Yes, that's what happens. Probably the pings stop eventually because ARP traffic cannot get through.
On Tue, Sep 20, 2011 at 09:49:57PM -0700, Aishwarya wrote: > Ok, I get it now. So, as soon as the first packet arrives, vswitch tells the > kernel module how to handle it, so that the kernel module will know how to > handle all subsequent packets. In that case, if I have just one flow, for > eg, if I continuously send some ping packets from one vm to another vm (I am > assuming these belong to the same flow?), once a few pings have gone > through, and the vswitchd dies, the remaining pings should still go through? > I don't see this happening though. The communication still stops. > > Thanks, > Aish > > On Tue, Sep 20, 2011 at 9:46 PM, Ben Pfaff <[email protected]> wrote: > > > Yes, ovs-vswitchd tells the kernel module what to do. > > > > On Tue, Sep 20, 2011 at 09:40:28PM -0700, Aishwarya wrote: > > > Ok. So what happens with other packets that belong to this same flow? Do > > > they get routed without the help of vswitchd then? If yes, how does this > > > happen? > > > > > > Thanks, > > > Aish > > > > > > On Tue, Sep 20, 2011 at 9:40 PM, Ben Pfaff <[email protected]> wrote: > > > > > > > On Tue, Sep 20, 2011 at 09:36:30PM -0700, Aishwarya wrote: > > > > > Thanks for your detailed reply. Does this mean that every time a > > packet > > > > > arrives, it is first sent to the user process vswitchd and then sent > > back > > > > > into the kernel for forwarding, which means there will be a > > > > kernel-to-user > > > > > space and back to kernel space context switch for every packet that > > > > arrives? > > > > > > > > No. There is one such transition for each new flow. > > > > > > _______________________________________________ discuss mailing list [email protected] http://openvswitch.org/mailman/listinfo/discuss
