On Mon, 2020-06-22 at 15:03 +0200, Florian Westphal wrote:
Paolo Abeni <pabeni(a)redhat.com> wrote:
> On Mon, 2020-06-22 at 13:18 +0200, Florian Westphal wrote:
> > Paolo Abeni <pabeni(a)redhat.com> wrote:
> > > I'm wondering if we should keep using an MPTCP-specific workqueue.
> > > Server side, we don't expect the work to trigger very often, but
> > > that ever happen, top will tell easily.
> > You mean switch msk->work from system_wq to a custom workqueue?
> yep, e.g. reusing and renaming 'pm_wq'. When the PM works a lot (in
> "ndiffports" mode as in issues/33), the pm_wq is visible in top - at
> least in dbg build, I never tested with regular kconfig - and I like
> being able to identify it easily. WDYT?
Can you do this?
I'll try to run the perf tests on non-debug build, but I see your
point. Let's keep the patch simple - as is now. We can look for a
workqueue only if mptcp workqueue usage proves to be relevant in non