Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220302211100.65264-1-paul.kocialkowski@bootlin.com>
 <20220302211100.65264-2-paul.kocialkowski@bootlin.com> <YiaTfsMDs7RGob2N@robh.at.kernel.org>
In-Reply-To: <YiaTfsMDs7RGob2N@robh.at.kernel.org>
From:   Saravana Kannan <saravanak@google.com>
Date:   Mon, 7 Mar 2022 19:34:22 -0800
Message-ID: <CAGETcx9u9RO_5nSp+=qgwDGY=jL_Q1hAcj+RfVN=q-H_8iuT4w@mail.gmail.com>
Subject: Re: [PATCH RFC v3 1/8] of: Mark interconnects property supplier as optional
To:     Rob Herring <robh@kernel.org>
Cc:     Paul Kocialkowski <paul.kocialkowski@bootlin.com>,
        devicetree@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
        linux-sunxi@lists.linux.dev, linux-kernel@vger.kernel.org,
        linux-media@vger.kernel.org, Chen-Yu Tsai <wens@csie.org>,
        Jernej Skrabec <jernej.skrabec@gmail.com>,
        Samuel Holland <samuel@sholland.org>,
        Michael Turquette <mturquette@baylibre.com>,
        Stephen Boyd <sboyd@kernel.org>,
        Frank Rowand <frowand.list@gmail.com>,
        Maxime Ripard <mripard@kernel.org>,
        Laurent Pinchart <laurent.pinchart@ideasonboard.com>,
        Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, Mar 7, 2022 at 3:21 PM Rob Herring <robh@kernel.org> wrote:
>
> +Saravana
>
> On Wed, Mar 02, 2022 at 10:10:53PM +0100, Paul Kocialkowski wrote:
> > In order to set their correct DMA address offset, some devices rely on
> > the device-tree interconnects property which identifies an
> > interconnect node that provides a dma-ranges property that can be used
> > to set said offset.
> >
> > Since that logic is all handled by the generic openfirmware and driver
> > code, the device-tree description could be enough to properly set
> > the offset.
> >
> > However the interconnects property is currently not marked as
> > optional, which implies that a driver for the corresponding node
> > must be loaded as a requirement. When no such driver exists, this
> > results in an endless EPROBE_DEFER which gets propagated to the
> > calling driver. This ends up in the driver never loading.
> >
> > Marking the interconnects property as optional makes it possible
> > to load the driver in that situation, since the EPROBE_DEFER return
> > code will no longer be propagated to the driver.
> >
> > There might however be undesirable consequences with this change,
> > which I do not fully grasp at this point.

Temporary NACK till I get a bit more time to take a closer look. I
really don't like the idea of making interconnects optional. IOMMUs
and DMAs were exceptions. Also, we kinda discuss similar issues in
LPC. We had some consensus on how to handle these and I noted them all
down with a lot of details -- let me go take a look at those notes
again and see if I can send a more generic patch.

Paul,

Can you point to the DTS (not DTSI) file that corresponds to this?
Also, if it's a builtin kernel, I'd recommend setting
deferred_probe_timeout=1 and that should take care of it too.

-Saravana

> >
> > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
> > ---
> >  drivers/of/property.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/of/property.c b/drivers/of/property.c
> > index 8e90071de6ed..ef7c56b510e8 100644
> > --- a/drivers/of/property.c
> > +++ b/drivers/of/property.c
> > @@ -1365,7 +1365,7 @@ static struct device_node *parse_interrupts(struct device_node *np,
> >
> >  static const struct supplier_bindings of_supplier_bindings[] = {
> >       { .parse_prop = parse_clocks, },
> > -     { .parse_prop = parse_interconnects, },
> > +     { .parse_prop = parse_interconnects, .optional = true,},
> >       { .parse_prop = parse_iommus, .optional = true, },
> >       { .parse_prop = parse_iommu_maps, .optional = true, },
> >       { .parse_prop = parse_mboxes, },
> > --
> > 2.35.1
> >
> >