Re: V2 API

[Jean Delvare <khali-PUYAD+kWke1g9hUCZPvPmw@public.gmane.org>]

On Fri, 30 Sep 2011 12:28:06 +0100, David Goodenough wrote:
> On Friday 30 Sep 2011, Jean Delvare wrote:
> > On Fri, 30 Sep 2011 11:34:38 +0100, David Goodenough wrote:
> > > On Wednesday 28 Sep 2011, David Goodenough wrote:
> > > > Doing a Google search looking to see if there was userland support
> > > > for SMBALERT messages I saw references to the V2 API which last year
> > > > was "being developed".  Did this happen, and is it defined anywhere?
> > > 
> > > I take it from the fact that no-one has replied, that the V2 API is
> > > not being worked on or has stalled.
> > 
> > I take it from the fact that no-one has replied, that "V2 API" is an
> > awfully vague subject line, so odds are that nobody read your post in
> > the first place. Try to come up with a better subject line next time.
> It was the only name I found for any effort to address this problem.

And...? You couldn't come up with a subject line at least mentioning
"SMBus" and/or "alert"?

> And that was from a message which I think was on this mailing list about
> a year ago.
> > 
> > > Is there any way using the
> > > existing API for userland code to be notified of an incomming
> > > SMBALERT?
> > 
> > Not that I know of, and I wasn't even aware that efforts had ever been
> > ongoing to achieve this.
> So I take it that the only way to do this would be to write a kernel
> driver?

At the moment, yes. It would technically be possible to do it in
user-space, I presume, by extending the i2c-dev driver to register a
proper driver with alert callback. But this would require a significant
redesign of the i2c-dev driver, which I am not up to for the time
being. An alternative would be a simple sysfs attribute in the i2c
adapter's directory in the device tree, which user-space could poll
until an event triggers. I don't know what the plan was.

-- 
Jean Delvare