On Thu, 2015-04-23 at 08:50 +0200, Jan Kiszka wrote:

On 2015-04-23 08:11, Mike Galbraith wrote:

quoted

@@ -103,6 +98,9 @@ EXPORT_SYMBOL_GPL(irq_work_queue_on);
 /* Enqueue the irq work @work on the current CPU */
 bool irq_work_queue(struct irq_work *work)
 {
+       bool realtime = IS_ENABLED(CONFIG_PREEMPT_RT_FULL);
+       bool raise = false;
+
        /* Only queue if not already pending */
        if (!irq_work_claim(work))
                return false;

@@ -110,25 +108,22 @@ bool irq_work_queue(struct irq_work *wor
        /* Queue the entry and raise the IPI if needed. */
        preempt_disable();
 
-#ifdef CONFIG_PREEMPT_RT_FULL
-       if (work->flags & IRQ_WORK_HARD_IRQ) {
+       if (realtime && (work->flags & IRQ_WORK_HARD_IRQ)) {
                if (llist_add(&work->llnode, 

this_cpu_ptr(&hirq_work_list)))

This boils down to

#ifdef CONFIG_X
some_type x;
#endif
...

if (IS_ENABLED(CONFIG_X) && ...)
      use(x);

And here we even have an indirection for IS_ENABLED via that local 
bool
variable. Is that pattern OK for Linux? Does it compile in all 
supported
optimization levels of all supported compilers?

I hope it all goes away, that being what IS_ENABLED() is there for.

        -Mike