Re: [RFC 0/4] dynamically allocate arch specific system vectors

[Dean Nelson <dcn@sgi.com>]

On Sun, Sep 14, 2008 at 05:35:22PM +0200, Ingo Molnar wrote:
> 
> * Dean Nelson <dcn@sgi.com> wrote:
> 
> > On Mon, Aug 11, 2008 at 12:39:22PM -0700, Eric W. Biederman wrote:
> > >
> > > Although I am not at all convinced that dynamic allocation of
> > > the vector number (instead of statically reserving it makes sense).
> > 
> > We (SGI) need somewhere around eight vectors.
> > 
> > There are two kernel modules, sgi-gru and sgi-xp (in drivers/misc), 
> > that each need two vectors. And there's the broadcast assist unit 
> > (BAU) that is involved in tlb shootdown on uv, which currently uses 
> > statically reserved vector 0xf8 (UV_BAU_MESSAGE -- see uv_bau_init()). 
> 
> while i understand the UV_BAU_MESSAGE case (TLB flushes are special), 
> why does sgi-gru and sgi-xp need to go that deep? They are drivers, they 
> should be able to make use of an ordinary irq just like the other 2000 
> drivers we have do.

The sgi-gru driver needs to be able to allocate a single irq/vector pair for
all CPUs even those that are not currently online. The sgi-xp driver has
similar but not as stringent needs.

The current __assign_irq_vector() restricts the allocation of the irq/vector
pair to a single CPU and its vector_allocation_domain().


> > I know of a debugger that also uses 0xf8 because it was previously 
> > available until UV_BAU_MESSAGE came along. The BAU would be happy with 
> > a dynamically allocated system vector. We have a couple of other 
> > things in the works that also need vectors.
> 
> which debugger is this?

KDB.