Re: [PATCH 3/3] bus: ti-sysc: Implement display subsystem reset quirk

From: Tony Lindgren <tony@atomide.com>
To: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: linux-omap@vger.kernel.org, "Andrew F . Davis" <afd@ti.com>,
	Dave Gerlach <d-gerlach@ti.com>, Faiz Abbas <faiz_abbas@ti.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Keerthy <j-keerthy@ti.com>, Nishanth Menon <nm@ti.com>,
	Peter Ujfalusi <peter.ujfalusi@ti.com>,
	Roger Quadros <rogerq@ti.com>, Suman Anna <s-anna@ti.com>,
	Tero Kristo <t-kristo@ti.com>,
	linux-kernel@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org, Jyri Sarha <jsarha@ti.com>,
	Laurent Pinchart <laurent.pinchart@ideasonboard.com>,
	dri-devel@lists.freedesktop.org
Subject: Re: [PATCH 3/3] bus: ti-sysc: Implement display subsystem reset quirk
Date: Tue, 3 Mar 2020 07:44:20 -0800	[thread overview]
Message-ID: <20200303154420.GS37466@atomide.com> (raw)
In-Reply-To: <8cadd536-668a-4309-1878-7db2362717d2@ti.com>

* Tomi Valkeinen <tomi.valkeinen@ti.com> [200303 15:36]:
> On 03/03/2020 17:13, Tony Lindgren wrote:
> > Hi,
> > 
> > * Tomi Valkeinen <tomi.valkeinen@ti.com> [200303 06:03]:
> > > On 24/02/2020 21:12, Tony Lindgren wrote:
> > > > +	/* Remap the whole module range to be able to reset dispc outputs */
> > > > +	devm_iounmap(ddata->dev, ddata->module_va);
> > > > +	ddata->module_va = devm_ioremap(ddata->dev,
> > > > +					ddata->module_pa,
> > > > +					ddata->module_size);
> > > 
> > > Why is this needed? The range is not mapped when sysc_pre_reset_quirk_dss()
> > > is called? This will unmap and remap twice, as this function is called
> > > twice. And then left mapped.
> > 
> > That's because by default we only ioremap the module revision, sysconfig
> > and sysstatus register are and provide the rest as a range for the child
> > nodes.
> > 
> > In the dss quirk case we need to tinker with registers also in the dispc
> > range, and at the parent dss probe time dispc has not probed yet.
> > 
> > We may be able to eventually move the reset quirk to dispc, but then
> > it won't happen in the current setup until after dss top level driver
> > has loaded.
> > 
> > We leave the module range ioremapped as we still need to access
> > sysconfig related registers for PM runtime.
> 
> Ok, makes sense. I guess a minor improvement would be to unmap & remap once
> in sysc_pre_reset_quirk_dss before calling sysc_quirk_dispc.

Yeah well we'd have to sprawl the module specific quirk checks
there too then.

I thought about using the whole module range for modules with a large
IO range, but so far DSS is the only one needing a quirk hadling
covering also child modules like this.

Regards,

Tony