Re: [PATCH v6.1-rt 00/15] net: Backport backlog-NAPI threads to resolve PREEMPT_RT softirq warning in SMP flush

[Clark Williams]

On Wed, Oct 14, 2026 at 05:08:53PM +0200, Sebastian Andrzej Siewior wrote:
> Please keep the networking folks out of the loop. This is a backport
> request which goes entirely via the RT-queue.
> 
> On 2026-10-14 01:58:34 [+0800], wen.yang@xxxxxxxxx wrote:
> > From: Wen Yang <wen.yang@xxxxxxxxx>
> > 
> > *** Overview ***
> > Backport the backlog-NAPI smpboot-thread mechanism and its prerequisite
> > NAPI/RPS/defer_list work from upstream (v6.4 - v6.9) to v6.1-rt-next.
> > 
> > *** Problem ***
> > On PREEMPT_RT, RPS steers packets to a remote CPU's backlog by sending an
> > IPI. The IPI handler (rps_trigger_softirq) calls ____napi_schedule(), which
> > raises NET_RX_SOFTIRQ. do_softirq_post_smp_call_flush() then sees a softirq
> > pending that was not pending before the flush, and NET_RX_SOFTIRQ is not
> > masked by SCHED_SOFTIRQ_MASK, so it fires:
> 
> I was about to point out that v6.1-RT should have duct tape for this but
> from reading 1/15 is see that this duct is not complete and is missing
> RPS. Indeed.
> 
> A backport of what is merged is my preference, too. However, we should
> get v6.6-rt fixed first then v6.1-rt. Otherwise users upgrading from
> v6.1 to v6.6 will have this case broken.
> 
> Clark, any objections here from your side?
> 
> Sebastian

I don't have a fundamental objection to the backport and agree we should
do v6.6-rt first, then move that back to v6.1-rt.

So it sounds like we need to pull in the RPS code first?

Clark

-- 
The United States Coast Guard
Ruining Natural Selection since 1790