In message <5liths4ie6.fsf@assaris.sics.se>, you write: 

-> --text follows this line--
-> lha@stacken.kth.se writes:
-> > 	Kernels (I tried with 1.5W) build with POOL_DIAGNOSTIC can't
-> > 	load lkm with built w/o.
-> 
-> The appended patch should fix this.  Any comments?

I thought POOL_DIAGNOSTIC was added as a special option because it
imposed quite a bit of overhead even people running kernels with the
DIAGNOSTIC option could do without... This makes *everyone* pay the
price (even in totally production-tuned configurations), and as such
seems wrong-headed.

Some alternate suggestions:
	(1) On POOL_DIAGNOSTIC configs, compile in a the non-diag pool_put
	    pool_get functions, and name then so non-POOL_DIAGNOSTIC LKMs
	    can find 'em (ie, without the underscores).  Ugly, but should
	    do the trick.

	(2) If we can do it and still pass the right file name/line number
	    info (this may not work) make the non-underscored versions into
	    inline functions in the headers.

--rafal

----
Rafal Boni                                                  rafal@mediaone.net