Messages in this thread |  | | | Date | Wed, 20 Aug 2008 15:48:49 +0200 | | From | Ingo Molnar <> | | Subject | Re: [PATCH 05/10] ldisc: new dcd_change() method for line disciplines. |
* Rodolfo Giometti <giometti@enneenne.com> wrote:
> On Wed, Aug 20, 2008 at 11:11:17AM +0100, Alan Cox wrote:
>
> > Can we get the basics right first - drop the IRQ_EVENTS and HARDPPS
> > hackery out of the patchset and the core code then looks quite clean.
> > After that is sorted/merged we can come back to working out the best way
> > to improve the IRQ and HARDPPS hacks.
>
> On Wed, Aug 20, 2008 at 12:31:15PM +0200, Ingo Molnar wrote:
>
> > hm, NACK on this line of approach. If then this instrumentation should
> > be done at the genirq level (kernel/irq/*.c). Also, the #ifdefs are ugly
> > - please hide them in include files.
>
> Ok, I'll propose a new patch set without the IRQ_EVENTS part (or with
> this support as last patch to be easily discarted) to get kernel
> inclusion, but we had to solve this issue soon since without this
> trick the PPS precision is quite low.
well, i didnt suggest you kill that aspect of the patchset - i suggested
you change it to hook into kernel/irq/*.c instead of
arch/x86/kernel/irq*.c.
That way all other architectures will benefit from it, not just x86.
Doing an x86-only thing for such features is completely unacceptable.
Ingo
|  |