On Sat, 2009-02-14 at 07:46 +1030, Rusty Russell wrote:
> On Thursday 12 February 2009 22:39:14 Peter Zijlstra wrote:
> > So it put in unconditionally, how about this?
> > 
> > 
> > --
> > Subject: generic-smp: remove single ipi fallback for smp_call_function_many()
> > 
> > In preparation of removing the kmalloc() calls from the generic-ipi code
> > get rid of the single ipi fallback for smp_call_function_many().
> > 
> > Because we cannot get around carrying the cpumask in the data -- imagine
> > 2 such calls with different but overlapping masks -- put in a full mask.
> 
> OK, if you really want this, please just change it to:
> 	unsigned long cpumask_bits[BITS_TO_LONGS(CONFIG_NR_CPUS)];
> 
> The 'struct cpumask' will be undefined soon when CONFIG_CPUMASK_OFFSTACK=y,
> which will prevent assignment and declaration on stack.
> 
> I'd be fascinated to see perf numbers once you kill the kmalloc.  Because
> this patch will add num_possible_cpus * NR_CPUS/8 bytes to the kernel which
> is something we're trying to avoid unless necessary.

You're free to make it a pointer and do node affine allocations from an
init section of choice and add a hotplug handler.

But I'm not quite sure how perf is affected by size overhead on
ridiculous configs.