Hi, The kernel version is 3.8.0-38 ( Linux ccpu1 3.8.0-38-generic #56~precise1-Ubuntu) I think you are right. it is the kernel datapath which is used by the patch ports. I am not too familiar with working you explained about the backlog queue. i will have to explore the suggestions that you have made. Thanks for providing them. i was hoping for a quick solution because 65 is not too high a number for limiting packet hops and also i dont know how the ovs keeps a record of these hops. I use a python script to generate my topology and there are no manual error in creating and setting up the links and flows etc.
The packet is dropped by the 66th switch as i posted with my original question. i have pasted the output of dump-ports on the 66th switch. that switch receives on one port but doesn't forward on another port even when there is a flow written to forward that. on a side note when i have a grid topology of 20 * 20 (400)switches all of them forwards the traffic. I have written the algo in a manner in which first switch forwards to the right and down direction and all the switches in first column forwards to the right. so whenever the packets reach the hop limit in any direction it stops forwarding packets. Let me know if anyone needs more info on this. Thanks KP On Wed, Aug 13, 2014 at 9:57 PM, Flavio Leitner <[email protected]> wrote: > > Are you using datapath in kernel or in userspace? > > What's the kernel version? > > If it's kernel DP, then patch ports are internal ports which during TX > inserts the packet (skb) into the CPU backlog queue (enqueue_to_backlog()) > for later processing. Later on, the backlog queue is processed > (process_backlog()) and the packet is actually received by the other port. > > The above code path contains trace points which you can hook using > systemtap or perf tools to see what is going on. Another approach is > using 'dropwatch' tool to find out where the packet might have been > dropped. > > I don't think it's related to TTL and I don't remember to have seen > any check like that. > > fbl > > > On Wed, Aug 13, 2014 at 01:23:13PM -0400, Krishna Pratap wrote: >> i dont know how i didnt recevie the first email from Duy. >> >> this is the original query >> http://openvswitch.org/pipermail/discuss/2014-August/014741.html >> >> >> On Wed, Aug 13, 2014 at 1:15 PM, Ben Pfaff <[email protected]> wrote: >> > Please don't drop the list. >> > >> > Open vSwitch does not have such a check. >> > >> > On Wed, Aug 13, 2014 at 01:06:46PM -0400, Krishna Pratap wrote: >> >> exactly, i have been wondering the same. do you know where this check >> >> is done the code? it doesnt forward anything after 65 hops >> >> >> >> On Wed, Aug 13, 2014 at 1:04 PM, Ben Pfaff <[email protected]> wrote: >> >> > On Wed, Aug 13, 2014 at 10:02 AM, Krishna Pratap >> >> > <[email protected]> wrote: >> >> >> TTL doesnt seems to be the problem. i have considered that. anyhow if >> >> >> you check the traffic in the output that i posted the TTL is set as >> >> >> 255. Also i dont think TTL should matter if we are forwarding packets >> >> >> at layer 2. >> >> >> >> >> >> I am guessing there is a check somewhere in OVS code which limits it. >> >> >> if anyone can point me out where it is or suggest a workaround, it >> >> >> would be great. I have not been able to find that check yet. >> >> > >> >> > I don't know how Open vSwitch would even know how many times a packet >> >> > had been forwarded. >> >> >> >> -- >> Best Regards, >> Krishna Pratap Singh >> _______________________________________________ >> discuss mailing list >> [email protected] >> http://openvswitch.org/mailman/listinfo/discuss >> -- Best Regards, Krishna Pratap Singh _______________________________________________ discuss mailing list [email protected] http://openvswitch.org/mailman/listinfo/discuss
