Re: [linuxdc-dev]do_map_probe

[M. R. B. <mr...@0x...>]

* Adrian McMenamin <ad...@mc...> on Sun, Aug 11, 2002:

>=20
> I have been reading the code and I am still deeply confused. If you idea =
is to=20
> map this into an area of the SH4's memory space I am not sure how that wo=
uld=20
> work.
>=20

I'm not too keen on how MTD works, and Paul isn't receiving the list
currently.  When you supply the low-level read/write byte hooks, that's
where you do all the nasty business of sending the Maple request.  I
*think* any virtual mapping is an illusion.  Even on other flash devices,
they may be mapped to a memory range, but there's a layer underneath that
sends commands to the flash controller, etc.  We just have a bit more
overhead.

> Is it not better to treat this as a block device - with all the async map=
le=20
> stuff underneath?

It goes along with the whole thing of a device should belong to the
subsystem to which it's most suited.  The VMU is a flash device, exhibits
all properties of flash devices, but it just sits on the awkwardly
accessible Maple bus.  Makes things complicated, but better in the long
run, IMO.

M. R.