Re: [PATCH v2 00/29] Remove the omapdrm and omapdss devices from platform code

From: Tony Lindgren <tony@atomide.com>
To: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>,
	Liam Girdwood <lgirdwood@gmail.com>,
	dri-devel@lists.freedesktop.org,
	Peter Ujfalusi <peter.ujfalusi@ti.com>,
	Mark Brown <broonie@kernel.org>,
	Laurent Pinchart <laurent.pinchart@ideasonboard.com>,
	Sebastian Reichel <sebastian.reichel@collabora.co.uk>,
	linux-omap@vger.kernel.org
Subject: Re: [PATCH v2 00/29] Remove the omapdrm and omapdss devices from platform code
Date: Fri, 12 May 2017 08:03:49 -0700	[thread overview]
Message-ID: <20170512150349.GF3489@atomide.com> (raw)
In-Reply-To: <4b7cadc5-f32b-f97f-29c3-a2de9927df6c@ti.com>

* Tomi Valkeinen <tomi.valkeinen@ti.com> [170512 00:32]:
> On 11/05/17 17:16, Tony Lindgren wrote:
> 
> >> pinctrl-single doesn't allow to freely set the bits, but requires the
> >> pins to have similar bit structure (function-mask). In CONTROL_DSIPHY,
> >> DSI1 and DSI2 have different bit structures.
> > 
> > OK if the register mixes different types of controllers that
> > can't be partitioned into separate 8 or 16 bit instances then
> > you're out of luck with pinctrl-single. If it does not fit, no
> > point trying to force it, then you need a custom pinctrl driver.
> 
> Writing a driver for a single register on a legacy SoC feels like an
> overkill... But I guess a generic pinctrl driver which allows free
> writes to registers would do the trick, but, then again, if so far we
> have a single register in a single SoC that needs this, maybe it's not
> worth the effort.

Yeah..

> >> I don't understand why pinctrl-single tries so hard to fit things into
> >> one mold...
> > 
> > Basically on many SoCs pinctrl is just the same exact control
> > register repeated for each pin on the SoC:
> 
> Right, I was just wondering why it forces one to have a function mask,
> versus allowing it to be left out and thus making it possible to handle
> also cases where the pins require different kinds of bit masks.

Some of the bits are not usable typically. I think what you're
describing could probably be done with a custom compatible plus
struct pcs_soc_data.

Regards,

Tony
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel