Re: [Patch 2/3] Prevent false sgi-xpc heartbeat failures.

[Andrew Morton <akpm@linux-foundation.org>]

On Tue, 7 Apr 2009 20:04:49 -0500 Robin Holt <holt@sgi.com> wrote:

> On Tue, Apr 07, 2009 at 12:49:33PM -0700, Andrew Morton wrote:
> > On Fri, 03 Apr 2009 11:04:42 -0500
> > holt@sgi.com wrote:
> > 
> > > +void (*xpc_allow_hb) (short partid);
> > > +void (*xpc_disallow_hb) (short partid);
> > > +void (*xpc_disallow_all_hbs) (void);
> > >  void (*xpc_heartbeat_init) (void);
> > >  void (*xpc_heartbeat_exit) (void);
> > >  void (*xpc_increment_heartbeat) (void);
> > 
> > The driver adds a huge number of globals.
> > 
> > As a pointless cleanup it might be nice to convert all this:
> 
> 
> Andrew,
> 
> Was this what you were thinking?  I just slapped this together.  It is not
> ready for inclusion yet.  I need to review the patch and test it first.
> 
> Index: 20090407-dcn-fixes/drivers/misc/sgi-xp/xpc.h
> ===================================================================
> --- 20090407-dcn-fixes.orig/drivers/misc/sgi-xp/xpc.h	2009-04-07 16:33:11.000000000 -0500
> +++ 20090407-dcn-fixes/drivers/misc/sgi-xp/xpc.h	2009-04-07 19:50:26.000000000 -0500
> @@ -780,6 +780,67 @@ struct xpc_partition {
>  
>  } ____cacheline_aligned;
>  
> +struct xpc_arch_operations {
> ...
> +};
>
> ...
>
> +extern struct xpc_arch_operations *xpc_arch_ops;

That adds an extra pointer dereference for each call.  I was thinking

  extern struct xpc_arch_operations xpc_arch_ops;

  static void
  xpc_start_hb_beater(void)
  {
	xpc_heartbeat_init();
	xpc_arch_ops.heartbeat_init();

which should generate the same code - it's just syntactic sugar.