Re: [PATCH] reset: thead: Scope TH1520 reset driver to VO subsystem

From: Yao Zi <ziyao@disroot.org>
To: Drew Fustini <fustini@kernel.org>,
	Michal Wilczynski <m.wilczynski@samsung.com>
Cc: Rob Herring <robh@kernel.org>, Conor Dooley <conor+dt@kernel.org>,
	Albert Ou <aou@eecs.berkeley.edu>,
	Alexandre Ghiti <alex@ghiti.fr>,
	devicetree@vger.kernel.org, linux-kernel@vger.kernel.org,
	Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>,
	Guo Ren <guoren@kernel.org>,
	Philipp Zabel <p.zabel@pengutronix.de>,
	Paul Walmsley <paul.walmsley@sifive.com>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Krzysztof Kozlowski <krzk+dt@kernel.org>,
	linux-riscv@lists.infradead.org, Fu Wei <wefu@redhat.com>
Subject: Re: [PATCH] reset: thead: Scope TH1520 reset driver to VO subsystem
Date: Wed, 13 Aug 2025 08:04:40 +0000	[thread overview]
Message-ID: <aJxGnnwlE1sWarXa@pie> (raw)
In-Reply-To: <aJrTq775sXTrsepl@x1>

On Mon, Aug 11, 2025 at 10:39:55PM -0700, Drew Fustini wrote:
> On Sun, Aug 10, 2025 at 11:14:19PM +0200, Michal Wilczynski wrote:
> > The reset controller driver for the TH1520 was using the generic
> > compatible string "thead,th1520-reset". However, the current
> > implementation only manages the resets for the Video Output (VO)
> > subsystem.
> 
> Looking at Section 5.4 on Page 451 of the TH1520 System User Manual [1],
> it does seem like we would ultimately need 6 separate nodes for reset
> controllers:

Yes, this is true. And another six nodes for clock controllers (there's
already one).

>  0xFF_EF01_4000: AP_SUBSYS
>  0xFF_EC02_C000: MISC_SUBSYS
>  0xFF_E404_0000: VI_SUBSYS
>  0xFF_EF52_8000: VO_SUBSYS
>  0xFF_ECC3_0000: VP_SUBSYS
>  0xFF_EF04_0000: DSP_SUBSYS
> 
> Maybe we should take this opportunity to document the bindings for all
> the resets that the REE (e.g. Linux) can control?

It's worth noting that with either mainline U-Boot or vendor U-Boot, no
core is configured to run as REE. IOW, AON_SUBSYS could be accessed by
AP cores as well.

I think introducing read-only clock support to Linux could help us to
correctly describe pvt clocks which are now replaced with a aonsys
placeholder and resolve issues like what is described in 0370395d45ca
("clk: thead: Mark essential bus clocks as CLK_IGNORE_UNUSED").

Futhermore, there may be downstream projects, e.g. U-Boot, make use of
these TEE-only devices, which could benefit if we have these devices
documented and described in devicetree, too. Thus I think the AON clock
and reset controllers should be documented as well if we're going to
document every reset/clock controller in a batch.

> It seemed like that was overkill for the 2 resets needed for the GPU,
> but, as Krzysztof noted in this thread, problems arise when bindings are
> introduced that are not complete.
> 
> Thanks,
> Drew
> 
> [1] https://git.beagleboard.org/beaglev-ahead/beaglev-ahead/-/blob/main/docs/TH1520%20System%20User%20Manual.pdf

Best regards,
Yao Zi

_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv