Re: RFC: gpio attachment for ppbus

[Hans Rosenfeld <rosenfeld%grumpf.hope-2000.org@localhost>]

On Tue, Apr 29, 2008 at 09:52:01AM -0400, der Mouse wrote:
> If that means it takes five syscalls to change five pins, I think I'd
> stick with hacking the lpt driver.  But if there's some way to batch
> changes, to send many of them across the kernel/user boundary at once,
> it'd be somewhat awkward but probably OK.

Thats what I was thinking about. The pins would still be changed
sequentially, but without the overhead of syscalls.

> > There is also no distinction between data pins, status pins and
> > control pins, they are all just gpio pins.
> 
> I wouldn't expect there'd be any difference at the API level, but there
> is a difference at the hardware level - as far as I can tell, you can
> change multiple data pins, or multiple control pins, simultaneously but
> you can't change data pins and control pins simultaneously, hence the
> question.

The whole gpio thing could be extended to support something like "pin
groups" that could be changed in one access, such a group would
correspond to a single device register. But this would also make the
gpio driver much more complicated.


-- 
%SYSTEM-F-ANARCHISM, The operating system has been overthrown