Re: [RFC patch 0/5] genirq: add infrastructure for threaded interrupthandlers

[Thomas Gleixner]

On Wed, 1 Oct 2008, Andrew Morton wrote:

> On Wed, 1 Oct 2008 16:29:50 -0700
> Arjan van de Ven <arjan@xxxxxxxxxxxxx> wrote:
> 
> > On Wed, 1 Oct 2008 16:23:33 -0700
> > Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> > 
> > > 
> > > I'm a bit surprised to see that there is no facility for per-cpu
> > > interrupt threads?
> > > 
> > 
> > per handler is the right approach (that way, if one dies, all other
> > interrupts will likely keep working)
> > 
> > now.. normally an interrupt only goes to one cpu, so effectively it is
> > per cpu already anyway
> 
> Yes, if a) the thread was asleep when it was woken up and b) if the
> scheduler does the right thing and wakes the thread on the CPU which
> called wake_up().
> 
> The ongoing sagas of tbench/mysql/volanomark regressions make me think
> that any behaviour which we "expect" of the scheduler should be
> triple-checked daily :(

Yup. I missed that detail when I dusted off the moldy patches. 

Of course we need to pin the thread to the affinity mask of the
hardware interrupt.

/me goes back to do home work :)

Thanks,

	tglx
--
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/