On Tue 11 Jun 09:41 PDT 2019, Georgi Djakov wrote:
> Register a platform device to handle the communication of bus bandwidth
> requests with the remote processor. The interconnect proxy device is part
> of this remote processor (RPM) hardware. Let's create a icc-smd-rpm proxy
> child device to represent the bus throughput functionality that is provided
> by the RPM.
>
> Signed-off-by: Georgi Djakov <georgi.djakov@xxxxxxxxxx>
> ---
>
> v3:
> - New patch.
>
> drivers/soc/qcom/smd-rpm.c | 17 ++++++++++++++++-
> 1 file changed, 16 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/soc/qcom/smd-rpm.c b/drivers/soc/qcom/smd-rpm.c
> index 9956bb2c63f2..a028250737ec 100644
> --- a/drivers/soc/qcom/smd-rpm.c
> +++ b/drivers/soc/qcom/smd-rpm.c
> @@ -27,12 +27,14 @@
> /**
> * struct qcom_smd_rpm - state of the rpm device driver
> * @rpm_channel: reference to the smd channel
> + * @icc: interconnect proxy device
> * @ack: completion for acks
> * @lock: mutual exclusion around the send/complete pair
> * @ack_status: result of the rpm request
> */
> struct qcom_smd_rpm {
> struct rpmsg_endpoint *rpm_channel;
> + struct platform_device *icc;
> struct device *dev;
>
> struct completion ack;
> @@ -201,6 +203,7 @@ static int qcom_smd_rpm_callback(struct rpmsg_device *rpdev,
> static int qcom_smd_rpm_probe(struct rpmsg_device *rpdev)
> {
> struct qcom_smd_rpm *rpm;
> + int ret;
>
> rpm = devm_kzalloc(&rpdev->dev, sizeof(*rpm), GFP_KERNEL);
> if (!rpm)
> @@ -213,11 +216,23 @@ static int qcom_smd_rpm_probe(struct rpmsg_device *rpdev)
> rpm->rpm_channel = rpdev->ept;
> dev_set_drvdata(&rpdev->dev, rpm);
>
> - return of_platform_populate(rpdev->dev.of_node, NULL, NULL, &rpdev->dev);
> + rpm->icc = platform_device_register_data(&rpdev->dev, "icc_smd_rpm", -1,
> + NULL, 0);
> + if (!IS_ERR(rpm->icc))
This will be IS_ERR() only if the struct platform_device couldn't be
allocated or registered, it does not relate to that driver's probe. As
such it makes sense to fail the probe if this failed.
So flip this around and return PTR_ERR() here.
> + platform_set_drvdata(rpm->icc, rpm);
It's possible that the device_register above finds the driver and calls
it (pending initcall ordering etc), in which case the child's drvdata
wouldn't yet be set.
In the other drivers where we do this we have the child to request the
drvdata of its parent, so I think you should do the same here.
Apart from this, this patch looks good!
Regards,
Bjorn
> +
> + ret = of_platform_populate(rpdev->dev.of_node, NULL, NULL, &rpdev->dev);
> + if (ret)
> + platform_device_unregister(rpm->icc);
> +
> + return ret;
> }
>
> static void qcom_smd_rpm_remove(struct rpmsg_device *rpdev)
> {
> + struct qcom_smd_rpm *rpm = dev_get_drvdata(&rpdev->dev);
> +
> + platform_device_unregister(rpm->icc);
> of_platform_depopulate(&rpdev->dev);
> }
>
