Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.32 as permitted sender) client-ip=23.128.96.32;
Date:   Fri, 1 Dec 2023 18:11:36 -0500
From:   Ben Wolsieffer <ben.wolsieffer@hefring.com>
To:     Mark Brown <broonie@kernel.org>
Cc:     linux-spi@vger.kernel.org,
        linux-stm32@st-md-mailman.stormreply.com,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org,
        Maxime Coquelin <mcoquelin.stm32@gmail.com>,
        Alexandre Torgue <alexandre.torgue@foss.st.com>,
        Alain Volmat <alain.volmat@foss.st.com>
Subject: Re: [PATCH] spi: stm32: enable controller before asserting CS
Message-ID: <ZWpoKEcM0ZeYAsBa@dell-precision-5540>
References: <20231201214014.2539031-1-ben.wolsieffer@hefring.com>
 <b070eb2a-05d7-4e6a-8de9-15179045d192@sirena.org.uk>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <b070eb2a-05d7-4e6a-8de9-15179045d192@sirena.org.uk>
Precedence: bulk

On Fri, Dec 01, 2023 at 09:50:33PM +0000, Mark Brown wrote:
> On Fri, Dec 01, 2023 at 04:40:14PM -0500, Ben Wolsieffer wrote:
> 
> This feels like it'd be a good fit for moving to runtime PM - that way
> we avoid bouncing the controller on and off between messages which is
> probably better anyway.  The driver already does pinctrl management for
> the device there.

Yes, that probably makes sense. There are a few bits that can only be
configured while the controller is disabled, but it doesn't look like
that applies to any of the ones set in stm32_spi_prepare_msg().

I'm a little hesitant to make big changes to the driver since I can only
test them on an STM32F7 though.

> It also occurs to me that this isn't going to work for devices which
> chip select inverted - for them we can't stop driving chip select at all
> since they need it held high when idle.  There aren't that many such
> devices and it'd loose us the PM which is rather awkward...  I guess
> that's an incremental issue with a more invasive fix though.

The driver only supports GPIO chip select rather than native, so I don't
think this is a problem. Also, I don't think there's any difference
between inverted or uninverted here. They both either need to be driven
all the time or have pull-up/downs.

Ben