Re: RT patch acceptance

From: Zwane Mwaikambo
Date: Wed Jun 01 2005 - 18:50:49 EST


On Wed, 1 Jun 2005, Nicolas Pitre wrote:

> On Wed, 1 Jun 2005, Andrea Arcangeli wrote:
>
> > On Wed, Jun 01, 2005 at 05:29:48PM -0400, Nicolas Pitre wrote:
> > > Actualy it's RTAI/rtlinux which is broken wrt the above IRQ disable.
> > > See for yourself when they're used and watch RTAI/rtlinux crash.
> >
> > Well it's not so clear so please elaborate since I'm curious. Especially
> > it'd be interesting to know if this is that an arm specific kernel
> > crash, or would it happen on x86 too?
>
> It would happen on any architecture supporting XIP from flash.
>
> The XIP code therefore polls the IRQ controller (with IRQs masked out)
> and whenever an IRQ is pending the flash operation is suspended and IRQs
> unmasked. In this case the hard IRQ latency is function of the flash
> suspend delay which is documented in the datasheet.
>
> > There sure can be arch dependencies where an hard_local_irq_disable can
> > be necessary in some places, but that's quite a separate topic, and on
> > x86 I don't see why it should crash.
>
> It would crash if the kernel is XIP. If not XIP then the
> local_irq_save() is never encountered in that case.

Hmm that's an interesting case, there are a few others, like MMU,TLB
update paths and IRQ controller register accesses. But overall these are
few and easy to identify.

Thanks Nicolas,
Zwane
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/