On Mon, Nov 04, 2019 at 12:44:53PM +0100, Lucas Stach wrote:
> F54 is IRQ capable, even if it is not used in the current driver
> implementation. The common driver code in rmi_create_function_irq always
> installs a irq handler for functions that are IRQ capable. Without a
> assigned attention handler, this means a NULL pointer being passed as
> the nested IRQ handler. This seems to work with some architecture
> implementations, but crashes on others like ARM64.
>
> Don't rely on implementation defined behavior and actually install
> a proper attention handler.
Instead of supplying dummy IRQ handler, can't we simply disable relevant
interrupts bits?
>
> Fixes: 24d28e4f1271 ("Input: synaptics-rmi4 - convert irq distribution
> to irq_domain")
> Signed-off-by: Lucas Stach <l.stach@xxxxxxxxxxxxxx>
> ---
> drivers/input/rmi4/rmi_f54.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/drivers/input/rmi4/rmi_f54.c b/drivers/input/rmi4/rmi_f54.c
> index 4841354af0d7..22390e89c680 100644
> --- a/drivers/input/rmi4/rmi_f54.c
> +++ b/drivers/input/rmi4/rmi_f54.c
> @@ -732,6 +732,11 @@ static void rmi_f54_remove(struct rmi_function *fn)
> v4l2_device_unregister(&f54->v4l2);
> }
>
> +static irqreturn_t rmi_f54_attention(int irq, void *ctx)
> +{
> + return IRQ_HANDLED;
> +}
> +
> struct rmi_function_handler rmi_f54_handler = {
> .driver = {
> .name = F54_NAME,
> @@ -740,4 +745,5 @@ struct rmi_function_handler rmi_f54_handler = {
> .probe = rmi_f54_probe,
> .config = rmi_f54_config,
> .remove = rmi_f54_remove,
> + .attention = rmi_f54_attention,
> };
> --
> 2.20.1
>
Thanks.
--
Dmitry
