Re: [PATCH] register_blkdev

[Christoph Hellwig <hch@infradead.org>]

On Sat, Mar 08, 2003 at 09:59:14PM +0100, Andries.Brouwer@cwi.nl wrote:
> > We need to get rid of the artifical major/minor split completly
> 
> I do not disagree with you, but your point of view seems
> to be that either we make everything perfect or we do nothing.
> I prefer slow progress.

I completly agree with you on the slow steps.  What we seem to
disagree about is the order of steps..

> Concerning this split - traces of it occur in a very large
> number of places.

Yupp.  And getting rid of those one by one is a a good thing
and I'm happy about every single patch you submit to get rid
of one.

> Let me just mention the raw device that
> I did this afternoon. How does one connect a raw device
> with a block device? Using a struct raw_config_request
> from user space. And look
> 
> struct raw_config_request
> {
>         int     raw_minor;
>         __u64   block_major;
>         __u64   block_minor;
> };
> 
> One of the many places that has a built-in major/minor split.
> Basically this split is unimportant. A dev_t is just a cookie.
> But as soon as you start looking at details this split is
> all over the place.

Yes.  And my opinion is that we need to sort these issues out
one for one before moving on to make dev_t bigger, not the other
way around.