[ovs-dev] [PATCH ovs-dev v1] netdev-offload: Flush offloaded rules when ports removed

William Tu u9012063 at gmail.com
Thu Jul 16 14:23:40 UTC 2020


> > > > Hi
> > > > Thanks for your review. I test it with openflow, It works fine because
> > > > the *revalidator
> > > > will clean up the flows(which may be offloaded).
> > > > > In case above scenario leaves ufids in OVS that cannot be removed, I'd suggest
> > > > > reverting of the previous patch that allowed offloading via add-flow since this
> > > > > is effectively a memory leak and even debugging commands should not produce
> > > > > a memory leak.
> > > > yes, we can revert the previous patch. But I think the command is
> > > > useful for us to
> > > > debug offload(rte flow offload and tc offload), and test the offload function.
> > > > For this issue, can we use this little patch to fix it ? or other better way ?
> > >
> > > I am confused. Which little patch?
> > Hi Simon
> > This patch [1]:
> > [1] - http://patchwork.ozlabs.org/project/openvswitch/patch/20200611103635.53367-1-xiangxia.m.yue@gmail.com/
> >
> > If we revert the patch which commit id is
> > e61984e781e6c7d621568428788cb87c11be8f1f
> > and we can't debug or install the flows with "ovs-app dpctl/add-flow",
> > for tc offload.
> > I think that tools are useful for us.
> Hi Simon, Ilya
> Do we have plan to apply this patch, or drop this patch ?

My understanding from reading the discussion is that
using ovs-appctl dpclt/add-flow is just for debugging purposes.
When using the right way, the ovs-ofctl add-flow, there is no such problem.

I don't have any strong opinion on drop or apply.
William


More information about the dev mailing list