Re: [PATCH 1/8] cpumask: Replace cpumask_of_cpu with cpumask_of_cpu_ptr

[Rusty Russell <rusty@rustcorp.com.au>]

On Friday 18 July 2008 23:43:07 Mike Travis wrote:
> Rusty Russell wrote:
> > On Wednesday 16 July 2008 07:14:30 Mike Travis wrote:
> >>   * This patch replaces the dangerous lvalue version of cpumask_of_cpu
> >>     with new cpumask_of_cpu_ptr macros.  These are patterned after the
> >>     node_to_cpumask_ptr macros.
> >
> > Hi Mike,
> >
> >    Should we just put cpumask_of_cpu_map[] in generic code and then have
> > cpumask_of_cpu() always return a cpumask_t pointer?  These macros which
> > declare things which may be one of two types is a real penalty for code
> > readability.
> >
> > Thanks,
> > Rusty.
>
> Hi,
>
> I wouldn't mind it at all, and since it's almost always calling a function
> that requires a cpumask_t pointer (like the cpu_* ops or
> set_cpus_allowed_ptr) then there shouldn't be too many "pointer
> dereference" penalties.  I'm just always a bit hesitant to make too many
> generic changes since I have only x86 and ia64 machines to test with.

The simple version is just a static array of [NR_CPUS] cpumask_t's.  Do that, 
with an override for smarter archs?

I really REALLY prefer that over the fairly tortuous macros.

> Another thought I had is perhaps cpumask.h should define something that
> indicates a "huge NR_CPUS count" that is used globally to trigger things
> like kmalloc of cpumask variables, instead of declaring them on the
> stack...?  Or (as has been discussed in the past), maybe a new cpumask_t
> type will be needed?

AFAICT the final answer has to be a get_cpu_mask()/put_cpu_mask(), which 
sleeps and doesn't nest (so we can use a pool allocator).  Of course, that 
kind of analysis is non-trivial, so I suggest that's not for this merge 
window...

Want me to try something and see if it boots?
Rusty.