Re: [PATCH 3/3] drm/i915/de: Implement register polling in the display code

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Fri, Mar 13, 2026 at 05:03:16PM +0200, Jani Nikula wrote:
> Another difference between i915/xe and poll_timeout_us() is the range
> for usleep_range(), which is *also* different from fsleep().
> 
> i915/xe have:
> 
> 	usleep_range(wait__, wait__ * 2);
> 
> iopoll has:
> 
> 	usleep_range((__sleep_us >> 2) + 1, __sleep_us);

I was pondering about this a bit and came to the conclusion that
what poll_timeout_us() does with the usleep_range() is a bad idea.
We might have some crappy hardware/firmware (*cough* pcode *cough*)
that doesn't like to be polled too frequently, so it seems much more
sensible to provide the minimum polling interval rather than the
maximum. Also the maximum won't be the actual maximum anyway due to 
random scheduling delays/etc. So I think we probably need to change
the behavior of poll_timeout_us() if we want to actually use it...

-- 
Ville Syrjälä
Intel