Re: [RFC] simple_char: New infrastructure to simplify chardev management

On Tue, Feb 10, 2015 at 4:44 PM, Greg Kroah-Hartman
[off-list ref] wrote:

On Tue, Feb 10, 2015 at 03:44:05PM -0800, Andy Lutomirski wrote:

quoted

This isn't adequately tested, and I don't have a demonstration (yet).
It's here for review for whether it's a good idea in the first place
and for weather the fully_dynamic mechanism is a good idea.

The current character device interfaces are IMO awful.

That's a total understatement.  Redoing the char interface has been in
my todo list for a decade now.  It's the complexity that happens to be
used by just a handful of drivers that have prevented me from doing the
rework in the past.  Creating a "new" interface that we then port code
to is a very good idea, as it can happen over time in a much more
orderly way.

And we can throw the kernel-janitors people at it once it's working, to
convert the rest of the tree, providing them a useful outlet for their
need for patch cleanups :)

So yes, I'm all for this, thanks so much for looking into this.  I'm at
a conference this week, but will go over it on the plane home and give
you some review comments.

It would be nice to make the reference counting cleaner, perhaps by
tying a chardev minor more directly to a struct device, but I wasn't
sure how to do that usefully.

--Andy

greg k-h

-- 
Andy Lutomirski
AMA Capital Management, LLC