Re: [PATCH]: Cleanup of __alloc_pages

[Paul Jackson <pj@sgi.com>]

Nick wrote:
> Because it is on the other side of an &&, which evaulates to a
> constant zero when !CONFIG_CPUSETS.

Ah so.

> Having __GFP_HIGH as its own flag gives some more flexibility. I
> don't think it has a downside?

With respect to GFP_ATOMIC, __GFP_HIGH has no flexibility, as they are
#defined to be the same thing.

With respect to __GFP_WAIT, if we only ever use it exactly when
we don't use __GFP_HIGH aka GFP_ATOMIC, then there is a definite
downside.  My old brain doesn't fold constants nearly as reliably or
rapidly as a compiler.  Every apparent degree of freedom that is unused
wastes a few of my remaining precious neurons understanding it.
It directly leads to such bugs as the one I noted in my last reply,
when I realized that checking cpusets in the 'ignoring mins' case
was bogus.

__GFP_HIGH has a second cost - it is easily confused with __GFP_HIGHMEM.

> That would be good. I'll send off a fresh patch with the
> ALLOC_WATERMARKS fixed after Rohit gets around to looking over
> it.

Good.

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401