Hi Jacopo,
On Thu, Jul 06, 2023 at 09:17:54AM +0200, Jacopo Mondi wrote:
> Hi Sakari
>
> On Thu, Jul 06, 2023 at 06:43:54AM +0000, Sakari Ailus wrote:
> > Hi Jacopo,
> >
> > On Thu, Jul 06, 2023 at 08:37:24AM +0200, Jacopo Mondi wrote:
> > > Hi Sakari, Hans
> > >
> > > On Wed, Jul 05, 2023 at 01:45:30PM +0000, Sakari Ailus wrote:
> > > > Hi Hans,
> > > >
> > > > On Thu, May 25, 2023 at 09:00:58PM +0200, Hans de Goede wrote:
> > > > > Switch the atomisp-gc0310 driver to v4l2 async device registration.
> > > > >
> > > > > After this change this driver no longer depends on
> > > > > atomisp_gmin_platform and all atomisp-isms are gone.
> > > > >
> > > > > Signed-off-by: Hans de Goede <hdegoede@xxxxxxxxxx>
> > > > > ---
> > > > > Changes in v2:
> > > > > - Drop v4l2_get_acpi_sensor_info() call in this patch
> > > > > - Wait for fwnode graph endpoint so that the bridge's ACPI
> > > > > parsing gets a chance to register the GPIO mappings
> > > > > before probing the sensor
> > > > > - Switch to endpoint matching
> > > > > ---
> > > > > .../media/atomisp/i2c/atomisp-gc0310.c | 29 ++++++++++++-------
> > > > > .../media/atomisp/pci/atomisp_csi2_bridge.c | 2 ++
> > > > > 2 files changed, 20 insertions(+), 11 deletions(-)
> > > > >
> > > > > diff --git a/drivers/staging/media/atomisp/i2c/atomisp-gc0310.c b/drivers/staging/media/atomisp/i2c/atomisp-gc0310.c
> > > > > index 1829ba419e3e..9a11793f34f7 100644
> > > > > --- a/drivers/staging/media/atomisp/i2c/atomisp-gc0310.c
> > > > > +++ b/drivers/staging/media/atomisp/i2c/atomisp-gc0310.c
> > > > > @@ -29,8 +29,6 @@
> > > > > #include <media/v4l2-ctrls.h>
> > > > > #include <media/v4l2-device.h>
> > > > >
> > > > > -#include "../include/linux/atomisp_gmin_platform.h"
> > > > > -
> > > > > #define GC0310_NATIVE_WIDTH 656
> > > > > #define GC0310_NATIVE_HEIGHT 496
> > > > >
> > > > > @@ -85,6 +83,7 @@ struct gc0310_device {
> > > > > struct mutex input_lock;
> > > > > bool is_streaming;
> > > > >
> > > > > + struct fwnode_handle *ep_fwnode;
> > > > > struct gpio_desc *reset;
> > > > > struct gpio_desc *powerdown;
> > > > >
> > > > > @@ -596,11 +595,11 @@ static void gc0310_remove(struct i2c_client *client)
> > > > >
> > > > > dev_dbg(&client->dev, "gc0310_remove...\n");
> > > > >
> > > > > - atomisp_unregister_subdev(sd);
> > > > > - v4l2_device_unregister_subdev(sd);
> > > > > + v4l2_async_unregister_subdev(sd);
> > > > > media_entity_cleanup(&dev->sd.entity);
> > > > > v4l2_ctrl_handler_free(&dev->ctrls.handler);
> > > > > mutex_destroy(&dev->input_lock);
> > > > > + fwnode_handle_put(dev->ep_fwnode);
> > > > > pm_runtime_disable(&client->dev);
> > > > > }
> > > > >
> > > > > @@ -613,19 +612,27 @@ static int gc0310_probe(struct i2c_client *client)
> > > > > if (!dev)
> > > > > return -ENOMEM;
> > > > >
> > > > > - ret = v4l2_get_acpi_sensor_info(&client->dev, NULL);
> > > > > - if (ret)
> > > > > - return ret;
> > > > > + /*
> > > > > + * Sometimes the fwnode graph is initialized by the bridge driver.
> > > > > + * Bridge drivers doing this may also add GPIO mappings, wait for this.
> > > > > + */
> > > > > + dev->ep_fwnode = fwnode_graph_get_next_endpoint(dev_fwnode(&client->dev), NULL);
> > > > > + if (!dev->ep_fwnode)
> > > > > + return dev_err_probe(&client->dev, -EPROBE_DEFER, "waiting for fwnode graph endpoint\n");
> > > > >
> > > > > dev->reset = devm_gpiod_get(&client->dev, "reset", GPIOD_OUT_HIGH);
> > > > > - if (IS_ERR(dev->reset))
> > > > > + if (IS_ERR(dev->reset)) {
> > > > > + fwnode_handle_put(dev->ep_fwnode);
> > > > > return dev_err_probe(&client->dev, PTR_ERR(dev->reset),
> > > > > "getting reset GPIO\n");
> > > > > + }
> > > > >
> > > > > dev->powerdown = devm_gpiod_get(&client->dev, "powerdown", GPIOD_OUT_HIGH);
> > > > > - if (IS_ERR(dev->powerdown))
> > > > > + if (IS_ERR(dev->powerdown)) {
> > > > > + fwnode_handle_put(dev->ep_fwnode);
> > > > > return dev_err_probe(&client->dev, PTR_ERR(dev->powerdown),
> > > > > "getting powerdown GPIO\n");
> > > > > + }
> > > > >
> > > > > mutex_init(&dev->input_lock);
> > > > > v4l2_i2c_subdev_init(&dev->sd, client, &gc0310_ops);
> > > > > @@ -645,6 +652,7 @@ static int gc0310_probe(struct i2c_client *client)
> > > > > dev->sd.flags |= V4L2_SUBDEV_FL_HAS_DEVNODE;
> > > > > dev->pad.flags = MEDIA_PAD_FL_SOURCE;
> > > > > dev->sd.entity.function = MEDIA_ENT_F_CAM_SENSOR;
> > > > > + dev->sd.fwnode = dev->ep_fwnode;
> > > >
> > > > Same for this one: leave as-is --- the v4l2_async_register_subdev()
> > > > function will set the device fwnode for this.
> > > >
> > > > >
> > > > > ret = gc0310_init_controls(dev);
> > > > > if (ret) {
> > > > > @@ -658,8 +666,7 @@ static int gc0310_probe(struct i2c_client *client)
> > > > > return ret;
> > > > > }
> > > >
> > > > This driver should (as well as ov2680) check for the link frequencies. This
> > > > is an old sensor so if all users eventually use this via firmware that
> > > > lacks this information, there's little benefit for adding the code. But
> > > > otherwise this is seen as a bug.
> > > >
> > > > <URL:https://hverkuil.home.xs4all.nl/spec/driver-api/camera-sensor.html#handling-clocks
> > > >
> > > > The raw cameras should use pixel rate and blanking controls for configuring
> > > > the frame interval. This one uses s_frame_interval instead, and it may be
> > > > difficult to find the information needed for the pixel rate based API.
> > > >
> > > > Cc Jacopo.
> > >
> > > Thanks
> > >
> > > If you intend to use these devices with libcamera be aware we mandate
> > > support for the following controls
> > >
> > > V4L2_CID_ANALOGUE_GAIN
> > > V4L2_CID_EXPOSURE
> > > V4L2_CID_HBLANK
> > > V4L2_CID_PIXEL_RATE
> > > V4L2_CID_VBLANK
> > >
> > > https://git.libcamera.org/libcamera/libcamera.git/tree/Documentation/sensor_driver_requirements.rst#n20
> > >
> > > Do you think it would be possible to at least expose them read-only ?
> > > This -should- be ok for libcamera. Your s_frame_interval() calls then
> > > need to update the value of the controls.
> >
> > Can libcamera use s_frame_interval or does it just mean the frame rate will
> > remain whatever it was when libcamera started?
>
> Currently if those 'mandatory' controls are not available libcamera
> refuses to detect the sensor at all.
>
> s_frame_interval could be considered for YUV/RGB sensors, but as both
> the gc0310 and the ov2680 are RAW sensors, frame rate control should
> really go through blankings and pixel rate. Read-only values are ok to
I don't disagree, I was just wondering what libcamera does. :-)
> start with, framerate will be fixed but that's ok I guess (also
> because as long as you don't have an IPA and do not support
> controllable durations, there is no way to change it anyway).
>
> Does this sound reasonable for you and Hans ?
Yes.
--
Regards,
Sakari Ailus
