Re: [PATCH v6] x86/apic: limit irq affinity

[Dimitri Sivanich <sivanich@sgi.com>]

On Fri, Dec 04, 2009 at 03:12:14PM -0800, Eric W. Biederman wrote:
> Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com> writes:
> 
> >> > > As a matter of fact, driver's allocating rings, buffers, queues on other nodes should optimally be made aware of the restriction.
> >> > 
> >> > The idea is that the driver will do its memory allocations for everything 
> >> > across nodes.  When it does that, it will use the kernel interface 
> >> > (function call) to set the corresponding mask it wants for those queue 
> >> > resources.  That is my end-goal for this code.
> >> > 
> >> 
> >> OK, but we will eventually have to reject any irqbalance attempts to send irqs to restricted nodes.
> >
> > See above.
> 
> Either I am parsing this conversation wrong or there is a strong
> reality distortion field in place.  It appears you are asking that we
> depend on a user space application to not attempt the physically
> impossible, when we could just as easily ignore or report -EINVAL to.
> 
> We really have two separate problems hear.
> - How to avoid the impossible.

The kernel does need to restrict attempts at the impossible.  However I see
nothing wrong with providing apps with information as to what the impossible
actually is.

> - How to deal with NUMA affinity.