Re: [PATCH/RFC] MTD: Striping layer core

["Jörn Engel" <joern@wohnheim.fh-wedel.de>]

On Fri, 31 March 2006 14:00:09 +0400, Artem B. Bityutskiy wrote:
> On Fri, 2006-03-31 at 11:40 +0200, Jörn Engel wrote:
> > On Fri, 31 March 2006 13:36:43 +0400, Artem B. Bityutskiy wrote:
> > > 
> > > No, mtd->type has to tell you the type of the MTD device. Ideally, this 
> > > has to me the only flash-specific field in the mtd_info structure. And 
> > > if users want to do some flash specific things, they have to look at 
> > > mtd->type, realize what is the subsystem which handles this flash, and 
> > > start working with this subsystem. For striping, this is the striping 
> > > subsystem. I don't know what for mtd->flags, probably this hast to go at 
> > > all.
> > 
> > Is this exported to userspace via mtdchar?
> > 
> I don't quite understand what is "this".
> 
> Ideally yes...

I take it that mtd->type is not exported to userspace yet.  Which is
good, because imho it shouldn't be.  Flash should me a step towards a
standard interface, something that hard disk have had for years.
Knowing every little detail about every single flash chip bears more
harm than good.

And as far as in-kernel users are concerned, JFFS2 is the only one
that really matters.  For JFFS2, my previous statement still stand
that it should base decisions on the existence of a feature, not a
type.

You could still have flash types, but merely as aggregates of
features, nothing else.  For example, NAND flash could have
FEATURE_OOB, FEATURE_WBUF, FEATURE_ECC etc.  Which of these feature
make sense and are really needed is another matter, but you should get
the idea.

Jörn

-- 
/* Keep these two variables together */
int bar;