Re: [dpdk-dev] [PATCH] eal: support using 0 as coremask for no-affinitization

[Bruce Richardson <bruce.richardson@intel.com>]

On Wed, Apr 14, 2021 at 09:29:38AM -0700, Stephen Hemminger wrote:
> On Wed, 14 Apr 2021 18:15:34 +0200
> David Marchand <david.marchand@redhat.com> wrote:
> 
> > On Tue, Feb 16, 2021 at 10:52 AM Bruce Richardson
> > <bruce.richardson@intel.com> wrote:
> > >
> > > On Tue, Feb 16, 2021 at 09:43:00AM +0000, Bruce Richardson wrote:  
> > > > Allow the user to specify that they don't want any core pinning from DPDK
> > > > by passing in the coremask of 0.
> > > > ---  
> > >
> > > Apologies, missed my signoff, will add in V2, but will wait for feedback on
> > > this V1 first.  
> > 
> > I did not see a v2, is this still for 21.05?
> > 
> > What are the usecases for this feature?
> > I could think of using this in combination with external threads in
> > applications like OVS... ?
> > 
> > 
> 
> It would cause lots of DPDK applications to discover the problems with
> preemption and priority inversion with DPDK spinlocks...

I'd actually be surprised. I would expect most apps that want this sort of
thing are already just using a single master core (perhaps with faked EAL
args) and adding their own threads on top of that. All this patch would do
is remove the requirement for that master lcore.

/Bruce