > 
> > > > Reading the ixgbe-3.3.9/2.0.44.13/2.0.44.14
> source, I
> > > see the line
> > > > (in ixgbe_main.c ixgbe_update_stats()):
> > > "net_stats->rx_dropped = 0;" 
> > > > 
> > > > So, does this mean that ixgbe always reports
> "0" for
> > > RX dropped?
> > > > 
> > > > Under what circumstances would
> /proc/net/dev's drop
> > > count for ixgbe
> > > > be incremented/changed from "0"?
> > > > 
> > > 
> > > What is your kernel version ?
> > > 
> > 
> > My kernel version is 2.6.18 PAE
> 
> I was asking because we added in 2.6.37 following patch
> that can let
> rx_dropped apparent values be incremented in core network
> stack, not
> because of a network device drop.
> 
> But of course your kernel is older ;)
> 
> commit   
> caf586e5f23cebb2a68cbaf288d59dbbf2d74052
> net: add a core netdev->rx_dropped counter
> 
> In various situations, a device provides a packet to our
> stack and we
> drop it before it enters protocol stack :
> - softnet backlog full (accounted in
> /proc/net/softnet_stat)
> - bad vlan tag (not accounted)
> - unknown/unregistered protocol (not accounted)
> 
> We can handle a per-device counter of such dropped frames
> at core level,
> and automatically adds it to the device provided stats
> (rx_dropped), so
> that standard tools can be used (ifconfig, ip link, cat
> /proc/net/dev)
> 
> This is a generalization of commit 8990f468a (net:
> rx_dropped
> accounting), thus reverting it.
> 
> 
> 

------------------------------------------------------------------------------
What Every C/C++ and Fortran developer Should Know!
Read this article and learn how Intel has extended the reach of its 
next-generation tools to help Windows* and Linux* C/C++ and Fortran 
developers boost performance applications - including clusters. 
http://p.sf.net/sfu/intel-dev2devmay
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel&#174; Ethernet, visit 
http://communities.intel.com/community/wired

Reply via email to