RE: [PATCH][UPSTREAM]net,RT:Remove preemption disabling in netif_rx()

[Peter Zijlstra <peterz@infradead.org>]

On Fri, 2012-05-25 at 18:43 -0400, Steven Rostedt wrote:
> On Sat, 2012-05-26 at 00:31 +0200, Thomas Gleixner wrote:
> > On Thu, 24 May 2012, Steven Rostedt wrote:
> > 
> > > On Thu, 2012-05-24 at 04:28 +0000, Jain Priyanka-B32167 wrote:
> > > > Waiting for review comments on this.
> > > > 
> > > 
> > > > diff --git a/net/core/dev.c b/net/core/dev.c index 452db70..4017820 100644
> > > > --- a/net/core/dev.c
> > > > +++ b/net/core/dev.c
> > > > @@ -2940,7 +2940,7 @@ int netif_rx(struct sk_buff *skb)
> > > >  		struct rps_dev_flow voidflow, *rflow = &voidflow;
> > > >  		int cpu;
> > > >  
> > > > -		preempt_disable();
> > > > +		migrate_disable();
> > > 
> > > I really want to avoid placing open coded migrate_disable() around the
> > > kernel. Perhaps we should use "get_cpu_light()" here too.
> > 
> > No. get_cpu_light() and migrate_disable() are different.
> > 
> > Following your argument we would have to replace preempt_disable()
> > with get_cpu() all over the place.
> > 
> 
> I didn't like the get_cpu_light either, but I thought it was Peter that
> was against a 'migrate-disable()' API leaking all over the kernel.
> 
> IIRC, he gave in when it was part of a locking internal infrastructure.
> Now it's coming to what he was against. Maybe he changed his mind. I'll
> let him speak for himself.

Ah, right, so I've mapped migrate_disable() to preempt_disable() for !
PREEMPT_RT. This to make sure people don't do stupid things with it :-)

Then again, people do lots of stupid things with preempt_disable() too,
but at least we don't grow actual migrate_disable() abuse.

I think I've done one or two migrate_disable() site like the proposed in
-rt already.

The thing I really want to avoid is introducing a migrate_disable() that
can schedule for !rt.