On Thu, May 03, 2012 at 05:00:40PM +0200, Johan Hovold wrote:
> On Thu, May 03, 2012 at 12:38:02PM +0100, Mark Brown wrote:

> > This is one of the reasons why we currently use tracepoints (they just
> > don't have this issue as they're trivial to filter), though
> > adding some sort of infrastructure for it ought not to be too difficult
> > even if it's just at the regmap level.

> So a /sys/kernel/debug/regmap/<device>/io_printk attribute (with a
> better name) to enable debug printks in io paths
> (regmap*{read,write,update} outside of mutex) in regmap.c would be
> acceptable?

Yes, that'd be totally fine for me - it's debugfs so we can always drop
it later if someone comes up with a better idea or something.

> > Actually, the other question I had but forgot to ask (or I think punted
> > on for your response) was why these are in sysfs at all - things like
> > which things are connected to the backlight are going to be a property
> > of the board design so should be defined by the machine not tweaked from
> > userspace.

> I agree with you and the reason is the same as for the max_current
> attribute (discussed in the other thread) -- it was an explicit request
> from the end customer.

> I could replace the boost attributes with a platform_data entry where it
> really belongs.

I really think this is much better for mainline.

> There is a use case (beyond testing/integration) for keeping the (lvled)
> outputs configurable from userspace, in that it provides a way to
> synchronise LED activity such as blinking. So I still want to keep those,
> at least for the lvleds.

I'm not sure exactly which control that is?