On Tue, May 15, 2012 at 10:29 AM, Ben Pfaff <b...@nicira.com> wrote: > On Tue, May 15, 2012 at 10:25:22AM -0700, Justin Pettit wrote: >> On May 15, 2012, at 9:58 AM, Ben Pfaff wrote: >> >> > On Tue, May 15, 2012 at 09:40:46AM -0700, Justin Pettit wrote: >> >> That's quite a packet. The change looks reasonable to me--and 200 >> >> bytes seems adequate to cover that packet. However, it might be worth >> >> having Jesse double-check that that's the correct encapsulation in the >> >> example >> > >> > I am always in favor of more checking, especially by careful developers >> > like Jesse, but perhaps I gave an incorrect impression that this value >> > depends on kernel code? It doesn't. ODPUTIL_FLOW_KEY_BYTES is the >> > maximum number of bytes that the userspace function >> > odp_flow_key_from_flow() can write to a buffer. >> >> Okay. However, I thought that function depended on the number and >> type of attributes that the kernel used to represent a flow. You and >> Jesse have the best understanding of what that interface is, and since >> this hypothetical flow is complicated, I suggested Jesse may want to >> look at it. If you're satisfied that it's correct (or I'm >> misunderstanding something), I obviously trust your judgment. > > I'd be delighted to have Jesse take a look. Jesse?
It looks correct to me. I would like it if we had a more automated way to compute this (even beyond the unit test that you added) since we have run into this problem a couple times now but this is clearly an improvement. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
