Robert Swindells <rjs@fdy2.demon.co.uk> writes:

> >> Why is the kernel mapped as pages not as a section?

[...]

> >playing with the pmap.  In theory we could use different sized
> >sections, 1MB, 64K, 4K and 1K as appropiate, but that'd get very
> >complex, and pmap is tricky enough as it is.
> 
> It currently works like that on arm32.
> 
> Take a look at map_chunk() in arch/arm32/machdep.c

How often does the kernel actually get mapped as a section in
practice? I ask because I did (actually I mapped all of physical
memory to its physical address with repeated map_section) and
pmap_extract blew up due to not handling L1 sections. Testing a fix
now.

Unfortunately it involves changing the pmap_pte_v macro to use its
argument twice: once to check if it's a section and once to
dereference it as a pointer. I'd personally like to turn it into an
inline function.

-- 

	http://ape.n3.net