Re: [PATCH] rtc-dev: stop periodic interrupts on device release

[David Brownell <david-b@pacbell.net>]

On Monday 28 July 2008, Tomáš Janoušek wrote:
> On Sat, Jul 26, 2008 at 01:50:55PM -0700, David Brownell wrote:
> > Hmm, I'd think that something like an rtc_dev_ioctl(PIE_OFF) would be

Typo, by the way ... I meant UIE_OFF.


> > preferable here ... so that it's not just UIE_EMUL logic which turns
> > off the one-per-second update IRQs.
> 
> I think it'd be more consistent if the framework only called the rtc api
> functions.

When they exist, sure.  But they currently don't, and I see no value
in adding them just to avoid a simple rtc_dev_ioctl(UIE_OFF) call.
It'd be different if any in-kernel code used update IRQs ... NTP sync?

(Regardless, that's a separate bug and appropriate for a different patch.)


> Like: if the driver doesn't export an op for it and handles it in 
> the ioctl op, it itself should be responsible to clear the irq in its release
> op. (I know there's no op for UIE, so we'd better add it instead of calling
> ioctl in the framework's release function.)

I still think the *existence* of a release() op is a problem.  It's
requiring the drivers to maintain history they should never need.

Surely you agree that having the framework shut down only *emulated*
update IRQs, not "real" ones, is inconsistent?  And hence undesirable?

- Dave