Messages in this thread | | | Date | Wed, 24 Nov 2004 04:27:57 +0100 | From | Ingo Molnar <> | Subject | Re: [patch] Real-Time Preemption, -RT-2.6.10-rc2-mm2-V0.7.29-0, and 30-9 |
| |
* Lee Revell <rlrevell@joe-job.com> wrote:
> On Tue, 2004-11-23 at 16:22 -0500, Steven Rostedt wrote: > > Just curious to why you scale the interrupts from 49 down to 25. What > > would be wrong with keeping all of them at 49 (or whatever). Being a > > FIFO, no interrupt would preempt another. Why would you want the first > > IRQs to be registered have higher priority than (and thus will preempt) > > irqs registered later. > > I raised this issue before. I agree that all interrupts should get > the same RT prio by default. Otherwise the default behavior is > arbitrary.
i agree that it's arbitrary. There are two reasons for the ordering:
1) _usually_ the IRQs that get registered first are the 'more important' ones. E.g. timer and keyboard interrupts will preempt the IDE interrupt. This is in no way a generic thing though.
2) testing: if all IRQs are at the same priority level then alot less inter-IRQ preemption occurs, and testing coverage is lower. With all irqs on different levels the bugs will trigger sooner.
To solve this cleanly some userspace policy code is needed that would take some settings (e.g. sound_highprio) through which the priority setup could be configured. It's not a simple task as that could would have to discover the type of devices that are in the system and their irqs - possibly a component of udev could do this?
Ingo - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |