On Thu, 25 Sep 2008, Sebastien Dugue wrote: > On Wed, 24 Sep 2008 17:45:36 +0200 (CEST) Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote: > > > On Tue, 23 Sep 2008, Sebastien Dugue wrote: > > > > > > > > handle_IRQ_event() reenables interrupts for threaded IRQ handlers, > > > provided said handler does not need to run with IRQF_DISABLED. > > > > > > Therefore, unless I'm missing something, in the following: > > > > > > if (!hardirq_count() || !(action->flags & IRQF_DISABLED)) > > > local_irq_enable(); > > > > > > the logical OR should be a logical AND: > > > > > > if (!hardirq_count() && !(action->flags & IRQF_DISABLED)) > > > local_irq_enable(); > > > > No. We don't want to run the threaded handler with interrupts disabled. > > > > !hardirq_count() tells us that we are in thread context. > > OK, but then if a driver asked to run with IRQ disabled it will not > be honored with threaded hardirq. Err, no. We switch all interrupts (except timer) to threaded with PREEMPT_RT=y. And we simply keep interrupts enabled across the thread the same way as we ignore the interrupt disable/enable of the spinlock_irqXX functions when we switch them to sleeping rt_mutex based locks. Thanks, tglx -- To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
