[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH v2 2/2] xen: move per-cpu area management into common code



On Wed, 2024-09-25 at 17:12 +0200, Jan Beulich wrote:
> On 24.09.2024 18:42, Oleksii Kurochko wrote:
> > --- a/xen/include/asm-generic/percpu.h
> > +++ b/xen/include/asm-generic/percpu.h
> > @@ -10,7 +10,14 @@
> >  extern char __per_cpu_start[];
> >  extern const char __per_cpu_data_end[];
> 
> Afaics the only users of these two are in the new common/percpu.c.
> These
> decls want to mover there then, I think, to limit exposure.
> 
> >  extern unsigned long __per_cpu_offset[NR_CPUS];
> 
> The definition for this moves to the new common/percpu.c. Hence this
> declaration wants to move to xen/percpu.c.
> 
> Overall with the generalization you do I'd expect asm-
> generic/percpu.h
> to go away altogether. That is, ...
> 
> > -void percpu_init_areas(void);
> > +
> > +#ifndef PARK_OFFLINE_CPUS
> > +/*
> > + * Do we, for platform reasons, need to actually keep CPUs online
> > when we
> > + * would otherwise prefer them to be off?
> > + */
> > +#define park_offline_cpus false
> > +#endif
> 
> ... this, for example, would likely also better be put in
> xen/percpu.h in
> the new model.
Agree, we could move everything to xen/percpu.h and just leave asm-
generic/percpu.h empty for Arm, PPC and RISC-V usage and define
asm/percpu.h only for x86.

I will stick to this approach.

Thanks.

~ Oleksii



 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.