Re: slab: setup allocators earlier in the boot sequence

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

> Yes, you're obviously right. I overlooked the fact that arch code have
> their own special slab_is_available() heuristics (yikes!).
> 
> But are you happy with the two patches I posted so I can push them to
> Linus?

I won't be able to test them until tomorrow. However, I think the first
one becomes unnecessary with the second one applied (provided you didn't
miss a case), no ?

I still prefer my approach of having a more fine grained control of what
bits to remove. First because applying a mask is less expensive than a
conditional branch (I used a negative mask because it would be too easy
to miss bits otherwise) and second, because it allows for masking of
other bits easily, for example, __GFP_IO for the suspend path etc...

Now, if you find it a bit too ugly, feel free to rename smellybits to
something else and create an accessor function for setting what bits are
masked out, but I still believe that the basic idea behind my patch is
saner than yours :-)

Cheers,
Ben.