[Xen-devel] Re: [RFC PATCH 15/35] subarch support for controlling interrupt delivery

["Martin J. Bligh" <mbligh@xxxxxxxxxx>]

Christian Limpach wrote:
> On Tue, May 09, 2006 at 07:49:42AM -0700, Martin J. Bligh wrote:
>
> > > +#define __cli()                                                         \
> > > +do {                                                                   \
> > > +       struct vcpu_info *_vcpu;                                        \
> > > +       preempt_disable();                                              \
> > > +       _vcpu = &HYPERVISOR_shared_info->vcpu_info[__vcpu_id];           \
> > > +       _vcpu->evtchn_upcall_mask = 1;                                       \
> > > +       preempt_enable_no_resched();                                    \
> > > +       barrier();                                                      \
> > > +} while (0)
> > Should be a real function
>
> Yes, except it's not trivially done because if __cli was an inline
> function, you need to have everything that is used in the declaration
> defined when the function is declared as opposed to when the #define
> gets used.  I'll give it another try, but it very quickly becomes
> #include hell.
>
> Anybody want to comment on the performance impact of making
> local_irq_* non-inline functions?
I wasn't concerned with inline vs non-inline - that's your choice.
Just the inherent foulness of multi-line macros ;-)

M.



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel