On Thu, May 25, 2023 at 10:01 PM Hans de Goede <hdegoede@xxxxxxxxxx> wrote:
>
> Add support for using v4l2-async sensor registration.
>
> This has been tested with both the gc0310 and the ov2680 sensor drivers.
>
> Drivers must add the ACPI HIDs they match on to the supported_sensors[]
> array in the same commit as that they are converted to
> v4l2_async_register_subdev_sensor().
>
> Sensor drivers also must check they have a fwnode graph endpoint and return
> -EPROBE_DEFER from probe() if there is no endpoint yet. This guarantees
> that the GPIO mappingss are in place before the driver tries to get GPIOs.
mappings
> For now it also is still possible to use the old atomisp_gmin_platform
> based sensor drivers. This is mainly intended for testing while moving
> other sensor drivers over to runtime-pm + v4l2-async.
...
> +struct acpi_device;
> struct atomisp_device;
> -struct v4l2_device;
> struct atomisp_sub_device;
> +struct v4l2_device;
I would group atomisp* separately
struct acpi_device;
struct v4l2_device;
struct atomisp_device;
struct atomisp_sub_device;
...
> +struct atomisp_csi2_bridge {
> + char csi2_node_name[14];
> + struct software_node csi2_node;
Wondering if swapping these two saves some code (due to potential use
of container_of() for the node member).
> + u32 data_lanes[CSI2_MAX_LANES];
> + unsigned int n_sensors;
> + struct atomisp_csi2_sensor sensors[ATOMISP_CAMERA_NR_PORTS];
> +};
...
> +static char *gmin_cfg_get_dsm(struct acpi_device *adev, const char *key)
> +{
> + union acpi_object *obj, *key_el, *val_el;
> + char *val = NULL;
> + int i;
> +
> + obj = acpi_evaluate_dsm_typed(adev->handle, &atomisp_dsm_guid, 0, 0,
> + NULL, ACPI_TYPE_PACKAGE);
> + if (!obj)
> + return NULL;
> +
> + for (i = 0; i < obj->package.count - 1; i += 2) {
> + key_el = &obj->package.elements[i + 0];
> + val_el = &obj->package.elements[i + 1];
> +
> + if (key_el->type != ACPI_TYPE_STRING || val_el->type != ACPI_TYPE_STRING)
> + break;
> +
> + if (!strcmp(key_el->string.pointer, key)) {
> + val = kstrdup(val_el->string.pointer, GFP_KERNEL);
> + dev_info(&adev->dev, "Using DSM entry %s=%s\n", key, val);
Do we really want to have "(null)" to be printed in case of the
kstrdup() failure?
Also this code may become a honeypot for all possible static analyzers
and even if we don't care about NULL it may become noisy activity.
Besides that since we have a handle, wouldn't it be better to use
acpi_handle_info() here?
> + break;
> + }
> + }
> +
> + ACPI_FREE(obj);
> + return val;
> +}
...
> + dev_info(&adev->dev, "Using DMI entry %s=%s\n", key, gv->val);
acpi_handle_info() ?
Note, I'm fine with dev_info() in both cases, just asking.
...
> + status = acpi_evaluate_object_typed(adev->handle, "_PR0", NULL, &buffer, ACPI_TYPE_PACKAGE);
> + if (!ACPI_SUCCESS(status))
ACPI_FAILURE()
> + return -ENOENT;
...
> + /*
> + * Get pmc-clock number from ACPI _PR0 method and compare this to
PMC ?
> + * the CsiPort 1 pmc-clock used in the CHT/BYT reference designs.
Ditto.
> + */
...
> + obj = acpi_evaluate_dsm_typed(adev->handle, &intel_sensor_module_guid,
> + 0x00, 0x01, NULL, ACPI_TYPE_STRING);
0x01 here...
> + if (obj) {
> + dev_info(&adev->dev, "Sensor module id: '%s'\n", obj->string.pointer);
> + ACPI_FREE(obj);
> + }
> +
> + /*
> + * First get the GPIO-settings count and then get count GPIO-settings
> + * values. Note the order of these may differ from the order in which
> + * the GPIOs are listed on the ACPI resources! So we first store them all
> + * and then enumerate the ACPI resources and match them up by pin number.
> + */
> + obj = acpi_evaluate_dsm_typed(adev->handle,
> + &intel_sensor_gpio_info_guid, 0x00, 1,
...and 1 here. Wouldn't it make sense to be consistent and use either
hex or decimal (looking into below code decimal looks more plausible)
in both cases?
> + NULL, ACPI_TYPE_INTEGER);
> + if (!obj)
> + return dev_err_probe(&adev->dev, -EIO, "No _DSM entry for GPIO pin count\n");
...
> + /* Since we match up by pin-number the pin-numbers must be unique */
> + for (i = 0; i < data.settings_count; i++) {
> + for (j = i + 1; j < data.settings_count; j++) {
> + if (INTEL_GPIO_DSM_PIN(data.settings[i]) !=
> + INTEL_GPIO_DSM_PIN(data.settings[j]))
> + continue;
Wondering if we can have pure pin numbers in some (bit)array, in that
case the uniqueness can be achieved by the test_bit() call in O(1).
> + return dev_err_probe(&adev->dev, -EIO, "Duplicate pin number %lu\n",
> + INTEL_GPIO_DSM_PIN(data.settings[i]));
> + }
> + }
...
> + for_each_acpi_dev_match(adev, cfg->hid, NULL, -1) {
> + if (!adev->status.enabled)
> + continue;
> +
> + if (bridge->n_sensors >= ATOMISP_CAMERA_NR_PORTS) {
> + dev_err(isp->dev, "Exceeded available CSI2 ports\n");
> + ret = -EINVAL;
EOVERFLOW ? EEXIST? ENOMEM
(EINVAL is fine, but to me it's too much use of the same code in the kernel)
> + goto err_put_adev;
> + }
> + }
...
> + /*
> + * This function is intended to run only once and then leave
> + * the created nodes attached even after a rmmod, therefor:
Some spellcheckers want "therefore" here.
> + * 1. The bridge memory is leaked deliberately on success
> + * 2. If a secondary fwnode is already set exit early.
> + */
--
With Best Regards,
Andy Shevchenko
