public inbox for linux-arm-kernel@lists.infradead.org
 help / color / mirror / Atom feed
From: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
To: Sakari Ailus <sakari.ailus@iki.fi>
Cc: Isaac Scott <isaac.scott@ideasonboard.com>,
	linux-media@vger.kernel.org, rmfrfs@gmail.com, martink@posteo.de,
	kernel@puri.sm, mchehab@kernel.org, shawnguo@kernel.org,
	s.hauer@pengutronix.de, kernel@pengutronix.de,
	festevam@gmail.com, imx@lists.linux.dev,
	linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH] imx-mipi-csis: Get the number of active lanes from mbus_config
Date: Tue, 19 Aug 2025 05:44:13 +0300	[thread overview]
Message-ID: <20250819024413.GJ5862@pendragon.ideasonboard.com> (raw)
In-Reply-To: <aJ8pKs_6YpAiPjlq@valkosipuli.retiisi.eu>

On Fri, Aug 15, 2025 at 12:33:46PM +0000, Sakari Ailus wrote:
> On Fri, Aug 15, 2025 at 02:36:33PM +0300, Laurent Pinchart wrote:
> > On Fri, Aug 15, 2025 at 11:25:24AM +0000, Sakari Ailus wrote:
> > > On Fri, Aug 15, 2025 at 01:32:05PM +0300, Laurent Pinchart wrote:
> > > > On Fri, Aug 15, 2025 at 09:18:13AM +0000, Sakari Ailus wrote:
> > > > > On Thu, Aug 14, 2025 at 12:37:01PM +0100, Isaac Scott wrote:
> > > > > > Although 4 lanes may be physically available, we may not be using all of
> > > > > > them. Get the number of configured lanes in the case a driver has
> > > > > > implemented the get_mbus_config op.
> > > > > > 
> > > > > > Signed-off-by: Isaac Scott <isaac.scott@ideasonboard.com>
> > > > > > 
> > > > > > ---
> > > > > > 
> > > > > > Currently, the imx-mipi-csis driver parses the device tree to determine
> > > > > > the number of configured lanes for the CSI receiver. This may be
> > > > > > incorrect in the case that the connected device only uses a subset of
> > > > > > lanes, for example. Allow the drivers for these cameras to create a
> > > > > > mbus_config to configure the number of lanes that are actually being
> > > > > > used.
> > > > > > 
> > > > > > If the driver does not support the get_mbus_config op, this patch will
> > > > > > have no functional change.
> > > > > > 
> > > > > > Compile tested against media-master (v6.17-rc1)
> > > > > > ---
> > > > > >  drivers/media/platform/nxp/imx-mipi-csis.c | 41 ++++++++++++++++++++++
> > > > > >  1 file changed, 41 insertions(+)
> > > > > > 
> > > > > > diff --git a/drivers/media/platform/nxp/imx-mipi-csis.c b/drivers/media/platform/nxp/imx-mipi-csis.c
> > > > > > index 2beb5f43c2c0..efe4e2ad0382 100644
> > > > > > --- a/drivers/media/platform/nxp/imx-mipi-csis.c
> > > > > > +++ b/drivers/media/platform/nxp/imx-mipi-csis.c
> > > > > > @@ -939,6 +939,43 @@ static struct mipi_csis_device *sd_to_mipi_csis_device(struct v4l2_subdev *sdev)
> > > > > >  	return container_of(sdev, struct mipi_csis_device, sd);
> > > > > >  }
> > > > > >  
> > > > > > +static int mipi_csis_get_active_lanes(struct v4l2_subdev *sd)
> > > > > > +{
> > > > > > +	struct mipi_csis_device *csis = sd_to_mipi_csis_device(sd);
> > > > > > +	struct v4l2_mbus_config mbus_config = { 0 };
> > > > > > +	int ret;
> > > > > > +
> > > > > > +	ret = v4l2_subdev_call(csis->source.sd, pad, get_mbus_config,
> > > > > > +			       0, &mbus_config);
> > > > > > +	if (ret == -ENOIOCTLCMD) {
> > > > > > +		dev_dbg(csis->dev, "No remote mbus configuration available\n");
> > > > > > +		return 0;
> > > > > > +	}
> > > > > > +
> > > > > > +	if (ret) {
> > > > > > +		dev_err(csis->dev, "Failed to get remote mbus configuration\n");
> > > > > > +		return ret;
> > > > > > +	}
> > > > > > +
> > > > > > +	if (mbus_config.type != V4L2_MBUS_CSI2_DPHY) {
> > > > > > +		dev_err(csis->dev, "Unsupported media bus type %u\n",
> > > > > > +			mbus_config.type);
> > > > > > +		return -EINVAL;
> > > > > > +	}
> > > > > > +
> > > > > > +	if (mbus_config.bus.mipi_csi2.num_data_lanes > csis->bus.num_data_lanes) {
> > > > > > +		dev_err(csis->dev,
> > > > > > +			"Unsupported mbus config: too many data lanes %u\n",
> > > > > > +			mbus_config.bus.mipi_csi2.num_data_lanes);
> > > > > > +		return -EINVAL;
> > > > > > +	}
> > > > > > +
> > > > > > +	csis->bus.num_data_lanes = mbus_config.bus.mipi_csi2.num_data_lanes;
> > > > 
> > > > There's a bug here, you override the number of lanes retrieved from DT,
> > > > which is the number of connected lanes, with the number of lanes used by
> > > > the source for its particular configuration. You will never be able to
> > > > then use more lanes in a different source configuration.
> > > > 
> > > > > > +	dev_dbg(csis->dev, "Number of lanes: %d\n", csis->bus.num_data_lanes);
> > > > > 
> > > > > None of the above is really specific to this driver. Could you instead
> > > > > implement a function that parses the information from the fwnode endpoint
> > > > > and uses mbus configuration on top?
> > > > 
> > > > That would need to parse the endpoint every time we start streaming, it
> > > > doesn't sound ideal.
> > > 
> > > Perhaps not, but does that matter in practice? Parsing the endpoint is,
> > > after all, fairly trivial. The advantage would be simplifying drivers.
> > 
> > It's trivial from a code point of view, but it's not a cheap operation.
> > I'd like to avoid making starting streaming more expensive.
> 
> How cheap is "not cheap"? I'd be surprised if parsing an endpoint took more
> time than e.g. an I²C register write. Of course it depends on the CPU...

Still, it's not cheap, and I think it can easily be avoided.

> > > Alternatively we could think of caching this information somewhere but I
> > > don't think it's worth it.
> > 
> > Drivers likely need to parse endpoints for other reasons. I'd cache the
> > value in drivers, like done today, and pass it to a get_active_lanes
> > helper.
> 
> Then drivers presumably would also validate this against the endpoint
> configuration, wouldn't they? That's extra code in every CSI-2 receiver
> driver.

Why so ? The number of connected lanes can be passed to the helper
function, which can use it to validate the number of lanes reported by
the source subdev.

> > > > > The function could take struct media_pad pointer as an argument, or struct
> > > > > v4l2_subdev pointer and the pad number.
> > > > > 
> > > > > I wonder if any other parameters could change dynamically but I can't think
> > > > > of that now, so perhaps just the number of lanes is what the function
> > > > > should indeed return.
> > > > > 
> > > > > > +
> > > > > > +	return 0;
> > > > > > +}
> > > > > > +
> > > > > >  static int mipi_csis_s_stream(struct v4l2_subdev *sd, int enable)
> > > > > >  {
> > > > > >  	struct mipi_csis_device *csis = sd_to_mipi_csis_device(sd);
> > > > > > @@ -965,6 +1002,10 @@ static int mipi_csis_s_stream(struct v4l2_subdev *sd, int enable)
> > > > > >  	format = v4l2_subdev_state_get_format(state, CSIS_PAD_SINK);
> > > > > >  	csis_fmt = find_csis_format(format->code);
> > > > > >  
> > > > > > +	ret = mipi_csis_get_active_lanes(sd);
> > > > > > +	if (ret < 0)
> > > > > > +		dev_dbg(csis->dev, "Failed to get active lanes: %d", ret);
> > > > > > +
> > > > > >  	ret = mipi_csis_calculate_params(csis, csis_fmt);
> > > > > >  	if (ret < 0)
> > > > > >  		goto err_unlock;

-- 
Regards,

Laurent Pinchart


  reply	other threads:[~2025-08-19  2:47 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-08-14 11:37 [PATCH] imx-mipi-csis: Get the number of active lanes from mbus_config Isaac Scott
2025-08-15  9:18 ` Sakari Ailus
2025-08-15 10:32   ` Laurent Pinchart
2025-08-15 11:25     ` Sakari Ailus
2025-08-15 11:36       ` Laurent Pinchart
2025-08-15 12:33         ` Sakari Ailus
2025-08-19  2:44           ` Laurent Pinchart [this message]
2025-09-02 12:28             ` Isaac Scott
2025-09-02 12:38               ` Laurent Pinchart
2025-09-02 12:48                 ` Sakari Ailus

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20250819024413.GJ5862@pendragon.ideasonboard.com \
    --to=laurent.pinchart@ideasonboard.com \
    --cc=festevam@gmail.com \
    --cc=imx@lists.linux.dev \
    --cc=isaac.scott@ideasonboard.com \
    --cc=kernel@pengutronix.de \
    --cc=kernel@puri.sm \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-media@vger.kernel.org \
    --cc=martink@posteo.de \
    --cc=mchehab@kernel.org \
    --cc=rmfrfs@gmail.com \
    --cc=s.hauer@pengutronix.de \
    --cc=sakari.ailus@iki.fi \
    --cc=shawnguo@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox