Re: [PATCH linux 2/6] hwmon: occ: Add sysfs interface

[Guenter Roeck]

On Tue, Jan 10, 2017 at 11:41:44AM -0600, Benjamin Herrenschmidt wrote:
> On Sat, 2017-01-07 at 09:15 -0800, Guenter Roeck wrote:
> > > Instead of the "online" attribute, what do you think about using the
> > > "bind"/"unbind" API to probe the device from user space once the system
> > > is powered on? All the hwmon registration would take place in the probe
> > > function, it would just occur some time after boot.
> > > 
> > 
> > A more common approach would be to have a platform driver. That platform
> > driver would need a means to detect if the OCC is up and running, and
> > instantiate everything else once it is.
> > 
> > A trigger from user space is problematic because there is no guarantee
> > that the OCC is really up (or that it even exists).
> > 
> > An alternative might be to have the hwmon driver poll for the OCC,
> > but that would be a bit more difficult and might require a kernel thread
> > or maybe asynchronous probing.
> 
> Hi Guenter !
> 
> I'm not sure I agree with you here ;-)
> 
> I'm don't know how much background you got (I missed the beginning of
> the discussion) but basically this driver runs on the BMC (system
> controller) of the POWER9 server, the OCC is inside the POWER9 chip
> itself.
> 
> So the presence/power state of the OCC doesn't depend on the BMC
> platform kernel code. The BMC userspace controls power up and down to the POWER9, and thus it's the one to know whether the remote is up.
> 
> If we were to create a "platform driver", all it would do is get input
> from userspace exactly like that sysfs file :-)
> 
> So if you don't like the sysfs file that registers/de-registers, which
> I sort-of understand, it's a bit of a hack (though a rather efficient
> one), I think the bind/unbind approach makes sense. However, I wonder
> whether the simplest and most efficient (remember this BMC has a really
> slow CPU) is an "online" file sysfs file, though rather than
> registering/de-registering the hwmon it would simply make it stop
> accessing the HW (and return some known "offline" values).
> 

I don't like that too much either; it still looks like a hack.
How about using bind/unbind then ?

Guenter