On Tue, 26 Jun 2012, Glauber Costa wrote:

quoted

Right, because I'm sure that __GFP_KMEMCG will be used in additional
places outside of this patchset and it will be a shame if we have to
always add #ifdef's.  I see no reason why we would care if __GFP_KMEMCG
was used when CONFIG_CGROUP_MEM_RES_CTLR_KMEM=n with the semantics that it
as in this patchset.  It's much cleaner by making it 0x0 when disabled.

What I can do, instead, is to WARN_ON conditionally to the config option in
the page allocator, and make sure no one is actually passing the flag in that
case.

I don't think adding a conditional to the page allocator's fastpath when 
CONFIG_CGROUP_MEM_RES_CTLR_KMEM=n is appropriate.  I don't understand why 
this can't be 0x0 for such a configuration, __GFP_KMEM certainly means 
nothing when we don't have it enabled so how is this different at all from 
kmemcheck?